Stabilized polypeptide compositions

专利摘要:

公开号:AU2007227292A1
申请号:U2007227292
申请日:2007-03-19
公开日:2007-09-27
发明作者:Lisa J. Croner；Stephen Demarest；Scott Glaser；Alexey Alexandrovich Lugovskoy；Brian Robert Miller；William B. Snyder；Norman Wang；Xiufeng Wu
申请人:Biogen Idec Inc；Biogen Idec MA Inc；
IPC主号:A61K39-395

专利说明:
WO 2007/109254 PCT/US2007/006883 STABILIZED POLYPEPTIDE COMPOSITIONS Related Applications The application claims the benefit of priority to the following U.S. provisional 5 patent applications, each of which is hereby incorporated by reference in its entirety for all purposes: 1) U.S. provisional patent application number 60/783,622, filed on March 17, 2006, entitled "STABILIZED POLYPEPTIDES AND METHODS FOR EVALUATING AND INCREASING THE STABILITY OF SAME"; 10 2) U.S. provisional patent application number 60/812,688, filed on June 9, 2006, entitled "STABILIZED POLYPEPTIDES AND METHODS FOR EVALUATING AND INCREASING THE STABILITY OF SAME"; 3) U.S. provisional patent application number 60/873,502, filed on December 8, 2006, entitled "METHODS FOR EVALUATING AND ENHANCING 15 BIOPHYSICAL PROPERTIES OF POLYPEPTIDES"; and 4) U.S. provisional patent application number 60/873,996, filed on December 8, 2006, entitled "STABILIZED POLYPEPTIDE COMPOSITIONS". The contents of any patents, patent applications, and references cited throughout this specification are hereby incorporated by reference in their entireties. 20 Background of the Invention Protein stability is now recognized as a central issue for the development and scale up of proteins, e.g, antibody molecules. The heavy and light chain variable domain sequences of antibodies are under selective pressure and can vary significantly 25 in sequence from one antibody to another (Wu and Kabat, 1970). There are a large number of functional germline variable heavy (-40; Tomlinson et al., 1992; Tomlinson et al., 1994; Matsuda and Honjo, 1996), variable kappa (-40; Meindl et al., 1989; Cox et al., 1994; Barbie and Lefranc, 1998) and variable lambda (-30; Williams et al., 1996; Kawasaki, et al., 1997) genes embedded within the human genome. The ability of the 30 immune system to create many combinations of heavy chain and light chain variable domain germlines is one mechanism for creating antibody diversity (Edelman, 1959; Frahiek, 1961). Additional diversity is derived by the insertion of a variable connecting peptide region, the J-connecting peptide (plus D-connecting peptide for heavy chains), 1 WO 2007/109254 PCT/US2007/006883 between the variable domains and the constant domains (Leder, 1982). Subsequent to germline antibody selection against a specific antigen, B-cells expressing a single antibody with selected variable domain germlines and (D-)J-connecting peptides undergo the process of hypersomatic mutation within the variable heavy and light chain 5 domains to increase antibody affinity towards the antigen (Leder, 1982; Tomlinson et al., 1996). Hypersomatic insertions or deletions within variable domains are also commonly observed, although at a much lower frequency than hypersomatic mutation (de Wildt et al., 1999). Thus, a mature antibody selected against a particular antigen will have highly unique variable domain sequences which are not necessarily optimized for 10 stability. Poor stability can affect the ability of an antibody or antibody domain to fold properly when expressed in various cellular systems, e.g., in bacterial or mammalian expression systems. Misfolding or poor stability can also result in fractional populations of non-functional material (Martsev et al., 1998) and/or antibodies with the tendency to 15 form large soluble aggregates which are potentially dangerous or immunogenic when used therapeutically. Other stability problems include impeded refolding, degradation, impaired avidity, aggregation, or loss of activity following storage. Stability problems are not limited to naturally occurring antibodies, but may occur in engineered antigen binding molecules as well, such as recombinant antibody 20 libraries (Hoogenboom, 2005), antibody fragments (Holt et al., 2003; Todorovska et al., 2001; Worn and Pl ickthun, 2001; Reiter and Pastan, 1996) and engineered or humanized therapeutic antibodies for manufacturing and clinical purposes (Ewert et al., 2004; Carter and Merchant, 1997). In addition, multivalent forms of antibodies, such as bispecific molecules may have stability problems. While bispecific antibodies are 25 desirable because of their therapeutic utility in humans, their production is unpredictable. For example, bispecific antibodies may result in substantial decrease in thermal stability, product yield, or in the formation of low molecular weight aggregates. Unnatural changes to the VH and VL domains may also arise when humanizing rodent antibodies. Humanization is generally performed by grafting rodent 30 complementarity determining regions (CDRs) onto the most similar human germline variable domain (Hurle and Gross, 1994). The germlines chosen for humanization may not be used with great frequency by the immune system and changes within the human frameworks are often necessary to achieve adequate antigen binding. 2 WO 2007/109254 PCT/US2007/006883 Unstable proteins suffer from many problems, including one or more of: unsuitability for scale-up production in bioreactors (e.g., because of low yield, significant levels of unwanted byproducts such as unassembled product, and/or aggregated material), difficulties in protein purification, and unsuitability for pharmaceutical 5 preparation and use (e.g., owing to significant levels of breakdown product, poor product quality, and/or unfavorable pharmacokinetic properties). Instability within the variable domains of antibodies, Fabs, Fvs, or single-chain Fvs (scFvs) may result in many of these problems. scFv constructs in particular have demonstrated problems with stability, solubility, expression, aggregation, breakdown products, and overall manufacturability 10 in both bacterial and in mammalian expression systems (see Worn and Pluckthun, 200 1). Additionally, incorporation of scFv molecules into otherwise stable recombinant antibody products often imparts these generally undesirable traits to the new recombinant design. Accordingly, there is a need for improved methods of predicting the stability of 15 proteins, such as antibodies, and for improved, stable antigen binding molecules that are suitable for scalable production. There is also a need for improved methods that allow for scalable production of a population of stable antigen binding molecules that are suitable for pharmaceutical applications. 20 Summary of the Invention The invention is based, at least in part, on the development of improved polypeptide compositions, e.g. improved binding molecules, comprising or consisting of scFv molecules with improved stability and methods for making the same. In one aspect, the invention pertains to a population of stabilized scFv molecules, 25 wherein the stabilized scFv molecules comprise i) a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH domain and an amino acid in the VL domain, and wherein the VH and VL domains of the scFv molecules have Tm values of greater than 55*C; or 30 ii) a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain, and wherein scFv molecules have a T 50 of greater than 49 0 C. 3 WO 2007/109254 PCT/US2007/006883 In another aspect, the invention pertains to stabilized scFv molecule wherein the stabilized scFv molecule comprises a. a scFv linker having the amino acid sequence (Gly 4 Ser), interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide 5 bond between VH amino acid 44 and VL amino acid 100; ii) a VH domain and a VL domain wherein the molecule has at least one substitution selected from a group consisting of: a) substitution of an amino acid at Kabat position 13 of VH; b) substitution of an amino acid at Kabat position 16 of VH; 10 c) substitution of an amino acid at Kabat position 46 of VL; d) substitution of an amino acid at Kabat position 49 of VL; e) substitution of an amino acid at Kabat position 50 of VL; f) substitution of amino acids at Kabat positions 49 and 50 of VL; g) substitution of amino acid at Kabat position 101 of VH; 15 h) substitution of amino acid at Kabat position 20 of VH; i) substitution of amino acid at Kabat position 48 of VH; j) substitution of amino acid at Kabat position 3 of VL; k) substitution of amino acid at Kabat position 55 of VH; 1) substitution of amino acid at Kabat position 67 of VH; 20 m) substitution of amino acid at Kabat position 6 of VH; n) substitution of amino acid at Kabat position 32 of VH; o) substitution of amino acid at Kabat position 49 of VH; p) substitution of amino acid at Kabat position 43 of VH; q) substitution of amino acid at Kabat position 72 of VH; 25 r) substitution of amino acid at Kabat position 79 of VH; s) substitution of amino acid at Kabat position 50 of VL; t) substitution of amino acid at Kabat position 75 of VL; u) substitution of amino acid at Kabat position 80 of VH; v) substitution of amino acid at Kabat position 83 of VH; or 30 iii) a VH domain and a VL domain with a stabilized interface between VH and VL of the scFv wherein the molecule has at least one substitution at an amino acid position directly within the interface and/or at an amino acid position that scaffolds the interaction between VH and VL, wherein the at least one amino acid position is selected from a 4 WO 2007/109254 PCT/US2007/006883 group consisting of: 47H, 37H, 45H, 46H, 51H, 59H, 109H, 14H, 26H, 27H, 40H, 69H, 103H, 29H, 38H, 44H, 49H, 55H, 63H, 54H, 68H, 72H, 74H, 79H, 82bH, 8H, 32H, 39H, 41H, 62H, 85H, 91H, 24H, 60H, 37L, 36L, 44L, 59L, 57L, 64L, 46L, 48L, 63L, 67L, 68L, 39L, 45L, 47L, 54L, 56L, 79L, 85L, 98L, 58L, 74L, 77L, 83L, 89L, and 5 104L, according to Kabat numbering, and wherein the stabilized scFv molecule has a T 5 o greater than that of a scFv molecule lacking the substitution. In another aspect, the invention pertains to a stabilized scFv molecule comprising a (Gly 4 Ser) 4 scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and 10 an amino acid in the VL domain. In another embodiment, the invention pertains to a stabilized scFv molecule comprising a VH domain and a VL domain wherein the molecule has at least one set of substitutions selected from the group consisting of: a) a substitution of an amino acid at Kabat position 16 of VH, e.g., with a 15 glutamate or glutamine and a substitution of an amino acid at Kabat position 46 of VL, e.g., with a lysine; b) a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine; a substitution of an amino acid at Kabat position 46 of VL, e.g., with a lysine; and a substitution of an amino acid at Kabat position 55 of VH, e.g., with 20 a glycine; and c) a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine; a substitution of an amino acid at Kabat position 46 of VL, e.g., with a lysine; a substitution of an amino acid at Kabat position 55 of VH, e.g., with a glycine; and substitution of an amino acid at Kabat position 101 of VH, e.g., with an 25 aspartic acid. In another embodiment, the stabilized scFv molecule has a stability equivalent to that of a conventional Fab fragment under conditions of thermal challenge. In one embodiment, the stability is assessed by measuring the ability to bind to a target molecule. 30 In one embodiment, the scFv molecule comprises an amino acid sequence encoded by a nucleotide sequence selected from the group consisting of SEQ ID NOs:9, 14, 16, and 18. In one embodiment, the scFv molecule comprises an amino acid sequence 5 WO 2007/109254 PCT/US2007/006883 selected from the group consisting of SEQ ID NOs: 10, 15, 17, and 19. In another embodiment, the invention pertains to a fusion protein comprising a stabilized scFv molecule. In one embodiment, the fusion protein comprises at least two antigen binding 5 sites. In another aspect, the invention pertains to a population of stabilized multivalent antigen binding molecules, wherein the stabilized binding molecules comprise at-least one stabilized scFv molecule comprising a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond 10 between an amino acid in the VH and an amino acid in the VL domain, and wherein the population of stabilized binding molecules comprise monomeric, soluble proteins of which not more than 10% is present in aggregated form. In one embodiment, the stabilized multivalent antigen binding molecules are expressed in CHO or NSO cells. 15 In another aspect, the invention pertains to a population of stabilized multivalent antigen binding molecules, wherein each multivalent antigen binding molecule comprises i) at least one scFv molecule comprising a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond 20 between an amino acid in the VH and an amino acid in the VL domain, and wherein the VH and VL domains of the at least one scFv molecule have a Tm of greater than 55*C; or ii) at least one scFv molecule comprising a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain, and wherein the at 25 least one scFv molecule of the multivalent antigen binding molecule has a T 50 of greater than 49 0 C. In another embodiment, the invention pertains to a stabilized multivalent antigen binding molecule comprising i) at least one stabilized scFv molecule, wherein the stabilized scFv molecule comprises 30 a (Gly 4 Ser)n scFv linker interposed between a VH domain and a VL domain and wherein the VH and VL domains are linked by a disulfide bond; ii) at least one stabilized scFv molecule comprising a scFv linker having the amino acid sequence (Gly 4 Ser), interposed between a VH domain and a VL domain, wherein the VH 6 WO 2007/109254 PCT/US2007/006883 and VL domains are linked by a disulfide bond between VH amino acid 44 and VL amino acid 100; iii) at least one stabilized scFv molecule, wherein the stabilized scFv molecule comprises at least one substitution selected from a group consisting of: 5 a) substitution of an amino acid at Kabat position 13 of VH; b) substitution of an amino acid at Kabat position 16 of VH; c) substitution of an amino acid at Kabat position 46 of VL; d) substitution of an amino acid at Kabat position 49 of VL; e) substitution of an amino acid at Kabat position 50 of VL; 10 f) substitution of amino acids at Kabat positions 49 and 50 of VL; g) substitution of amino acid at Kabat position 101 of VH; h) substitution of amino acid at Kabat position 20 of VH; i) substitution of amino acid at Kabat position 48 of VH; j) substitution of amino acid at Kabat position 3 of VL; 15 k) substitution of amino acid at Kabat position 55 -of VH; 1) substitution of amino acid at Kabat position 67 of VH; m) substitution of amino acid at Kabat position 6 of VH; n) substitution of amino acid at Kabat position 32 of VH; o) substitution of amino acid at Kabat position 49 of VH; 20 p) substitution of amino acid at Kabat position 43 of VH; q) substitution of amino acid at Kabat position 72 of VH; r) substitution of amino acid at Kabat position 79 of VH; s) substitution of amino acid at Kabat position 50 of VL; t) substitution of amino acid at Kabat position 75 of VL; 25 u) substitution of amino acid at Kabat position 80 of VH; v) substitution of amino acid at Kabat position 83 of VH; or iv) a VH domain and a VL domain with a stabilized interface between VH and VL of the scFv wherein the molecule has at least one substitution at an amino acid position directly within the interface and/or at an amino acid position that scaffolds the interaction 30 between VH and VL, wherein the at least one amino acid position is selected from a group consisting of: 47H, 37H, 45H, 46H, 51H, 59H, 109H, 14H, 26H, 27H, 40H, 69H, 103H, 29H, 38H, 44H, 49H, 55H, 63H, 54H, 68H, 72H, 74H, 79H, 82bH, 8H, 32H, 39H, 41H, 62H, 85H, 91H, 24H, 60H, 37L, 36L, 44L, 59L, 57L, 64L, 46L, 48L, 63L, 7 WO 2007/109254 PCT/US2007/006883 67L, 68L, 39L, 45L, 47L, 54L, 56L, 79L, 85L, 98L, 58L, 74L, 77L, 83L, 89L, and 104L, according to Kabat numbering, In another aspect, the invention pertains to a stabilized multivalent antigen binding molecule comprising at least one stabilized scFv molecule comprising a (Gly 4 5 Ser) 4 scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain. ii) at least one stabilized scFv molecule comprising a scFv linker having the amino acid sequence (Gly 4 Ser)n interposed between a VH domain and a VL domain, 10 wherein the VH and VL domains are linked by a disulfide bond between VH amino acid 44 and VL amino acid 100; or iii) at least one stabilized scFv molecule, wherein the stabilized scFv molecule comprises at least one substitution selected from a group consisting of: a) substitution of an amino acid (e.g., glutamine) at Kabat position 3 of VL, e.g., 15 with an alanine, a serine, a valine, an aspartic acid, or a glycine; b) substitution of an amino acid (e.g., serine) at Kabat position 46 of VL, e.g., with leucine; c) substitution of an amino acid (e.g., serine) at Kabat position 49 of VL, e.g., with tyrosine or seine; 20 d) substitution of an amino acid (e.g., seine or valine) at Kabat position 50 of VL, e.g., with seine, threonine, and arginine, aspartic acid, glycine, or lysine; e) substitution of amino acids (e.g., seine) at Kabat position 49 and (e.g., serine) at Kabat position 50 of VL, respectively with tyrosine and serine; tyrosine and threonine; tyrosine and arginine; tyrosine and glysine; serine and arginine; or 25 serine and lysine; f) substitution of an amino acid (e.g., valine) at Kabat position 75 of VL, e.g., with isoleucine; g) substitution of an amino acid (e.g., proline) at Kabat position 80 of VL, e.g., with serine or glycine; 30 h) substitution of an.amino acid (e.g., phenylalanine) at Kabat position 83 of VL, e.g., with serine, alanine, glycine, or threonine; i) substitution of an amino acid (e.g., glutamic acid) at Kabat position 6 of VH, e.g., with glutamine; 8 WO 2007/109254 PCT/US2007/006883 j) substitution of an amino acid (e.g., lysine) at Kabat position 13 of VH, e.g., with glutamate; k) substitution of an amino acid (e.g., serine) at Kabat position 16 of VH, e.g., with glutamate or glutamine; 5 1) substitution of an amino acid (e.g., valine) at Kabat position 20 of VH, e.g., with an isoleucine; m) substitution of an amino acid (e.g., asparagine) at Kabat position 32 of VH, e.g., with serine; n) substitution of an amino acid (e.g., glutamine) at Kabat position 43 of VH, e.g, 10 with lysine or arginine; o) substitution of an amino acid (e.g., methionine) at Kabat position 48 of VH, e.g., with an isoleucine or a glycine; p) substitution of an amino acid (e.g., serine) at Kabat position 49 of VH, e.g, with glycine or alanine; 15 q) substitution of an amino acid (e.g., valine) at Kabat position 55 of VH, e.g., with a glycine; r) substitution of an amino acid (e.g., valine) at Kabat position 67 of VH, e.g., with an isoleucine or a leucine; s) substitution of an amino acid (e.g., glutamic acid) at Kabat position 72 of VH, 20 e.g., with aspartate or asparagine; t) substitution of an amino acid (e.g., phenylalanine) at Kabat position 79 of VH, e.g., with serine, valine, or tyrosine; and u) substitution of an amino acid (e.g., proline) at Kabat position 101 of VH, e.g., with an aspartic acid. 25 In another aspect, the invention pertains to a multivalent antigen binding molecule comprising at least one stabilized scFv molecule, wherein the stabilized scFv molecule comprises at least one set of substitutions selected from the group consisting of: a) a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine and a substitution of an amino acid at Kabat position 46 of VL, 30 e.g., with a lysine; b) a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine; a substitution of an amino acid at Kabat position 46 of VL, e.g., with a lysine; and a substitution of an amino acid at Kabat position 55 of VH, e.g., with 9 WO 2007/109254 PCT/US2007/006883 a glycine; and c) a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine; a substitution of an amino acid at Kabat position 46 of VL, e.g., with a lysine; a substitution of an amino acid at Kabat position 55 of VH, e.g., with a 5 glycine; and substitution of an amino acid at Kabat position 101 of VH, e.g., with an aspartic acid. In one embodiment, at least one stabilized scFv molecule is genetically fused to an antibody. In another embodiment, two stabilized scFv molecules are genetically fused to an antibody. 10 In one embodiment, at least one stabilized scFv molecule is genetically fused to the carboxy terminus of the light or heavy chain of the antibody. In one embodiment, at least one stabilized scFv molecule is genetically fused to the amino terminus of a light chain or heavy chain of the antibody. In one embodiment, a binding molecule of the invention comprises at least one 15 binding site that binds to a molecule preferentially expressed on cancer cells In one embodiment, the VH domain of a scFv molecule of the invention is derived from the BHAIO antibody. In one embodiment, the VL domain of a scFv molecule of the invention is derived from the BHA 10 antibody. 20 In one embodiment a binding molecule of the invention comprises at least one binding site that is specific for a molecule selected from the group consisting of HERI, HER3, CD80, CD86, PD-1, CTLA4, B7-H4, RON, CD200, CD4, BAF R, EGFR, IGFR, VEGFR, a member of the TNF family of receptors, a Tie receptor, MET, IGF1, IGF2, TNF, a TNF ligand, IL-6, TWEAK, Fn14, CD20, CD23, CRIPTO, HGF, alpha4betal 25 integrin, alpha5betal integrin, alpha6beta4 integrin,, and alphaVbeta6 integrin. In one embodiment, a binding molecule of the invention comprises at least one binding site that binds to a molecule involved in modulating immune responses. In one embodiment, the molecule is an Fc receptor. In one embodiment, a binding molecule of the invention comprises at least one 30 binding site that binds to a molecule involved in modulating angiogenesis. In one embodiment, a binding molecule of the invention binds to VEGF or angiopoitin. In one embodiment, a binding molecule of the invention comprises at least one binding site that binds to a neurological target. 10 WO 2007/109254 PCT/US2007/006883 In one embodiment, a binding molecule of the invention is multispecific. In one embodiment, a binding molecule of the invention is bispecific. In another embodiment, a binding molecule of the invention binds to two members of the TNF receptor family. In one embodiment, a binding molecule of the 5 invention binds to LTpR and Trail R2. In one embodiment, a binding molecule of the invention comprises a BHA1O scFv molecule genetically fused a 14A2 antibody. In one embodiment, a binding molecule of the invention comprises a heavy chain having a sequence selected from the group consisting of SEQ ID NO:33, SEQ ID NO:35, SEQ ID NO:37, SEQ ID NO:47, SEQ ID NO:49, SEQ ID NO:51, SEQ ID 10 NO:53, and SEQ ID NO:55. In one embodiment, a binding molecule of the invention comprises a nucleotide sequence encoding a polypeptide of the molecule of any one of claims 4-8, 11, 12, and 18-22. In one embodiment, the invention pertains to a nucleic acid molecule comprising 15 a nucleotide sequence which encodes the stabilized scFv molecule or binding molecule comprising a stabilized scFv molecule of the invention. In one embodiment, the nucleic acid molecule is in a vector. In another embodiment, the invention pertains to a host cell comprising such a vector. In one embodiment, the host cell is a mammalian host cell, e.g., a CHO cell or an 20 NSO cell. In one embodiment, the invention pertains to a population of binding molecules comprising scFv molecules, the population having at least 10% fewer aggregates relative to a host cell which expresses a population of binding molecules comprising conventional scFv molecules. 25 In one aspect, the invention pertains to a method of producing a stabilized binding molecule, comprising culturing a host cell under conditions such that the stabilized binding molecule is produced. In one embodiment, at least 10 mg of the stabilized binding molecule is produced for every liter of the host cell culture medium and wherein not more than 10% of the 30 binding molecule is present in aggregate form. In one aspect, the invention pertains method of treating a subject that would benefit from treatment with a binding molecule of the invention comprising administering the binding molecule to the subject such that treatment occurs. 11 WO 2007/109254 PCT/US2007/006883 In one embodiment, the subject is suffering from a disease or disorder selected from the group consisting of cancer, an autoimmune disease or disorder, and a neurological disease or disorder. In one aspect, the invention pertains to a method of stabilizing an scFv molecule 5 comprising genetically fusing a VH domain and a VL domain using a (Gly 4 Ser), scFv linker and engineering the VH domain to comprise a cysteine at amino acid 44 and engineering the VL domain to comprise a cysteine at amino acid 100 to form a stabilized scFv molecule. In one aspect, the invention pertains to a method of making a stabilized 10 multivalent antibody comprising a stabilized scFv molecule, the method comprising genetically fusing a stabilized scFv molecule to the amino terminus or the carboxy terminus of a light or heavy chain of an antibody molecule. In one embodiment, the invention pertains to a method of making a stabilized scFv molecule comprising substituting at least one amino acid in the VH/VL interface of 15 the scFv molecule and/or at least one amino acid that scaffolds the VH/VL interface of the scFv molecule substitution(s) simultaneously improve the thermal stability of both the VH and VL domains of the scFv molecule as compared to a conventional scFv molecule. In yet another embodiment, the invention pertains to a method of making a 20 stabilized scFv molecule comprising substituting at least one amino acid in the VH domain or VL domain that covaries with two or more amino acids at the interface between the VH and VL domains. In yet another embodiment, the invention pertains to a method of making a stabilized fusion protein comprising an scFv molecule comprising substituting at least 25 one amino acid in the VH/VL interface of the scFv molecule and/or at least one amino acid in the VH/VL scaffold of the scFv molecule and genetically fusing the scFv molecule to a polypeptide to thereby make a stabilized fusion protein. In yet another embodiment, the invention pertains to a method of improving the stability of a multivalent molecule comprising at least one scFv molecule, the method 30 comprising introducing at least one stabilizing mutation into the at least one scFv .molecule to thereby improve the stability of the multivalent molecule. In yet another embodiment, the invention pertains to a method of improving the stability of an scFv molecule comprising introducing at least one stabilizing mutation 12 WO 2007/109254 PCT/US2007/006883 into the scFv molecule to thereby improve the stability of an scFv molecule. In yet another aspect, the invention pertains to a method for large scale manufacture of a stabilized fusion protein, the method comprising: (a) conducting a thermal stability test of an scFv molecule to determine the 5 thermal stability of a candidate scFv molecule; (b) comparing the thermal stability of the candidate scFv molecule with a suitable control to identify stabilized scFv molecules having increased thermal stability relative to the control, (c) selecting the stabilized scFv molecules identified in step (b); 10 (d) genetically fusing at least one stabilized scFv molecule to protein to form a stabilized fusion protein; (e) transfecting a mammalian host cell with a nucleic acid molecule encoding the stabilized fusion protein, (g) culturing the host cell of step (f) under conditions such that the stabilized 15 fusion protein is expressed; wherein the stabilized fusion protein is expressed with not more than 10% protein aggregation when grown in 1OL or more of culture medium. In yet another aspect, the invention pertains to a method of determining the thermal stability of a protein fold from an immunoglobulin (Ig) superfamily polypeptide 20 comprising a candidate sequence of amino acids, the method comprising the steps of: a) providing an alignment of a curated reference set of sequences corresponding to an Ig fold of the polypeptide; b) calculating covariation between amino acid residues of the sequences of the alignment to generate covariation data; 25 c) determining a sequence position specific covariation score for amino acid positions within the candidate sequence from covariations within the covariation data; and d) storing or outputting the amino acid position specific sequence covariation score as a measure of the stability of the polypeptide. 30 In one embodiment, the method comprises repeating at least steps c) and d) for a plurality of immunoglobulin (Ig) superfamily polypeptides. In yet another embodiment, the method further comprises the step of selecting from amongst the plurality of immunoglobulin (Ig) superfamily polypeptides an Ig superfamily polypeptide for 13 WO 2007/109254 PCT/US2007/006883 production, based on the sequence position specific covariation scores of the candidate sequences. In another embodiment, the method further comprises the step of producing the selected Ig superfamily polypeptide and preparing it for therapeutic use. 5 In another embodiment, the Ig superfamily polypeptide derived from a protein selected from the group consisting of an immunoglobulin or antigen-binding fragment thereof, an immunoglobulin receptor, a cell adhesion protein, an integrin, an allergen; a T-cell receptor, and a major histocompatibility complex (MHC). In one embodiment, the Ig superfamily polypeptide is selected from the group 10 consisting of a heavy chain variable region (VH), a light chain variable region (VL), and a single chain antibody (scFv). In another embodiment, the Ig fold is selected from the group consisting of a V class fold, an I-class fold, a Cl-class fold, and a C2-class fold. In one embodiment, the curated reference set or the alignment has a diversity of 15 at least 50%. In one embodiment, each sequence of the alignment has less than 90% identity with every other sequence of the alignment. In another embodiment, at least one sequence is from a mammalian species and at least one sequence is from a non-mammalian species. 20 In one embodiment, residues of the candidate sequence are assigned positive position specific covariation scores for satisfying positive covariations found at corresponding positions in the alignment. In one embodiment, the residues of the candidate sequence are assigned negative position specific covariation scores for satisfying negative covariations found at 25 corresponding positions in the alignment. In one embodiment, the positive covariation has a phi association coefficient (CD) of about +0.25 to about +1.0. In one embodiment, the negative covariation has a phi association coefficient (<D) of about -0.25 to about -1.0. 30 In one embodiment, the alignment is selected from the group consisting of a structure-based sequence alignment, a sequence-based sequence alignment, and a structure-based structural alignment. In one embodiment, the position specific covariation score is determined for all possible residue pairs at each residue position of the alignment. 35 In one embodiment, the alignment is obtained by (i) generating a Hidden Markov Model (HMM) from an initial alignment; (ii) acquiring additional sequences with the HMM; and (iii) aligning the additional sequences with the initial alignment. 14 WO 2007/109254 PCT/US2007/006883 In one aspect, the invention pertains to a computer program or computer program product or computer readable medium having a set of instructions executable by a processor which when executed cause the processor to perform a method of the invention. 5 In yet another embodiment, the invention pertains to a computer program or computer program product or computer readable medium suitable for use in an electronic device comprising data corresponding to the alignment of the method of the invention. In yet another embodiment, the invention pertains to a computer program or 10 computer program product or computer readable medium suitable for use in an electronic device comprising the covariation data of a method of the invention. In one embodiment, the invention pertains to a device arranged to carry out the method of the invention. In one aspect, the invention pertains to a system in a computing device, 15 comprising: (i) an initial sequence collection process; (ii) a storage location capable of holding sequence data from the alignment of any one of claims 58-74; (iii) an analysis facility, which programmatically analyses covariation between 20 pairs of residues within the alignment to obtain covariation data; (iv) an output device interfaced with said electronic device, said output device outputting said covariation data. In one embodiment, the invention pertains to a medium in a computing device holding executable steps for carrying out a method comprising any of the steps of the 25 method of the invention. In yet another embodiment, the display of residue usage frequency data comprises a matrix comprising (i) contiguous residue positions corresponding to the polypeptide sequence; and (ii) residue usage frequencies for a residue type at each residue position. 30 In yet another embodiment, the residue usage frequency is symbolically depicted. In one embodiment, all 20 natural amino acids, gaps, and ambiguous residues are represented. In one embodiment, residue use frequency is correlated with symbol size. In one embodiment, residue use frequency is positively correlated with symbol 35 size. 15 WO 2007/109254 PCT/US2007/006883 In one embodiment, (i) the residue positions are displayed in columns; and (ii) the residue usage frequencies are displayed in rows. In another embodiment, covariation among the covariant residues is depicted by a network overlay. 5 In one aspect, the invention pertains to a graphical user interface comprising a graphical display of: (i) a grid layout comprising a collection of residue usage frequency data for a reference set of polypeptide sequences; and (ii) a co variation overlay. 10 In one embodiment, the interface further comprises: (iii) a display of a sequence of interest. In one aspect, the invention pertains to a method for producing an immunoglobulin (Ig) superfamily polypeptide with an improved biophysical property, said method comprising the steps of: 15 a) providing an alignment of a curated reference set of sequences corresponding to an Ig fold of the polypeptide; b) calculating covariation between residues of the sequences of the alignment to identify covarying residues; c) substituting a residue of the polypeptide that fails to satisfy a covariation 20 with the covarying residue found at a corresponding position in the alignment. In another embodiment, the biophysical property is selected from the group consisting of thermal stability, pH unfolding profile, stable removal of glycosylation, solubility, biochemical function, and combinations thereof. In one embodiment, the biochemical function is selected from the group 25 consisting of recognition of a protein target or a chemical moiety, which chemical moiety is selected from the group consisting of phosphate, methyl, acetyl, lipid, detergent, metal ion, halogen, RNA, DNA, oligonucleotide, and oligonucleoside. In one embodiment, the Ig superfamily polypeptide is derived from a protein selected from the group consisting of an antibody or antigen-binding fragment thereof, 30 an immunoglobulin receptor, a cell adhesion protein, an integrin, an allergen, a T-cell receptor, and a major histocompatibility complex (MHC) molecule. In one embodiment, the Ig superfamily polypeptide is selected from the group consisting of a heavy chain variable region (VH), a light chain variable region (VL), and a single chain antibody (sFv). 16 WO 2007/109254 PCT/US2007/006883 In one embodiment, the Ig fold is selected from the group consisting of a V-class fold, an I-class fold, a Cl-class fold, and a C2-class fold. In another embodiment, the Ig superfamily polypeptide is a modified antibody selected from the group consisting of a domain antibody, a humanized antibody, a 5 human antibody, a non-human monoclonal antibody, a chimeric antibody, a bispecific antibody, a scFv-containing antibody, and a domain-deleted antibody. In one embodiment, the covarying residues are part of a structural feature selected from the group consisting of a disulfide bond, a salt bridge, a portion of a ligand binding pocket or surface, and a network of van Der Waals, hydrogen bond, and/or 10 charge-charge interactions. In another embodiment, the invention pertains to a method for producing an antibody or modified antibody with an improved stability, said method comprising the steps of: a) providing a high-resolution structural model of a template antibody that is 15 experimentally validated as stable; b) using the structural model to identify a VH/VL interface and/or interface scaffolding residue in the template antibody; c) using a homology model to identify a corresponding VH/VL interface residue of the antibody or modified antibody that is important for stabilizing the interface; and 20 d) substituting the corresponding VHNL interface residue in the antibody or modified antibody with the interface residue from the template protein. In one embodiment, the interface residue of the template protein buries at least 10 A 2 of surface area in the interface. In one embodiment, the modified antibody is selected from the group consisting 25 of a domain antibody, a humanized antibody, a human antibody, a non-human monoclonal antibody, a chimeric antibody, a bispecific antibody, a scFv-containing antibody, and a domain-deleted antibody. In another embodiment, the invention pertains to a polypeptide produced by the method of the invention. 30 In one embodiment, the polypeptide exhibits an improvement in a biophysical property selected from the group consisting of thermal stability, pH unfolding profile, stable removal of glycosylation, solubility, biochemical function, and combinations thereof. In one embodiment, the biochemical function is ligand binding affinity or 35 specificity. 17 WO 2007/109254 PCT/US2007/006883 In another aspect, the invention pertains to a method for producing a library comprising stabilized scFv molecules, the method comprising: (a) providing a reference set of sequences corresponding to a variable domain sequence of a scFv molecule; 5 (b) identifying an amino acid within the variable domain which is absent or found at low frequencies at a corresponding position in the reference set; and (c) combining this information with covariation data suggesting that modifying the amino acid identified with low frequency in the reference set may satisfy existing covariation restraints within the variable domain; and 10 (d) substituting the amino acid of step (b) with a candidate stabilizing amino acid to thereby produce a library comprising stabilized scFv molecules. In one embodiment, the amino acid of step (b) has a consensus score of less than 0.5. In another embodiment, the amino acid of step (b) is present within less than 10% of the sequences in the reference set. In another embodiment, the amino acid of step (b) is a 15 non-consensus residue. In one embodiment, the candidate stabilizing amino acid is found at a corresponding position within the reference set. In yet another embodiment, the candidate stabilizing amino acid is a consensus amino acid. 20 In one embodiment, the candidate stabilizing amino acid is identified by an analysis of a 3-D structure of the variable region sequence. In another embodiment, the stabilizing amino acid is identified by a side-chain repacking calculation. In yet another aspect, the invention pertains to a method for producing a stabilized scFv molecule, the method comprising: 25 (a) providing a scFv library designed according to the method of claim 56, (b) screening the scFv library with a thermal challenge assay to identify candidate scFv molecules, (c) comparing the thermal stability of a candidate scFv molecule of the scFv library with a suitable control, 30 whereby an increase in thermal stability of a candidate scFv molecule relative to the control identifies the candidate scFv molecule as a stabilized scFv molecule. In another embodiment, the invention pertains to a stabilized scFv molecule identified using a method of the invention or a fusion protein comprising a stabilized 18 WO 2007/109254 PCT/US2007/006883 scFv molecule of the invention. In another aspect, the invention pertains to a method for predicting the stability of a candidate protein comprising an amino acid sequence having candidate domain sequences, the method comprising: 5 (a) providing a reference set of amino acid sequences corresponding to a candidate domain sequence of the candidate protein; (b) determining residue frequencies at individual amino acid positions within the test domain sequence to obtain a consensus score; and (c) using the consensus score to predict the stability of the candidate protein, 10 wherein the consensus score is correlated with the stability of the candidate protein. In another aspect, the invention pertains to a method for predicting the stability of a candidate protein, the method comprising: (a) providing a reference set of sequences corresponding to a test domain 15 sequence of the candidate protein; (b) determining residue frequencies at individual amino acid positions within the test domain sequence to obtain a consensus score; and (c) using the consensus score to predict the stability of the candidate protein, wherein the consensus score is correlated with the stability of the candidate 20 protein. In another embodiment, the invention pertains to a method for predicting the stability of a candidate protein, the method comprising: (a) providing a reference set of sequences corresponding to a test domain sequence of the candidate protein; 25 (b) determining residue frequencies at amino acid positions within the test domain sequence to obtain a consensus score; (c) determining residue frequencies at corresponding amino acid positions within the sequences of the reference set to determine an average consensus score; (d) comparing the consensus score with the average consensus score to 30 determine a sequence score; and (e) using the sequence score to predict the stability of the candidate protein, wherein the consensus score is directly correlated with the stability of the candidate protein. 19 WO 2007/109254 PCT/US2007/006883 In yet another embodiment, the invention pertains to a method for predicting the stability of a candidate antibody or candidate modified antibody, the method comprising: (a) providing a reference set of VH domain sequences corresponding to a test VH domain sequence of the candidate antibody or modified antibody; 5 (b) determining residue frequencies at amino acid positions within the test VH domain sequence to obtain a consensus score; (c) determining residue frequencies at corresponding amino acid positions within the sequences of the reference set to determine an average consensus score; (d) comparing the consensus score with the average consensus score to 10 determine a sequence score; and (e) using the sequence score to predict the stability of the candidate antibody or modified antibody, wherein the consensus score is directly correlated with the stability of the candidate antibody or modified antibody. 15 In one embodiment, the reference set comprises protein sequences from proteins with the same class of protein fold as the candidate protein. In another embodiment, the reference set comprises orthologous sequences. In one embodiment, the reference set comprises human sequences, e.g., human VH sequences, e.g., of the same Kabat class. 20 In one embodiment, the reference set comprises human VH germline sequences. In one embodiment, the test domain sequence is a portion of the domain sequence of the candidate protein. In one embodiment, the consensus score is determined for each position of the test sequence to obtain a total consensus score, wherein the total consensus score is 25 correlated with protein stability. In another embodiment, the total consensus score is calculated using the formula: score = Z ) c,(r) wherein: c;(r) equals the consensus residue frequency at an amino acid position of the 30 consensus sequence, h,(r) equals the test residue frequency of an amino acid position of the test sequence, and 20 WO 2007/109254 PCT/US2007/006883 i equals the number of amino acid positions within the test sequence. In one embodiment, the stability of the protein is determined by comparing the .consensus score of the candidate protein with a consensus score of a suitable control. In another embodiment, the stability of the protein is determined by comparing 5 the consensus score of the candidate protein with the perfect consensus score of the candidate protein. In yet another embodiment, stability is determined by comparing the sequence score of the candidate protein with a suitable control. In another embodiment, stability is determined by comparing the sequence score 10 of the candidate antibody or modified antibody with a suitable control. In one embodiment, the protein is a multi-domain protein, and wherein the target domain sequence is derived from the least stable domain of the multi-domain protein. In still another embodiment, the protein is a modified antibody selected from the group consisting of a camelid antibody, humanized antibody, a human antibody, a non 15 human monoclonal antibody, a chimeric antibody, a bispecific antibody, a scFv containing antibody, and a domain-deleted antibody. In another embodiment, the invention pertains to a method for selecting a candidate protein for expression, comprising determining the stability of the candidate protein using the method of the invention, wherein the candidate protein is selected for 20 expression if its consensus score or sequence score is predictive of high stability. In another embodiment, the invention pertains to a method for selecting a candidate protein for expression, comprising determining the stability of the candidate protein using the method of the invention, wherein the candidate protein is selected for expression if its sequence position specific covariation score is predictive of high 25 stability. In another embodiment, the invention pertains to a method for selecting an acceptor immunoglobulin variable region test sequence for use in the humanization of a donor antibody, the method comprising determining the stability of the candidate sequence using the method of the invention, wherein the candidate sequence is selected 30 if its sequence score is predictive of high stability. In yet another embodiment, the invention pertains to a method for selecting an acceptor immunoglobulin variable region test sequence for use in the humanization of a donor antibody, the method comprising determining the stability of the candidate 21 WO 2007/109254 PCT/US2007/006883 sequence using the method of the invention, wherein the candidate sequence is selected if its position specific covariation score is predictive of high stability. In another embodiment, the test sequence is a human germline sequence. 5 Brief Description of the Drawings Figure IA shows the DNA sequence (SEQ ID NO:3) encoding a conventional BHAl0 scFv construct comprising a (Gly4Ser)3 linker (indicated in bold type). Figure lB shows the amino acid sequence (SEQ ID NO:4) of the conventional BHA10 scFv 10 construct. Figure 2 depicts the results of differential scanning calorimetry (DSC) measurements with purified conventional Fab or scFv fragments of BHA 10. Figure 2A depicts the results of a DSC study comparing the unfolding of conventional Fab and 15 scFv fragments of BHA10. Figure 2B depicts the results of DSC scans of the purified BHA10 scFv at scan rates of 1 'C/min and 2*C/min. Figure 3A depicts the DNA sequence (SEQ ID NO:9) encoding a stabilized VH44/VLIOO disulfide-stabilized BHA10 scFv construct comprising a (Gly4Ser)3 20 linker (indicated in bold type). Figure 3B shows the amino acid sequence (SEQ ID NO: 10) of the VH44NL 100 disulfide-stabilized BHA10 scFv construct. The cysteine residues forming the VH44/VL100 disulfide bond are indicated in bold and italicized type. 25 Figure 4A depicts the DNA sequence (SEQ ID NO: 14) encoding a stabilized BHA 10 scFv comprising a (Gly 4 Ser)4 linker (indicated in bold type). Figure 4B shows the amino acid sequence (SEQ ID NO: 15) of the (Gly 4 Ser) 4 BHA 10 scFv construct. Annotation is the same as in Figure 4A. 30 Figure 5A depicts the DNA sequence (SEQ ID NO:16) encoding a BHA10 scFv comprising a (Gly4Ser)5 linker (indicated in bold type). Figure 5B depicts the amino acid sequence (SEQ ID NO:17) of the (Gly 4 Ser) 5 BHA10 scFv construct. Annotation is the same as in Figure 5A. 22 WO 2007/109254 PCT/US2007/006883 Figure 6 depicts the results of a Western Blot analysis to compare the expression levels of a conventional BHA 10 scFv (lane 1) and the stablilized BHA 10 scFv molecules of the invention (lanes 2-4). Each sample was electrophoresed under both 5 reducing (left panel) and non-reducing (right panel) conditions. Figure 7A depicts the DNA sequence (SEQ ID NO: 18) encoding a VH44/VL100 disulfide-stabilized BHA 10 scFv comprising a (Gly 4 Ser) 4 linker (indicated in bold type). Figure 7B depicts the amino acid sequence (SEQ ID NO: 19) of the VH44NL 100 10 disulfide + (Gly 4 Ser) 4 -stabilized BHA10 scFv. Annotation is the same as in Figure 7A. The cysteine residues forming the VH44/VL100 disulfide bond are indicated in bold and italicized type. Figure 8 depicts the results of a thermal challenge assay in which the thermal 15 stabilities of the stabilized BHA10 scFv molecules of the invention were compared with that of a conventional BHA1O scFv molecule. The temperature at which 50% of the scFv molecules retain their binding activity (T50) is indicated in the figure legend. Figure 9 depicts the results of Differential Scanning Calorimetry (DSC) analyses 20 performed with a conventional BHA 10 scFv and a BHA 10 VH44:VLI 00/(Gly 4 Ser) 4 scFv. Figure 10 shows results of binding of the hydrophobic fluorescent dye I -anilino 8-naphthaline sulfonate (ANS) to conventional BHA10 scFv, BHAI 0 (Gly 4 Ser) 4 scFv, 25 and BHA10 VH44:VLIOO/(Gly4Ser)4 scFv. Figure II depicts the results of residue frequency analysis of the BHA1O VH and VL domains. Figure 11 A lists BHA10 VH library positions for screening based on residue frequency analysis of IgG variable domain sequences. Figure 11 B lists BHA10 30 VL library positions for screening based on residue frequency analysis of IgG variable domain sequences. Figure 12 depicts the results of a thermal challenge assay in which the effects of stabilizing BHA10 scFv mutations on thermal stability and the additivity of stabilizing 23 WO 2007/109254 PCT/US2007/006883 mutations were assessed. (GS3 deonotes (G 4 S)3 and GS4 denotes (G 4 S)4; SS denotes a disulfide bond). Figure 13 depicts exemplary stabilized LTpR/TRAIL-R2 bispecific antibodies 5 ("Hercules" antibodies) of the invention. The Hercules antibodies are formed by the fusion of a stabilized BHA10 scFv molecule of the invention to an 14A2 IgG antibody. The scFv molecule may be fused to the C-terminus.or N-terminus of the heavy chain (C Hercules or NH-Hercules) or to the N-terminus of the light chain (NL-Hercules). 10 Figure 14 is a schematic depicting the sequential PCR reactions used for fusing conventional and engineered BHAIO scFvs to the amino terminus (Figure 14A) or carboxyl terminus (Figure 14B) of a 14A2 heavy chain. Figure 15A depicts the DNA sequence (SEQ ID NO:28) of a chimeric 14A2 light 15 chain comprising signal peptide (underlined). Figure 15B shows the amino acid sequence (SEQ ID NO:29) of chimeric 14A2 light chain. Figure 16 depicts the DNA sequence (SEQ ID NO:30) of the heavy chain of a conventional BHA 10 scFv NH-Hercules. 20 Figure 17 depicts the amino acid sequence (SEQ ID NO:31) of the heavy chain of a conventional BHA10 scFv NH-Hercules. Figure 18 depicts the DNA sequence (SEQ ID NO:32) of the heavy chain of a 25 BHAlO scFv (Gly4Ser)4 NH- Hercules. Figure 19 depicts the amino acid sequence (SEQ ID NO:33) of the heavy chain of a BHA10 scFv (Gly 4 Ser) 4 NH-Hercules. 30 Figure 20 depicts the DNA sequence (SEQ ID NO:34) of the heavy chain of a BHA10 scFv VH44:VL1OO N-Hercules. Figure 21 depicts the amino acid sequence (SEQ ID NO:35) of the heavy chain of a BHA10 scFv VH44:VL10 0 NH-Hercules. 24 WO 2007/109254 PCT/US2007/006883 Figure 22 depicts the DNA sequence (SEQ ID NO:36) of the heavy chain of a BHA 10 scFv VH 4 4 :VLI 00/(Gly 4 Ser) 4 NH-Hercules. 5 Figure 23 depicts the amino acid sequence (SEQ ID NO:37) of the heavy chain of a BHA 10 scFv VH44:VL1 00/(Gly 4 Ser) 4 N-Hercules. Figure 24 depicts the DNA sequence (SEQ ID NO:44) of the heavy chain of a conventional BHA10 scFv C-Hercules. 10 Figure 25 depicts the amino acid sequence (SEQ ID NO:45) of the heavy chain of a conventional BHAI 0 scFv C-Hercules. Figure 26 depicts the DNA sequence (SEQ ID NO:46) of the heavy chain of a 15 BHA10 scFv (Gly 4 Ser) 4 C- Hercules. Figure 27 depicts the amino acid sequence (SEQ ID NO:47) of the heavy chain of a BHA 10 scFv (Gly 4 Ser)4 C-Hercules. 20 Figure 28 depicts the DNA sequence (SEQ ID NO:48) of the heavy chain of a BHA 10 scFv VH44:VL100 C-Hercules. Figure 29 depicts the amino acid sequence (SEQ ID NO:49) of the heavy chain of a BHA 10 scFv VH4 4 :VLI00 C-Hercules. 25 Figure 30 depicts the DNA sequence (SEQ ID NO:50) of the heavy chain of a BHA10 scFv VH44:V L100/(Gly 4 Ser) 4 C-Hercules. Figure 31 depicts the amino acid sequence (SEQ ID NO:51) of the heavy chain 30 ofa BH A10 scFv VH44:VL100/(Gly4Ser)4 C-Hercules. Figure 32 shows results from Western Blot analyses of transiently expressed C and NH- Hercules bispecific antibodies in CHO cells. The left panel is analyzed under 25 WO 2007/109254 PCT/US2007/006883 non-reducing conditions and the right panel under reducing conditions. Figure 33 depicts the results of an ELISA to evaluate the binding activity of the stabilized Hercules antibodies of the invention to TRAIL R2 receptors. Figure 33A 5 depicts the results with stabilized C-Hercules antibodies. Figure 33B depicts the results with stabilized NH Hercules antibodies. "wt" is conventional NH Hercules antibody. "GS4" is a stablilized NH Hercules antibody comprising a stabilized (Gly4Ser)4 scFv. "ds" is a stablilized NH Hercules antibody comprising a stabilized scFv comprising the VH44/VL100 disulfide linker. 10 Figure 34 depicts the results of an ELISA to evaluate the binding activity of the stabilized Hercules antibodies of the invention to LTpR receptors. Figure 34A depicts the results with stabilized C-terminal Hercules antibodies. Figure 34B depicts the results with stabilized NH Hercules antibodies. "wt" is conventional NH Hercules antibody. 15 "GS4" is a stabilized NH Hercules antibody comprising a stabilized (Gly 4 Ser) 4 scFv. "ds" is a stabilized NH Hercules antibody comprising a stabilized scFv comprising the VH44/VL 100 disulfide linker. Figure 35 shows the results of SEC analysis of stabilized C-Hercules bispecific 20 antibodies following Protein A chromatography. Figure 36 depicts the results of SDS-PAGE analysis of purified stabilized Hercules bispecific antibodies. Figure 36A depicts results with NH-Hercules bispecific antibodies. Figure 36B depicts results with C-Hercules bispecific antibodies. 5 ug 25 sample were loaded in each lane. Figure 37 shows the results of analytical SEC analysis of purified stabilized N Hercules bispecific antibodies. Panel A shows the profile of N-Hercules with VH44:VL100 BHAIO scFv following Protein A chromatography and Panel B following 30 preparative SEC. Panel C shows the profile of N-Hercules with VH44:VLOO/(G4S)4 BHA10 scFv following Protein A chromatography and Panel D following preparative SEC. 26 WO 2007/109254 PCT/US2007/006883 Figure 38 shows the results of analytical SEC analysis of purified stabilized C Hercules bispecific antibodies. Panel A shows the profile of C-Hercules with VH44:VLIOO BHA10 scFv following Protein A chromatography and Panel B following preparative SEC. Panel C shows the profile of C-Hercules with VH44:VL100I/(G4S)4 5 BHA 10 scFv following Protein A chromatography and Panel D following preparative SEC. Figure 39 shows results of Differential Scanning Calorimetry (DSC) analyses performed with C-Hercules with VH44:VL1OO/(G4S)4 BHAIO scFv and C-Hercules with 10 conventional BHA10 scFv. Figure 40 depicts the results of an ELISA to evaluate the bispecific binding activity of the stabilized Hercules antibodies of the invention to TRAIL-R2 and LTpR Ig receptors. "N-"Hercules" Cys and C-"Hercules"Cys are Hercules antibodies comprising 15 NH- or C-terminal scFvs having the VH44NL100 disulfide bond. "N-"Hercules"Cys and N-"Hercules" DM" are Hercules antibodies comprising the NH- or C-terminal scFv comprising the VH44/VLIOO disulfide and the (Gly4Ser)4 linker. CBE 11 pentamer is a control antibody with monospecific LTpR binding activity. 20 Figure 41 depicts the effects of stabilized N- and C-Hercules bispecific and monospecific antibodies on tumor cell growth. Figures 41A-41D depict effects on the growth of WiDr tumor cells, Mel 80 tumor cells, MDA231 tumor cells, and HUVEC cells, respectively. 25 Figure 42 depicts a schematic diagram of a stabilized two chain dimeric minibody comprising stabilized scFvs. The exemplary minibody contains a first chain portion comprising a stabilized scFv with binding specificity for TRAIL R2 antigen and a second chain portion comprising a stabilized scFv with binding specificity for the LTpR antigen. The orientation of the VH and VL domains in the scFv may be changed 30 and the respective binding specificities may be altered. Figure 43 depicts a schematic diagram of an exemplary stabilized bispecific two chain dimeric tetravalent minibody (stablized Bispecific N-scFv tetravalent minibody) 27 WO 2007/109254 PCT/US2007/006883 comprising stabilized scFv fragments of the invention appended to the amino termini. The exemplary stabilized bivalent tetravalent minibody contains a first chain portion with comprising 2 stabilized scFvs with binding specificity for TRAIL R2 antigen and a second chain portion comprising 2 stabilized scFvs with binding specificity for the 5 LTsR antigen. Other configurations are also possible, for example, the bispecific tetravalent minibody can also be constructed such that each chain portion contains 2 stabilized scFv fragments with different specificities. In another embodiment, the orientation of the VH and VL domains in the scFv may be changed. In another embodiment, fewer than all of the scFvs are stabilized. 10 Figure 44 depicts a schematic diagram of an exemplary stabilized bispecific two chain dimeric tetravalent minibody (stabilized Bispecific C-scFv tetravalent minibody) comprising stabilized scFv fragments appended to both carboxyl termini of a bivalent minibody. The exemplary stabilized bivalent tetravalent minibody contains a first chain 15 portion with comprising 2 stabilized scFvs with binding specificity for TRAIL R2 antigen and a second chain portion comprising 2 stabilized scFvs with binding specificity for the LTpR antigen. Other configurations are also possible, for example, the bispecific two-chain dimeric tetravalent minibodies can also be constructed such that each chain contains 2 stabilized scFv fragments with different specificities. In another 20 embodiment, the orientation of the VH and VL domains in the scFv may be changed. In another embodiment, fewer than all of the scFvs are stabilized. Figure 45 shows a schematic diagram of a stabilized bispecific four chain dimeric diabody comprising stabilized scFvs of the invention. Other configurations are 25 also possible, for example, the stabilized bispecific two-chain dimeric tetravalent minibody can also be constructed such that each arm contains stabilized scFv fragments with different specificities. The orientation of the VH and VL domains may be changed. In another embodiment, fewer than all of the scFvs are stabilized. 30 Figure 46 shows a schematic diagram of a stabilized bispecific four chain dimeric tetravalent scFv antibody (stabilized C-scFv tetravalent antibody) comprising a stabilized scFv appended to the carboxyl terminus of CH3 and a hinge connecting peptide. The orientation of the VH and VL domains in the stabilized scFv may be 28 WO 2007/109254 PCT/US2007/006883 changed. Alternatively, the stabilized scFv fragments can be appended to the amino termini of either the heavy or light chain portions to form NH-scFv tetravalent antibodies or NL-scFv tetravalent antibodies, respectively. 5 Figure 47 shows a schematic diagram of a stablized four-chain tetravalent scFv CH2 domain deleted bispecific antibody (stablized C-scFv tetravalent CH2 domain deleted bispecific antibody) comprising a stabilized scFv appended to the carboxyl terminus of CH3 and a hinge connecting peptide. Each heavy chain portion of the bispecific antibody contains a Fv region with binding specificity for the TRAIL R2 10 antigen and a stabilized scFv region with binding specificity for the LT$R antigen. The orientation of the VH and VL domains in the stabilized scFv may be changed and the respective antigen binding specificities may be altered. In another embodiment, fewer than all of the scFvs are stabilized. 15 Figure 48 shows a schematic diagram of a stabilized four-chain tetravalent scFv CH2 domain deleted bispecific antibody (stabilized N-scFv tetravalent CH2 domain deleted bispecific antibody) comprising a stabilized scFv appended to the amino terminus of VH and comprising a hinge connecting peptide. Each heavy chain portion of the bispecific antibody contains an Fv region with binding specificity for the TRAIL 20 R2 antigen and a stabilized scFv region with binding specificity for the LTpR antigen. The orientation of the VH and VL domains in the stabilized scFv may be changed and the respective antigen binding specificities may be altered. In another embodiment, fewer than all of the scFvs are stabilized. 25 Figure 49 shows a schematic diagram of a stabilized four-chain tetravalent scFv CH2 domain deleted bispecific antibody (stabilized NL-scFv tetravalent CH2 domain deleted bispecific antibody) comprising a stabilized scFv appended to the amino terminus of VL and comprising a hinge connecting peptide. Each heavy chain portion of the bispecific antibody contains an Fv region with binding specificity for the TRAIL R2 30 antigen and a stabilized scFv region with binding specificity for the LTpR antigen. The orientation of the VH and VL domains in the stabilized scFv may be changed and the respective antigen binding specificities may be altered. In another embodiment, fewer than all of the scFvs are stabilized. 29 WO 2007/109254 PCT/US2007/006883 Figure 50 depicts T50 curves of wild-type BHA 10 scFv (closed circles) versus BHAIO scFVs containing the VLS46L (Figure 15A, open circles), VHV55G (Figure 15B, open circles), and VH PIOD (open circles) (Figure 15C) stabilizing mutations. 5 Figure 51 depicts DSC curves of the wild-type VH-(GS)4-VL-6His BHA 10 scFv (grey line) versus BH10 scFvs containing the VL S46L (black line) (Figure 51A) and VH V55G (black line) (Figure 51B) stabilizing mutations. 10 Figure 52 depicts DSC curves of the wild-type VH-(GS)4-VL-6His BHA 10 scFv (grey line) versus a BH10 scFv containing containing the VH_PlOD stabilizing mutation (black line). Figure 53 depicts an SDS-Page analysis of the wild-type ("WT (G4S)X3") 15 and stabilized mutant scFvs. All mutant scFvs contain a gly-ser connecting peptide of the formula (Gly4 Ser) 4 between the N-terminal VH and C-terminal VL ("(G4S)X4"). The identity of the scFv in each lane is depicted. Lanes 12-14 contain BHAlO scFv having a disulfide bond between VH44 and VLI 10. The (G4S)X4/disulfide stabilized scFv are denoted 'DM' for 'double mutant'. The final two lanes include designed 20 mutations that were also stabilizing to disulfide linked scFvs. Figure 54A depicts DSC curves of the wild-type ("VH-(GS)4-VL-6His") BHA10 scFv (dashed line), VH S 1 6E mutant scFv (thin grey line), VL S46L mutant scFv (thin black line), VH V55G mutant scFv (thick grey line), and VH PIOID mutant scFv (thick 25 black line). Figure 54B depicts temperature dependent fluorescence of 10 mM ANS in PBS (open circles) or in the presence of the wild-type ("VH-(GS)4-VL-6His") BHA1O scFv (closed grey circles), VH S 16E mutant scFv (diamonds), VL S46L mutant scFv (squares), VH V55G mutant scFv (inverted triangles), and VH P101D mutant scFv (triangles). The lines through the curves are the fits to a two-state unfolding model. 30 Figure 55 depicts ANS fluorescence at 15 *C in the presence of wild-type and mutant scFvs. The ANS fluorescence induced by each scFv is plotted against the TM of the VH domain of the same scFv as determined by DSC (Figure 55A) and the 30 WO 2007/109254 PCT/US2007/006883 temperature-dependent increase in ANS fluorescence due to the unfolding transition of each scFv (Figure 55B). Figure 56 depicts size-exclusion chromatography (SEC) of high concentrations 5 of stabilized N-Hercules (XWUO28; BHA1O scFv VH44:VLIOO/(Gly 4 Ser) 4 N- Hercules; Figure 56A), stabilized C-Hercules (XWU036; (BHA 10 scFv VH44 :VLI0O/(Gly 4 Ser) 4 C-Hercules; Figure 56B) and the wild-type BHA1O antibody (Figure 56C) at six time points (T=0, T=1 week, T=2 weeks, T=1 mo, T=2 mo, and T=3 mo) and low temperatures (2-8 *C). The x-axis is time (minutes) and the y-axis is mAU (absorbance 10 units) at 280 nm. No significant changes in the elution profile were observed over the time course. Figure 57 depicts SDS-PAGE analysis of XWUO28 and XWU036 before and after 3 months storage at 2-8 *C. Panel A. Samples were unreduced (designated as 15 "NR"). The expected MW, -200 kDa, is observed for the bispecific samples. BHA10 runs at its expected MW of -150 kDa. Lanes 1-4 are XWU028 samples. Lanes 5-8 are XWU036 samples. Lanes 9-10 are BHA1O IgG samples. Panel B. Samples were reduced with DTT (designated as "Red" for reduced). The expected MWs of the heavy chain, -75 kDa, and the light chain, -25 kDa, are observed for the bispecific samples 20 under reducing conditions. The BHA10 heavy and light chains run at their expected MWs of 50 kDa and 25 kDa, respectively, under reducing conditions. Lanes 1, 2, 6, and 7 are XWUO28 samples. Lanes 3, 4, 8, and 9 are XWU036 samples. Lanes 5 andlO are BHA10 IgG samples. 25 Figure 58 depicts intact mass analyses of XWU028 (Panel A) and XWU036 (Panel B) at T=0, T= 1 mo, and T=3 mo. Multiple peaks are observed for each protein due to various levels of sialylation of the N-linked carbo-hydrates in the CH2. The carbohydrate distribution of the bispecifics is typical of standard IgGI proteins. 30 Figure 59 depicts the results of a sandwich ELISA assay measuring bispecific binding of serum samples containing N- terminal Hercules (XWUO28; Figure 56A) or C- terminal Hercules (XWU036; Figure 56B) to TRAIL-R2 and LTpR receptors. 31 WO 2007/109254 PCT/US2007/006883 Figure 60 depicts the results of an experiment evaluating the relative in vivo activity of stabilized bispecific antibodies (XWUO28 (diamonds) and XWU036 (open squares)), unstabilized bispecific antibodies (hCBE1I1 (closed triangle)) and monospecific antibodies administered alone (hBHA10 (inverted triangle) and chl4A2 5 (diamond)) and in combination (open triangle) against a tumor xenograft mouse model. Dosing began at day 13. hCBE1 1 v. VC: P<0.001; hBHAIO v. VC: P<0.001; chl4A2 v. VC: P<0.01; Hercules-Il XWU028 v. VC. : P<0.001; Hercules-II XWUO28 v.hBHA1O: P<0.05; Hercules-II XWUO28 v. chl4A2: P<0.005 ; Hercules-II XWU036 v. VC: P<0.001; Hercules-II XWUO36 v. hBHA1O: P<0.01; Hercules-II XWU036 v. 10 chl4A2: P<0.05; combo v. V.C.: P<0.001 Figure 61 depicts the heavy chain DNA sequence (SEQ ID NO:52) for C Hercules BHA 10 scFv VH S16E + VL S46L bispecific antibody. 15 Figure 62 depicts the heavy chain amino acid sequence (SEQ ID NO:53) for C Hercules BHAIO scFv VH S16E + VL S46L bispecific antibody. Figure 63 depicts the heavy chain DNA sequence (SEQ ID NO:54) for C Hercules BHA1O scFv VH44-VLIOO/VH S16E + VL S46L bispecific antibody. 20 Figure 64 depicts the heavy chain amino acid sequence (SEQ ID NO:55) for C Hercules BHA10 scFv VH44-VLI0VH S16E + VL S46L bispecific antibody. Figure 65 depicts the results of analytical size exclusion chromatography (SEC) 25 of Protein A eluates from supernatants containing C-Hercules with the stabilized VH S16E + VL S46L BHA10 scFv (XWU054) and C- Hercules with conventional ("WT") BHA 10 scFv. Figure 66 depicts SDS-PAGE gels of purified C-terminal Hercules with the VH 30 S16E + VL S46L BHA10 scFv (Figure 66A) and purified C-terminal Hercules with the VH44:VLIOO/VH S16E + VL S46L BHA10 scFv (Figure 66B). Figure 67 shows the analytical SEC elution profiles of C-terminal Hercules with 32 WO 2007/109254 PCT/US2007/006883 the VH S16E + VL S46L BHA 10 scFv (Figure 67A) and C-terminal Hercules with the V H44:VLIOOVH S16E + VL S46L BHA10 scFv (Figure 67B) subsequent to the initial protein A purification step. 5 Figure 68 shows the results of in vitro tumor cell proliferation assays of the activity of C-terminal Hercules with the VH S16E + VL S46L BHAIO scFv (XWU054) and C-terminal Hercules with the VH44:VLIOO/VH S16E + VL S46L BHA1O scFv (XWU055). Both MDA231 (Figure 68A) and WiDr cells (Figure 68B) were administed antibody. Activity for both Hercules antibodies was compared with tabilized XWU036 10 bispecific antibody and corresponding monospecific antibodies (hBHA10 and ch14A2) administered alone or in combination. Figure 69A depicts a DSC unfolding curve of BIIB7 IgG 1 taken under conditions identical to those used for the remaining 18 BIIB human(ized) antibodies 15 (solid line). The individual Fab, CH2 and CH 3 transitions apparent within the DSC curve are labeled and sketched with dotted lines. Figure 69B depicts the DSC unfolding curves of the BIIB7 antibody in both the IgG I and IgG4 formats. Figure 70 depicts thermal unfolding curves of four human(ized) IgGI antibodies, 20 BIIB 16, BIIB6, BIIB4 and BIIB1. The unfolding transitions of the CH 2 and CH3 domains for all four IgG 1 constructs are identical, while the Fab unfolding transitions are highly variable. Figure 71 depicts consensus scoring results for 182 human antibody variable 25 domain reference sequences and the individual subclass consensus sequences. A. VH subclass sequence scores derived by residue frequency analysis against a mammalian VH database. B. Distribution of scores from reference VH sequence selected from NCBI. VH sequence scores are clustered according to subclass. C. V, subclass consensus sequence scores derived by residue frequency analysis against a mammalian VK database. D. 30 Distribution of scores from reference VK sequence selected from NCBI. V, sequence scores are clustered according to subclass. Figure 72 depicts 2-dimensional plots comparing sequence scores for 18 BIIB 33 WO 2007/109254 PCT/US2007/006883 antibodies with the corresponding Fab TMs as measured by DSC. Figure 72A depicts BIIB1-18 VH scores plotted against the Fab Tms. The identity of each point can be found in Table 20. The VHs of BIIB15 and BIIB 16 contain unusual insertions/deletions in their CDRs. The results for BIIB15 and BIIB16 are displayed as squares. BIIB18's 5 TM was artificially labeled as 57.2 *C (to match the lowest measured TM from BIIB 17) since its own Tm was immeasurable due to the lack of expression. BIIB18's TM is likely much lower based on its extremely low sequence score and its inability to express. The Fab results for BIIB 18 are displayed as a triangle. The remaining BIIB antibody VH results are displayed as circles. Figure 72B depicts BIIB1-18 VK scores plotted against 10 the Fab TMs. The symbols are the same as described for plot (A), except the scores are for V. Figure 73 depicts representative structures for the V-class, C-class, and I-class of Ig folds. 15 Figure 74 depicts a histogram of the sequence lengths of V-, I-, Cl-, and C2 class Ig-Folds obtained from SCOP. Figure 75 depicts a histogram of f-value correlation coefficients calculated for 20 every residue combination within the C-class dataset. Figure 76 shows the output of the NAPMAP Visual Tool which includes a portion of the alignment platform on which the covariation analysis is overlayed. The platform displays columns containing all of the amino acids including gaps (residue 25 type) for each position within the V region (residue position #). Figure 77 depicts an overlay of Ig-Fold covariation statistics onto the NAPMAP template. The sequence of interest is shown as a curved line through the relevant amino acids. Covariations are shown as straight bars between the relevant covarying amino 30 acid residues. Figure 78 shows a detail from the covariation analysis of BHA1O VL highlighting the covariation between Tyr at Kabat position 36 (residue position # 48) and the 34 WO 2007/109254 PCT/US2007/006883 mutation Ser-+Leu at Kabat position 46 (residue position # 67) as described in Example 7. Figure 79 shows a detail from the covariation analysis of BH A10 VH highlighting the conflicting covariation between Val at Kabat position 67 (residue 5 position # 100), Thr at Kabat position 70 (residue position # 104), and the mutation Met-Leu at Kabat position 80 (residue position # 116). Figure 80 (SEQ ID NO:63) shows the single-stranded DNA sequence of heavy chain C-terminal tetravalent PRIMATIZED@ p5E8 antibody comprising a conventional 10 scFv. Figure 81 (SEQ ID NO:64) shows the amino acid sequence of heavy chain C terminal tetravalent PRIMATIZED@ p5E8 antibody comprising a conventional scFv. 15 Figure 82A (SEQ ID NO:65) shows the single-stranded DNA sequence of PRIMATIZED@ p5E8 light chain. Signal peptide sequence is shown as underlined. Figure 82B (SEQ ID NO:66) shows the amino acid sequence of PRIMATIZED@ p5E8 light chain. 20 Figure 83A (SEQ ID NO:67) shows single-stranded DNA sequence of conventional PRIMATIZED@ p5E8 (VL/VH) scFv. Signal peptide sequence is shown as underlined. Figure 83B (SEQ ID NO:68) shows amino acid sequence of conventional PRIMATIZED@ p5E8 (VL/VH) scFv. 25 Figure 84A (SEQ ID NO:69) shows single-stranded DNA sequence of conventional PRIMATIZED@ p5E8 (VHNL) scFv. Figure 84B (SEQ ID NO:70) shows amino acid sequence of conventional PRIMATIZED@ p5E8 (VH/VL) scFv. Figure 85 depicts T50 curves of conventional p5E8 (VH/VL) scFv (filled circles) 30 versus conventional p5E8 (VL/VH) scFv (open circles). Figure 86 depicts the results of an experiment evaluating the relative in vivo activity of stabilized bispecific "Hercules-II"antibodies (XWU036 (closed circle)) and 35 WO 2007/109254 PCT/US2007/006883 monospecific antibodies administered alone (hBHA1O (inverted closed triangle) and chl4A2 (closed diamond)) and in combination (open inverted triangle) against a tumor xenograft mouse model of human breast cancer (MDA-MB-23 1). 5 Figure 87 depicts VH residues involved in Covariation networks that are important for interface maintenance (Figure 87A) and interface residues involved in making direct contacts with the VHNL interface and that are mapped directly to the surface of VH domain (Figure 87B). Arrows indicate residue positions outside the actual interface that covariation analysis informs to be important for stabilizing the VHNL interaction and 10 for VH/VL stability. The location of hypervariable CDR3 residues that make direct contacts at the interface between VH and VL are enclosed within a box. Figure 88 depicts VL residues involved in Covariation networks important for interface maintenance (see Figure 88A) and interface residues involved in making direct 15 contacts with the VH/VL interface and that are mapped directly to the surface of VL (Figure 88B). Arrows indicate residue positions outside the actual interface that covariation analysis informs to be important for stabilizing the VHNL interaction and for VH/VL stability. The location of hypervariable CDR3 residues that make direct contacts at the interface between VH and VL are enclosed within a box. 20 Figure 89 depicts an exemplary environment suitable for practicing an embodiment of the invention. Figure 90 is a flowchart of an exemplary sequence of steps that may be followed 25 by an embodiment of the invention in order to determine a sequence covariation score that may be used as a measure of the stability of a polypeptide. Figure 91 is a flowchart of an exemplary sequence of steps that may be followed by an embodiment of the invention in order to determine and use a consensus score to 30 predict the stability of a candidate protein. Figure 92 is a flowchart of an exemplary sequence of steps that may be followed by an embodiment of the invention in order to determine and use an average consensus 36 WO 2007/109254 PCT/US2007/006883 score as a measure of the stability of a candidate protein. Figure 93 is a flowchart of an exemplary series of steps that may be followed by an embodiment of the invention in order to use a substitute residue of a polypeptide to 5 determine an improved sequence of the polypeptide. Figures 94A and 94B shows the amino acid -sequences of the stabilized BHA1O scFVs containing the S46L(VL) stabilizing mutation (SEQ ID NO: 137), and the V55G(VH) stabilizing mutation (SEQ ID NO:138). The stabilizing mutation is 10 indicated by the boxed residue. The leader sequence, gly/ser connecting peptide, and CHI domain are indicated by the underlined, bolded, and italicized residues, respectively. Detailed Description of the Invention 15 The invention is based, at least in part, on the development of stabilized target binding molecules that consist of or comprise a stabilized scFv molecule and methods for making such stabilized binding molecules. In addition, the instant invention provides methods for designing stable proteins, such as antibody molecules. The stabilized scFv molecules of the instant invention are especially useful in 20 producing stable multispecific, e.g., bispecific molecules. The stabilized binding molecules of the invention, e.g., multispecific binding molecules can be stably expressed in culture, are suitable for large scale production, and are stable in vivo. The invention is also based, at least in part, on the development of stabilized binding molecules that consist of or comprise a stabilized scFv molecule and methods 25 for making such stabilized binding molecules. In addition, the instant invention provides methods for designing stable proteins and for predicting the stability of proteins, such as antibody molecules. Before further description of the invention, for convenience, certain terms are described below: 30 I. Definitions As used herein the term "scFv molecule" includes binding molecules which 37 WO 2007/109254 PCT/US2007/006883 consist of one light chain variable domain (VL) or portion thereof, and one heavy chain variable domain (VH) or portion thereof, wherein each variable domain (or portion thereof) is derived from the same or different antibodies. scFv molecules preferably comprise an scFv linker interposed between the VH domain and the VL domain. scFv 5 molecules are known in the art and are described, e.g., in US patent 5,892,019, Ho et al. 1989. Gene 77:51; Bird et al. 1988 Science 242:423; Pantoliano et al. 1991. Biochemistry 30:10117; Milenic et al. 1991 .-Cancer Research 51:6363; Takkinen et al. 1991. Protein Engineering 4:837. A "scFv linked" as used herein refers to a moiety interposed between the VL and 10 VH domains of the scFv. scFv linkers preferably maintain the scFv molecule in a antigen binding conformation. In one embodiment,an scFv linker comprises or consists of an scFv linker peptide. In certain embodiments,an scFv linker peptide comprises or consists of a gly-ser connecting peptide. In other embodiments,an scFv linker comprises a disulfide bond. 15 As used herein, the term "gly-ser connecting peptide" refers to a peptide that consists of glycine and seine residues. An exemplary gly/ser connecting peptide comprises the amino acid sequence (Gly 4 Ser)n. In one embodiment, n=1. In one embodiment, n=2. In another embodiment, n=3. In a preferred embodiment, n=4, i.e., (Gly4 Ser) 4 . In another embodiment, n=5. In yet another embodiment, n=6. 20 Another exemplary gly/ser connecting peptide comprises the amino acid sequence Ser(Gly 4 Ser).. In one embodiment, n=l. In one embodiment, n=2. In a preferred embodiment, n=3. In another embodiment, n=4. In another embodiment, n=5. In yet another embodiment, n=6. As used herein the term "disulfide bond" refers to the covalent bond formed 25 between two sulfur atoms. The amino acid cysteine comprises a thiol group that can form a disulfide bond or bridge with a second thiol group. As used herein the term "conventional scFv molecule" refers to an scFv molecules which is not a stabilized scFv molecule. For example, a typical conventional lacks stabilizing mutations and comprises a VH and a VL domain linked by a (G 4 S)3 30 linker. A "stabilized scFv molecule" of the invention is an scFv molecule comprising at least one change or alteration as compared to a conventional scFv molecule which results in stabilization of the scFv molecule. As used herein, the term "stabilizing 38 WO 2007/109254 PCT/US2007/006883 mutation" includes a mutation which confers enhanced protein stability (e.g. thermal stability) to the scFv molecule and/or to a larger protein comprising said scFv molecule. In one embodiment, the stabilizing mutation comprises the substitution of a destabilizing amino acid with a replacement amino acid that confers enhanced protein stability (herein 5 a "stabilizing amino acid"). In one embodiment, the stabilizing mutation is one in which the length ofan scFv linker has been optimized. In one embodiment, a stabilized scFv molecule of the invention comprises one or more amino acid substitutions. For example, in one embodiment, a stabilizing mutation comprises a substitution of at least one amino acid residue which substitution results in an increase in stability of the VH and VL 10 interface of an scFv molecule. In one embodiment, the amino acid is within the interface. In another embodiment, the amino acid is one which scaffolds the interaction between VH and VL. In another embodiment, a stabilizing mutation comprises substituting at least one amino acid in the VH domain or VL domain that covaries with two or more amino acids at the interface between the VH and VL domains. In another 15 embodiment, the stabilizing mutation is one in which at least one cysteine residue is introduced (i.e., is engineered into one or more of the VH or VL domain) such that the VH and VL domains are linked by at least one disulfide bond between an amino acid in the VH and an amino acid in the VL domain. In certain preferred embodiments, a stabilized scFv molecule of the invention is one in which both the length of the scFv 20 linker is optimized and at least one amino acid residue is substituted and/or the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain. In one embodiment, more than one of the stabilizing mutations described herein may be made in an scFv molecule. In one embodiment, one or more stabilizing mutations made to an scFv molecule 25 simultaneously improves the thermal stability of both the VH and VL domains of the scFv molecule as compared to a conventional scFv molecule. Preferably, a population of one or more of the stabilized scFv molecules of the invention is expressed as a population of monomeric, soluble proteins. In one embodiment, no more than 10% is present in aggregated form. In one embodiment, the 30 stabilized scFv molecules of the population may comprise the same stabilizing mutation or a combination of stabilizing mutations. In other embodiments, the individual stabilized scFv molecules of the population comprise different stabilizing mutations. The subject stabilized scFv molecules may be used alone to bind to a target 39 WO 2007/109254 PCT/US2007/006883 molecule or may be linked to another polypeptide to form stabilized binding molecules which comprise a stabilized scFv molecule. For example, a binding molecule of the invention may comprise an scFv molecule linked to a second scFv molecule or a non scFv molecule, e.g., that imparts target binding specificity, such as an antibody. 5 As used herein the term "protein stability" refers to an art-recognized measure of the maintainance of one or more physical properties of a protein in response to an environmental condition (e.g. an elevated or lowered temperature). In one embodiment, the physical property is the maintenance of the covalent structure of the protein (e.g. the absence of proteolytic cleavage, unwanted oxidation or deamidation). In another 10 embodiment, the physical property is the presence of the protein in a properly folded state (e.g. the absence of soluble or insoluble aggregates or precipitates). In one embodiment, stability of a protein is measured by assaying a biophysical property of the protein, for example thermal stability, pH unfolding profile, stable removal of glycosylation, solubility, biochemical function (e.g., ability to bind to a protein (e.g., a 15 ligand, a receptor, an antigen, etc.) or chemical moiety, etc.), and/or combinations thereof. In another embodiment, biochemical function is demonstrated by the binding affinity of an interaction. In one embodiment, a measure of protein stability is thermal stability, i.e., resistance to thermal challenge. Stability can be measured using methods known in the art and/or described herein. 20 The VL and VH domains of an scFv molecule are derived from one or more antibody molecules. It will also be understood by one of ordinary skill in the art that the variable regions of the scFv molecules of the invention may be modified such that they vary in amino acid sequence from the antibody molecule from which they were derived. For example, in one embodiment, nucleotide or amino acid substitutions leading to 25 conservative substitutions or changes at amino acid residues may be made (e.g., in CDR and/or framework residues). Alternatively or in addition, mutations may be made to CDR amino acid residues to optimize antigen binding using art recognized techniques. The binding molecules of the invention maintain the ability to bind to antigen. As used herein the term "derived from" a designated protein refers to the origin 30 of the polypeptide. In one embodiment, the polypeptide or amino acid sequence which is derived from a particular starting polypeptide is a variable region sequence (e.g. a VH or VL) or sequence related thereto (e.g. a CDR or framework region). In one embodiment, the amino acid sequence which is derived from a particular starting 40 WO 2007/109254 PCT/US2007/006883 polypeptide is not contiguous. For example, in one embodiment, one, two, three, four, five, or six CDRs are derived from a starting antibody. In one embodiment, the polypeptide or amino acid sequence that is derived from a particular starting polypeptide or amino acid sequence has an amino acid sequence that is essentially identical to that of 5 the starting sequence or a portion thereof, wherein the portion consists of at least 3-5 amino acids, 5-10 amino acids, at least 10-20 amino acids, at least 20-30 amino acids, or at least 30-50 amino acids, or which is otherwise identifiable to one of ordinary skill in the art as having its origin in the starting sequence. An isolated nucleic acid molecule encoding a stabilized scFv molecule or a 10 portion thereof can be created by introducing one or more nucleotide substitutions, additions or deletions into the nucleotide sequence of a conventional scFv molecule or an immunoglobulin from which it is derived such that one or more amino acid substitutions, additions or deletions are introduced into the encoded protein. Mutations may be introduced by standard techniques, such as site-directed mutagenesis and PCR 15 mediated mutagenesis. In one embodiment, conservative amino acid substitutions are made at one or more non-essential amino acid residues. A "conservative amino acid substitution" is one in which the amino acid residue is replaced with an amino acid residue having a similar side chain. Families of amino acid residues having similar side chains have been defined in the art, including basic side chains (e.g., lysine, arginine, 20 histidine), acidic side chains (e.g., aspartic acid, glutamic acid), uncharged polar side chains (e.g., glycine, asparagine, glutamine, serine, threonine, tyrosine, cysteine), nonpolar side chains (e.g., alanine, valine, leucine, isoleucine, proline, phenylalanine, methionine, tryptophan), beta-branched side chains (e.g., threonine, valine, isoleucine) and aromatic side chains (e.g., tyrosine, phenylalanine, tryptophan, histidine). Thus, an 25 amino acid residue in an immunoglobulin polypeptide may be replaced with another amino acid residue from the same side chain family. In another embodiment, a string of amino acids can be replaced with a structurally similar string that differs in order and/or composition of side chain family members. In another embodiment, a mutation is introduced in order to introduce at least one cysteine molecule into the VH and into the 30 VL domain and, thereby, introduce a disulfide bond into the scFv molecule. In another embodiment, an amino acid of a conventional scFv molecule may be substituted with an amino acid having similar physical (e.g., spatial) or functional properties. Preferably, amino acids substituted into conventional scFv molecules are compatible with the 41 WO 2007/109254 PCT/US2007/006883 integrity of the VL/VH interface, CDR conformations, and VH and/or VL folding. Alternatively, in another embodiment, mutations may be introduced randomly along all or part of the immunoglobulin coding sequence. The stabilized scFv molecules of the invention or polypeptides comprising the 5 stabilized scFv molecules are binding molecules, i.e., they bind to a target molecule of interest, e.g., an antigen. When a stabilized scFv molecule of the invention is fused to a second molecule, the second molecule may also impart a binding specificity to the fusion protein. The binding molecules of the invention consist of scFv molecules (e.g., a VH 10 and a VL domain joined by an scFv linker) or comprise a stabilized scFv molecule of the invention. In one embodiment, the binding molecules of the invention are monovalent, i.e., comprise one target binding site (e.g., as in the case of an scFv molecule). In one embodiment, the binding molecules of the invention are multivalent, i.e., comprise more 15 than one target binding site. In another embodiment, the binding molecules comprise at least two binding sites. In one embodiment, the binding molecules comprise two binding sites. In one embodiment, the binding molecules comprise three binding sites. In another embodiment, the binding molecules comprise four binding sites. In another embodiment, the binding molecules comprise greater than four binding sites. 20 In one embodiment, the binding molecules of the invention are monomers. In another embodiment, the binding molecules of the invention are multimers. For example, in one embodiment, the binding molecules of the invention are dimers. In one embodiment, the dimers of the invention are homodimers, comprising two identical monomeric subunits. In another embodiment, the dimers of the invention are 25 heterodimers, comprising two non-identical monomeric subunits. The subunits of the dimer may comprise one or more polypeptide chains. For example, in one embodiment, the dimers comprise at least two polypeptide chains. In one embodiment, the dimers comprise two polypeptide chains. In another embodiment, the dimers comprise four polypeptide chains (e.g., as in the case of antibody molecules). 30 As used herein the term "valency" refers to the number of potential target binding sites in a polypeptide. Each target binding site specifically binds one target molecule or specific site on a target molecule. When a polypeptide comprises more than one target binding site, each target binding site may specifically bind the same or 42 WO 2007/109254 PCT/US2007/006883 different molecules (e.g., may bind to different molecules, e.g., different antigens, or different epitopes on the same molecule). The term "specificity" refers to the ability to specifically bind (e.g., immunoreact with) a given target. A polypeptide may be monospecific and contain one or more 5 binding sites which specifically bind a target or a polypeptide may be multispecific (e.g., bispecific or trispecific) and contain two or more binding sites which specifically bind the same or different targets. Specific binding may be imparted by a stabilized scFv molecule of the invention and/or a non-scFv moiety to which a stabilized scFv molecule of the invention is linked. 10 In one embodiment, a binding molecule of the invention is multispecific. For example, in one embodiment, a multispecific binding molecule of the invention is a bispecific molecule (e.g., antibody, minibody, domain deleted antibody, or fusion protein, single domain antibodies (e.g., camelid, shark, human) comprising at least one stabilized scFv molecule) having binding specificity for at least two targets, e.g., more 15 than one target molecule or more than one epitope on the same target molecule. In one embodiment, a multispecific molecule has at least one binding site specific for a molecule targeted for reduction or elimination and a target molecule on a cell. In another embodiment, a multispecific molecule has at least one target binding site specific for a molecule targeted for reduction or elimination and at least one binding site 20 specific for a drug. In yet another embodiment, a multispecific molecule has at least one binding site specific for a molecule targeted for reduction or elimination and at least one binding site specific for a prodrug. In one embodiment, a multispecific molecule comprises one specificity for a soluble molecule and one specifity for a cell surface molecule. In another embodiment, 25 a multispecific molecule has two binding specificities for two targets present on one or more soluble molecules. In another embodiment, a multispecific molecule has two binding specificities for two targets present on one or more cell surface molecules (which may be present on one or more cells). In one embodiment, the binding molecules have at least one target binding site 30 specific for a molecule which mediates a biological effect (e.g., which modulates cellular activation (e.g., by binding to a cell surface receptor and resulting in transmission or inhibition of an activating or inhibitory signal), which results in death of the cell (e.g., by a cell signal induced pathway, by complement fixation or exposure to a 43 WO 2007/109254 PCT/US2007/006883 payload present on the binding molecule), or which modulates a disease or disorder in a subject (e.g., by mediating or promoting cell killing, by promoting lysis of a fibrin clot or promoting clot formation, or by modulating the amount of a substance which is bioavailable (e.g., by enhancing or reducing the amount of a ligand such as TNFa. in the 5 subject)). In another embodiment, the binding molecules of the invention bind at least one target that transduces a signal to a cell, e.g., by binding to a cell surface receptor, such as a TNF family receptor. By "transduces a signal" it is meant that by binding to the cell, the binding molecule converts the extracellular influence on the cell surface receptor 10 into a cellular response, e.g., by modulating a signal transduction pathway. In one embodiment, the binding molecules bind at least one target binding site specific for a molecule targeted for reduction or elimination, e.g., a cell surface antigen or a soluble antigen. In one embodiment, the binding of the binding molecule to the target results in reduction or elimination of the target, e.g., from a tissue or from the 15 circulation. In another embodiment, the binding molecules have at least one binding site specific for a molecule that can be used to detect the presence of a target molecule (e.g., to detect a contaminant or diagnose a condition or disorder). In yet another embodiment, a binding molecule of the invention comprises at least one binding site that targets the binding molecule to a specific site in a subject (e.g., to a tumor cell or blood 20 clot). In a preferred embodiment, a multispecific molecule is a tetravalent antibody that has four binding sites. A tetravalent molecule may be bispecific and bivalent for each specificity. Further description of exemplary bispecific molecules is provided below. Preferred binding molecules of the invention comprise framework and constant 25 region amino acid sequences derived from a human amino acid sequence. However, binding polypeptides may comprise framework and/or constant region sequences derived from another mammalian species. For example, binding molecules comprising murine sequences may be appropriate for certain applications. In one embodiment, a primate framework region (e.g., non-human primate), heavy chain portion, and/or hinge 30 portion may be included in the subject binding molecules. In one embodiment, one or more murine amino acids may be present in the framework region of a binding polypeptide, e.g., a human or non-human primate framework amino acid sequence may comprise one or more amino acid back mutations in which the corresponding murine 44 WO 2007/109254 PCT/US2007/006883 amino acid residue is present and/or may comprise one or mutations to a different amino acid residue not found in the starting murine antibody. Preferred binding molecules of the invention are less immunogenic than murine antibodies. A "fusion" or chimeric protein comprises a first amino acid sequence linked to a 5 second amino acid sequence with which it is not naturally linked in nature. The amino acid sequences may normally exist in separate proteins that are brought together in the fusion polypeptide or they may normally exist in the same protein but are placed in a new arrangement in the fusion polypeptide. A fusion protein may be created, for example, by chemical synthesis, or by creating and translating a polynucleotide in which 10 the peptide regions are encoded in the desired relationship. The term "heterologous" as applied to a polynucleotide or a polypeptide, means that the polynucleotide or polypeptide is derived from a genotypically distinct entity from that of the entity to which it is being compared. For instance, a heterologous polynucleotide or antigen may be derived from a different species, different cell type of 15 an individual, or the same or different type of cell of distinct individuals. The term "ligand binding domain" or "ligand binding portion" as used herein refers to any native receptor (e.g., cell surface receptor) or any region or derivative thereof retaining at least a qualitative ligand binding ability, and preferably the biological activity of a corresponding native receptor. 20 The term "receptor binding domain" or "receptor binding portion" as used herein refers to any native ligand or any region or derivative thereof retaining at least a qualitative receptor binding ability, and preferably the biological activity of a corresponding native ligand. In one embodiment, the binding molecules of the invention are stabilized 25 "antibody" or "immunoglobulin" molecules, e.g., naturally occurring antibody or immunoglobulin molecules (or an antiben binding fragment thereof) or genetically engineered antibody molecules that bind antigen in a manner similar to antibody molecules and that comprise an scFv molecule of the invention. As used herein, the term "immunoglobulin" includes a polypeptide having a combination of two heavy and 30 two light chains whether or not it possesses any relevant specific immunoreactivity. "Antibodies" refers to such assemblies which have significant known specific immunoreactive activity to an antigen of interest (e.g. a tumor associated antigen). Antibodies and immunoglobulins comprise light and heavy chains, with or without an 45 WO 2007/109254 PCT/US2007/006883 interchain covalent linkage between them. Basic immunoglobulin structures in vertebrate systems are relatively well understood. As will be discussed in more detail below, the generic term "immunoglobulin" comprises five distinct classes of that can be distinguished biochemically. All five 5 classes of antibodies are within the scope of the present invention, the following discussion will generally be directed to the IgG class of immunoglobulin molecules. With regard to IgG, immunoglobulins comprise two identical light polypeptide chains of molecular weight approximately 23,000 Daltons, and two identical heavy chains of molecular weight 53,000-70,000. The four chains are joined by disulfide bonds in a "Y" 10 configuration wherein the light chains bracket the heavy chains starting at the mouth of the "Y" and continuing through the variable region. Both the light and heavy chains are divided into regions of structural and functional homology. The terms "constant" and "variable" are used functionally. In this regard, it will be appreciated that the variable domains of both the light (VL) and heavy 15 (VH) chain portions determine antigen recognition and specificity. Conversely, the constant domains of the light chain (CL) and the heavy chain (CH1, CH2 or CH3) confer important biological properties such as secretion, transplacental mobility, Fc receptor binding, complement binding, and the like. By convention the numbering of the constant region domains increases as they become more distal from the antigen 20 binding site or amino-terminus of the antibody. The N-terminus is a variable region and at the C-terminus is a constant region; the CH3 and CL domains actually comprise the carboxy-terminus of the heavy and light chain, respectively. Stabilizing mutations to scFv molecules may be made to amino acids in the CDR and/or in the framework regions of an scFv variable heavy and/or variable light chaing. 25 As used herein the term "variable region CDR amino acid residues" includes amino acids in a CDR or complementarity determining region as identified using sequence or structure based methods. As used herein, the term "CDR" or "complementarity determining region" means the noncontiguous antigen combining sites found within the variable region of both heavy and light chain polypeptides. These particular regions have 30 been described by Kabat et al., J. Biol. Chem. 252, 6609-6616 (1977) and Kabat et al., Sequences of protein of immunological interest. (1991), and by Chothia et al., J. Mol. Biol. 196:901-917 (1987) and by MacCallum et al., J. Mol. Biol. 262:732-745 (1996) where the definitions include overlapping or subsets of amino acid residues when 46 WO 2007/109254 PCT/US2007/006883 compared against each other. The amino acid residues which encompass the CDRs as defined by each of the above cited references are set forth for comparison. Preferably, the term "CDR" is a CDR as defined by Kabat based on sequence comparisons. CDR Definitions Kabat' Chothia 2 MacCallum 3 VH CDRI 31-35 26-32 30-35 VH CDR2 50-65 53-55 47-58 VHCDR3 95-102 96-101 93-101 VLCDRI 24-34 26-32 30-36 VLCDR2 50-56 50-52 46-55 VLCDR3 89-97 91-96 89-96 5 'Residue numbering follows the nomenclature of Kabat et al., supra 2Residue numbering follows the nomenclature of Chothia et al., supra 3 Residue numbering follows the nomenclature of MacCallum et al., supra As used herein the term "variable region framework (FR) amino acid residues" 10 refers to those amino acids in the framework region of an Ig chain. The term "framework region" or "FR region" as used herein, includes the amino acid residues that are part of the variable region, but are not part of the CDRs (e.g., using the Kabat definition of CDRs). Therefore, a variable region framework is between about 100-120 amino acids in length but includes only those amino acids outside of the CDRs. For the 15 specific example of a heavy chain variable region and for the CDRs as defined by Kabat et al., framework region 1 corresponds to the domain of the variable region encompassing amino acids 1-30; framework region 2 corresponds to the domain of the variable region encompassing amino acids 36-49; framework region 3 corresponds to the domain of the variable region encompassing amino acids 66-94, and framework region 4 20 corresponds to the domain of the variable region from amino acids 103 to the end of the variable region. The framework regions for the light chain are similarly separated by each of the light chain variable region CDRs. Similarly, using the definition of CDRs by Chothia et al. or McCallum et al. the framework region boundaries are separated by the respective CDR termini as described above. In preferred embodiments, the CDRs are as 25 defined by Kabat. 47 WO 2007/109254 PCT/US2007/006883 In naturally occurring antibodies, the six CDRs present on each monomeric antibody are short, non-contiguous sequences of amino acids that are specifically positioned to form the antigen binding site as the antibody assumes its three dimensional configuration in an aqueous environment. The remainder of the heavy and light variable 5 domains show less inter-molecular variability in amino acid sequence and are termed the framework regions. The framework regions largely adopt a p-sheet conformation and the CDRs form loops which connect, and in some cases form part of, the P-sheet structure. Thus, these framework regions act to form a scaffold that provides for positioning the six CDRs in correct orientation by inter-chain, non-covalent interactions. 10 The antigen binding site formed by the positioned CDRs defines a surface complementary to the epitope on the immunoreactive antigen. This complementary surface promotes the non-covalent binding of the antibody to the immunoreactive antigen epitope. The position of CDRs can be readily identified by one of ordinary skill in the art. 15 As previously indicated, the subunit structures and three dimensional configuration of the constant regions of the various immunoglobulin classes are well known. As used herein, the term "VH domain" includes the amino terminal variable domain of an immunoglobulin heavy chain. As used herein, the term "VL domain" includes the amino terminal variable domain of an immunoglobulin light chain. 20 The term "fragment" refers to a part or portion of a polypeptide (e.g., an antibody or an antibody chain) comprising fewer amino acid residues than an intact or complete polypeptide. The term "antigen-binding fragment" refers to a polypeptide fragment of an immunoglobulin or antibody that binds antigen or competes with intact antibody (i.e., with the intact antibody from which they were derived) for antigen binding (i.e., specific 25 binding). As used herein, the term "fragment" of an antibody molecule includes antigen-binding fragments of antibodies, for example, an antibody light chain (VL), an antibody heavy chain (VH), a single chain antibody (scFv), a F(ab')2 fragment, a Fab fragment, an Fd fragment, an Fv fragment, and a single domain antibody fragment (DAb). Fragments can be obtained, e.g., via chemical or enzymatic treatment of an intact 30 or complete antibody or antibody chain or by recombinant means. As used herein, the term "binding site" comprises a region of a polypeptide which is responsible for selectively binding to a target molecule of interest (e.g. an antigen, ligand, receptor, substrate or inhibitor). Binding domains comprise at least one 48 WO 2007/109254 PCT/US2007/006883 target binding site. Exemplary binding domains include an antibody variable domain, a receptor binding domain of a ligand, a ligand binding domain of a receptor or an enzymatic domain. Binding molecules of the invention can be made using techniques that are known 5 in the art. In one embodiment, the polypeptides of the invention are "recombinantly produced," i.e., are produced using recombinant DNA technology. Exemplary techniques for making such molecules are discussed in more detail below. In one embodiment, a binding molecule of the invention is a naturally occurring antibody to which a stabilized scFv molecule has been fused. In one embodiment, a 10 binding molecule of the invention is a modified antibody to which a stabilized scFv molecule has been fused. As used herein, the term "modified antibody" includes synthetic forms of antibodies which are altered such that they are not naturally occurring. In another embodiment, a binding molecule of the invention is a fusion protein comprising at least one scFv molecule. 15 In preferred embodiments, a polypeptide of the invention will not elicit a deleterious immune response in a human. In one embodiment, a binding molecule of the invention comprises a constant region, e.g., a heavy chain constant region. In one embodiment, such a constant region is modified compared to a wild-type constant region. That is, the polypeptides of the 20 invention disclosed herein may comprise alterations or modifications to one or more of the three heavy chain constant domains (CH 1, CH2 or CH3) and/or to the light chain constant region domain (CL). Exemplary modifications include additions, deletions or substitutions of one or more amino acids in one or more domains. Such changes may be included to optimize effector function, half-life, etc. 25 As used herein, the term "malignancy" refers to a non-benign tumor or a cancer. As used herein, the term "cancer" includes a malignancy characterized by deregulated or uncontrolled cell growth. Exemplary cancers include: carcinomas, sarcomas, leukemias, and lymphomas. The term "cancer" includes primary malignant tumors (e.g., those whose cells have not migrated to sites in the subject's body other than the site of 30 the original tumor) and secondary malignant tumors (e.g., those arising from metastasis, the migration of tumor cells to secondary sites that are different from the site of the original tumor). As used herein the term "engineered" includes manipulation of nucleic acid or 49 WO 2007/109254 PCT/US2007/006883 polypeptide molecules by synthetic means (e.g. by recombinant techniques, in vitro peptide synthesis, by enzymatic or chemical coupling of peptides or some combination of these techniques). Preferably, the binding molecules of the invention are made using such methods. 5 As used herein, the terms "linked," "fused" or "fusion" are used interchangeably. These terms refer to the joining together of two more elements or components, by whatever means including chemical conjugation or recombinant means. Preferably the polypeptides which are fused are genetically fused, i.e., are fused using recombinant DNA technology. An "in-frame fusion" refers to the joining of two or more open 10 reading frames (ORFs) to form a continuous longer ORF, in a manner that maintains the correct reading frame of the original ORFs. Thus, the resulting fusion protein is a single protein containing two ore more segments that correspond to polypeptides encoded by the original ORFs (which segments are not normally so joined in nature.) Although the reading frame is thus made continuous throughout the fused segments, the segments may 15 be physically or spatially separated by, for example, in-frame scFv linker sequence. In the context of polypeptides, a "linear sequence" or a "sequence" is an order of amino acids in a polypeptide in an amino to carboxyl terminal direction in which residues that neighbor each other in the sequence are contiguous in the primary structure of the polypeptide. 20 As used herein, the phrase "subject that would benefit from administration of a binding molecule" includes subjects, such as mammalian subjects, that would benefit from administration of a binding molecule used, e.g., for detection of an antigen recognized by a binding molecule (e.g., for a diagnostic procedure) and/or from treatment with a binding molecule to reduce or eliminate the target recognized by the 25 binding molecule. For example, in one embodiment, the subject may benefit from reduction or elimination of a soluble or particulate molecule from the circulation or serum (e.g., a toxin or pathogen) or from reduction or elimination of a population of cells expressing the target (e.g., tumor cells). As described in more detail herein, the binding molecule can be used in unconjugated form or can be conjugated, e.g., to a 30 detectable moiety, a drug, prodrug, or an isotope. The term "TNF receptor" or "TNF receptor family member" refers to any receptor belonging to the Tumor Necrosis Factor ("TNF") superfamily of receptors. Members of the TNF Receptor Superfamily ("TNFRSF") are characterized by an 50 WO 2007/109254 PCT/US2007/006883 extracellular region with two or more cysteine-rich domains (-40 amino acids each) arranged as cysteine knots (see Dempsey et al.,Cytokine Growth Factor Rev. (2003). 14(3-4):193-209). Upon binding their cognate TNF ligands, TNF receptors transduce signals by interacting directly or indirectly with cytoplasmic adapter proteins known as 5 TRAFs (TNF receptor associate factors). TRAFs can induce the activation of several kinase cascades that ultimately lead to the activation of signal transduction pathways such as NF-KappaB, JNK, ERK, p38 and PI3K, which in turn regulate cellular processes ranging from immune function and tissue differentiation to apoptosis. The nucleotide and amino acid sequences of several TNF receptors family 10 members are known in the art and include at least 29 human genes: TNFRSF1A (TNFRI, also known as DRI, CD120a, TNF-R-I p55, TNF-R, TNFRI, TNFAR, TNF R55, p55TNFR, p55R, or TNFR60, GenBank GI No. 4507575; see also US 5,395,760)), TNFRSF I B (CD 1 20b, also known as p75, TNF-R, TNF-R-II, TNFR80, TNFR2,TNF R75, TNFBR , or p75TNFR; GenBank GI No. 4507577), TNFRSF3 (Lymphotoxin 15 Beta Receptor (LTpR), also known as TNFR2-RP, CD18, TNFR-RP, TNFCR, or TNF R-I1I; GI Nos. 4505038 and 20072212), TNFRSF4 (OX40, also known as ACT35, TXGP1L, or CD134 antigen; GI Nos. 4507579 and 8926702), TNFRSF5 (CD40, also known as p50 or Bp50; GI Nos. 4507581 and 23312371), TNFRSF6 (FAS, also known as FAS-R, DcR-2, DR2, CD95, APO-1, or APTI; GenBank GI Nos. 4507583, 20 23510421, 23510423, 23510425, 23510427, 23510429, 23510431, and 23510434)), TNFRSF6B (DcR3, DR3; GenBank GI Nos. 4507569, 23200021, 23200023, 23200025, 23200027, 23200029, 23200031, 23200033, 23200035, 23200037, and 23200039), TNFRSF7 (CD27, also known as Tp55 or S 152; GenBank GI No. 4507587), TNFRSF8 (CD30, also known as Ki-1, or D IS166E; GenBank GI Nos. 4507589 and 23510437), 25 TNFRSF9 (4-1-BB, also known as CD137 or ILA; GI Nos. 5730095 and 728738), TNFRSF10A (TRAIL-RI, also known as DR4 or Apo2; GenBank GI No. 21361086), TNFRSF OB (TRAIL-R2,, also known as DR5, KILLER, TRICK2A, or TRICKB; GenBank GI Nos. 22547116 and 22547119), TNFRSF1OC (TRAIL-R3, also known as DcR1, LIT, or TRID; GenBank GI No. 22547121), TNFRSFIOD (TRAIL-R4, also 30 known as DcR2 or TRUNDD), TNFRSFI IA (RANK; GenBank GI No. 4507565; see US Patent Nos. 6,562,948; 6,537,763; 6,528,482; 6,479,635; 6,271,349; 6,017,729), TNFRSF1 lB (Osteoprotegerin (OPG), also known as OCIF or TR1; GI Nos. 38530116, 22547122 and 33878056), TNFRSF12 (Translocating chain-Association Membrane 51 WO 2007/109254 PCT/US2007/006883 Protein (TRAMP), also known as DR3, WSL-1, LARD, WSL-LR, DDR3, TR3, APO-3, Fnl4, or TWEAKR; GenBank GI No. 7706186; US Patent Application Publication No. 2004/0033225A1), TNFRSF12L (DR3L), TNFRSFI3B (TACI; GI No. 6912694), TNFRSF13C (BAFFR; GI No. 16445027), TNFRSF14 (Herpes Virus Entry Mediator 5 (HVEM), also known as ATAR, TR2, LIGHTR, or HVEA; GenBank GI Nos. 23200041, 12803895, and 3878821), TNFRSF16 (Low-Affinity Nerve Growth Factor Receptor (LNGFR), also known as Neurotrophin Receptor or p75(NTR); GenBank GI Nos. 128156 and 4505393), TNFRSF17 (BCM, also known as BCMA; GI No. 23238192), TNFRSF18 (AITR, also known as GITR; GenBank GI Nos. 4759246, 10 23238194 and 23238197), TNFRSF19 (Troy/Trade, also known as TAJ; GenBank GI Nos. 23238202 and 23238204), TNFRSF20 (RELT, also known as FLJ14993; GI Nos. 21361873 and 23238200), TNFRSF21 (DR6), TNFRSF22 (SOBa, also known as Tnfrh2 or 2810028K06Rik), and TNFRSF23 (mSOB, also known as Tnfrhl). Other TNF family members include EDARI (Ectodysplasin A Receptor, also known as Downless 15 (DL), ED3, ED5, ED IR, EDA3, EDA IR, EDA-A IR; GenBank GI No. 11641231; US Patent No. 6,355,782), XEDAR (also known as EDA-A2R; GenBank GI No. 11140823); and CD39 (GI Nos. 2135580 and 765256). The term "TNF ligand" or "TNF ligand family member" refers to a ligand belonging to the Tumor Necrosis Factor (TNF) superfamily. TNF ligands bind to 20 distinct receptors of the TNF receptor superfamily and exhibit 15-25% amino acid sequence homology with each other (Gaur et al., Biochem. Pharmacol. (2003), 66(8):1403-8). The nucleotide and amino acid sequences of several TNF Receptor (Ligand) Superfamily ("TNFSF") members are known in the art and include at least 16 human genes: TNFSFI (also known as Lymphotoxin-a (LTA), TNFp or LT, GI 25 No.:34444 and 6806893), TNFSF2 (also known as TNF, TNFa, or DIF; GI No. 25952111), TNFSF3 (also known as Lymphotoxin-p (LTB), TNFC, or p33), TNFSF4 (also known as OX-40L, gp3 4 , CD I34L, or tax-transcriptionally activated glycoprotein 1, 34kD (TXGP 1); GI No. 4507603), TNFSF5 (also known as CD40LG, IMD3, HIGMI, CD40L, hCD40L, TRAP, CD154, or gp39; GI No. 4557433), TNFSF6 (also 30 known as FasL or APTILGI; GenBank GI No. 4557329), TNFSF7 (also known as CD70, CD27L, or CD27LG; GI No. 4507605), TNFSF8 (also known as CD30LG, CD30L, or CD153; GI No. 4507607), TNFSF9 (also known as 4-1BB-L or ILA ligand; GI No. 4507609), TNFSF10 (also known as TRAIL, Apo-2L, or TL2; GI No. 4507593), 52 WO 2007/109254 PCT/US2007/006883 TNFSF1 I (also known as TRANCE, RANKL, OPGL, or ODF; GI Nos. 4507595 and 14790152), TNFSF12 (also known as Fn14L, TWEAK, DR3LG, or APO3L; GI Nos. 4507597 and 23510441), TNFSF13 (also known as APRIL), TNFSF14 (also known as LIGHT, LTg, or HVEM-L; GI Nos. 25952144 and 25952147), TNFSF15 (also known 5 as TL1 or VEGI), or TNFSF16 (also known as AITRL, TL6, hGITRL, or GITRL; GI No. 4827034). Other TNF ligand family members include EDARI & XEDAR ligand (EDI; GI No. 4503449; Monreal et al. (1998) Am JHum Genet. 63:380), Troy/Trade ligand, BAFF (also known as TALLI; GI No. 5730097), and NGF ligands (e.g. NGF-3 (GI No. 4505391), NGF-2/NTF3; GI No. 4505469), NTF5 (GI No. 5453808)), BDNF 10 (GI Nos. 25306267, 25306235, 25306253, 25306257, 25306261, 25306264; IFRD1 (GI No. 4504607)). The term "Tm", also referred to as the "transition temperature", is the temperature at which 50% of a macromolecule, e.g., binding molecule, becomes denatured, and is considered to be the standard parameter for describing the thermal 15 stability of a protein. As used herein the term "scaffolding residue" refers to amino acid residues or residue positions that are not in an interface (e.g., the VH/VL interface) but that are important in maintaining the interface. These amino acid residues do not physically interact with the interface residues on the opposing domain or contribute surface area to 20 the interface, but are nonetheless important for providing proper structural context for interface residues. Such amino acid residues scaffold the interaction between VH and VL. Two or more amino acid residue positions within a candidate polypeptide sequence that normally occur together are said to "covary" ("covarying residue 25 positions" or "covariant residue positions"). Covariance between two or more amino acid positions is observed when the type of amino acid found at a first amino acid position is dependent on the type of amino acid found at another amino acid position. That is, when one particular amino acid is found at a first position within a sequence, a second particular amino acid is usually found at a second position within the sequence. 30 As used herein the term "Ig fold" includes a protein domain found in proteins belonging to the immunoglobulin superfamily of proteins. As is well known in the art, the Ig fold is a distinguishing feature of the immunoglobulin superfamily (see, e.g. Bork, P., Holm, L. & Sander, C. 1994. The Immunoglobulin Fold. J. Mot. Biol. 242, 309-320). 53 WO 2007/109254 PCT/US2007/006883 Representative structures for each class of Ig fold are depicted in Figure 73. II. Stabilized scFv Molecules In one embodiment, a binding molecule of the invention is a stabilized 5 scFv molecule. The stabilized scFv molecules of the invention may comprisean scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain. In other embodiments, the stabilized scFv molecules of the invention comprisean scFv linker having an optimized length or composition. In yet 10 other embodiments, the stabilized scFv molecules of the invention comprise a VH or VL domain having at least one stabilizing amino acid substitution(s). In yet another embodiment, a stabilized scFv molecule of the invention comprises at least two of the above listed stabilizing features. The stabilized scFv molecules of the invention have improved stability. 15 In one embodiment, populations of the stabilized scFv molecules of the invention or polypeptides comprising the same are expressed as a monomeric, soluble protein of which is no more than 10% in dimeric, tetrameric, or otherwise aggregated form. In another embodiment, populations of the stabilized scFv molecules of the invention have VH and VL domains with Tm-values greater than 55*C. In another embodiment, 20 populations of the stabilized scFv molecules of the invention have a T50 of greater than 49*C. In another embodiment, populations of the stabilized scFv molecules of the invention have a T50 of greater than 40, 41, 42, 43, 44, 45, 46, 47, or 48*C. In yet another embodiment, populations of the stabilized scFv molecules of the invention have a T50 of greater than 50, 51, 52, 53, 54, 55, 56, 57, 58, or 49"C. 25 The scFv molecules of the invention bind to a target molecule of interest. The VH and VL domains used to makean scFv may be derived from the same or from different antibodies. In another embodiment, a VH or VL for use in a stabilized scFv of the invention may comprise one or more CDRs which bind to a target of interest, while the remainder of the VH or VL domain is derived from a different antibody or is 30 synthetic. In a preferred embodiment, a binding molecule of the invention comprises at least one CDR of an antibody, e.g., an antibody known in the art to bind to a target of interest. In another embodiment, a binding molecule of the invention comprises at least two CDRs of a given antibody. In another embodiment, a binding molecule of the 54 WO 2007/109254 PCT/US2007/006883 invention comprises at least three CDRs of a given antibody. In another embodiment, a binding molecule of the invention comprises at least four CDRs of a given antibody. In another embodiment, a binding molecule of the invention comprises at least five CDRs of a given antibody. In another embodiment, a binding molecule of the invention 5 comprises at least six CDRs of a given antibody. In a preferred embodiment, a binding molecule of the invention comprises at least one VH domain of an antibody, e.g., an antibody known in the art to bind to a target of interest. In a preferred embodiment, a binding molecule of the invention comprises at least one VL domain of a given antibody. In another preferred embodiment, a binding molecule of the invention 10 comprises at least one VH domain and one VL domain of an antibody known in the art to bind a target of interest. scFv molecules can be constructed in a VH-linker-VL orientation or VL-linker-VH orientation. The stability of scFv molecules of the invention or fusion proteins comprising them can be evaluated in reference to the biophysical properties (e.g., thermal stability) 15 of a conventional (non-stabilized) scFv molecule or a binding molecule comprising a conventional scFv molecule. In one embodiment, the binding molecules of the invention have a thermal stability that is greater than about 0.1, about 0.25, about 0.5, about 0.75, about 1, about 1.25, about 1.5, about 1.75, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, or about 10 degrees Celsius than a control binding 20 molecule (eg. a conventional scFv molecule). Stabilized scFv molecules of the invention include those identified using methods of the invention as described in Section VIII infra. In other embodiments, the stabilized scFv molecules of the invention comprise an scFv linker with an optimized length and/or amino acid composition. Preferred scFv 25 linkers of the invention improve the thermal stability of a binding molecule of the invention by at least about 2 0 C or 3"C as compared to a conventional binding molecule. In one embodiment, a binding molecule of the invention has a 1 "C improved thermal stability as compared to a conventional binding molecule. In another embodiment, a binding molecule of the invention has a 2 "C improved thermal stability as compared to a 30 conventional binding molecule. In another embodiment, a binding molecule of the invention has a 4, 5, 6*C improved thermal stability as compared to a conventional binding molecule. Comparisons can be made, for example, between the scFv molecules of the invention and scFv molecules made using prior art methods or between scFv 55 WO 2007/109254 PCT/US2007/006883 molecules and fab fragments of an antibody from which the scFv VH and VL were derived. Thermal stability can be measured using methods known in the art. For example, in one embodiment, Tm can be measured. Methods for measuring Tm and other methods of determining protein stability are described in more detail below. 5 In one embodiment, the scFv linker consists of the amino acid sequence (Gly 4 Ser) 4 or comprises a (Gly 4 Ser) 4 sequence. Other exemplary linkers comprise or consist of (Gly 4 Ser) 3 and (Gly 4 Ser)s sequences. scFv linkers of the invention can be of varying lengths. In one embodiment,an scFv linker of the invention is from about 5 to about 50 amino acids in length. In another embodiment,an scFv linker of the invention 10 is from about 10 to about 40 amino acids in length. In another embodiment,an scFv linker of the invention is from about 15 to about 30 amino acids in length. In another embodiment,an scFv linker of the invention is from about 17 to about 28 amino acids in length. In another embodiment,an scFv linker of the invention is from about 19 to about 26 amino acids in length. In another embodiment,an scFv linker of the invention is from 15 about 21 to about 24 amino acids in length. scFv linkers can be introduced into polypeptide sequences using techniques known in the art. For example, in one embodiment, PCR mutagenesis can be used. Modifications can be confirmed by DNA sequence analysis. Plasmid DNA can be used to transform host cells for stable production of the polypeptides produced. 20 In certain embodiments, the stabilized scFv molecules of the invention comprise at least one disulfide bond which links an amino acid in the VL domain with an amino acid in the VH domain. Cysteine residues are necessary to provide disulfide bonds. Disulfide bonds can be included inan scFv molecule of the invention, e.g., to connect FR4 of VL and FR2 of VH or to connect FR2 of VL and FR4 of VH. Exemplary 25 positions for disulfide bonding include: 43, 44, 45, 46, 47, 103, 104, 105, and 106 of VH and 42, 43, 44, 45, 46, 98, 99, 100, and 101 of VL, Kabat numbering. Exemplary combinations of amino acid positions which are mutated to cysteine residues include: VH44-VL 100, VH 105-VL43, VH 105-VL42, VH44-VL 101, VH106-VL43, VH 104 VL43, VH44-VL99, VH45-VL98, VH46-VL98, VH103-VL43, VH103-VL44, and 30 VH103-VL45. In one embodiment, a disulfide bond links VH amino acid 44 and VL amino acid 100. Modifications of the genes which encode the VH and VL domains may be 56 WO 2007/109254 PCT/US2007/006883 accomplished using techniques known in the art, for example, site-directed mutagenesis. In one embodiment, a stabilized scFv molecule of the invention comprisesan scFv linker having the amino acid sequence (Gly4 Ser) 4 interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond 5 between an amino acid in the VH at amino acid position 44 and an amino acid in the VL at amino acid position 100. In other embodiments the stabilized scFv molecules of the invention comprise one or more stabilizing mutations within a variable domain (VH or VL) of the scFv. In one embodiment, the stabilizing mutation is selected from the group consisting of: 10 a) substitution of an amino acid (e.g., glutamine) at Kabat position 3 of VL, e.g., with an alanine, a serine, a valine, an aspartic acid, or a glycine; b) substitution of an amino acid (e.g., serine) at Kabat position 46 of VL, e.g., with leucine; c) substitution of an amino acid (e.g., serine) at Kabat position 49 of VL, e.g., with 15 tyrosine or serine; d) substitution of an amino acid (e.g., serine or valine) at Kabat position 50 of VL, e.g., with serine, threonine, and arginine, aspartic acid, glycine, or lysine; e) substitution of amino acids (e.g., serine) at Kabat position 49 and (e.g., serine) at Kabat position 50 of VL, respectively with tyrosine and seine; tyrosine and 20 threonine; tyrosine and arginine; tyrosine and glysine; serine and arginine; or serine and lysine; f) substitution of an amino acid (e.g., valine) at Kabat position 75 of VL, e.g., with isoleucine; g) substitution of an amino acid (e.g., proline) at Kabat position 80 of VL, e.g., with 25 serine or glycine; h) substitution of an amino acid (e.g., phenylalanine) at Kabat position 83 of VL, e.g., with serine, alanine, glycine, or threonine; i) substitution of an amino acid (e.g., glutamic acid) at Kabat position 6 of VH, e.g., with glutamine; 30 j) substitution of an amino acid (e.g., lysine) at Kabat position 13 of VH, e.g., with glutamate; k) substitution of an amino acid (e.g., serine) at Kabat position 16 of VH, e.g., with glutamate or glutamine; 57 WO 2007/109254 PCT/US2007/006883 1) substitution of an amino acid (e.g., valine) at Kabat position 20 of VH, e.g., with an isoleucine; m) substitution of an amino acid (e.g., asparagine) at Kabat position 32 of VH, e.g., with serine; 5 n) substitution of an amino acid (e.g., glutamine) at Kabat position 43 of VH, e.g, with lysine or arginine; o) substitution of an amino acid (e.g., methionine) at Kabat position 48 of VH; e.g., with an isoleucine or a glycine; p) substitution of an amino acid (e.g., serine) at Kabat position 49 of VH, e.g, with 10 glycine or alanine; q) substitution of an amino acid (e.g., valine) at Kabat position 55 of VH, e.g., with a glycine; r) substitution of an amino acid (e.g., valine) at Kabat position 67 of VH, e.g., with an isoleucine or a leucine; 15 s) substitution of an amino acid (e.g., glutamic acid) at Kabat position 72 of VH, e.g., with aspartate or asparagine; t) substitution of an amino acid (e.g., phenylalanine) at Kabat position 79 of VH, e.g., with serine, valine, or tyrosine; and u) substitution of an amino acid (e.g., proline) at Kabat position 101 of VH, e.g., 20 with an aspartic acid. v) In an exemplary embodiment, a stabilized scFv molecule of the invention comprises two or more of the stabilizing mutations described in a) through u) above. In an exemplary embodiment, a stabilized scFv molecule of the invention 25 comprises a substitution of an amino acid at Kabat position 49 of VL, e.g., with tyrosine or serine and a substitution of an amino acid at Kabat position 50 of VL, e.g., with, threonine, and arginine, aspartic acid, glycine, or lysine. In another exemplary embodiment, a stabilized scFv molecule of the invention comprises a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate 30 or glutamine and a substitution of an amino acid at Kabat position 46 of VL, e.g., with a lysine. In another embodiment a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine; a substitution of an amino acid at Kabat position 46 of 58 WO 2007/109254 PCT/US2007/006883 VL, e.g., with a lysine; and a substitution of an amino acid at Kabat position 55 of VH, e.g., with a glycine. In another embodiment a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 16 of VH, e.g., with a glutamate or glutamine; a substitution of an amino acid at Kabat 5 position 46 of VL, e.g., with a lysine; a substitution of an amino acid at Kabat position 55 of VH, e.g., with a glycine; and substitution of an amino acid at Kabat position 101 of VH, e.g., with an aspartic acid. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 6 of VH, e.g., with glutamine. In another embodiment, a 10 binding molecule of the invention comprises a substitution of an amino acid at Kabat position 13 of VH, e.g., with glutamate. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 16 of VH, e.g., with glutamate or glutamine. In another embodiment, a binding molecule of the invention comprises a substitution of an 15 amino acid at Kabat position 20 of VH, e.g., with an isoleucine; In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 32 of VH, e.g., with serine. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 43 of VH, e.g, with lysine or arginine. In another 20 embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 48 of VH, e.g., with an isoleucine or a glycine. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position substitution 49 of VH, e.g, with glycine or alanine. In another embodiment, a binding molecule of the invention 25 comprises a substitution of an amino acid at Kabat position 55 of VH, e.g., with a glycine. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 67 of VH, e.g., with an isoleucine or a leucine. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 72 of VH, 30 e.g., with aspartate or asparagines. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat position 79 of VH, e.g., with serine, valine, or tyrosine. In another embodiment, a binding molecule of the invention comprises a substitution of an amino acid at Kabat 59 WO 2007/109254 PCT/US2007/006883 position 101 of VH, e.g., with an aspartic acid. In another exemplary embodiment, a stabilized scFv molecule of the invention comprises one or more of the stabilizing amino acid substitutions described herein andan scFv linker with an optimized length or composition (e.g. (Gly 4 Ser) 4 ). In another 5 exemplary embodiment, a stabilized scFv molecule of the invention comprises one or more of the amino acid substitution described herein and a disulfide bind which links an amino acid in the VL domain with an amino acid in the VH domain (e.g. VH44-VL100). In yet another exemplary embodiment, a stabilized scFv molecule of the invention comprises one or more of the amino acid substitutions described herein, an scFv linker 10 with an optimized length or composition (e.g. (Gly 4 Ser) 4 ), and a disulfide bind which links an amino acid in the VL domain with an amino acid in the VH domain (e.g. VH44 VL100). Stabilized scFv molecules may be expressed using art recognized techniques. For example, in one embodiment, such molecules may be expressed using an expression 15 vector appropriate for expression in a cellular expression system, e.g., a bacterial or mammalian expression system. In one embodiment, scFv molecules may be expressed in E. coli, e.g., using a vector appropriate for periplasmic expression. Additional sequences may be included to optimize expression, e.g., a signal sequence and/or a tag to facilitate purification and/or 20 detection of the scFv. In one embodiment, additives, such as 1-2% triton (e.g., triton x-100) or 1-2% glysine or a combination thereof (e.g., 1% glycine and 1% triton) may be added to facilitate secretion from the periplasm into the medium. 25 III. Methods for Predicting/Determining Protein Stability In certain aspects, the invention provides methods for predicting, a priori, potentional biophysical problems with proteins selected for large-scale expression, e.g. therapeutic proteins or industrial enzymes. In certain exemplary aspects, the methods of the invention allow one skilled in the art to avoid expressing protein sequences that are 30 predicted to have inherently poor stability when recombinantly expressed, for example, in mammalian cells. In alternative aspects, the methods of the invention may be employed to identify a variant of a protein sequence that is predicted to have improved biophysical properties, including, but not limited to improved stability, improved 60 WO 2007/109254 PCT/US2007/006883 stability to changes in pH, and enhanced biochemical function. In certain aspects, the invention provides computational methods for predicting a biophysical property of (e.g., thermal stability) a candidate proteins, or a sequence variant or homolog thereof, based on the polypeptide sequence (e.g. amino acid 5 sequence) of the candidate protein. In other aspects, the methods of the invention employ structural modeling to predict the stability of the candidate protein or to identify homologs or variants of the candidate protein with known exemplary stability that can be used to improve the stability of a candidate protein, e.g., an scFv molecule. 10 a. Covariation Analysis In certain exemplary aspects, the invention provides an improved computational method termed "covariation analysis" for predicting a biophysical property (e.g.protein stability). As used herein, "covariation analysis" refers to a computational method for identifying two or more amino acid residue positions within a candidate polypeptide 15 sequence that normally occur together, or covary, in homologs of the candidate sequence (herein "covarying residue positions" or "covariant residue positions"). Covariance between two or more amino acid positions is observed when the type of amino acid found at a first amino acid position is dependent on the type of amino acid found at another amino acid position. That is, when one particular amino acid is found at a first 20 position within a sequence, a second particular amino acid is usually found at a second position within the sequence. Since covariant residues are likely to have coevolved, an observation of covariation indicates that the compatibility of amino acid residues residing at normally covarying positions is likely to be important for functional or structural reasons. 25 Accordingly, the covariation method of the invention can be used to identify one or more covarying residues or groups of residues.(e.g. covarying residue pairs) within a polypeptide sequence (e.g. a candidate polypeptide sequence). As used herein, a "covarying residue" (also termed a "linked residue") is an amino acid which is statistically prevalent at a covarying residue position within a polypeptide sequence. 30 In certain embodiments, the covariation methods of the invention can be used to identify important functional residues in a candidate sequence. Moreover, since the number of covarying residues within a candidate polypeptide is predictive of its stability, the covariation method of the invention can be used to predict the stability candidate 61 WO 2007/109254 PCT/US2007/006883 protein. In other certain embodiments, the covariation methods of the invention can be employed to guide successful protein designs. In one exemplary embodiment, the covariation method of the invention can be used to identify covarying amino acids 5 within a candidate sequence. If present, the covarying residues are preferably retained during subsequent engineering of the protein sequence. In another exemplary embodiment, the covariation methods of the invention can be used to identify the amino acid positions of non-covarying residues within a candidate polypeptide sequence wherein corresponding amino acids in other proteins (e.g., proteins corresponding to the 10 sequences of a reference set) are normally covarying amino acids. Once identified, non covarying residues can be substituted (e.g. using recombinant DNA methodology) with corresponding covarying residues to enhance function or improve the biophysical properties (e.g., stability) of the candidate polypeptide sequence. Covariant residue positions may be identified by a statistical analysis (e.g., a 15 correlation analysis) of residue positions within a database of related polypeptide sequences (e.g., an aligned reference set or multiple sequence alignment). In preferred embodiments, the covariation analysis involves a statistical comparison of candidate polypeptide sequences against a curated database of diverse polypeptide sequences having the same structural fold (e.g. an Ig fold). 20 In the novel covariation analysis described herein, the following basic steps are taken to predict the stability of a candidate polypeptide: 1. Providing a set of homologous sequences (herein, a "reference set") corresponding to a sequence of a candidate polypeptide or a domain or portion thereof (herein a "test domain sequence"). 25 2. Aligning the sequences of the reference set to generate an aligned set of homologous sequence (herein, an "aligned set). 3. Determining covariation between two or more residue positions (e.g. at least one pair of amino acids) within the reference set to generate covariation data or a set of covariation data ("covariation dataset"). 30 4. Using the covariation data to predict the stability of the candidate polypeptide. Step 1: Providing a Reference Set The covariation methods of the invention employ a set of of sequences (a 62 WO 2007/109254 PCT/US2007/006883 "reference set") which are homologous to (i.e., related to) a sequence of interest or "test sequence". The reference set may comprise a set of homologous sequences having moderate to high degrees of similarity with the test sequence. In certain embodiments, the homologous sequences of the reference set have a moderate to high degree of amino 5 acid sequence similarity. In more preferred embodiments, the homologous sequences of the reference set have a moderate to high degree of structural similarity. In still more preferred embodiments, the homologous sequences of the reference set have a high degree of structural similarity (e.g. they share at least one protein domain or fold). The set of homologous sequences need not be large, as long as a reasonably unbiased 10 selection is obtained. In exemplary embodiments, the covariation method of the invention employs a curated reference set. As used herein, the term "curated reference set" refers to a reference set in which component sequences have been deleted or additional sequences 15 added according to certain selection criteria. Accordingly, whereas a non-curated reference is generated by unbiased selection of component sequences, a curated database is generated by biased selection. For example, the reference set of homologous sequences may be culled to eliminate certain sequences, e.g. redundant sequence. Alternatively, a curated database may be expanded, e.g. to include a suitable number of 20 non-redundant or non-identical sequences. In certain embodiments, the curated database comprises at least one hundred sequences (e.g. 100, 150, 200, 250, 300, 400, 500, 750, or more sequences). In one embodiment, the curated database comprises at least one thousand sequences (e.g. 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, or more sequences). In another 25 embodiment, the curated database comprises at least ten thousand sequences (e.g., 10,000, 20,000, 50,000, 100,000, 250,000, 500,000, 750,000 or more sequence. In one embodiment the curated database comprises at least I million individual sequences (e.g., 1 million, 2 million, 5 million, 10 million, or more sequences). In preferred embodiments, the covariation method of the invention employs a 30 curated reference set having at least 50% diversity (e.g. at least 60%, at least 70%, at least 75%, at least 80%, at least 90%, at least 95% diversity, at least 97%, at least 98%, at least 99% or more diversity) in order to minimize artificial or non-informative covariations. As used herein, the term "% diversity" refers to the percentage of 63 WO 2007/109254 PCT/US2007/006883 sequences in the reference set that are non-redundant or non-identical. As used herein, the term "non-redundant" refers to a sequence which has less than 100% sequence identity with every other sequence in the reference set, i.e., a sequence which differs at least in one amino acid position from every other sequence in the reference set. In 5 certain embodiments, a sequence of the reference set has less than 99%, less than 95%, less than 90%, or less than 85% sequence identity with every other sequence in the reference set. In preferred embodiments, a sequence of the reference set -has less than 80% sequence identity with every other sequence in the reference set (e.g. less than 75%, less than 70%, less than 65%, less than 60%, or less than 50% sequence identity). 10 In particularly preferred embodiments, every sequence of the reference set has less than than 80% sequence identity with every other sequence in the reference set. In exemplary embodiments, the sequences of the reference set are curated for the preferred diversity using a filtering tool which culls sequences containing more common sequence types (e.g., a Henikoff sorting tool). When two sequences are equal in length, sorting by 15 Henikoff weight ensures that sequences with rare residue types are retained while sequences with more common residue types are discarded (Henikoff et aL, J Mo. Biol., 243: 574-578 (1994)). In other preferred embodiments, the reference set comprises sequences encoded by genes which are not evolutionarly close to genes encoding other sequences in the 20 database. For example, to increase diversity, the reference set may comprise one or more ortholog sequences, that is a sequence which is from a different species (e.g., different mammalian species)) but which has the same or similar biochemical function. In one exemplary embodiment, the reference set may comprise one human sequence and at least one non-human sequence (e.g., a non-human mammalian sequence). In another 25 exemplary embodiment, the reference set may comprise at least one sequence for a mammalian species (e.g,. a human, chimpanzee, dog, cow, pig, cat, rat, or mouse) and at least one non-mammalian sequence (e.g. a non-mammalian vertebrate sequence (e.g., a teleost or avian sequence), an invertebrate sequence (e.g., an insect (e.g., Drosophila) or nematode (e.g., C. elegans) sequence), or a fungal, plant, bacterial, or virus sequence). 30 In other embodiments, the reference set may comprise one or more paralog sequences, that is sequences from the same species as a first sequence, but which has the same or similar biochemical function. In yet other embodiments, the sequences of the reference set are above a certain 64 WO 2007/109254 PCT/US2007/006883 threshold length. Preferably, the sequences of the reference set are at least 20 residues in length (e.g., 25, 30, or 40 residues in length). More preferably, the sequences of the reference set are at least 50 residues in length (e.g., 60, 65, 70, 75, 80, 85, 90, or 95 residues in length). Still more preferred are sequences of at least 100 residues in length 5 (e.g., 110, 120, 130, 140, 150, 160, 170, 180, or more residues in length). In exemplary embodiments, the sequences of the reference set are curated for the preferred length by filtering using a filtering tool (e.g., a non-gap residue count). Ranking by decreasing non-gap residue count ensures that shorter sequences (e.g., gapped sequences) are filtered out over longer ones. 10 The sequences can be obtained through sequence-based homology searches for any of the sequence databases known in the art (e.g. NCBI, TIGR databases). In an exemplary embodiment, a standard BLAST search with default parameters may be performed with the test sequence to retrieve homologous sequences of interest. Sequences with a minimum percentage of sequence identity (e.g. greater than 25%, 15 30%, 35%, 40%, 45% or 50% sequence identity, preferably greater than 30% sequence identity) to the test sequence may be selected for inclusion in the reference set. In certain preferred embodiments, the reference set is curated to comprise sequences from proteins having the same class of protein fold (e.g., proteins having SH3, TPR, GPCR, serine protease, aspartyl protease, globin, immunoglobulin, or other 20 folds). Such proteins or structures may be identified and collected by bioinformatics techniques known in the art (e.g., be structure-based searching using the Conserved Domain Database (CDD) of the National Institutes of Health), or by searching the ASTRAL database or the RCSB protein database (PDB) using a desired SCOP classification). Preferably, the sequences are derived from proteins whose 3 25 dimensional structures have been solved at high resolution, for example, by X-ray crystallography. In preferred embodiments, sequences corresponding to collected structures are filtered using art-recognized techniques to remove erroneously categorized, incomplete, redundant, or domain-swapped structures. In one embodiment, the structures are 30 visually inspected (e.g., using the SwissPDB Viewer) for breaks in the corresponding sequence due to unresolved densities or domain swapping. The PDB files of faulty structures are manually removed from the corresponding set of structures. In another embodiment, sequences (e.g. FASTA sequences) corresponding to the collected 65 WO 2007/109254 PCT/US2007/006883 structures are filtered to remove any sequences that are 100% identical, or perfect match substrings of remaning sequences. In yet other embodiments, PDB structures with aberrantly long or short amino acid sequences (a hallmark of erroneous structural categorization) are culled from the structure datasets. The length cutoff critera may be 5 determined, for example, by examining a histogram of all sequence lengths of sequences of the reference set. In still other embodiments, the structures are visually inspected (e.g., using the Swiss PDB viewer) for misfolding. Any structures that do not conform with the standard topology of the protein fold are preferably discarded. In exemplary embodiments, the curated reference set comprises sequences 10 corresponding to an immunoglobulin-type fold ("Ig fold") of a protein belonging to the immunoglobulin superfamily of proteins. As is well known in the art, the Ig fold is a distinguishing feature of the immunoglobulin superfamily (see, e.g. Bork, P., Holm, L. & Sander, C. 1994. The Immunoglobulin Fold. J Mol. Biol. 242, 309-320). The Ig fold appears often in mammalian proteins and serves as a platform for various functions 15 particularly protein-protein interactions. For example, all immunoglobulins and most immunoglobulin receptors are composed of multiple Ig-fold domains. Ig folds may be subdivided into several subfamilies including Cl, C2, I, and V. Although members of all the subfamilies comprise a two- sheet, greek-key topology, these subfamilies differ by the number of P-strands on each sheet and by the connections across p-sheets (see, 20 e.g., AF Williams, Immunology Today, 8, (1987)). Representative structures for each class of Ig fold are depicted in Figure 73. In one embodiment, the reference set comprises an Ig-fold sequence of an immunoglobulin superfamily protein selected from the group consisting of a cell adhesion protein, a integrin, an allergen, a T-cell receptor, a major histocompatibility 25 complex protein (e.g., a MHC Class I or MHC Class II protein), an immunoglobulin receptor (e.g., an Fc gamma receptor (e.g. FcyRl, FcyRlla)), and an immunoglobulin (e.g., an IgG, IgM, IgA, or IgE immunoglobulin). In certain preferred embodiments, all of the sequences of the reference set correspond to an Ig-fold or portion thereof. In one embodiment, the Ig-fold is a Cl 30 fold. In another embodiment, the Ig-fold is a C2-fold. In another embodiment, the Ig fold is a V-fold. In yet another embodiment, the Ig-fold is an I-fold. In another embodiment, all of the sequences of the reference set correspond to a Cl fold. In another embodiment, all of the sequences of the reference set correspond to a C2-fold. 66 WO 2007/109254 PCT/US2007/006883 In another embodiment, all of the sequences of the reference set correspond to an I-fold. In certain preferred embodiments, the sequences of Ig-fold reference set are within about 75 to about 150 residues in length. Step 2: Aligning the Reference Set 5 In the second step, the sequences of the reference set are precisely aligned to generate an aligned set of sequences (or alignment). In one embodiment, the sequences of the reference set are aligned using sequence alignment algorithms (e.g. LALIGN or BLAST and other art-recognized alignments described supra) to create an aligned set of sequences (herein a "sequence-based sequence alignment"). In another embodiment, 10 strucutures corresponding to the sequences of the reference set are aligned using a structural alignment algorithm (e.g. Secondary Structure Matching (SSM) using the Schr6dinger structalign package) to create an aligned set of structures (herein, a "structure alignment"). Preferably, the structure alignment algorithm ensures that core regions (e.g., 3 strands) of each structure are aligned, instead of intervening loops. In 15 yet another embodiment, the sequences of the reference set are aligned by structure based methods (e.g., by matching amino acids from one superimposed structure to that of another superimposed structure based on the shortest distance between the a-carbons of the polypeptide backbones, e.g., using the Schrodinger package). In another embodiment, the sequences of the reference set are aligned using both 20 sequence-based and structure-based alignment algorithms. Preferably, the structures corresponding to the sequences of the reference set are first aligned using a structure based alignment algorithm or other structure-guided alignment tools, followed by alignment of the corresponding sequences with a sequence-based algorithm. More preferably, the structures corresponding to the sequences of the reference set are first 25 aligned, followed by alignment of the sequences using a structure-based sequence alignment. In certain optional embodiments, the aligned set of sequences is further curated. For example, the aligned set of sequences can be expanded (e.g., by non-gap sorting or by Henikoff sorting) to increase its diversity (e.g., to obtain very large sets of homolgous 30 sequences having the same fold). In other embodiments, the aligned set of sequences can be further subjected to several rounds of curation and, optionally, further alignment (e.g. structure-based alignment). In certain preferred embodiments, the invention provides methods for curating an 67 WO 2007/109254 PCT/US2007/006883 aligned set containing a relatively small number of sequences in order to enhance its diversity. By increasing the number of sequence in the alignment, the robustness of the covariation analysis is increased. In one embodiment, the methods may employ a heuristically-derived model or profile to search for additional homologous sequences 5 from a large, publically-available, sequence database (e.g., the NR database maintained by the NCBI). The heuristically-derived models may be produced using one or more regression-based algorithms selected from, e.g., a partial least squares regression, a multiple linear regression, an inverse least squares regression, a principal component regression, a variable importance for projection, or the like. In other embodiments, the 10 heuristically-derived model is produced using one or more pattern-based algorithm selected from, e.g., a Hidden Markov model, a Smith Waterman algorithm, neural network, a classification and regression tree, a multivariate adaptive regression spline, or the like. In an exemplary embodiment, the pattern recognition algorithm is a Hidden Markov Model. An HMM is a statistical model where the system being modeled 15 contains hidden parameters which can be extracted using observable parameters and employed for analysis of pattern recognition. For example, art-recognized HMMs may employ predicted secondary structures to search for antibodies of the same fold as a given antibody, rather than antibodies having purely sequence-based relatedness (see e.g., Proteins 36(1), 68-76). MetaFam (Silverstein et al. Nucleic Acids Res 29(1), 49-51 20 (2001)), Interpro (Apweiler, et al., Nucleic Acids Res, 29(1), 37-40, (2001)), and HMMER (Bateman et al, Nucleic Acids Res 27(1), 260-2 (1999)) are exemplary HMM based pattern recognition tools. In certain additional embodiments, sequences extracted with a HMM algorithm may be validated for their proper structural class assignment using art-recognized 25 bioinformatics software. An exemplary validation tool is the PFAM classification tool maintained by the Wellcome Trust Sanger Institute and is pubicly available on the internet. For example, the PFAM tool 'pfamverify' may be applied to each extracted sequence to confirm that it was correctly classified by the class-specific HMM created from the structure-based structure alignment (Finn et al., Nucleic Acids Res, 30 24(Database issue), D247-51, (2006). HMM profiles corresponding to a PFAM clan (or related family of proteins, e.g. an Ig-fold clan) may be downloaded from the PFAM website to facilitate scoring of the extracted sequences. Extracted sequences whose scores lay below a recommended cutoff are preferably removed from the reference set. 68 WO 2007/109254 PCT/US2007/006883 In other embodiments, sequences extracted with the HMM algorithm may also be aligned using the HMMC algorithm. Since these HMM algorithms are based upon careful structural alignments, this process insures the structure guided alignment of the additional extracted sequences. For example, the HMMER package may be utilized to 5 generate 'mapali' alignments in FASTA output format. In preferred embodiments, the method of the invention provides a novel H1MM which not only finds sequences similar to those of the aligned set, but aligns the new sequences to insure the correct core region alignment. For example, the novel HMM algorithm of the invention can search for separate domains (e.g., Ig-fold domains) within 10 a single protein (e.g., an immunoglobulin superfamily member), extract each domain separately, and add it to the aligned set of sequences. HMM algorithms or profiles may be built from structure-based sequence alignments (e.g., using the HMMER software package which is publically available on the internet, e.g., at www.psc.edu). For each structure-specific HMM search, hit sequences above a criterion score threshold are 15 preferably retained as candidate members of the structural class whose HMM was used. For those hit regions that are subsequences of hit sequence, the exact subsequence hit may be extracted from the full sequence. In other exemplary embodiments, the invention provides methods for curating the aligned set of sequences in order to reduce its redundancy. For example, while the 20 large number of sequences in a structurally aligned set (e.g., an Ig-fold aligned set) . represents a rich source of sequences, some subfamilies (e.g., certain Ig-fold subclasses) may be highly over-represented, while other are highly under-represented. Over- or under- representation results in significant bias due to close common ancestry and limits the overall usefulness of covariation analyses. Accordingly, in preferred embodiments, 25 the alignment is subjected to further filtering steps to reduce alignment redundancy. In preferred embodiments, at least one sequence with greater than 90% identity to another sequence (e.g., 91%, 92%, 93%, 95%, 96%, 97%, 98%, 99% or more sequence identity) is removed from the alignment. In particularly preferred embodiments, at least one sequence sharing greater than 80% sequence identity with another sequence (e.g., 81%, 30 82%, 83%, 84%, 87%, 88%, 89%, or 90% sequence identity) is removed from the alignment. In exemplary embodiments, the invention provides a novel method to reduce redundancy within the alignment. Such methods comprise use of a heuristic algorithm 69 WO 2007/109254 PCT/US2007/006883 to find and rank sequences of desired sequence identity cutoff. Such methods comprise one of more of the following sequential steps: (1) calculating percent identities for all pairs of sequences in the alignment; (2) grouping identity values into one or more bins of percent identity; (3) ranking the sequences of each bin by descreasing non-gap residue 5 count and/or by Henikoff sequence weight; and (4) removing redundant sequences according to a cutoff criterion (e.g.,-a % identity cutoff level). Rernoval of the redundant sequences can be performed'in a number of ways. In one embodiment, ranked sequences are grouped into multiple bins of sequence identity (e.g., a 99% bin, a 98% bin, a 97% bin, etc.). Sequences are then systematically 10 removed by rank (e.g., highest ranked sequences first) from the highest % identity bin (ie., the 99% bin, followed by the 98% bin, followed by the 97% bin, etc.). Identity calculation, grouping and/or ranking steps (Steps (1)-(3)) are optionally repeated after each removal step until the final bin that meets the cutoff criterion is eliminated. In another embodiment, a single bin of ranked sequences is created at the cutoff criterion 15 and sequences are systematically removed by rank from the bin. Identity calculation and/or ranking steps (Step (1) and (3)) are optionally repeated after each removal step. In preferred embodiments, the cutoff criterion is 90% or higher (e.g., 91%, 92%, 93%, 95%, 96%, 97%, 98%, 99% or more sequence identity). In particularly preferred embodiments, the cutoff criterion is 80% or higher (e.g., 81%, 82%, 83%, 84%, 87%, 20 88%, 89%, or 90% sequence identity). In another optional step, the lengths of sequences in the aligned reference set may be truncated. In one embodiment, gapped regions within the alignment may be removed to avoid calculation on these less informative regions. For example, columns that are not match states in an HMM profile used to find the sequences may be removed. 25 In another embodiment, sequences within the alignment which overlap the consensus length of the alignment are cropped to remove the overhanging portion of the sequence. In other embodiment, overlapping sequences are removed from the alignment. Step 3: Covariation Analysis of the Aligned Set 30 Once the sequences of the aligned set are compiled, covariation analysis is performed on one or more pairs of amino acid residues in the alignment set. The covariation analysis results in the generation of a novel dataset (herein, a "covariation dataset") describing the statistical significance of correlated residues within the 70 WO 2007/109254 PCT/US2007/006883 alignments. In certain embodiments, the covaration dataset lists correlations between every possible pair of residues within the alignment. In preferred embodiments, computation of covariation is a computer implemented process (e.g., a Java-executable process). Computation of covariation may 5 result in several art-recognized statistical parameters. In one embodiment, the statistical significance (e.g. a X 2 -value) of covariation is calculated using a Chi-square analysis. In another embodiment, the statistical strength (e.g. * value) of covariation is calculated. For example, a negative $ value may be predictive of a negative correlation. That is, the presence of one amino acid at a first position in the alignment favors the absence of 10 another specific amino acid at a second position. Conversely, a positive 4 value may be predictive of a positive correlation, thereby indicating that the amino acid at the first position favors the presence of the amino acid at the second position. In one embodiment, a # value is calculated using formula I: 15(aib * -b ) -(aib; * (bI) V(a,b, +ibj )* (ajbj +Zb;) * (a,bj + a bj) * (Ub + Ub;) wherein aibj is the number of times residues of type "a" or "b" are found in the same 20 sequence at positions i and j, respectively; d56 is the number of times both residue types are absent from the same sequence; a5j is the number of times a is found present while b is absent; and Hibj is the number of times a is absent while b is present. 25 In one embodiment, a 4 value is calculated using the following formula II: 0(abj =(c *f)-(d *e)0) 30 where c through j are defined by the matrix: ai di Total 71 WO 2007/109254 PCT/US2007/006883 bjc d g 5e f h Total i j 5 and c = abj, d =ibj, eai5 , f = i 5 .g = c + d, h = e + f, i = c + e, andj =d + f. In certain embodiments, a X 2 -value may be calculated using a "frequency-of 10 occurrence" based formula. In other embodiments, a x 2 -value is calculated using an "event-based" formula (ie. number of occurrences). An exemplary event-based formula is set forth as Formula III: 2 [c(i, j)-(p(i)- p(j) c(t))] 2 = p(i)-p(j)-c(t) 15 wherein p(i) and p(j) are the residue frequencies of any two residue types of interest at positions i and j, respectively, in the aligned set of sequences; c(ij) is number of times p(i) and p(j) are observed in the same sequence; and c(t) is the number of total sequences in the alignment; 20 and wherein residue frequencies are defined as the number of times a residue type is observed at a specific position in an alignment divided by the total number of sequences in the alignment. In certain embodiments, only a natural amino acid residue is considered a "residue type" for the purposes of a covariation calculation. In preferred embodiments, 25 however, a residue type may also include a gap as a distinct residue type, since gaps (especially in loop portion) often help to discriminate one motif from another. In certain embodiments, the covariation calculation may include a diversity weighting function (e.g., a Henikoff weighing scheme). In other embodiments, Sequence Average Identities (SAIs) may be used to filter 30 out covarying pairs. For example, SAIs may be used to filter out covariation from pairs with greater than average sequence identity in order to remove potential artifactual covariations arising from closely related sequences. In embodiments where the alignment has been subjected to curation by a high identity threshold (e.g.. 80% identity 72 WO 2007/109254 PCT/US2007/006883 as cutoff criterion), SAI calculations are preferably dispensed with. In certain embodiments, covarying pairs are not reported by the calculation unless they are observed a minimum number of times (herein, an "event cutoff"). In one embodiment, the event cutoff is 10 or more events. In a more preferred embodiment, 5 the event cutoff about 2 or more, and less than 10, events (e.g., 9, 8, 7, 6, or 5 events). Step 4: Using Covariation Data to Predict Protein Stability In certain embodiments, the statistically significant covariations within the covariation dataset are used to search for corresponding covariations within the 10 candidate or test sequence. In particular, the conservation of one or more covariant amino acids of the reference set at corresponding amino acid positions within the candidate or test sequence indicates that these residues are functionally important and are predictive of favorable protein stability. Accordingly, the number of covariant amino acids that are conserved within a candidate or test sequence may be used to 15 predict the stability of the sequence. In certain exemplary aspects, the relavent covariations may be visualized using a graphical user interface (GUI) of the invention (e.g., the NAPMAP tool described in Section V infra) for analysis of the covariation dataset. Because interpretation of data in the covariation dataset may be cumbersome, the graphical user interface may ease data 20 analysis and facilitate rapid protein designs and functional analyses. In exemplary embodiments, the presence or absence of covariant residues within the reference set or alignment that are present or absent within the candidate sequence can be used to establish a score which is correlative with protein stability. For example, where a covariant (e.g., positively covariant) pair of residues is found to be retained 25 within the candidate sequence, a first score (e.g. a positive score) can be assigned to the candidate sequence. Conversely, where a covariant (e.g., positively covariant) pair of residues is missing from the candidate sequence, a second score of oppositive sign to the first score (e.g., a negative score) can be assigned to the candidate sequence. In one preferred embodiment, the absence of a negatively covariant pair of residues within a 30 test sequence is assigned a positive score, while the presence of the negatively covariant pair of residues is assigned a negative score. In another preferred embodiment, the presence of a positively covariant pair of residues of a test sequence is assigned a positive score, while the abasence of a positively covariant pair test sequence residues is 73 WO 2007/109254 PCT/US2007/006883 assigned a negative score. In certain embodiments, covariation scores are generated only for covariations that satisfy a threshold level of statisitical significance. In one embodiment, covariation scores are generated only for covariations above (or below) a certain X2 -value or * value. 5 For example, the cutoff for designation of negative covariation score may be a covaration with a phi association coefficient (0) of less than +0.2 (about +0.2 to about 1.0), and more preferable less than -0.2. In another exemplary embodiment, a covariation is assigned a positive covariation score if it has a phi association coefficient of greater than -0.2, preferably greater than +0.2. In preferred embodiments, a negative 10 covariation score is assigned to a covariation with a phi association coefficient (0) of less than -0.5 (e.g., -0.5, -0.6, -0.7, -0.8, -0.9, or -1.0), and a positive covariation score is assigned to a covaration with a phi association coefficient (0) of greater than +0.5 (e.g., +0.5, +0.6, +0.7, +0.8, +0.9, or +1.0), while all other covariations are masked by assigning a neutral score (e.g., zero). 15 In certain embodiments, the negative or positive covariation score can be weighted to reflect the statistical significance or strength of the corresponding covariation. In one exemplary embodiment, a positive covariation can be assigned a high positive covariation score if its phi association coefficient is also high or a lower positive covariation score its * value is low. For example, a candidate sequence having 20 first and second covariations with respecitive ) values of +0.5 and +1.0 may be assigned a low positive covariation score (e.g., +1) for the first covariation and a higher positive covariation score for the second covariation (e.g., +2). In other embodiments, the covariation score is assigned only to those covariations which are validated by a means other than calculated statistical 25 significance. Several non-statistical criteria may be used to validate that the calculated covariations are significant and biologically meaningful. In one embodiment, the covariation is validated by examining whether there is correlation or trend between residues that covary with each other and their promixity to each other within a structural model of the corresponding protein. If said residues are close in proximity (e.g. 30 A or 30 less, preferably 10 A or less) to each other in the structure, the predicted covariation is validated as a true covariation. In another embodiment, the covariation is validated by determining whether the covarying amino acids form a connection which is already known to exist within a subset of proteins corresponding to sequences of the reference 74 WO 2007/109254 PCT/US2007/006883 set (e.g., a known disulfide bond). For example, residues 6-10 at the N-terminus of human or murine IgG variable heavy chain folds are known to adopt very specific conformations based on the conservation of covarying pairs of amino acids (Ewert et al., Methods, 34: 184-199 (2004)). 5 In yet other embodiments, a covariation score of an amino acid in a candidate sequence can be weighted to reflect the number of covariations within the reference set or alignment that is satisfies (or violates). In one embodiment, the candidate sequence is assigned a positive covariation score for each covariation within the reference set or alignment that it satisfies, and a negative covariation score for each covariation within 10 the reference set that it violates. In preferred embodiments, the covariation score for a particular amino acid of a test sequence is the sum total of the positive covariation scores offset by the sum total of the negative covariation scores. For example, where 2 negative covariations and 9 positive covariations are observed within the reference set at amino acid positions corresponding to the amino acid of the candidate sequence, the 15 candidate sequence amino acid may be assigned a total covariation score of 7. In certain embodiments, a total covariation scores are summed for all the amino acids in a candidate sequence (or portion thereof) to obtain a "sequence covariation score". This sequence covariation score may be employed to predict the stability of the sequence. Generally, a negative sequence covariation score is predictive of low protein 20 stability, whereas a positive sequence covariation score is predictive of high protein stability. b. Consensus Scoring 25 In other embodiments, the computational methods of the invention employ a novel scoring technique termed "consensus-based scoring" which compares candidate polypeptide sequences against a database of related polypeptide sequences to identify proteins with potentially low stability. As used herein "consensus-based scoring" refers to calculation of the number of non-consensus amino acids within a protein (e.g., a test 30 protein) based on the information available from sequence compilations of related proteins. The calculation results in a "consensus score" for the protein. As shown in the Examples infra, the consensus score of a protein is highly correlative with its empirically-determined protein stability. Accordingly, an object of the invention is to employ consensus-based scoring to predict protein stability. The greater the deviation of 75 WO 2007/109254 PCT/US2007/006883 a test protein sequence from a consensus sequence, the higher the consensus score, and the more likely the test protein is to contain amino acids detrimental to its stability. In the novel consensus scoring approach described herein, the following basic steps are taken to predict the stability of a candidate or test protein: 5 1. Providing a set of homologous sequences (herein, a "reference set") corresponding to a sequence of a domain (or portion thereof) of the candidate protein (herein a "test domain sequence"). 2. Determining the residue frequency at every residue position within the test domain sequence (or portion thereof) to obtain a consensus score. 10 3. Using the consensus score to predict the stability of the candidate or test protein. In certain exemplary embodiments, the following optional steps are employed to increase the predictive value of the consensus score: 4. Determining the residue frequency at every corresponding residue position 15 within every sequence of the reference set (or a subset thereof) obtain an average score. 5. Comparing the consensus score to the average consensus score to determine a sequence score. 6. Using the sequence score to predict the stability of the candidate or test protein. 20 Step 1: Providing a Reference Set The consensus scoring based methods of the invention employ a set of of sequences (a "reference set") which are homologous to (i.e., related to) a sequence of interest or "test sequence". The reference set may comprise a set of homologous 25 sequences having moderate to high degrees of similarity with the test sequence. The set of homologous sequences may be determined by one skilled in the art to include a suitable number of non-redundant sequences. The set of homologous sequences need not be large, as long as a reasonably unbiased selection is obtained. It is not the number of sequences, but the ratio of frequencies that is important. 30 Such sequences can be obtained through homology-based searches of any of the sequence databases known in the art (e.g. NCBI, TIGR databases). In an exemplary embodiment, a standard BLAST search with default parameters may be performed with the test sequence to retrieve homologous sequences of interest. Sequences with a 76 WO 2007/109254 PCT/US2007/006883 minimum percentage of sequence identity (e.g. greater than 25%, 30%, 35%, 40%, 45% or 50% sequence identity, preferably greater than 30% sequence identity) may be selected for inclusion in the reference set. In certain embodiments, those sequences with a high degree of sequence identity (e.g., greater than 85%, 90%, 95%, or 98% sequence 5 identity, preferably greater than 95% sequence identity) are excluded from the reference set to avoid bias. In certain embodiments, the reference set may be curated to include only those sequences which satisfy one or more criteria for inclusion in the set. For example, in certain embodiments, the reference set may comprise ortholog sequences (e.g. sequences 10 which are from different species (e.g., different mammalian species) but which have the same biochemical function). In other embodiments, the reference set comprises human sequences. In an exemplary embodiment, the set of homologous sequences includes only mammalian germline sequences. In preferred embodiments, the reference set includes sequences from proteins having the same class of protein fold (e.g. a similar 15 protein domain) to the test protein (e.g. proteins having an immunoglobulin-type fold). Such proteins may be identified by bioinformatics techniques known in the art (e.g., be searching the Conserved Domain Database (CDD) of the National Institutes of Health). Step 2: Determining the Consensus Score of the Test Sequence 20 Once the sequences of the reference set are compiled, the consensus sequence of the reference set is determined in order to facilitate the consensus scoring method of the invention. As used herein, the term "consensus sequence" refers to a sequence wherein the residue (e.g. amino acid residue) at each position within the sequence corresponds to the most common residue at that position in an aligned set of related sequences (i.e., the 25 reference set). To determine the consensus sequence, the sequences within the reference set may first be aligned using sequence alignment algorithms known in the art. A preferred, non-limiting example of a local alignment algorithm utilized for the comparison of sequences is the algorithm of Karlin and Altschul (1990) Proc. Natl. Acad. Sci. USA 30 87:2264-68, modified as in Karlin and Altschul (1993) Proc. Natl. Acad. Sci. USA 90:5858-77. Such an algorithm is incorporated into the BLAST programs (version 2.0) of Altschul, et aL. (1990) J Mol. Biol. 215:403-10. In another embodiment, the alignment is optimized by introducing appropriate gaps and percent identity is 77 WO 2007/109254 PCT/US2007/006883 determined over the length of the aligned sequences (i.e., a gapped alignment). To obtain gapped alignments for comparison purposes, Gapped BLAST can be utilized as described in Altschul et al., (1997) Nucleic Acids Res. 25(17):3389-3402. In another embodiment, the alignment is optimized by introducing appropriate gaps and percent 5 identity is determined over the entire length of the sequences aligned (i.e., a global alignment). A preferred, non-limiting example of a mathematical algorithm utilized for the global comparison of sequences is the algorithm of Myers and-Miller, CABIOS (1989). Such an algorithm is incorporated into the ALIGN program (version 2.0) which is part of the GCG sequence alignment software package. When utilizing the ALIGN 10 program for comparing amino acid sequences, a PAM 120 weight residue table, a gap length penalty of 12, and a gap penalty of 4 can be used. To determine the consensus sequence of the aligned set of references, the most frequently occurring residue at each position within the aligned set of references is determined. The consensus sequence may be determined visually or it may be 15 determined by computational anaysis using a publically available bioinformatics program (e.g., the EMBOSS CONS tool available from the Helix Systems of the National Institute of Health). The consensus sequence is employed in the methods of the invention to determine a consensus score. In certain embodiments, a consensus score is a numerical 20 value which equates the residue frequency or occurrence within the reference set of the consensus residue at each amino acid position in the consensus sequence (ie. the "consensus residue frequency") with the residue frequency or occurrence within the reference set of the amino acid at the corresponding position in the test sequence (ie. the "test residue frequency"). Residue frequencies at a position within the test or consensus 25 sequence are calculated by summing the number of times the amino acid at that position is present at the corresponding position in the aligned reference set, and dividing the summed value by the total number of sequences within the reference set. In an exemplary embodiment, the consensus score at each amino acid position is calculated by dividing the test residue frequency by the consensus residue frequency to give a non 30 zero score of between 1 and 0. For example, a residue which is identical to the consensus amino acid at that position is assigned a perfect score of 1, while a consensus score approaches zero, the less common the amino acid is at that position. In certain preferred embodiments, the consensus score may be summed for each 78 WO 2007/109254 PCT/US2007/006883 position of the test sequence (or a portion thereof) to obtain a total consensus score. For example, a perfect total consensus score is the number of amino acids in the test sequence, indicating that the test sequence is identical to its corresponding consensus sequence. 5 An exemplary formula for calculating the total consensus score of a protein is set forth below: h, (r) score = c,(r) wherein c;(r) equals the consensus residue frequency at an amino acid position of the 10 consensus sequence, h 1 (r) equals the test residue frequency of an amino acid position of the test sequence, and i equals the number of amino acid positions within the test sequence. 15 Step 3: Using the Consensus Score to Predict Protein Stability The consensus score may then be employed to predict the stability of the test protein. In general, the higher the value of the consensus score, the greater the probability that the test protein will have adequate stability for its intended use. In certain embodiments, the consensus score of the test protein may be compared with the 20 consensus score of a control sequence. In one exemplary embodiment, the control sequence is from a protein which is known to have poor or undesired stability properties (herein, a "negative control"). Accordingly, the test protein is predicted to have improved stability if its consensus score is higher than that of the negative control. In another exemplary embodiment, the control sequence is from a protein which is known 25 to have desirable stability properties (herein, a "positive control"). Accordingly, the test protein is predicted to have improved stability if its consensus score is substantially similar or higher than that of the positive control. In other embodiments, the consensus score of the test sequence may be compared with its hypothetical perfect consensus score, a value which corresponds to the total number of amino acids in the test sequence. 30 Step 4: Determining the Average Consensus Score of the Reference Set In certain embodiments, average consensus scores at some or all of the amino 79 WO 2007/109254 PCT/US2007/006883 acid positions of the sequences within the reference set are determined for comparison with the consensus scores at corresponding positions within the test sequence. In preferred embodiments, the average total consensus score for the reference set is determined for comparison with the total consensus score of the of the test sequence. 5 The average total consensus score for the reference set is the sum of the residue frequencies for all amino acid positions of a sequence within the reference set, divided by the total number of sequences in there reference set. Step 5: Determining the Sequence Score 10 In certain embodiments, the sequence score is determined by comparing the average consensus score of the reference set with the consensus score of the test sequence. The differential of these values is the "sequence score" of the test sequence (also referred to herein as the "A score"). The sequence score may be determined by subtracting the consensus score from the average consensus score. 15 Step 6: Using the Sequence Score to Predict Protein Stability The sequence score provides a measure of the deviation of the consensus score of the test sequence from the average consensus score of the reference set. The Examples infra demonstrate that the sequence score of a protein is highly correlated with the 20 stability (e.g. thermal stability) of the protein. Accordingly, in certain embodiments, the sequence score may also be used to predict the stability of the protein. If the A score of the protein is a negative value (ie. the consensus score is lower than the average consensus score), the prediction would be that the stability of the protein of interest will have a lower stability than the majority of proteins within the reference set. 25 In certain embodiments, the test sequences employed in the methods of the invention are single-domain proteins. As used herein, a "single domain protein" is a protein comprising one or more domains, wherein each of the domains are of the same type (e.g. a variable heavy (VH) domain). 30 In other embodiments, the test sequences employed in the methods of the invention are multi-domain proteins. As used used herein, a "multi-domain protein" is a protein comprising two or more domains, where at least two of the domains are of a 80 WO 2007/109254 PCT/US2007/006883 different type. For example, antibodies are proteins typically comprising six domains (i.e., VH, VL, CL, CHI, CH2, and CH3). In preferred embodiments, the stability of the least stable domain of a multi-domain protein is predicted using the methods of the invention. The Examples infra demonstrate that the methods of the invention are 5 particularly effective in predicting thermal stability when the least stable domain (or portion thereof) of a multi-domain protein (e.g. the VH domain of an antibody) is employed as a test sequence. If the least stable domain is not known a priori, it can be determined using experimental methods that are known in the art (e.g. using any of the methods described 10 supra, such as differential scanning calorimetry (DSC), temperature dependent circular dichroism (CD), or fluorescence measurements). Such methods allow for the determination of multiple thermal unfolding transitions where the least stable domain either unfolds first (see Figure 69A) or limits the overall stability threshold of a multidomain unit that unfolds cooperatively (i.e. a multidomain protein which exhibits a 15 single unfolding transition). The least stable domain can be identified in a number of additional ways. Mutagenesis can be performed to probe which domain limits the overall stability. Additionally, protease resistance of a multidomain protein can be performed under conditions where the least stable domain is known to be intrinsically unfolded via DSC or other spectroscopic methods (Fontana, et al., Fold. Des., 2: R17 20 26, 1997). Once the least stable domain is identified, the sequence encoding this domain (or a portion thereof) may be employed as a test sequence in the methods of the invention. In certain exemplary embodiments, the multi-domain proteins evaluated using the methods of the invention are antibodies. The methods of the invention allow one 25 skilled in the art to exclude the selection of inappropriate variable region domains from among the vast pool of variable region sequences created by the diversity mechanisms of the immune system. In exemplary embodiments, e.g., scFv molecules or antibody molecules comprising variable region (Fv) domains with a relatively high stability (e.g. thermal stability) may be identified using the methods of the invention, and selected. In 30 other exemplary embodiments, human immunoglobulin variable region (Fv) domains having acceptable stability may be identified using the methods of the invention and selected for use as acceptor immunoglobulin chains in the humanization of an antibody. In other exemplary embodiments, the methods of the invention may be employed to 81 WO 2007/109254 PCT/US2007/006883 screen candidate engineered antibody variable sequences (e.g., humanized VL or VH sequences) for appropriate stability before proceeding with the synthesis of a binding molecule comprising the candidate sequence. As is known in the art, protein stability varies from one antibody to another 5 based on the many possible variations in natural or engineered antibodies. For example, the selective pressure of the immune system to create diversity within the variable domains of natural antibody can negatively affect their stability or foldability in recombinant expression systems (Knappik and Pluckthun, Protein Engng., 8: 81-89, (1995); Wall and Pluckthun, Protein Engng, 12: 605-11 (1999); Knappik et al., J. Mol. 10 Biol., 296: 57-86, (2000); Ewert et al., J. Mol. Biol., 3325: 531-33, (2003)). Antibodies consisting of the same isotype and subclass, human IgGI being the most commonly used for antibody therapeutics, vary.in stability from one another primarily based on differences in their variable domains since the remainder of their sequences are identical. Thermal unfolding of antibody Fab fragments predominantly occurs in a single apparent 15 transition. In most cases, the antibody Fv region limits the overall thermostability of the Fab except in the rare cases where Fv (or single VH or VL) stability is extremely low or extremely high and the Fv unfolding transition is decoupled from that of the CHl/CL region (Shimba, et al., 1995; R6thlisberger et al., 2005). In certain embodiments, the methods of the invention employ a variable region 20 sequence of an antibody as a test sequence. In certain preferred embodiments, the methods of the invention employ the heavy chain variable domain (VH) (or a portion thereof) as a test sequence. The Examples infra demonstrate that the VH domain is highly predictive of the stability of an antibody comprising said domain. In certain embodiments, the variable domain test sequence is compared with a 25 reference set comprising, or consisting solely of, variable domain sequences. Said variable domain sequences may be compiled from the Kabat database of immunoglobulin sequences (Kabat et al., "Distribution Files of the Fifth Edition of Sequences of Proteins of Immunological Interest", 1992). In certain preferred embodiments, the reference set may comprise or consist of human immunoglobulin 30 sequences. In other preferred embodiments, the reference set may comprise, or consist solely of, human germline sequences. In one embodiment, the reference comprises variable domain sequences from the same Kabat subclass of antibody sequences. 82 WO 2007/109254 PCT/US2007/006883 In other embodiments, a portion of variable domain (e.g. VH domain) is used as a test sequence in the methods of the invention. An exemplary variable domain portion is a sequence comprising less than all of the framework regions or CDRs of the variable region sequence (e.g. a VH sequence that is truncated to remove CDR3 and FR4). In 5 one embodiment, the VH domain sequences are of the VHl Kabat germline subclass. In another embodiment, the variable domain sequences are of the VH2 Kabat germline subclass. In another embodiment, the variable domain sequences are of the VH3 Kabat germline subclass. In another embodiment, the variable domain sequences are of the VH4 Kabat germline subclass. In another embodiment, the variable domain sequences 10 are of the VH5 Kabat germline subclass. In another embodiment, the variable domain sequences are of the VH7 Kabat germline subclass. The methods of invention may be employed with proteins known in the art. In one embodiment, the protein is an antibody or portion thereof. In certain embodiments, the antibody is a humanized antibody. In other embodiments, the antibody is a human 15 antibody. In other embodiments, the antibody is a non-human antibody (e.g., mouse monoclonal antibody). In other embodiments, the antibody is a single-domain antibody (e.g., camelid, shark, human). In yet other embodiments, the antibody is a chimeric antibody. Other exemplary modified antibodies for use in the methods of the invention include domain-deleted antibodies and bispecific binding molecules. Other exemplary 20 antibodies include the scFv-containing antibodies or modified antibodies described supra. In preferred embodiments, the stability of the aforementioned binding proteins is evaluated using a heavy chain variable sequence thereof (or portion thereof) as the test sequence in the methods of the invention. In one embodiment, of the invention, the data obtained for a given polypeptide 25 are stored or outputted as a measure of the stability of the polypeptide. In one embodiment, the methods described herein can be repeated for a plurality of polypeptides. In one embodiment, a selection of a polypeptide may be made based on the data obtained. In another embodiment, a selected polypeptide may be formulated for therapeutic use. 30 IV. Methods for Desianing/Producing Proteins with Enhanced Biophysical Properties (e.g., Protein Stability) 83 WO 2007/109254 PCT/US2007/006883 (a) Covariation Design In another aspect, the invention provides methods for designing protein variants with an enhanced biophysical property relative to the parent molecule from which it is 5 derived using the results of a covariation analysis. There are many types of protein engineering efforts where covariation data may be used to facilitate design, including but not limited to enhanced protein stability designs, modification of pH unfolding profiles, stable removal of glycosylation, and alteration of biochemical function. In certain embodiments, the invention provides methods for designing proteins 10 with enhanced stability designs based on the identification of covariation scores (or covariations themselves) that are predictive of enhanced or reduced protein stability according the method of Section III supra. For example, if a sequence belonging to the protein of interest is missing or violates several key covariations, protein designs can be made to correct for this. Modifications can also be included which improve the 15 calculated number of strong covariations observed within a single sequence. A number of designs have been developed for stabilizing the BHA10 scFv VH and VL domains. In particular, initial library screening results for two scFv BHA 10 domains validate the ability of the Covariation Analysis Tool for protein stability design. The first example involves the successful prediction of a stabilizing mutation (S46L) 20 within the BHA 10 VL domain that imparts a significant improvement in thermal stability
(AT
5 o of + 10 "C) as measured by a thermal challenge assay. Mutation from Ser to Leu at position 46 (Kabat numbering system) leads to a positive connection to an existing Tyr at residue 36 which the Tool shows covaries strongly with Leu 46 (Figure 78). A second example of the utility of the Covariation Analysis Tool is the analytical 25 agreement of the negative effects of a BHA 10 VH Q6E mutation. Single residue frequency analysis suggested that mutation to the much more commonly observed Glu at this position would lead to an increase in stability. However, the Covariation Analysis Tool indicated that single mutation to Glu violates several existing covariations present within the BHA1 0 VH sequence. To obtain an improvement in stability one must 30 replace several amino acids that preferentially stabilize Glu at this position. Another example of the predictive value of the Covariation Analysis Tool is shown in Figure 79. Met 80 was mutated to Leu as part of a single residue library design. It was thought that this single mutation would be highly stabilizing, as Leu is 84 WO 2007/109254 PCT/US2007/006883 the most frequent amino acid observed at this position within the sequence database. However, covariation analyses indicate that two other amino acids must be mutated (V67F and T70S) in order to achieve covariation harmony. 5 (b) Interface Desian In yet other aspects, the invention provides methods for enhancing a biophysical property (e.g., protein stability) of a second or candidate protein (e.g., an antibody) using a first protein as a template protein for rational design. As used herein, the term "template protein", refers to a protein with a desirable 10 biophysical property (e.g., high protein thermal stability) which is used to model the same property in a second protein for an improvement of the biophysical property is also desirable. In preferred embodiments, the template protein is of the same structural class as the second protein (e.g., if the second protein is an antibody, an antibody is employed as a template protein). 15 In certain embodiments, the template protein is identified as having desirable biophysical property (e.g., protein stability) according to a method of the invention. In one exemplary embodiment, the template protein with desirable stability properties may have been identified from a group of proteins (Such as BIIB1-4, described in Table 20, supra) using experimental methods such as those described in Example 15. This 20 template protein would then serve as a sequence platform for design. In one exemplary embodiment, the template protein is predicted to have an improved biophysical property (e.g. protein stability) by a covariation analysis method of the invention outlined supra. In other exemplary embodiments, the template protein is predicted to be a desirable protein (e.g., a stabilized protein) according to a consensus scoring method as outlined 25 supra. In yet another exemplary embodiment, the template protein is identified according to a screening method of the invention. In other embodiments, the template protein is a protein designed according to a method of the invention. In one exemplary embodiment, the template protein is designed to have an improved biophysical property (e.g., enhanced protein stability) 30 according to a covariation method of the invention. In another exemplary embodiment, the template protein is designed to have an improved biophysical property according a consensus scoring method. The methods of this aspect of the invention involve the following steps: 85 WO 2007/109254 PCT/US2007/006883 (1) providing structural models of the template protein and the candidate protein; (2) identifying residues in the template protein that are important for stability; and (3) substituting residues in the candidate protein that correspond to the important 5 residues template protein. In one embodiment, the template protein structural model is a crystal structure (e.g., an X-ray crystal structure) and the candidate protein is modeled by homology modeling using the structural model of the template protien. In another embodiment, an important residue is a VH/VL interface residue. As used 10 herein, an "interface residue" is a residue that participates in domain-domain interactions within or among proteins. Preferably, an interface residue resides at the boundary, i.e., the "interface" between two protein domains and contributes its surface area to at least one of the domains (preferably both). The residues that bury significantly more surface area than others at the interface, i.e., the residues which do not expose significantly more 15 of their surface area to the surface of the folded protein, are considered to be more crucial to interface stability. The surface area that each residue contributes to the interface can be determined using art-recognized techniques (e.g., analysis using MOLMOL software). In certain embodiments, the interface residue buries at least 10 A 2 of surface area (e.g. 10 A 2 , 15 A 2 , 20 A 2 , 25 A 2 or more of buried surface area). 20 Residues of the candidate protein located at amino acid positions corresponding to those of the important residues (e.g., buried interface residues) of the template protein are preferably substituted if they are of a non-identical type. In certain embodiments, only non-conservative substitutions are made. In other embodiments, only conservative substitutions are made. 25 In exemplary embodiments, the template and candidate proteins are binding molecules (e.g., antibodies or variants thereof (e.g. scFv molecules)). In one embodiment, the interface that is examined is that between one Fe domain (e.g., hinge, CH2, and CH3) and another Fc domain. In another embodiment, the interface is that formed between the VH domain and the VL domain. 30 In preferred embodiments, the designs are focused on the interface between the VH and VL domains of an antibody. Biophysical data (see DSC data - Example 15) suggest that the affinity of VH and VL for one another is marginal in the scFv format (i.e. in the absence of the CHI and CL domains of the Fab). In particular, VH and VL 86 WO 2007/109254 PCT/US2007/006883 interactions are not always strong enough (i.e. do not have sufficient affinity to lead to a cooperatively folded single entity. Therefore, improving the stability of the VHNL interface (ie. increasing the affinity between the VH and VL domains) may be a preferred route to enhancing antibody/scFv stability. By stabilizing the interface, the 5 folding transition of antibody can be turned into a single event, raise the affinity of the domains for one another, and allowing for less dynamic fluctuations that may lead to exposed hydrophobic surface area (as observed by ANS-binding experiments). In yet other embodiments, the designs may be validated using diffierential scanning calorimetry (DSC) experiments which allow one skilled in the art to non 10 quantitatively observe real affinity increases between the VH and VL. (c) Combined Covariation and Interface Design In yet other aspects of the invention, two or more of the sequence based design tools described herein may be employed for the design of optimized protein variants. In 15 one exemplary embodiment, interface design is employed together with covariation analysis to determine residues or residue positions which are important for protein structure and function, including, for example, protein stability. In one exemplary embodiment, covariation analysis is used to determine residues or residue positions outside an interface (herein, termed "scaffolding residues") that do not physically 20 interact with the interface residues on the opposing domain or contribute surface area to the interface, but are nonetheless important for providing proper structural context for interface residues. For example, covariation analysis may be conducted on residues located within the interface to identify scaffolding residues which covary with the interface residues. Preferably, interface residues which bury a high degree of surface 25 area (e.g., residues that bury at least 40 A 2 of surface area) are selected for covariation analysis. Scaffolding residues which covary with the selected interface residues are then identified using the covariation methods of the invention. Preferably, scaffolding residues which strongly covary (e.g., phi-value of at least 0.25) are selected. In another preferred embodiment, scaffolding residues are selected if they covary with at least two 30 interface residues (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10 or more interface residues). The greater the number of covariations, the higher the covariation score may be attributed to the scaffolding residue, since the number of covariations is predictive of the contribution made by the scaffolding residue to stabilization of the VH/VL interface. 87 WO 2007/109254 PCT/US2007/006883 The methods of this aspect of the invention involve the following steps: (1) providing structural models of a template protein and a candidate protein; (2) identifying interface residues (e.g., VH/VL) in the template protein that are important for stability; 5 (3) identifying scaffolding residues which covary with the interface residues of step (2), e.g., using the Covariation Analysis Tool described herein; (4) substituting one or more interface residues or scaffolding residues in the candidate protein with the corresponding residues or scaffolding residues identified in steps (2) and (3). 10 Residues of the candidate protein located at amino acid positions corresponding to those of the important residues (e.g., interface or scaffolding residues) of the template protein are preferably substituted if they are of a non-identical type. In certain embodiments, only non-conservative substitutions are made. In other embodiments, only conservative substitutions are made. 15 In one embodiment, of the invention, the improved sequence of a given polypeptide is stored or outputted. In another embodiment, a selected polypeptide may be formulated for therapeutic use. 20 V. Systems. Interfaces, and Computer-Implemented Methods One aspect of the invention pertains to methods (e.g., computer-implemented methods), apparatus, and software for implementing the selection or design methods of the invention. In exemplary embodiments, the invention provides a computer 25 implemented method and apparatus for identifying amino acid residues (e.g., covariant amino acid residues) in a candidate sequence. In another embodiment, the invention provides computer-implemented methods and apparatus for ranking one or more test or candidate sequences by a score (e.g., a consensus score or covariation score) which is predictive of a biophysical property (e.g., protein stability, catalytic activity, therapeutic 30 activity, resistance to a pathogen or toxin, toxicity, etc). The computer implemented methods aspect may be described by the following sequence of operations: (a) receiving data (e.g., sequence or structural data) characterizing a reference set and a test sequence; (b) from the data, calculating a score 88 WO 2007/109254 PCT/US2007/006883 (e.g., a covariation score or consensus score); (c) ranking the scores to identify one or more sequences or amino acid residues of interest. In some embodiments, the scores are ranked to identify one or more amino acid residues that are to remain fixed in the test or candidate sequence. In other 5 embodiments, the scores are ranked to identify one or more amino acid residues which are candidates for substitution. In one exemplary embodiment, the computer implemented method ranks residue positions (or specific residues at certain positions) in order of covariation score. In another exemplary embodiment, the computer implemented method ranks candidate sequences in order of consensus score. 10 Yet another aspect of the invention pertains to apparatus and machine-readable media on which are provided program instructions and/or arrangements of data for implementing the methods described above. Frequently, the program instructions are provided as code for performing certain method operations. Data, if employed to implement features of this invention, may be provided as data structures, database tables, 15 data objects, or other appropriate arrangements of specified information. Any of the methods or systems of this invention may be represented, in whole or in part, as such program instructions and/or data provided on machine-readable media. The computer-implemented methods of the invention may include an output device that displays information to a user (e.g., a CRT display, an LCD, a printer, a 20 communication device such as a modem, audio output, and the like). In addition, instructions (e.g. an algorithm) for carrying out the calculation, in part or in whole, can be conferred to a medium suitable for use in an electronic device for carrying out the instructions. Thus, the methods of the invention are amenable to a high throughput approach comprising software (e.g., computer-readable instructions) and hardware (e.g., 25 computers, robotics, and chips). The computer-implemented process is not limited to a particular computer platform, particular processor, or particular high-level programming language. In certain embodiments, the calculations used in the methods of the invention are carried out by a computer algorithm suitable for use with a computer programming 30 language (e.g. PERL). In exemplary embodiments, the computer algorithm is designed to recognize (i) a sequence alignment of a reference set and/or (ii) one or more test sequences as input for the algorithm. 89 WO 2007/109254 PCT/US2007/006883 (i) Computer-implemented Consensus Scoring In one aspect of the invention pertains to computer-implemented methods, apparatus, and software for conducting the consensus scoring method of the invention described supra using a computer-implemented algorithm. In exemplary embodiments, 5 the algorithm is designed to perform one or more of the following steps: (1) cutting out each residue position of the alignment as a column and calculating the consensus residue frequency at that position; (2) cutting off each residue position of the test sequence(s) and calculating the test residue frequency at that position, and (3) dividing the test residue frequency by the consensus residue frequency to give a consensus score; and/or 10 (4) summing the consensus score at each position to give a total consensus score. In other embodiments, the algorithm may be designed to perform one or more of the following additional steps: (5) calculating the average consensus score of the reference set, and/or (6) the sequence score of the test sequence. These aspects of the invention are also embodied in a system for conducting the 15 consensus scoring method of the invention. The system includes (a) a computer that includes a database capable of storing at least one population of reference sets, and (b) system software. The system software includes one or more logic instructions for conducting any of steps (1)-(6) of the computer-implemented consensus scoring method. The invention also provides a computer program product for conducting the 20 consensus scoring method of the invention. The computer program product includes a computer-readable medium having one or more logic instructions for conducting any of steps (1)-(6). (ii) Computer-implemented Covariation Analysis 25 Another aspect of the invention pertains to computer-implemented methods for conducting the covariation analysis method of the invention as described supra using a computer-implemented algorigthm. In exemplary embodiments, the methods perform one or more of the following steps: (1) cutting out each residue position of the aligned set as a column; (2) arranging columns in a matrix; (3) calculating correlation among 30 various columns in the matrix to derive a correlative term; (4) adding correlative terms to linear terms, which correspond to amino acid residues, to generate an expanded predictor matrix; (5) generating a heuristically-derived model (e.g. a hidden Markov model) from the expanded predictor matrix to identify important correlations. 90 WO 2007/109254 PCT/US2007/006883 These aspects of the invention are also embodied in a system for conducting the covariation analysis method of the invention. The system includes (a) a computer that includes a database capable of storing at least one population of character string libraries, and (b) system software. The system software includes one or more logic 5 instructions for conducting any of steps (1)-(5) of the computer-implemented method covariation scoring method. The invention also provides a computer program product for conducting the covariation scoring method of the invention. The computer program product includes a computer-readable medium having one or more logic instructions for conducting any of 10 steps (1)-(5). (iii) Graphical User Interfaces for Covariation Analysis In certain exemplary aspects, the invention provides a novel graphical user interface (GUI) for use with the covariation methods of the invention. This covariation 15 analysis tool, termed NAPMAP, assists the user in visualizing covariations in the context of aligned set of sequences. The novel graphical user interface is computer implemented using art-recognized programming languages (e.g. in Java 1.4.2 utilizing the Processing library from processing.org). 20 (a) Grid Layouts In one aspect, the graphical user interface of the invention comprises a graphical depiction of an alignment (e.g., a sequence alignment created according to the methods of the invention supra) or data corresponding thereto. A simple form of this display is the grid layout. As used herein, the term "grid layout" (also known as an alignment 25 platform) refers to a matrix in which the properties of an alignment are represented. In one embodiment, the grid layout graphically represents residue positions of the sequences of an alignment. In another embodiment, the grid layout graphically represents residue types of sequences within the alignment. In another embodiment, the grid layout graphically represents the frequency of residue types (or "residue usage 30 frequency") within the alignment. In yet other embodiments, the grid layout represents one or more properties of residues within the alignment (e.g., hydrophobicity, charge, size, pKa, etc.). In certain exemplary embodiments, the grid layout comprises a representative 91 WO 2007/109254 PCT/US2007/006883 display of (i) contiguous residue positions corresponding to a sequence of the alignment; and (ii) residue usage frequencies for residue types at each residue position of the alignment sequence. In one embodiment, the residue usage frequency is symbolically depicted (e.g., by a circle or square). In other embodiments, all 20 natural amino acids, 5 gaps, and ambiguous residues are represented. In certain embodiments, the residue positions are displayed in rows and the residue usage frequencies are displayed in columns. In certain embodiments, the residue positions are displayed in columns and the residue usage frequencies are displayed in rows. In one embodiment, the residue use frequency is correlated (e.g., positively or 10 negatively correlated) with symbol size (e.g., size of circle), so that more conserved residues can be discriminated from less frequently observed residues. For example, residue frequency may be represented as circles of various sizes with circle area proportional to the frequency of its usage., The radius of each circle (R) may be used for drawing the node using an ellipse function calculated according to the following formula 15 (IV) N R- M (IV) wherein: 20 A = desired average area of all nodes N = number of sequences using this residue M = the average number of sequences using each residue In an exemplary embodiment, the grid layout comprises a matrix of contiguous 25 columns and rows wherein (i) contiguous columns graphically represent corresponding contiguous residue positions of the alignment; (ii) each of the contiguous rows graphically represent a specific amino acid type; and (iii) the residue usage frequency at each residue position is graphically represented in each cell of the matrix. Preferably all residue types (e.g., all 20 amino acids and/or possible gaps) are represented by a separate 30 row in the grid layout. More preferably, residue usage frequency at each residue position of the alignment is graphically represented in each cell of the matrix with a symbol (e.g., a circle) whose size is correlated (e.g., positively correlated) with residue 92 WO 2007/109254 PCT/US2007/006883 usage frequency. An exemplary grid layout is depicted in Figure 76. (b) Covariation Overlays In other aspects, the graphical user interface of the invention comprises a 5 graphical depiction of one or more covariations (e.g., covariations identified according to the methods of the invention supra) or data corresponding thereto. In certain embodiments, the graphical user interface of the invention comprises graphical depicts of both (i) a grid layout; and (ii) one or more covariations. In these embodiments, the grid layout may serve as alignment platform for the display of 10 covariation data. Covariation data may be displayed, for example, by overlaying the grid layout with covariation data. Covariation data arranged in this fashion are termed a "covariation overlay". For example, covariations may be depicted in the covariation overlay as lines or network of lines connecting two or more covarying amino acids that are graphically depicted within the grid layout. 15 The lines of the covariation overlay may be of various thicknesses (e.g., thin lines or bars) or continuities (e.g., dashed or solid, straight or jagged), so long as the connection between two or more covarying amino acids is apparent to the user (see, e.g., Figure 77, wherein covariations are depicted as semi-transparent bars that connect two covarying amino acids). In certain exemplary embodiments, the graphical 20 representation of the covariation may further represent the statistical significance of the covariation (e.g., its * or X v aluee. In one exemplary embodiment, the thickness of a line within the covariation overlay may be proportional to the statistical significance of the covariation. For example, the line thickness of the covariation overaly may be drawn in proportion to * value according to formula V: 25 W =1I+ (M. -QN)V wherein: 30 W = line thickness = phi (correlation coefficient) N = exaggeration multiplier M = max line thickness 93 WO 2007/109254 PCT/US2007/006883 In another exemplary embodiment, positive and negative covariations may be graphically discriminated from each other (e.g., by different colored lines or thicknesses). In yet other exemplary embodiments, only covariations above a certain 5 threshold level of statistical significance are displayed to the user. For example, in certain embodiments, ) value cutoffs may be applied to view only covariations with desired correlation strengths. (c) Sequence Traces 10 In other optional aspects, the graphical user interface comprises a graphical depiction of a sequence of interest to aid with possible protein design efforts. The sequence of interest may be the test or candidate sequence (ie., an input sequence) or it may be a sequence in which desirable covarations are integrated (ie., an output sequence). The sequence of interest may be graphically represented as a line. 15 In one embodiment, the graphical interface comprises both (i) a depiction of a sequence of interest; and (ii) a grid layout (e.g., the sequence of interest is superimposed on the grid layout). In another embodiment, the graphical interface comprises both (i) a depiction of a sequence of interest; and (ii) a covariation overlay (e.g., the sequence of interest is superimposed on the covariation overlay). In yet other embodiments, the 20 graphical interface comprises (i) a depiction of a sequence of the interest; (ii) a covariation overlay; and (iii) a grid layout (e.g., the sequence of interest is superimposed (e.g., traced) on the covariation overlay which is in turn superimposed on the grid layout). A sequence of interest may be displayed on a grid layout, for example, as a curve 25 connecting adjacent amino acids that are graphically depicted within the grid layout. A sequence of interest displayed in this fashion is termed a "sequence trace". In preferred embodiments, the sequence trace should pass or thread through only one cell (or graphical respresentation of an amino acid type) for a particular row or column of the grid layout that represents a particular residue position. More preferably, the sequence 30 trace passes through the cell which graphically represents the residue type occurring at a particular residue position within the sequence of interest. The sequence trace may pass though every cell of the grid layout that represent a residue within the sequence of interest, or it may pass through only the cells that represent a portion of the sequence of 94 WO 2007/109254 PCT/US2007/006883 interest. An exemplary sequence trace of a sequence of interest is depicted in Figure 77. Sequence traces depicted therein (catmull-rom splines) were drawn using a bezier function. Given two nodes to draw a sequence transition bezier curve, the node pair's previous and next nodes are used as anchor points and the two nodes themselves are 5 used as the first and second control points. For nodes in the first column (representing the first residues of amino acid sequences), the first control point doubles as its anchor point; for nodes in the last column, its second control point double as its anchor point. In preferred embodiments, the sequence trace plot is rendered interactive for manipulation by the user. For example, the cells within the grid layout (representing can 10 be rendered interactive for selection by the user. In one embodiment, the selected cell can be assigned a particular selection state (e.g., a positive, neutral, or negative selection state). In one exemplary embodiment, a cell representing a particular residue can be positively selected by the user so that the sequence trace can be retraced to pass though the selected cell. For example, if a cell is positively selected by the user, multiple 15 sequence traces can be generated for all sequences of interest (e.g., all sequences within the reference set or alignment) that employ the residue represented by the positively selected cell. In an alternative embodiment, a cell representing a particular residue can be negatively selected by the user so that the sequence trace does not pass through the selected cell. Accordingly, sequence traces can be generated for every sequence of 20 interest (e.g., all sequences within the reference set or alignment) that excludes the residue represented by the negatively selected cell. Multiple sequences of interest can be graphically displayed, so long as they can be distinguished by the user (e.g., using different line thicknesses or continuities). For example, to compare different sequence alignments, sequences traces representing 25 multiple sequences of interest from two or more alignments can be overlaid on the grid layout (e.g., with different colors, preferably opposing colors) to enable visualization of differential residue usage on a global scale. (d) Display Modes 30 The graphical user interface may facilitate any art-recognized functionality. For example, in certain embodiments, the graphical user interface may be capable of panning or zooming on different portions of the grid layout. For example, the viewport size of the grid layout can be changed depending on the user's computer screen 95 WO 2007/109254 PCT/US2007/006883 resolution and visualization can be zoomed out or panned for viewing an alignment that does not fit in the view port initially. In preferred embodiments, data masking is employed, for example, to emphasize the most important covariation pairs and/or networks across several pairs. The graphical 5 user interace of the invention may employ one or more display modes for the viewing of covariations (e.g., statistically-signicant covariations). Preferably, said display modes are available in an interative plot when viewing a sequence trace -of a sequence of interest. Exemplary Viewermodes are as follows: Viewermode #1: Only covariations between residues that are also present in the 10 sequence of interest are displayed. These covariations can be thought of as hypothesized residue-residue interactions (e.g., those found by statistical residue pair frequency analysis) that hold the corresponding protein molecule together and may be important for function. Viewermode #2: The only covariations displayed are those between residue 15 pairs having only one residue of the residue pair present in the sequence of interest. These covariations are residue-residue interactions that tend to be conserved but are not present in the sequence of interest, and whose absence may be detrimental to protein stability. This viewermode is particularly preferred when the Covariation Analysis Tool is employed for protein design (see Section IV supra). 20 Viewermode #3: The only covariations displayed are those between residue pairs having neither amino acid member of the covarying pair present in the sequence of interest. In certain embodiments, all viewermodes are displayed to the user but each in a unique display format. In another embodiment, all viewermodes are separately 25 displayed (e.g., in separate display windows). Figure 89 depicts an exemplary environment suitable for practicing an embodiment of the invention. A computing device 8900 supports an initial sequence collection process 8902 and analysis facility 8904. The computing device 8900 may be 30 a workstation, server, laptop, mainframe, PDA or other computing device equipped with one or more processors and able to support the initial sequence collection process 8902 and analysis facility 8904. The computing device 8900 may have a single processor or multiple processors and each of the processors may have one core or multiple cores. The 96 WO 2007/109254 PCT/US2007/006883 analysis facility 8904 is implemented in software and programmatically analyzes covariation between pairs of residues within alignment to obtain covariation data. The computing device 8900 is in communication with, and outputs data to, an output display device 8910. The Output display device 8910 displays a graphical user interface (GUI) 5 8912 generated by the analysis facility 8904. The GUI 8912 may include a grid layout 8914 and a covariation overlay 8916. The computing device 8900 may also be in communication over a network 8920 with one or more storage locations 8930 and 8940 that respectively hold sequence data 8932 and 8942. The initial sequence collection process 8902 programmatically retrieves candidate sequences from the sequence data 10 8932 and 8942 based on user-supplied and/or pre-determined parameters. The retrieved sequence data is analyzed by the analysis facility 8904. It will be appreciated by those skilled in the art that although the initial sequence collection process 8902 and analysis facility 8904 are depicted separately in Figure 89, they may be implemented as part of an integrated application, process or plug-in. Similarly, it should be appreciated that the 15 initial sequence collection process 8902 and analysis facility may each separately be a task, thread, process, application or plug-in. The computing device 8900 may communicate over the network 8920 with the storage locations 8930 and 8940 using a number of different mediums and configurations. For example, the network 8920 may be arranged as a Local Area Network (LAN), a Wide Area Network (WAN), an intranet, 20 the Internet, and/or may be a wireless network and/or a telephone line, or some other type of network allowing the computing device 8900 to communicate with the storage locations 8930 and 8940. The storage locations 8930 and 8940 may be databases or some other type of storage datastructure hosted by computing devices that are accessible over the network 8920. It should be noted that although components of the present 25 invention are depicted in a particular configuration in Figure 89, other component configurations are also possible within the scope of the present invention. For example, the initial sequence collection process 8902 and analysis facility 8904 may be located on separate computing devices and/or the sequence data 8932 and 8942 may be located on one or more databases hosted by the computing device 8900. 30 Figure 90 is a flowchart of an exemplary sequence of steps that may be followed by an embodiment of the invention in order to determine a sequence covariation score that may be used as a measure of the stability of a polypeptide. The sequence begins 97 WO 2007/109254 PCT/US2007/006883 with the initial sequence collection process 8902 providing alignment of a curated reference set of sequences corresponding to an Ig fold of a polypeptide (step 9000). The analysis facility 8904 then calculates covariation between residues of the sequence of the alignment as described herein to generate covariation data (step 9002). The analysis 5 facility 8904 uses the covariation data to determine a sequence covariation score for a candidate sequence from covariations within the covariation data (step 9004). The determined sequence covariation score provides an indicator of the measure of stability of the polypeptide and may be stored and/or output to a user via the GUI 8912 or in some other manner (step 9006). 10 As noted above, the analysis facility 8904 may also be used to determine a consensus score. Figure 91 is a flowchart of an exemplary sequence of steps that may be followed by an embodiment of the invention in order to determine and use a consensus score to predict the stability of a candidate protein. The sequence begins by using the 15 initial sequence collection process 8902 to provide a reference set of sequences corresponding to a test domain sequence of a candidate protein (step 9100). The analysis facility 8904 then determines residue frequencies at amino acid positions within the test domain sequence to obtain a consensus score (step 9102). The determined consensus score may then be stored and/or output to a user to provide a prediction of the 20 stability of the candidate protein (step 9104). The analysis facility 8904 may also be used to determine an average consensus score. Figure 92 is a flowchart of an exemplary sequence of steps that may be followed by an embodiment of the invention in order to determine and use an average consensus 25 score as a measure of the stability of a candidate protein. The sequence begins by using the initial sequence collection process 8902 to provide a reference set of sequences corresponding to a test domain sequence of a candidate protein (step 9200). The analysis facility 8904 then determines residue frequencies at amino acid positions within the test domain sequence to obtain a consensus score (step 9202). The analysis facility 30 8904 also determines residue frequencies within the sequences of the reference set to determine an average consensus score (step 9204). The consensus score is compared with the average consensus score in order to determine a sequence score (step 9206) and the determined sequence score may then be stored and/or output to a user to provide a 98 WO 2007/109254 PCT/US2007/006883 prediction of the stability of the candidate protein (step 9206). Figure 93 is a flowchart of an exemplary series of steps that may be followed by an embodiment of the invention in order to identify and substitute an amino acid residue 5 that fails to satisfy a covariation with the covarying residue found at a corresponding position in the alignment to determine an improved amino acid sequence of the polypeptide. The sequence begins by the initial sequence collection process 8902 providing alignment of a curated reference set of sequences corresponding to an Ig fold of a polypeptide (step 9300). The analysis facility 8904 then calculates covariation 10 between residues of the sequence of the alignment to identify covarying residues (step 9302). The analysis facility 8904 may identify a particular residue of the polypeptide that fails to satisfy a covariation and may substitute a covarying residue found at a corresponding position in the alignment (step 9304) in order to improve the sequence of the polypeptide. The improved 15 sequence may then be stored or output to a user (step 9306). The present invention may be provided as one or more computer-readable programs embodied on or in one or more mediums. The mediums may be a floppy disk, a hard disk, a compact disc, a digital versatile disc, a flash memory card, a PROM, an 20 MRAM, a RAM, a ROM, or a magnetic tape. In general, the computer-readable programs may be implemented in any programming language. Some examples of languages that can be used include FORTRAN, C, C++, C#, Python or Java. The software programs may be stored on or in one or more mediums as object code. Hardware acceleration may be used and all or a portion of the code may run on a FPGA, 25 an ASIP, or an ASIC. The code may run in a virtualized environment such as in a virtual machine. Multiple virtual machines running the code may be resident on a single processor. 30 VI. Methods of Evaluating Protein Stability The stability properties of the compositions of the invention can be analyzed using methods known in the art. Stability parameters acceptable to those in the art may be employed. Exemplary parameters are described in more detail below. In exemplary 99 WO 2007/109254 PCT/US2007/006883 embodiments, thermal stability is evaluated. In preffered embodiments, the expression levels (e.g., as measured by %yield) of the compositions of the invention are evaluated. In other preferred embodiments, the aggregration levels of the compositions of the invention are evaluated. 5 In certain embodiments, the stability properties of a composition of an invention are compared with that of a suitable control. Exemplary controls include conventional scFv molecule. A particularly preferred control is a (Gly 4 Ser) 3 scFv molecule. In one embodiment, one or more parameters described below are measured. In one embodiment, one or more of these parameters is measured following expression in a 10 mammalian cell. In one embodiment, one or more parameters described below are measured under large scale manufacturing conditions (e.g., expression of scFvs or molecules comprising scFvs in a bioreactor). 15 a) Thermal Stability The thermal stability of the compositions of the invention may be analyzed using a number of non-limiting biophysical or biochemical techniques known in the art. In certain embodiments, thermal stability is evaluated by analytical spectroscopy. An exemplary analytical spectroscopy method is Differential Scanning 20 Calorimetry (DSC). DSC employs a calorimeter which is sensitive to the heat absorbances that accompany the unfolding of most proteins or protein domains (see, e.g. Sanchez-Ruiz, et al., Biochemistry, 27: 1648-52, 1988). To determine the thermal stability of a protein, a sample of the protein is inserted into the calorimeter and the temperature is raised until the Fac or scFv unfolds. The temperature at which the protein 25 unfolds is indicative of overall protein stability. Another exemplary analytical spectroscopy method is Circular Dichroism (CD) spectroscopy. CD spectrometry measures the optical activity of a composition as a function of increasing temperature. Circular dichroism (CD) spectroscopy measures differences in the absorption of left-handed polarized light versus right-handed polarized 30 light which arise due to structural asymmetry. A disordered or unfolded structure results in a CD spectrum very different from that of an ordered or folded structure. The CD spectrum reflects the sensitivity of the proteins to the denaturing effects of increasing temperature and is therefore indicative of a protein's thermal stability (see van Mierlo 100 WO 2007/109254 PCT/US2007/006883 and Steemsma, J. Biotechnol., 79(3):281-98, 2000). Another exemplary analytical spectroscopy method for measuring thermal stability is Fluorescence Emission Spectroscopy (see van Mierlo and Steemsma, supra). Yet another exemplary analytical spectroscopy method for measuring thermal stability is 5 Nuclear Magnetic Resonance (NMR) spectroscopy (see, e.g. van Mierlo and Steemsma, supra). In other embodiments, the thermal stability of a composition of the invention is measured biochemically. An exemplary biochemical method for assessing thermal stability is a thermal challenge assay. In a "thermal challenge assay", a composition of 10 the invention is subjected to a range of elevated temperatures for a set period of time. For example, in one embodiment, test scFv molecules or molecules comprising scFv molecules are subject to an range of increasing temperatures, e.g., for 1 -1.5 hours. The activity of the protein is then assayed by a relevant biochemical assay. For example, if the protein is a binding protein (e.g.an scFv or scFv-containing polypeptide of the 15 invention) the binding activity of the binding protein may be determined by a functional or quantitative ELISA. In one embodiment, such an assay may be done in a high-throughput format. In another embodiment, a library of scFv variants may be created using methods known in the art. scFv expression may be induced an scFvs may be subjected to thermal 20 challenge. The challenged test samples may be assayed for binding and those scFvs which are stable may be scaled up and further characterized. In certain embodiments, thermal stability is evaluated by measuring the melting temperature (Tm) of a composition of the invention using any of the above techniques (e.g. analytical spectroscopy techniques). The melting temperature is the temperature at 25 at the midpoint of a thermal transition curve wherein 50% of molecules of a composition are in a folded state. In other embodiments, thermal stability is evaluated by measuring the specific heat or heat capacity (Cp) of a composition of the invention using an analytical calorimetric technique (e.g. DSC). The specific heat of a composition is the energy (e.g. 30 in kcal/mol) required to raise by 1 C, the temperature of I mol of water. As large Cp is a hallmark of a denatured or inactive protein composition. In certain embodiments, the change in heat capacity (ACp) of a composition is measured by determining the specific heat of a composition before and after its thermal transition. In other embodiments, 101 WO 2007/109254 PCT/US2007/006883 thermal stability may be evaluated by measuring or determining other parameters of thermodynamic stability including Gibbs free energy of unfolding (AG), enthalpy of unfolding (AH), or entropy of unfolding (AS). In other embodiments, one or more of the above biochemical assays (e.g. a 5 thermal challenge assay) is used to determine the temperature (ie. the Tc value) at which 50% of the composition retains its activity (e.g. binding activity). b) % Aggregation In certain embodiments, the stability of a composition of the invention is 10 determined by measuring its propensity to aggregate. Aggregation can be measured by a number of non-limiting biochemical or biophysical techniques. For example, the aggregation of a composition of the invention may be evaluated using chromatography, e.g. Size-Exclusion Chromatograpy (SEC). SEC separates molecules on the basis of size. A column is filled with semi-solid beads of a polymeric gel that will admit ions and 15 small molecules into their interior but not large ones. When a protein compostion is applied to the top of the column, the compact folded proteins (ie. non-aggregated proteins) are distributed through a larger volume of solvent than is available to the large protein aggregates. Consequently, the large aggregates move more rapidly through the column, and in this way the mixture can be separated or fractionated into its 20 components. Each fraction can be separately quantified (e.g. by light scattering) as it elutes from the gel. Accordingly, the % aggregation of a composition of the invention can be determined by comparing the concentration of a fraction with the total concentration of protein applied to the gel. Stable compositions elute from the column as essentially a single fraction and appear as essentially a single peak in the elution 25 profile or chromatogram. In preferred embodiments, SEC is used in conjunction with in-line light scattering (e.g. classical or dynamic light scattering) to determine the % aggregation of a composition. In certain preferred embodiments, static light scattering is employed to measure the mass of each fraction or peak, independent of the molecular shape or elution 30 position. In other preferred embodiments, dynamic light scattering is employed to measure the hydrodynamic size of a composition. Other exemplary methods for evaluating protein stability include High-Speed SEC (see e.g. Corbett et al., Biochemistry. 23(8):1888-94, 1984). 102 WO 2007/109254 PCT/US2007/006883 In a preferred embodiment, the % aggregation is determined by measuring the fraction of protein aggregates within the protein sample. In a preferred embodiment, the % aggregation of a composition is measured by determining the fraction of folded protein within the protein sample. 5 c) %Yield In other embodiments, the stability of a composition of the invention is evaluated 10 by measuring the amount of protein that is recovered (herein the "% yield") following expression (e.g. recombinant expression) of the protein. For example, the %yield can be measured by determining milligrams of protein recovered for every ml of host culture media (ie. mg/mi of protein). In a preferred embodiment the % yield is evaluated following expression in a mammalian host cell (e.g. a CHO cell). 15 d) % Loss In yet other embodiments, the stability of a composition of the invention is evaluated by monitoring the loss of protein at a range of temperatures (e.g. from -80 to 25 *C) following storage for a defined time period. The amount or concentration of 20 recovered protein can be determined using any protein quantification method known in the art, and compared with the initial concentration of protein. Exemplary protein quantification methods include SDS-PAGE analysis or the Bradford assay for (Bradford, et al., Anal. Biochem. 72, 248, (1976)). A preferred method for evaluating % loss employs any of the analytical SEC methods described supra. It will be appreciated that 25 % Loss measurements can be determined under any desired storage condition or storage formulation, including, for example, lyophilized protein preparations. e) % Proteolysis In still other embodiments, the stability of a composition of the invention is 30 evaluated by determining the amount of protein that is proteolyzed following storage under standard conditions. In an exemplary embodiment, proteolysis is determined by SDS-PAGE a sample of the protein wherein the amount of intact protein is compared with the amount of low-molecular weight fragments which appear on the SDS-PAGE 103 WO 2007/109254 PCT/US2007/006883 gel. In another exemplary embodiment, proteolysis is determined by Mass Spectrometry (MS), wherein the amount of protein of the expected molecular weight is compared with the amount of low-molecular weight protein fragments within the sample. 5 f) Binding Affinity In still other embodiments, the stability of a composition of the invention may be assessed by determining its target binding affinity. A wide variety- of methods for determining binding affinity are known in the art. An exemplary method for determining binding affinity employs surface plasmon resonance. Surface plasmon 10 resonance is an optical phenomenon that allows for the analysis of real-time biospecific interactions by detection of alterations in protein concentrations within a biosensor matrix, for example using the BlAcore system (Pharmacia Biosensor AB, Uppsala, Sweden and Piscataway, NJ). For further descriptions, see Jbnsson, U., et aL. (1993) Ann. Biol. Clin. 51:19-26; Jonsson, U., et aL. (1991) Biotechniques 11:620-627; 15 Johnsson, B., et aL. (1995) J. Mol. Recognit. 8:125-131; and Johnnson, B., et aL. (1991) AnaL Biochem. 198:268-277. g) Other Binding Studies In yet other embodiments, the stability of a composition of the invention may be 20 asessed by quantifying the binding of a labeled compound to denatured or unfolded portions of a binding molecule. Such molecules are preferably hydrophobic, as they preferably bind or interact with large hydrophobic patches of amino acids that are normally buried in the interior of the native protein, but which are exposed in a denatured or unfolded binding molecule. An exemplary labeled compound is the 25 hydrophobic fluorescent dye, 1 -anilino-8-naphthaline sulfonate (ANS). VII. Methods for Selecting Stable Proteins The methods described supra for the prediction of protein stability can be employed to select a candidate protein (e.g. antibodies) for further use. In certain 30 embodiments, the methods of the invention are employed to select a protein for expression. In other embodiments, the methods of the invention are used to select a candidate protein for modification. In exemplary embodiments, the prediction methods of the invention can be employed in the humanization of a non-human donor antibody 104 WO 2007/109254 PCT/US2007/006883 (e.g. to select an acceptor immunoglobulin). A candidate protein may be selected based on its total consensus score or sequence score as determined using the methods of the invention. In certain embodiments, the candidate protein is selected if its consensus score is greater than a 5 suitable negative control (e.g. greater than 5%, preferably greater than 10%, more preferably greater than 20%). In other embodiments, the candidate protein is selected if its consensus score is substantially similar to (e.g. within 20%, -10%, .or 5%) or greater than (e.g. greater than 5%, preferably greater than 10%, more preferably greater than 20% ) a suitable positive control. 10 In other embodiments, a candidate protein is selected if its consensus score is substantially similar to its ideal or perfect consensus score (e.g. within 30%, preferably within 20%, more preferably within 10%). In other embodiments, a candidate protein is selected if its sequence score is greater than zero. In one embodiment, the candidate protein is selected if its sequence 15 score is greater than 0.5. In another embodiment, the candidate protein is selected if its sequence score is greater than 1. In other embodiment, the candidate protein is selected if its sequence sequence score is greater than 2. In a preferred embodiment, the candidate protein is selected if its sequence score is greater than 3. In other embodiments, a candidate protein is selected if its A score is at 20 least -3, at least -2, or at least -1, preferably at least 0, more preferably at least 1. a. Selecting an Acceptor Immunoglobulin for Antibody Humanization Humanized antibodies can be produced using recombinant DNA technology, see 25 for example, e.g., Queen et al., Proc. Natl. A cad. Sci. USA, (1989), 86:10029-10033; Jones et al., Nature, (1986), 321:522-25; Riechmann et al., Nature, (1988), 332:323-27; Verhoeyen et al., Science, (1988), 239:1534-36; Orlandi et al., Proc. Nat. Acad. Sci. USA, (1989), 86:3833-37; US Patent Nos. US 5,225,539; 5,530,101; 5,585,089; 5,693,761; 5,693,762; 6,180,370. When a preferred nonhuman donor antibody has been 30 selected for humanization, an appropriate human acceptor antibody may be obtained, e.g., from sequence databases of expressed human antibody genes, from germline Ig sequences or a consensus sequence of several human antibodies. The substitution of nonhuman CDRs into a human variable domain framework is most likely to result in 105 WO 2007/109254 PCT/US2007/006883 retention of their correct spatial orientation if the human variable domain framework adopts the same or similar conformation to the nonhuman variable framework from which the CDRs originated. This is achieved by obtaining the human variable domains from human acceptor antibodies whose framework sequences exhibit a high degree of 5 sequence identity with the nonhuman variable framework domains from which the CDRs were derived. The heavy and light chain variable framework regions can be derived from the same or different human antibody sequences. Preferably the human acceptor antibody retains the canonical and interface residues of the donor antibody. Additionally, the human acceptor antibody preferably has substantial similarity in the 10 length of CDR loops. See Kettleborough et al., Protein Engineering 4:758 (1991); Kolbinger et al., Protein Engineering 6:971 (1993) and Carter et al., WO 92/22653. Having identified the CDRs of the donor antibody and appropriate human acceptor antibody, the next step is to determine which, if any, residues from these components should be substituted to optimize the properties of the resulting humanized 15 antibody. Typically, some or all of the amino acids of the nonhuman, donor immunoglobulin light or heavy chain that are required for antigen binding (e.g., one or more CDRs) are used to substitute for the corresponding amino acids from the light or heavy chain of the human acceptor antibody. The human acceptor antibody retains some or all of the amino acids that are not required for antigen binding. When 20 necessary, one or more residues in the human framework regions can be changed to residues at the corresponding positions in the murine antibody so as to preserve the binding affinity of the humanized antibody to the antigen. This change is sometimes called "back mutation." Certain amino acids from the human variable region framework residues are selected for back mutation based on their possible influence on CDR 25 conformation and/or binding to antigen. In some cases, however, the humanization process results in unnatural changes to the variable domain which confers undesirable instability to the humanized antibody. For example, the acceptor immunoglobulin may have rare sequence variations that confer low stability. This is particularly true of certain germline sequences which may 30 not be used with great frequency by the immune system. The methods of the invention provide improved methods for humanization. In particular, the methods of the invention allow the skilled artisan to predict which test sequence (e.g. test germline sequence) with a reference set of candidate acceptor 106 WO 2007/109254 PCT/US2007/006883 sequences (e.g. germline sequences) are suitably stable for use in the humanization of the acceptor sequence. Test acceptor sequences (e.g. germline sequences) which have a score that is predictive of acceptable stability (e.g. have a high consensus score as compared to the average consensus score of the reference set) may be selected as an 5 acceptor immunoglobulin. VIII. Library Based Methods for Identifying Stabilized scFv molecules In another aspect, the present invention provides methods of identifying scFv 10 molecules with improved protein stability, e.g. improved thermal stability. The methods of the invention comprise (i) providing a library comprising candidate scFv molecules; and (ii) screening the library to identify candidate scFv molecules with improved protein stability relative to a suitable control (e.g. a control scFv molecule). As used herein, a "candidate scFv molecule" isan scFv molecule formed by introducing one or more 15 candidate stabilizing mutations into the scFv molecule, wherein the effect of the candidate stabilizing mutation on the stability of the scFv molecule is not known a priori, i.e.,an scFv molecule into which a mutation has been introduced and which has to be evaluated to determine whether the mutation results in an scFv molecule with increased stability. In one embodiment, a candidate scFv molecule is one formed by 20 introducing one or more candidate stabilizing mutations into a conventional (ie. non stabilized) scFv molecule. In another embodiment, a candidate scFv molecule is one formed by introducing one or more candidate stabilizing mutations into one of the stabilized scFv molecules described in Section II, supra. By a "library of candidate scFv molecules" herein is meant at least two non 25 redundant candidate scFv molecules, with at least about 10 being preferred, at least about 100 being particularly preferred, and at least about 1000 being especially preferred (e.g. at least about 104, 10', 106, 10', or 108 scFv molecules). In one embodiment, the library is randomized, with unspecified mutations generated randomly at any position. In another embodiment, the scFv library is a partially randomized or designed. That is, a 30 specified amino acid residue or class of amino acid residues is introduced randomly at any position of the scFv, or a specified amino acid position is selected for mutagenesis with an unspecified amino acid, a particular class of amino acids, or a specified amino amino acid. In a preferred embodiment,an scFv library is designed by substituting a 107 WO 2007/109254 PCT/US2007/006883 selected residue of an scFv molecule with an amino acid of a defined class, for example, a hydrophobic amino acid, a basic amino acid, a hydrophilic amino acid, a charged amino acid, a sterically biased (either small or large) amino acid, or a cysteine capable of disulfide bond formation. 5 In certain embodiments, the scFv library comprises candidate scFv molecules with candidate stabilizing mutation(s) that are introduced within a variable region (VL arid/or VH) of the scFv molecule. In other embodiments, the scFv library comprises candidate scFv molecules with candidate stabilizing mutation(s) that are introduced withinan scFv linker of the scFv molecule. In other embodiments, the scFv library 10 comprises candidate scFv molecules with candidate stabilizing mutation(s) that are introduced withinan scFv linker of the scFv molecule and candidate stabilizing mutation(s) that are introduced within a variable region (VL and/or VH). a. Design of scFv Libraries 15 Methods for designing the scFv library may be aided by molecular or computational modeling. In certain embodiments, library design may be carried out in silico. In certain embodiments, the scFv library may be designed by the following two step method: 1) identifying target amino acid residues in the scFv that when altered by a mutation 20 (e.g, by amino acid substitution), are predicted to result in improved scFv stability (herein "candidate destabilizing residues"); and 2) substituting the target amino acid residues with candidate stabilizing amino acid residues. 25 Step #1: Identifying Candidate Destabilizing Residues In certain embodiments, candidate destabilizing amino acid residues may be identified by sequence-based analysis. For example, candidate destabilizing amino acid residues may be identified by comparing a variable region sequence ofan scFv with a reference set of variable region sequences, e.g. variable region sequences from naturally 30 occurring human antibodies, and selecting those variable region amino acid residues of the scFv which are unusual or rare at their corresponding amino acid positions within the reference set. In preferred embodiments, only the framework regions of a variable region sequence of the scFv are analysed, while the complementarity determining 108 WO 2007/109254 PCT/US2007/006883 regions (CDRs) of the variable region are conserved in order to avoid disrupting the binding activity of the scFv molecule. In certain embodiments, a candidate destabilizing amino acid residue is one which is absent or found at a low frequency at a corresponding position within a 5 reference set of homologous variable region sequences. Methods for compiling reference sets are described in Sections III-V, supra. In preferred embodiments, the candidate destabilizing amino acid residue is one which is present within less than 10% of the sequences at corresponding positions within the reference set. In more preferred embodiments, the candidate destabilizing amino acid residue is one which is present 10 within less than 5% of the sequences at corresponding positions within the reference set. In particularly preferred embodiments, the candidate destabilizing amino acid residue is one which is present within less than 2% (e.g. 0.5%, 0.75%, or 1%) of the sequences at corresponding positions within the reference set. In another embodiment, a candidate destabilizing amino acid residue is one 15 which differs from the amino acid present at the corresponding position in a consensus sequence of the reference set of variable region sequences (ie. the consensus amino acid residue). Methods for determining the consensus sequence of the reference set are described in Section III and V, supra. In another embodiment, a candidate destabilizing amino acid residue is one 20 which has a low consensus score. Methods for determining a consensus score of an amino acid are described in Section III and V, supra. In one embodiment, the candidate destabilizing amino acid residue is one with a consensus score of less than 0.5 (e.g. less than 0.4, less than 0.3, less than 0.2, or less than 0.1). In a preferred embodiment, the candidate destabilizing amino acid residue is one with a consensus 25 score of less than 0.3. Step #2: Selection of Candidate Stabilizing Mutations Having identified one or more candidate destabilizing amino acids, the next step is to select one or more candidate stabilizing mutations for each destabilizing amino acid. 30 The scFv library may then be designed to include a representative candidate scFv molecule for each candidate stabilizing mutation that is selected. In certain embodiments, every natural amino acid variant of a particular destabilizing amino acid is selected as a candidate stabilizing mutation (ie. 19 candidate 109 WO 2007/109254 PCT/US2007/006883 stabilizing mutations for every destabilizing amino acid). In more preferred embodiments, a subset of the natural amino acid variants of a particular destabilizing amino acid are selected as candidate stabilizing mutations (ie. 1 18 candidate stabilizing amino acids for every destabilizing amino acid). In one 5 embodiment, the subset of candidate stabilizing mutations comprise substitutions with an amino acid of a defined class, for example, hydrophobic amino acids, basic amino acids, hydrophilic amino acids, charged amino acids; or sterically biased amino acids (either small or large) amino acids. In one embodiment, a subset of candidate stabilizing mutations include 10 substitutions with amino acids which are present at high frequencies at a position corresponding to that of the destabilizing amino acid within a reference set of homologous variable region sequences. In preferred embodiments, a candidate stabilizing mutation comprises substitution with an amino acid that is present within the database at a frequency of greater than 10%. In more preferred embodiments, the amino 15 acid is present at a frequency greater than 15%. In still more preferred embodiments, the amino acid is present at a frequency of greater than 20%. In yet more preferred embodiments, the amino acid is present at a frequency of greater than 25%. In another embodiment, a candidate stabilizing mutation is a substitution with the consensus amino acid (ie. most frequent residue) found at the position of the destabilizing 20 amino acid within the reference set. In another embodiment, a subset of candidate stabilizing mutations comprise substitutions with every amino acid that is found within a reference set of homologous variable region sequences at the position of the destabilizing amino acid. Accordingly,an scFv library may then be designed to include a representative candidate scFv molecule 25 for each candidate stabilizing mutation that is represented in the reference set. In other embodiments, the subset of candidate stabilizing mutations may be identified or prioritized for screening by an analysis (e.g. visual inspection or computational analysis) of a three-dimensional structure or model of a variable region of the scFv molecule. The three-dimensional structure of a polypeptide influences its 30 biological activity and stability, and that structure can be determined or predicted in a number of ways. Tertiary structure can be predicted using model building of three dimensional structures of one or more homologous proteins (or protein complexes) that have a known three-dimensional structure. X-ray crystallography is perhaps the best 110 WO 2007/109254 PCT/US2007/006883 known way of determining protein structure (accordingly, the term "crystal structure" may be used in place of the term "structure"), but estimates can also be made using circular dichroism, light scattering, or by measuring the absorption and emission of radiant energy. Other useful techniques include neutron diffraction, nuclear magnetic 5 resonance (NMR), and homology modeling. All of these methods are known to those of ordinary skill in the art, and they have been well described in standard textbooks (see, e.g., Physical Chemistry, 4th Ed., W.J. Moore, Prentiss-Hall, N.J., 1972, or Physical Biochemistry, K.E. Van Holde, Prentiss-Hall, N.J., 1971)) and numerous publications. Any of these techniques can be carried out to determine the structure of a molecule 10 comprising a variable region ofan scFv molecule (e.g. an antibody, a Fab, oran scFv molecule itself). The structure of the variable region may be modeled in silico. For example, the compatability of a candidate stabilizing mutation with the three-dimensional structure may be analyzed by computationally modeling the substitution of a destabilizing 15 mutation with a candidate stabilizing mutation. The candidate stabilizing mutation may be selected for inclusion in the scFv library if it is compatable with the overall structure of the scFv molecule. In one embodiment, the candidate stabilizing mutation may be selected if it does not perturb the native folding or conformation of the variable region of the scFv molecule or one or more complementarity determining regions (CDRs) thereof. 20 In another embodiment, the candidate stabilizing mutation may be selected if it does not perturb the ability of the variable region to form a native VL/VH interface. In certain embodiments, a candidate stabilizing mutation may be selected by applying a sidechain repacking technique to a structure (e.g. the crystal structure or model) of the variable region. In a sidechain repacking calculation, the candidate 25 stabilizing residues can be modified computationally, and the stability of the resulting mutants is evaluated computationally. The sidechain repacking calculation generates a ranked list of the mutants that have altered stability (i.e., altered intramolecular energy). The number of protein mutants that is evaluated computationally can be very large, since every variable amino acid position can be mutated into all 20 standard amino acids. 30 Exemplary computational algorithms used to rank the results of the computational analysis include dead-end elimination and tree search algorithms (see for example, Lasters et al. (Protein Eng. .:815-822, 1995), Looger and Hellinga (J. Mol. Biol. 307:429-445, 2001), and Dahiyat and Mayo (Protein Sci. 5:895-903, 1996)).
III
WO 2007/109254 PCT/US2007/006883 Accordingly,an scFv library may then be designed to include a representative candidate scFv molecule for each of the top-ranked candidate stabilizing mutations in the ranked list of mutations generated by the sidechain repacking calculation. In certain embodiments, at least the top ranked mutation is selected (e.g., the top ranked, the top 5 two ranked, the top three ranked, the top four ranked, or the top five ranked mutations are selected). b. Construction of scFv libraries 10 Having determined the candidate stabilizing mutations to include in the scFv library, one can use any of a variety of available methods to produce candidate scFv molecules comprising the mutations. Such polypeptides can, for example, be produced by recombinant methods. Moreover, because of the degeneracy of the genetic code, a variety of nucleic acid sequences can be used to encode each desired scFv. 15 Exemplary art recognized methods for making a nucleic acid molecule encoding a candidate scFv molecule include, but are not limited to, preparation by site-directed (or oligonucleotide-mediated) mutagenesis, PCR mutagenesis, and cassette mutagenesis of an earlier prepared DNA encoding the candidate scFv. Site-directed mutagenesis is a preferred method for preparing substitution 20 variants. This technique is well known in the art (see, e.g., Carter et al. Nucleic Acids Res. 13:4431-4443 (1985) and Kunkel et al., Proc. Natl. Acad. Sci. USA 82:488 (1987)). Briefly, in carrying out site-directed mutagenesis of DNA, the parent DNA is altered by first hybridizing an oligonucleotide encoding the desired mutation to a single strand of such parent DNA. After hybridization, a DNA polymerase is used to synthesize an entire 25 second strand, using the hybridized oligonucleotide as a primer, and using the single strand of the parent DNA as a template. Thus, the oligonucleotide encoding the desired mutation is incorporated in the resulting double-stranded DNA, which may then be ligated into a plasmid or other suitable vector. PCR mutagenesis is also suitable for making candidate scFv molecules. See 30 Higuchi, in PCR Protocols, pp.177-183 (Academic Press, 1990); and Vallette et al., Nuc. Acids Res. 17:723-583 (1989). Briefly, when small amounts of template DNA are used as starting material in a PCR, primers that differ slightly in sequence from the corresponding region in a template DNA can be used to generate relatively large 112 WO 2007/109254 PCT/US2007/006883 quantities of a specific DNA fragment that differs from the template sequence only at the positions where the primers differ from the template. The PCR primers may also be designed to incorporate restriction sites, such that the DNA product of the PCR reaction can be can then be directly ligated into a plasmid or other suitable vector. 5 Another method for preparing variants, cassette mutagenesis, is based on the technique described by Wells et al., Gene 34:315-323 (1985). The starting material is the plasmid (or other vector) comprising the starting polypeptide DNA to be mutated. The, codon(s) in the parent DNA to be mutated are identified. There must be a unique restriction endonuclease site on each side of the identified mutation site(s). If no such 10 restriction sites exist, they may be generated using the above-described oligonucleotide mediated mutagenesis method to introduce them at appropriate locations in the starting polypeptide DNA. The plasmid DNA is cut at these sites to linearize it. A double stranded oligonucleotide encoding the sequence of the DNA between the restriction sites but containing the desired mutation(s) is synthesized using standard procedures, wherein 15 the two strands of the oligonucleotide are synthesized separately and then hybridized together using standard techniques. This double-stranded oligonucleotide is referred to as the cassette. This cassette is designed to have 5' and 3' ends that are compatible with the ends of the linearized plasmid, such that it can be directly ligated to the plasmid. Representative nucleic acids for each of the candidate scFv molecules may be 20 generated by the above methods. The nucleic acids may then be cloned into expression vectors to form an expression vector library. Host cells may then be transformed with the resulting library of vectors, and the host cells cultured under the appropriate conditions in order to express each candidate scFv molecule. 25 c. Screening Methods A scFv library the invention may be screened in an assay (e.g. a high-throughput assay) to identify candidate scFv molecules with desired protein stability. Such an assay may employ any of the methods for evaluating protein stability described in Section VI, supra. A particularly preferred method is a thermal challenge assay. 30 Such assay methods generally involve comparing the thermal stability of a candidate scFv molecule with that of a suitable control and selecting the candidate sdFv molecule if the thermal stability is greater than that of the control. Exemplary suitable controls include conventional scFv molecules, e.g. a (Gly4Ser)3 scFv molecule. Candidate scFv 113 WO 2007/109254 PCT/US2007/006883 molecules are may be selected if they have a thermal stability that is greater than about 0.1, about 0.25, about 0.5, about 0.75, about 1, about 1.25, about 1.5, about 1.75, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, or about 10 degrees Celsius than that of the control. In an exemplary embodiment, a candidate scFv 5 molecule is selected if it has a thermal stability than is greater than about 3 degrees Celsius than that of the control. scFv Libraries may be presented in different assay-formats. For example, scFv molecules may be presented in solution (e.g., Houghten (1992) Biotechniques 13:412 421), on beads (Lam (1991) Nature 354:82-84), chips (Fodor (1993) Nature 364:555 10 556), bacteria (Ladner USP 5,223,409), spores (Ladner USP '409), or on phage (Scott and Smith (1990) Science 249:386-390); (Devlin (1990) Science 249:404-406); (Cwirla et aL. (1990) Proc. Natl. Acad. Sci. 87:6378-6382); (Felici (1991) J. Mol. Biol. 222:301 310); (Ladner supra.). In certain exemplary embodiments, the candidate scFv molecules are assayed in 15 a solution format. In one embodiment, each sample comprises an aliquout of a solution having candidate scFv molecules with the same candidate stabilizing mutation or mutations. Such a solution can generated by isolating individual host cell colonies from a library of host cells transformed with expression plasmid library and culturing the host cell colony in an appropriate vessel under conditions which facilitate expression of the 20 scFv molecule. In one embodiment, the candidate scFv molecule may be purified from the host cell and resolubilized in an appropriate assay solution. In a more preferred embodiment, the candidate scFv molecule is fused to a cleavable signal peptide sequence such that the scFv molecule is secreted by the host cell into the host cell culture medium. In yet other preferred embodiments, the host cell may be cultured under 25 conditions such that candidate scFv molecule is released into the media, along with host cell proteins. It may be desirable to automate any of the above assay formats. For example, robotics may be employed to isolate each member of the scFv library (e.g. as individual host cell colonies) in rapid succession so they can be assayed in separate vessels. 30 Examples of assay vessels include microtiter plates (e.g. 96-well microtiter plates), test tubes, and micro-centrifuge tubes. d. Further Optimization of Stabilized scFv molecules 114 WO 2007/109254 PCT/US2007/006883 A stable scFv molecule identified by the above screening methods can be re modeled and further optimized to further improve its protein stability. Thus, the steps described above can be repeated with a stable scFv molecule identified in an initial round of optimization. Alternatively, the stabilizing mutations from two or more 5 stabilized scFv molecules may be combined in a single scFv molecule to further improve protein stability. In certain embodiments, the stable scFv molecule identified by the methods of the invention can be further optimized. For example, on or more of the following additional alterations may be made. In one embodiment, a stabilized scFv molecule may 10 be further stabilized by introducing a disulfide bond which links an amino acid in the VL domain with an amino acid in the VH domain. Exemplary disulfide bonds include any the disulfide bonds described in Section II supra. A particularly preferred disulfide bond is VH44-VL100. In other embodiments, the stable scFv molecule identified by the methods of the 15 invention is further optimized by introducingan scFv linker with an optimized length or composition. Exemplary scFv linkers are described in Section II supra. A particularly preferred scFv linker is (Gly 4 Ser) 4 . In another embodiment, the stable scFv molecule identified by the methods of the invention is further optimized by introducing a stabilizing mutation into at least one 20 of the VH or VL domain. IX. Methods for Stabilizing Binding Molecules In other aspects, the invention provides methods for improving the stability properties of binding molecules. These methods generally involve incorporating or 25 appending a stabilized scFv molecule of the invention to the binding molecule. Surprisingly, as shown in the working examples herein, scFv molecules of the invention are not only stable on their own, they also confer improved stability to binding molecules into which they are incorporated. Accordingly, the methods of the invention provide a convenient and reliable means for improving the stability of commercially 30 valuable binding molecules for which large scale manufacturing is often limited by poor protein stability (e.g. multispecific antibodies (e.g. bispecific antibodies) and other modified antibodies). Stabilized scFv molecules may be incorporated into binding molecules using 115 WO 2007/109254 PCT/US2007/006883 protein conjugation methodology that is known in the art. In one embodiment, the stabilized scFv is fused directly to an N- or C-terminus of a polypeptide, e.g., an antibody molecule. In another embodiment, a non-peptide linker is employed to link the stabilized scFv to an N- or C-terminus of a polypeptide. In yet other embodiments, a 5 connecting peptide is used to link the stabilized scFv to a polypeptide. In an exemplary embodiment, the connecting peptide is a short gly/ser rich peptide. Exemplary Gly/Ser rich peptides are listed in Table 1 below. Other exemplary connecting peptides are known in the art (see, e.g., International PCT Application Nos.WO 2005/000898 and WO 2005/000899). In one embodiment, a stabilized scFv of the invention is linked to 10 the C-terminal end of a binding molecule, e.g., an antibody molecule, using a S(G 4
S)
3 linker. In another embodiment, a stabilized scFv of the invention is linked to the N terminal end of a binding molecule, e.g., an antibody molecule, using a (G 4
S)
5 linker. Table 1: Connecting Peptides Linker SEQ ID DNA or Amino Acid Sequence NO (Gly 4 Ser)s 132 5' GGCGGTGGAGGGTCCGGTGGAGGGGGCTCTGGA GGGGGCGGTTCAGGGGGCGGTGGATCGGGGGG AGGTGGCTCC-3' 133 GGGGSGGGGSGGGGSGGGGSGGGGS Ser(Gly 4 Ser) 3 42 5' TCCGGCGGGGGTGGATCCGGTGGAGGGGGCTCC GGCGGTGGCGGGTCC-3' 43 SGGGGSGGGGSGGGGS 15 In one embodiment, at least one stabilized scFv molecule is appended to an antibody molecule to make a bispecific molecule. In another embodiment, two stabilized scFv molecules are appended to an antibody molecule to make a bispecific molecule. 20 In certain embodiments, stabilized binding molecules of the invention result in increased yield as compared to conventional scFv molecules or binding molecules comprising conventional scFv molecules. Methods for evaluating yield are described in Section VI, supra. In one embodiment, a stabilized binding molecule produced by the methods of the invention has an increase in yield of at least 1% relative to the 25 unstabilized binding molecule. In other embodiments, the stabilized binding molecule has an increase in yield of at least 2%, at least 5%, at least 10%, at least 20%, at least 116 WO 2007/109254 PCT/US2007/006883 30%, at least 50%, at least 75%, or at least 100%, relative to the unstabilized binding molecule. In certain embodiments, binding molecules of the invention result in reduced aggregation as compared to conventional scFv molecules or binding molecules 5 comprising conventional scFv molecules. Methods for evaluating aggregation are described in Section VI, supra. In one embodiment, a stabilized binding molecule produced by the methods of the invention has a decrease in aggregation of at least 1% relative to the unstabilized binding molecule. In other embodiments, the stabilized binding molecule has a decrease in aggregation of at least 2%, at least 5%, at least 10%, 10 at least 20%, at least 30%, at least 50%, at least 75%, or at least 100%, relative to the unstabilized binding molecule. In other embodiments, binding molecules of the invention result in increased long-term stability or shelf-life as compared to conventional scFv molecules or binding molecules comprising conventional scFv molecules. Methods for evaluating shelf-life 15 include % loss or % proteolysis as described in Section VI, supra. In one embodiment, a stabilized binding molecule produced by the methods of the invention has an increase in shelf life of at least 1 day relative to the unstabilized binding molecule. This means that a preparation of binding molecules has substantially the same amount of stable binding molecules as present on the previous day. In other embodiments, the stabilized binding 20 molecule has an increase in shelf life of at least 2 days, at least 5 days, at least I week, at least 2 weeks, at least 1 month, at least 2 months, at least 6 months, or at least 1 year, relative to the unstabilized binding molecule. In other embodiments, binding molecules of the invention result improved stability e.g., when expressed in a particular host cell type, as compared to conventional 25 scFv molecules or binding molecules comprising conventional scFv molecules. . In exemplary embodiments, the methods of the invention result in the production of binding molecules which have increased stability (e.g. increased yield) when the binding molecule is expressed in a host cell, e.g., a bacterial or eukaryotic (e.g., yeast or mammalian) host cell. Exemplary mammalian host cells include Chinese Hamster 30 Ovary (CHO) cells, HELA (human cervical carcinoma) cells, CVI (monkey kidney line) cells, COS (a derivative of CVI with SV40 T antigen) cells, R1610 (Chinese hamster fibroblast) cells, BALBC/3T3 (mouse fibroblast) cells, HAK (hamster kidney line) cells, SP2/O (mouse myeloma) cells, BFA-lcIBPT cells (bovine endothelial cells), RAJI 117 WO 2007/109254 PCT/US2007/006883 (human lymphocyte) cells and 293 cells (human kidney). In a preferred embodiment, two stabilized scFv molecules are appended to an antibody molecule to create a stabilized bispecific molecule for secretion in CHO cells. In other embodiments, host cells capable of expressing stabilized binding 5 molecules can be screened to select for single cell isolates that are capable of expressing high levels of solubilized and properly-folded stabilized binding molecules (e.g. binding molecules exhibiting less than 10% aggregation). Such methods may employ fluorescence-activated cell sorting (FACS) techniques (see, for example, Brezinky et al., J Immunol Meth (2003). 277:141-155). In one embodiment, the single cell isolate is 10 adapted to serum-free conditions to establish a stable producer cell line. The stable producer cell line may then be cultured to facilitate large-scale manufacture of a stabilized binding molecule of the invention. In other embodiments, the methods of the invention are employed to improve the stability of a binding molecule that is expressed from a host cell in a large volume of 15 culture media. In exemplary embodiments, the methods of the invention result in an increased stability (e.g. increased yield) when the binding molecule is expressed in at least I liter of culture media. In other embodiments, the methods of the invention are used to produce a stabilized binding molecule which has an increased stability (e.g. yield) when expressed from a host cell in at least 2 liters, at least 10 liters, at least 20 20 liters, at least 50 liters, at least 75 liters, at least 100 liters, at least 200 liters, or at least 500 liters of culture media. In an exemplary embodiment, the methods of the invention are used to produce at least 10 mg of a stabilized binding molecule for every liter of culture media. 25 X. Stabilized Binding Molecules Comprising Stabilized scFv Molecules In one embodiment, a stabilized binding molecule of the invention is a fusion protein. For example, a stabilized scFv molecule of the invention may be linked to a second scFv molecule or a non-scFv molecule. In one embodiment, a non-scFv molecule to which a stabilized scFv molecule of the invention may be linked provides at 30 least one additional binding site. For example, exemplary binding sites that can be included in a binding molecule of the invention include: the receptor binding portion of a ligand, the ligand binding portion of a receptor, the substrate binding portion of an enzyme, the enzyme binding portion of a substrate, or one or more antigen binding 118 WO 2007/109254 PCT/US2007/006883 portions of an antibody. scFv molecules may be linked, e.g., to antibody molecules to form modified antibody molecules or to other polypeptides to form fusion proteins. Some examples are described below. 5 A. Modified Antibody Molecules In one embodiment, a stabilized scFv molecule of the invention is linked to an antibody or fragment thereof to form a stabilized binding protein which is a modified antibody. In another embodiment, a stabilized scFv of the invention is linked to a modified antibody, i.e., non-naturally occurring antibody molecule, to form a stabilized 10 binding protein. Preferred modified antibody constructs are described in more detail below. As used herein, the term "modified antibody" includes synthetic forms of antibodies which are altered such that they are not naturally occurring, e.g., antibodies that comprise at least two heavy chain portions but not two complete heavy chains (such as, domain deleted antibodies or minibodies); multispecific forms of antibodies (e.g., 15 bispecific, trispecific, etc.) altered to bind to two or more different antigens or to different epitopes on a single antigen). In addition, the term "modified antibody" includes multivalent forms of antibodies (e.g., trivalent, tetravalent, etc., antibodies that bind to three or more copies of the same antigen). It will be understood that when discussing the binding molecules of the 20 invention, the exemplary binding specificities described herein may be imparted by a stabilized scFv molecule of the invention, a binding molecule comprising an scFv molecule of the invention or both. In one embodiment, the stabilized binding proteins of the present invention may be immunoreactive with one or more tumor antigens or antigens associated with immune 25 disorders. For example, for neoplastic disorders, the binding site (i.e. the variable region or immunoreactive fragment or recombinant thereof) of the disclosed binding molecules binds to a selected tumor associated antigen at the site of the malignancy. Similarly, in one embodiment, a binding molecule may bind to at least one selected marker present on immune cells. Given the number of reported antigens associated with 30 neoplasias and immune disorders, and the number of related antibodies, those skilled in the art will appreciate that the presently disclosed binding molecules may therefore be derived from any one of a number of whole antibodies. More generally, polypeptides useful in the present invention may be obtained or derived from any antibody (including 119 WO 2007/109254 PCT/US2007/006883 those previously reported in the literature) that reacts with a molecule or marker associated with the selected condition. Further, the parent or precursor antibody, or fragment thereof, used to generate the stabilized binding molecules of the invention may be murine, human, chimeric, humanized, non-human primate or primatized. 5 As used herein, "tumor associated antigens" includes antigens which are generally associated with tumor cells, e.g., expressed on tumor cells. More generally, tumor associated antigens comprise antigens that provide for the localization of immunoreactive antibodies at a neoplastic cell irrespective of its expression on non malignant cells. Such antigens may be relatively tumor specific, e.g., limited in 10 expression to the surface of malignant cells. Alternatively, such antigens may be found on both malignant and non-malignant cells. For example, CD20 is a pan B antigen that is found on the surface of both malignant and non-malignant B cells that has proved to be an extremely effective target for immunotherapeutic antibodies for the treatment of non-Hodgkin's lymphoma. In this respect, pan T cell antigens such as CD2, CD3, CD5, 15 CD6 and CD7 also comprise tumor associated antigens within the meaning of the present invention. Still other exemplary tumor associated antigens comprise but not limited to MAGE-1, MAGE-3, MUC-1, HPV 16, HPV E6 & E7, TAG-72, CEA, L6 Antigen, CD19, CD22, CD37, CD52, HLA-DR, EGF Receptor and HER2 Receptor. In many cases immunoreative antibodies for each of these antigens have been reported in 20 the literature. Those skilled in the art will appreciate that each of these antibodies may serve as a precursor for antibodies of the invention in accordance with the present invention. In one embodiment, the stabilized binding molecules of the present invention preferably associate with, and bind to, tumor or immune associated antigens as described 25 above. Accordingly, as will be discussed in some detail below the stabilized binding molecules of the present invention may be derived, generated or fabricated from any one of a number of antibodies that react with tumor associated antigens. In certain embodiments the stabilized binding molecules of the invention are domain deleted antibodies that are derived using common genetic engineering techniques whereby at 30 least a portion of one or more constant region domains are deleted or altered so as to provide the desired biochemical characteristics such as reduced serum half-life. More particularly, one skilled in the art may readily isolate the genetic sequence corresponding to the variable and/or constant regions of the subject stabilized binding molecule and 120 WO 2007/109254 PCT/US2007/006883 delete or alter the appropriate nucleotides to provide polypeptides of the invention for use as monomeric subunits in accordance with the instant invention. Previously reported antibodies that react with tumor associated antigens may be stabilized as described herein to provide the stabilized binding molecules of the present 5 invention. Exemplary antibodies that may be used to provide antigen binding regions to generate or derive the disclosed stabilized binding molecules include, but are not limited to 2B8 and C2B8 (Zevalin@ and Rituxan@, IDEC Pharmaceuticals Corp., San Diego), Lym I and Lym 2 (Techniclone), LL2 (Immunomedics Corp., New Jersey), HER2 (Herceptin@, Genentech Inc., South San Francisco), B 1 (Bexxar@, Coulter Pharm., San 10 Francisco), Campath@ (Millennium Pharmaceuticals, Cambridge), abagovomab (Menarini, Italy), CEA-ScanTM (Immunomedics, Morris Plains, NJ), capromab (Prostascint@, Cytogen Corp.), edrecolomab (Panorex@, Johnson & Johnson, New Brunswick, NJ), igovomab (CIS Bio Intl., France), mitumomab (BEC2, Imclone Systems, Somerville, NJ), nofetumomab (Verluma@, Boehringer Ingleheim, Ridgefield, 15 CT), OvaRex (Altarex Corp., Waltham, MA), satumomab (Onoscint@, Cytogen Corp.), cetuximab (Erbitux@, Imclone Systems, New York, NY), bevacizumab (AVASTIN@, Genentech Inc., S. San Francisco, CA), apolizumab (REMITOGEN TM, Protein Design Labs, Fremont, CA), labetuzumab (CEACIDE TM, Immunomedics Inc., Morris Plains, NJ) , pertuzumab (OMNITARG TM, Genentech Inc., S. San Francisco, CA), MBI, BH3, 20 B4, B72.3 (Cytogen Corp.), CC49 (National Cancer Institute) and 5E10 (University of Iowa). Other binding sites that can be incorporated into the subject binding molecules include those found in: Orthoclone OKT3 (CD3), ReoPro (GpIIb/gIIa), Zenapax (C25), Remicade (TNF-a), Simulect (CD25), Synagis (RSV), Mylotarg (CD33), and Campath (CD52). In preferred embodiments, the stabilized binding molecules of the present 25 invention will bind to the same tumor associated antigens as the antibodies enumerated immediately above. In particularly preferred embodiments, the stabilized binding molecules will be derived from or bind the same antigens as 2B8, C2B8, CC49 and C5E10 and, even more preferably, will lack all or part of a CH2 domain. In one embodiment, a binding molecule of the invention may have one or more 30 binding sites derived from one or more of the following antibodies. tositumomab (BEXXAR@), muromonab (ORTHOCLONE@) and ibritumomab (ZEVALIN@), cetuximab (ERBITUXTM), rituximab (MABTHERA@ / RITUXAN@), infliximab (REMICADE@), abciximab (REOPRO®) and basiliximab (SIMULECT®), efalizumab 121 WO 2007/109254 PCT/US2007/006883 (RAPTIVA@, bevacizumab (AVASTIN@), alemtuzumab (CAMPATH@), trastuzumab (HERCEPTIN@), gemtuzumab (MYLOTARG@), palivizumab (SYNAGIS@), omalizumab (XOLAIR@), daclizumab (ZENAPAX@), natalizumab (TYSABRI@) and ranibizumab (LUVENTIS@), adalimumab (HUM IRA@) and panitumumab 5 (VECTIBIX@). In one embodiment, a stabilized binding molecule of the invention binds to CD23 (U.S. patent 6,011,138).- In a preferred embodiment, a stabilized binding molecule of the invention binds to the same epitope as the 5E8 antibody. In another embodiment, a binding molecule of the invention comprises at least one CDR (e.g., at 10 least 1, 2, 3, 4, 5, or 6 CDRs) from an anti-CD23 antibody, e.g., the 5E8 antibody. In one embodiment, a stabilized binding molecule of the invention binds to a TNF receptor. In one exemplary embodiment, a stabilized binding molecule of the invention binds to a LTpR. In another exemplary embodiment, a stabilized binding molecule of the invention binds to a TRAIL receptor. In another embodiment, a 15 stabilized binding molecule of the invention comprises at least one CDR(e.g., at least 1, 2, 3, 4, 5, or 6 CDRs) from an anti-TRAIL-R2 antibody (e.g. murine or chimeric 14A2). In another embodiment, a stabilized binding molecule of the invention comprises at least one CDR from an anti-LTPR antibody. Examples of anti-LTPR antibodies include BKA 11, CDH 10, BCG6, AGH1, BDA8, CBE1 1 and BHA10. 20 In one embodiment, a stabilized binding molecule of the invention binds to the CRIPTO-I antigen (WO02/088170A2 or WO03/083041A2). In a preferred embodiment, a stabilized binding molecule of the invention binds to the same epitope as the B3F6 antibody. In another embodiment, a stabilized binding molecule of the invention comprises at least one CDR from an anti-CRIPTO-I antibody, e.g., the B3F6 25 antibody. In one embodiment, the stabilized binding molecule will bind to the same tumor associated antigen as Rituxan@. Rituxan@ (also known as, rituximab, IDEC-C2B8 and C2B8) was the first FDA-approved monoclonal antibody for treatment of human B-cell lymphoma (see U.S. Patent Nos. 5,843,439; 5,776,456 and 5,736,137 each of which is 30 incorporated herein by reference). Y2B8 (90Y labeled 2B8; Zevalin@; ibritumomab tiuxetan) is the murine parent of C2B8. Rituxan@ is a chimeric, anti-CD20 monoclonal antibody which is growth inhibitory and reportedly sensitizes certain lymphoma cell lines for apoptosis by chemotherapeutic agents in vitro. The antibody efficiently binds 122 WO 2007/109254 PCT/US2007/006883 human complement, has strong FcR binding, and can effectively kill human lymphocytes in vitro via both complement dependent (CDC) and antibody-dependent (ADCC) mechanisms (Reff et al., Blood 83: 435-445 (1994)). Those skilled in the art will appreciate that dimeric variants (homodimers or heterodimers) of C2B8 or 2B8, 5 modified according to the instant disclosure, may be used in conjugated or unconjugated forms to effectively treat patients presenting with CD20+ malignancies. More generally, it must be reiterated that the stabilized binding molecule disclosed herein may be used in either a "naked" or unconjugated state or conjugated to a cytotoxic agent to effectively treat any one of a number of disorders. 10 In other preferred embodiments of the present invention, the stabilized binding molecule of the. invention will be derived from, or bind to, the same tumor associated antigen as CC49. CC49 binds human tumor associated antigen TAG-72 which is associated with the surface of certain tumor cells of human origin, specifically the LS174T tumor cell line. LS174T [American Type Culture Collection (herein ATCC) 15 No. CL 188] is a variant of the LS180 (ATCC No. CL 187) colon adenocarcinoma line. It will further be appreciated that numerous murine monoclonal antibodies have been developed which have binding specificity for TAG-72. One of these monoclonal antibodies, designated B72.3, is a murine IgGI produced by hybridoma B72.3 (ATCC No. HB-8108). B72.3 is a first generation monoclonal antibody developed using a 20 human breast carcinoma extract as the immunogen (see Colcher et al., Proc. Natl. Acad. Sci. (USA), 78:3199-3203 (1981); and U.S. Pat. Nos. 4,522,918 and 4,612,282 each of which is incorporated herein by reference). Other monoclonal antibodies directed against TAG-72 are designated "CC" (for colon cancer). As described by Schlom et al. (U.S. Pat. No. 5,512,443 which is incorporated herein by reference) CC monoclonal 25 antibodies are a family of second generation murine monoclonal antibodies that were prepared using TAG-72 purified with B72.3. Because of their relatively good binding affinities to TAG-72, the following CC antibodies have been deposited at the ATCC, with restricted access having been requested: CC49 (ATCC No. HB 9459); CC 83 (ATCC No. HB 9453); CC46 (ATCC No. HB 9458); CC92 (ATTCC No. HB 9454); 30 CC30 (ATCC No. HB 9457); CCII (ATCC No. 9455); and CC15 (ATCC No. RB 9460). U.S.P.N. 5,512,443 further teaches that the disclosed antibodies may be altered into their chimeric form by substituting, e.g., human constant regions (Fc) domains for mouse constant regions by recombinant DNA techniques known in the art. Besides 123 WO 2007/109254 PCT/US2007/006883 disclosing murine and chimeric anti-TAG-72 antibodies, Schlom et al. have also produced variants of a humanized CC49 antibody as disclosed in PCT/US99/25552 and single chain constructs as disclosed in U.S. Pat. No. 5,892,019 each of which is also incorporated herein by reference. Those skilled in the art will appreciate that each of the 5 foregoing antibodies, constructs or recombinants, and variations thereof, may be modified and used to provide polypeptides in accordance with the present invention. In addition to the anti-TAG-72 antibodies discussed above, various-groups have also reported the construction and partial characterization of domain-deleted CC49 and B72.3 antibodies (e.g., Calvo et al. Cancer Biotherapy, 8(1):95-109 (1993), Slavin 10 Chiorini et al. Int. J. Cancer 53:97-103 (1993) and Slavin-Chiorini et al. Cancer. Res. 55:5957-5967 (1995 Still other preferred embodiments of the present invention comprise binding sites that are derived from or bind to the same tumor associated antigen as C5E 10. As set forth in co-pending application 09/104,717, C5EIO is an antibody that recognizes a 15 glycoprotein determinant of approximately 115 kDa that appears to be specific to prostate tumor cell lines (e.g. DU145, PC3, or NDl). Thus, in conjunction with the present invention, stabilized binding molecules (e.g. CH2 domain-deleted antibodies) that specifically bind to the same tumor associated antigen recognized by C5E10 antibodies could be produced and used in a conjugated or unconjugated form for the 20 treatment of neoplastic disorders. In particularly preferred embodiments, the stabilized binding molecule will be derived or comprise all or part of the antigen binding region of the C5E10 antibody as secreted from the hybridoma cell line having ATCC accession No. PTA-865. The resulting stabilized binding molecule could then be conjugated to a radionuclide as described below and administered to a patient suffering from prostate 25 cancer in accordance with the methods herein. In certain embodiments, the stabilized binding molecules of the invention have at least one of the binding specificities described herein, e.g., in Section C infra. In another embodiment, a stabilized binding molecule of the invention may bind to a target molecule of interest, e.g., described herein, e.g., in section B or C, infra. 30 Previously reported antibodies that react with antigens associated with immune cell disorders (e.g., B-cell disorders) may be stabilized as described herein to provide the stabilized binding molecules of the present invention. Exemplary antibodies that may be used to provide antigen binding regions to generate or derive the disclosed stabilized 124 WO 2007/109254 PCT/US2007/006883 binding molecules include, but are not limited to an anti-TNFa antibody (e.g., infliximab (Remicade@, Centocor, Horsham, PA); MAK195-F (Abbott Labs., Abbott Park, IL); adalimumab (Humira@, Abbott Labs, Abbott Park, IL); an anti-CD3 antibody (e.g., Orthoclone (OKT3@, OrthoBiotech, Bridgewater, NJ); MEDI-500 (Medimmune, 5 Gaithersburg, MD); visilizumab (NUVION@, Protein Design Labs (Fremont, CA, USA)), an anti-IgE antibody (e.g., omalizumab, XOLAIR@ , Genentech, South San Francisco, CA), an anti-VLA-4 antibody (e.g.; TYSABRI@, Biogenldec, Cambridge, MA), an anti-CD 147 antibody (e.g., ABX-CBL (Abgenix, Fremont, CA)), an anti-CD25 antibody (e.g., basiliximab, Simulect@ (East Hanover, NJ); Inolimomab (OPI, France), 10 an anti-CD 18 antibody (e.g., odulimomab, Antilfa@, Pateur Meriuex, France), anti NCA90 (e.g., sulesomab, Leukoscan@, Immunomedics, Morris Plains, NJ), an anti GpIIb/gIIa antibody (e.g., abciximab, ReoPro®, Centocor, Horsham, PA), an anti-C25 antibody (e.g., Zenapax), an anti-CD33 antibody (e.g., Mylotarg), and an anti-CD25 antibody (e.g., alemtuzumab, Campath@ (Milleneum Pharmaceuticals, Cambridge, 15 MA). In preferred embodiments, the stabilized binding molecules of the present invention will bind to the same immune-cell ssociated antigens as the antibodies enumerated immediately above. In one embodiment, the term, "modified antibody" according to the present invention includes immunoglobulins, antibodies, or immunoreactive fragments or 20 recombinants thereof, in which at least a fraction of one or more of the constant region domains has been deleted or otherwise altered so as to provide desired biochemical characteristics such as the ability to non-covalently dimerize, increased ability to localize at the site of a tumor, or reduced serum half-life when compared with a whole, unaltered antibody of approximately the same immunogenicity. In a preferred 25 embodiment, the polypeptides of the present invention are domain deleted antibodies which comprise a polypeptide chain similar to an immunoglobulin heavy chain, but which lack at least a portion of one or more heavy chain domains. In one embodiment, one entire domain of the constant region of the stabilized binding protein will be deleted. In another embodiment, all or part of the CH2 domain will be deleted. 30 In one embodiment, the stabilized binding proteins of the invention are minibodies. Minibodies are dimeric molecules made up of two polypeptide chains each comprising a stabilized scFv molecule (a single polypeptide comprising one or more antigen binding sites, e.g., a VL domain linked by a flexible linker to a VH domain 125 WO 2007/109254 PCT/US2007/006883 fused to a CH3 domain via a connecting peptide. Minibodies can be made by constructing an scFv component and connecting peptide-CH3 component using methods described in the art (see, e.g., US patent 5,837,821 or WO 94/09817A]). These components can be isolated from separate 5 plasmids as restriction fragments and then ligated and recloned into an appropriate vector. Appropriate assembly can be verified by restriction digestion and DNA sequence analysis. In another embodiment, a tetravalent minibody can be constructed. Tetravalent minibodies can be constructed in the same manner as minibodies, except that two scFv 10 molecules are linked using a flexible linker, e.g., having an amino acid sequence
(G
4
S)
4
G
3 AS. In one embodiment, tetravalent antibodies can be produced by combining a DNA sequence encoding an antibody with a scFv molecule. For example, in one embodiment, these sequences are combined such that the scFv molecule is linked at its N-terminus to 15 the CH3 domain of the antibody via a flexible linker (e.g., a gly/ser linker such as (Gly 4 Ser) 3 . In another embodiment a tetravalent antibody can be made by fusing n stabilized scFv molecule to a connecting peptide, which is fused to a CHI domain to construct a stabilized scFv - Fab tetravalent molecule (Coloma and Morrison. 1997. Nature 20 Biotechnology. 15:159; WO 95/09917). In one embodiment a stabilized binding molecule of the invention comprisies a tetravalent or bispecific tetravalent antibody withan scFv appended to the N-terminus of the light chain. In another embodiment of the invention, a binding molecule comprises a tetravalent or bispecific tetravalent CH2 domain-deleted antibody withan scFv appended 25 to the N-terminus of the heavy chain. In one embodiment, the attachment of the scFv to the N-terminus results in reduced aggregation of the molecules as compared to molecules in which the scFv is attached at the carboxy-terminus. Antibodies or fragments thereof for use in a stabilized binding molecule of the invention may be obtained using art recognized protocols, for example, antibodies are 30 preferably raised in mammals by multiple subcutaneous or intraperitoneal injections of the relevant antigen (e.g., purified tumor associated antigens or cells or cellular extracts comprising such antigens) and an adjuvant. This immunization typically elicits an immune response that comprises production of antigen-reactive antibodies from activated 126 WO 2007/109254 PCT/US2007/006883 splenocytes or lymphocytes. While the resulting antibodies may be harvested from the serum of the animal to provide polyclonal preparations, it is often desirable to isolate individual lymphocytes from the spleen, lymph nodes or peripheral blood to provide homogenous preparations of monoclonal antibodies (MAbs). Preferably, the lymphocytes 5 are obtained from the spleen. In this well known process (Kohler et al., Nature, 256:495 (1975)) the relatively short-lived, or mortal, lymphocytes from a mammal which has been injected with antigen are fused with an immortal tumor cell line (e.g. a myeloma cell line), thus, producing hybrid cells or "hybridomas" which are both immortal and capable of producing the 10 genetically coded antibody of the B cell. The resulting hybrids are segregated into single genetic strains by selection, dilution, and regrowth with each individual strain comprising specific genes for the formation of a single antibody. They produce antibodies which are homogeneous against a desired antigen and, in reference to their pure genetic parentage, are termed "monoclonal." 15 Hybridoma cells thus prepared are seeded and grown in a suitable culture medium that preferably contains one or more substances that inhibit the growth or survival of the unfused, parental myeloma cells. Those skilled in the art will appreciate that reagents, cell lines and media for the formation, selection and growth of hybridomas are commercially available from a number of sources and standardized protocols are well established. 20 Generally, culture medium in which the hybridoma cells are growing is assayed for production of monoclonal antibodies against the desired antigen. Preferably, the binding specificity of the monoclonal antibodies produced by hybridoma cells is determined by immunoprecipitation or by an in vitro assay, such as a radioimmunoassay (RIA) or enzyme-linked immunoabsorbent assay (ELISA). After hybridoma cells are identified that 25 produce antibodies of the desired specificity, affinity and/or activity, the clones may be subcloned by limiting dilution procedures and grown by standard methods (Goding, Monoclonal Antibodies: Principles and Practice, pp 59-103 (Academic Press, 1986)). It will further be appreciated that the monoclonal antibodies secreted by the subclones may be separated from culture medium, ascites fluid or serum by conventional purification 30 procedures such as, for example, protein-A, hydroxylapatite chromatography, gel electrophoresis, dialysis or affinity chromatography. In another embodiment, DNA encoding desired monoclonal antibodies may be readily isolated and sequenced using conventional procedures (e.g., by using 127 WO 2007/109254 PCT/US2007/006883 oligonucleotide probes that are capable of binding specifically to genes encoding the heavy and light chains of murine antibodies). The isolated and subcloned hybridoma cells serve as a preferred source of such DNA. Once isolated, the DNA may be placed into expression vectors, which are then transfected into prokaryotic or eukaryotic host cells such as E. coli 5 cells, simian COS cells, Chinese Hamster Ovary (CHO) cells or myeloma cells that do not otherwise produce immunoglobulins. More particularly, the isolated DNA (which may be synthetic as described herein) may be used to clone constant and variable region sequences for the manufacture antibodies as described in Newman et aL., U.S. Pat. No.. 5,658,570, filed January 25, 1995, which is incorporated by reference herein. Essentially, this entails 10- extraction of RNA from the selected cells, conversion to cDNA, and amplification by PCR using Ig specific primers. Suitable primers for this purpose are also described in U.S. Pat. No. 5,658,570. As will be discussed in more detail below, transformed cells expressing the desired antibody may be grown up in relatively large quantities to provide clinical and commercial supplies of the immunoglobulin. 15 Those skilled in the art will also appreciate that DNA encoding antibodies or antibody fragments (e.g., antigen binding sites) may also be derived from antibody phage libraries, e.g., using pd phage or Fd phagemid technology. Exemplary methods are set forth, for example, in EP 368 684 BI; U.S. patent. 5,969,108, Hoogenboom, H.R. and Chames. 2000. Immune. Today 21:371; Nagy et al. 2002. Nat. Med. 8:801; Huie et al. 20 2001. Proc. Natl. Acad. Sci. USA 98:2682; Lui et al. 2002. J. Mo. Biol. 315:1063, each of which is incorporated herein by reference. Several publications (e.g., Marks et al. BiolTechnology 10:779-783 (1992)) have described the production of high affinity human antibodies by chain shuffling, as well as combinatorial infection and in vivo recombination as a strategy for constructing large phage libraries. In another embodiment, Ribosomal 25 display can be used to replace bacteriophage as the display platform (see, e.g., Hanes et al. 2000. Nat. Biotechnol. 18:1287; Wilson et al. 2001. Proc. Nat. Acad. Sci. USA 98:3750; or Irving et al. 2001 J. Immune. Methods 248:31. In yet another embodiment, cell surface libraries can be screened for antibodies (Boder et al. 2000. Proc. Nat. Acad. Sci. USA 97:10701; Daugherty et al. 2000 J. Immune. Methods 243:211. Such procedures 30 provide alternatives to traditional hybridoma techniques for the isolation and subsequent cloning of monoclonal antibodies. In another embodiment of the present invention a binding site of a binding molecule of the invention may be provided by a human or substantially human antibody. 128 WO 2007/109254 PCT/US2007/006883 Human or substantially human antibodies may be made in transgenic animals (e.g., mice) that are incapable of endogenous immunoglobulin production (see e.g., U.S. Pat. Nos. 6,075,181, 5,939,598, 5,591,669 and 5,589,369 each of which is incorporated herein by reference). For example, it has been described that the homozygous deletion 5 of the antibody heavy-chain joining region in chimeric and germ-line mutant mice results in complete inhibition of endogenous antibody production. Transfer of a human immunoglobulin gene array to such germ line mutant mice will result in the production of human antibodies upon antigen challenge. Another preferred means of generating human antibodies using SCID mice is disclosed in U.S. Pat. No. 5,811,524 which is 10 incorporated herein by reference. It will be appreciated that the genetic material associated with these human antibodies may also be isolated and manipulated as described herein. Yet another highly efficient means for generating recombinant antibodies is disclosed by Newman, Biotechnology, 10: 1455-1460 (1992). Specifically, this 15 technique results in the generation of primatized antibodies that contain monkey variable domains and human constant sequences. This reference is incorporated by reference in its entirety herein. Moreover, this technique is also described in commonly assigned U.S. Pat. Nos. 5,658,570, 5,693,780 and 5,756,096 each of which is incorporated herein by reference. 20 In another embodiment, lymphocytes can be selected by micromanipulation and the variable genes isolated. For example, peripheral blood mononuclear cells can be isolated from an immunized mammal and cultured for about 7 days in vitro. The cultures can be screened for specific IgGs that meet the screening criteria. Cells from positive wells can be isolated. Individual Ig-producing B cells can be isolated by FACS 25 or by identifying them in a complement-mediated hemolytic plaque assay. Ig-producing B cells can be micromanipulated into a tube and the VH and VL genes can be amplified using, e.g., RT-PCR. The VH and VL genes can be cloned into an antibody expression vector and transfected into cells (e.g., eukaryotic or prokaryotic cells) for expression. Moreover, genetic sequences useful for producing the binding molecules of the 30 present invention may be obtained from a number of different sources. For example, as discussed extensively above, a variety of human antibody genes are available in the form of publicly accessible deposits. Many sequences of antibodies and antibody-encoding genes have been published and suitable antibody genes can be chemically synthesized 129 WO 2007/109254 PCT/US2007/006883 from these sequences using art recognized techniques. Oligonucleotide synthesis techniques compatible with this aspect of the invention are well known to the skilled artisan and may be carried out using any of several commercially available automated synthesizers. In addition, DNA sequences encoding several types of heavy and light 5 chains set forth herein can be obtained through the services of commercial DNA synthesis vendors. The genetic material obtained using any of the foregoing methods may then be altered or synthetic to provide obtain polypeptides of the-present invention. Alternatively, antibody-producing cell lines may be selected and cultured using techniques well known to the skilled artisan. Such techniques are described in a variety of 10 laboratory manuals and primary publications. In this respect, techniques suitable for use in the invention as described below are described in Current Protocols in Immunology, Coligan et al., Eds., Green Publishing Associates and Wiley-Interscience, John Wiley and Sons, New York (1991) which is herein incorporated by reference in its entirety, including supplements. 15 It will further be appreciated that the scope of this invention encompasses stabilized binding molecules comprising alleles, variants and mutations of art recognized antigen binding DNA sequences and a stabilized scFv molecule of the invention. As is well known, RNA may be isolated from the original hybridoma cells or from other transformed cells by standard techniques, such as guanidinium isothiocyanate 20 extraction and precipitation followed by centrifugation or chromatography. Where desirable, mRNA may be isolated from total RNA by standard techniques such as chromatography on oligo dT cellulose. Suitable techniques are familiar in the art. In one embodiment, cDNAs that encode the light and the heavy chains of the antibody may be made, either simultaneously or separately, using reverse transcriptase 25 and DNA polymerase in accordance with well known methods. PCR may be initiated by consensus constant region primers or by more specific primers based on the published heavy and light chain DNA and amino acid sequences. As discussed above, PCR also may be used to isolate DNA clones encoding the antibody light and heavy chains. In this case the libraries may be screened by consensus primers or larger 30 homologous probes, such as mouse constant region probes. DNA, typically plasmid DNA, may be isolated from the cells using techniques known in the art, restriction mapped and sequenced in accordance with standard, well known techniques set forth in detail, e.g., in the foregoing references relating to 130 WO 2007/109254 PCT/US2007/006883 recombinant DNA techniques. Of course, the DNA may be synthetic according to the present invention at any point during the isolation process or subsequent analysis. Exemplary antibodies or fragments thereof for use in the binding molecules of the invention include antibodies that recognize the targets set forth herein. 5 In certain embodiments, antigen binding fragments of antibodies can be produced using techniques well known in the art. In one embodiment, a binding molecule of the invention may comprise a complete antibody molecule and a stabilized scFv molecule. In another embodiment, a binding molecule of the invention may comprise a portion of an antibody molecule and a 10 a stabilized scFv molecule. In one embodiment, a binding molecule of the invention comprises a fragment or portion of an antibody and a stabilized scFv molecule. For example, in one embodiment, a binding molecule of the invention may comprise a domain deleted antibody and a stabilized scFv molecule. Domain deleted antibodies are antibody 15 molecules in which one or more domains are partially or entirely deleted. In especially preferred embodiments compatible stabilized binding molecules will comprise domain deleted constructs or variants wherein the entire CH2 domain has been removed (ACH2 constructs). For other preferred embodiments a short connecting peptide may be substituted for the deleted domain to provide flexibility and freedom of movement for 20 the variable region. Those skilled in the art will appreciate that such constructs are particularly preferred due to the regulatory properties of the CH2 domain on the catabolic rate of the antibody. Domain deleted constructs can be derived using a vector (e.g., from IDEC Pharmaceuticals, San Diego) encoding an IgG 1 human constant domain (see, e.g., WO 25 02/060955A2 and W002/096948A2). It will be noted that these exemplary constructs were engineered to fuse the CH3 domain directly to a hinge region of the respective polypeptides of the invention. In other constructs it may be desirable to provide a peptide spacer between the hinge region and the synthetic CH2 and/or CH3 domains. For example, compatible constructs could be expressed wherein the CH2 domain has 30 been deleted and the remaining CH3 domain (synthetic or unsynthetic) is joined to the hinge region with a 5 - 20 amino acid spacer. Such a spacer may be added, for instance, to ensure that the regulatory elements of the constant domain remain free and accessible or that the hinge region remains flexible. For example, a domain deleted B3F6 construct 131 - WO 2007/109254 PCT/US2007/006883 having a short amino acid spacer substituted for the CH2 domain and the lower hinge region (B3F6.ACH2 [gly/ser]) can be used. Other exemplary connecting peptides are known in the art (see, e.g., International PCT Application Nos.WO 2005/000898 and WO 2005/000899). These connecting peptides can be used in connection with the 5 binding molecules of the invention. Preferably, the connecting peptides are used with a polypeptide lacking a CH2 heavy chain domain. Preferably, any connecting peptide compatible with the instant invention will be relatively non-immunogenic and not inhibit the non-covalent association of the polypeptides of the invention. In one embodiment, a binding molecule of the invention comprises an 10 immunoglobulin heavy chain having deletion or substitution of a few or even a single amino acid as long as it permits the desired covalent or non-covalent association between the monomeric subunits. For example, the mutation of a single amino acid in selected areas of the CH2 domain may be enough to substantially reduce Fe binding and thereby increase tumor localization. Similarly, it may be desirable to simply delete that part of one 15 or more constant region domains that control the effector function (e.g. complement binding) to be modulated. Such partial deletions of the constant regions may improve selected characteristics of the antibody (serum half-life) while leaving other desirable functions associated with the subject constant region domain intact. Moreover, as alluded to above, the constant regions of the disclosed antibodies may be synthetic through the 20 mutation or substitution of one or more amino acids that enhances the profile of the resulting construct. In this respect it may be possible to disrupt the activity provided by a conserved binding site (e.g. Fc binding) while substantially maintaining the configuration and immunogenic profile of the stabilized binding molecule. Yet other preferred embodiments may comprise the addition of one or more amino acids to the constant region 25 to enhance desirable characteristics such as effector function or provide for more cytotoxin or carbohydrate attachment. In such embodiments it may be desirable to insert or replicate specific sequences derived from selected constant region domains. It is known in the art that the constant region mediates several effector functions. For example, binding of the CI component of complement to antibodies activates the 30 complement system. Activation of complement is important in the opsonisation and lysis of cell pathogens. The activation of complement also stimulates the inflammatory response and may also be involved in autoimmune hypersensitivity. Further, antibodies bind to cells via the Fc region, with a Fc receptor site on the antibody Fc region binding 132 WO 2007/109254 PCT/US2007/006883 to a Fc receptor (FcR) on a cell. There are a number of Fc receptors which are specific for different classes of antibody, including IgG (gamma receptors), IgE (epsilon receptors), IgA (alpha receptors) and IgM (mu receptors). Binding of antibody to Fc receptors on cell surfaces triggers a number of important and diverse biological 5 responses including engulfment and destruction of antibody-coated particles, clearance of immune complexes, lysis of antibody-coated target cells by killer cells (called antibody-dependent cell-mediated cytotoxicity, or ADCC), release of inflammatory mediators, placental transfer and control of immunoglobulin production. In one embodiment, effector functions may be eliminated or reduced by using a 10 constant region of an IgG4 antibody, which is thought to be unable to deplete target cells, or making Fc variants, wherein residues in the Fc region critical for effector function(s) are mutated using techniques known in the art, for example, U.S. Pat. No. 5,585,097. For example, the deletion or inactivation (through point mutations or other means) of a constant region domain may reduce Fc receptor binding of the circulating 15 stabilized binding molecule thereby increasing tumor localization. In other cases it may be that constant region modifications consistent with the instant invention moderate compliment binding and thus reduce the serum half life and nonspecific association of a conjugated cytotoxin. Yet other modifications of the constant region may be used to modify disulfide linkages or oligosaccharide molecules that allow for enhanced 20 localization due to increased antigen specificity or antibody flexibility. More generally, those skilled in the art will realize that antibodies modified as described herein may exert a number of subtle effects that may or may not be readily appreciated. However the resulting physiological profile, bioavailability and other biochemical effects of the modifications, such as tumor localization, biodistribution and serum half-life, may easily 25 be measured and quantified using well know immunological techniques without undue experimentation. In one embodiment, modified forms of antibodies can be made from a whole precursor or parent antibody using techniques known in the art. Exemplary techniques are discussed in more detail herein. 30 B. Modified Fusion proteins In certain embodiments, a stabilized binding molecule of the invention is a modified fusion protein. In one exemplary embodiment, a binding molecule of the 133 WO 2007/109254 PCT/US2007/006883 invention is a fusion protein comprising a stabilized scFv molecule linked to the ligand binding region of a receptor, an adhesion molecule, a ligand, or an enzyme. In another exemplary embodiment, a binding molecule of the invention is fusion protein comprising a stabilized scFv molecule linked to a receptor binding portion of ligand. 5 For example a binding molecule of the invention is fusion protein comprsing a stabilized scFv molecule to one or more of the following molecules: Cytokines and Cytokine Receptors 10 Cytokines have pleiotropic effects on the proliferation, differentiation, and functional activation of lymphocytes. Various cytokines, or receptor binding portions thereof, can be utilized in the fusion proteins of the invention. Exemplary cytokines include the interleukins (e.g. IL-1, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-11, IL- 12, IL- 13, and IL- 18), the colony stimulating factors (CSFs) (e.g. granulocyte CSF 15 (G-CSF), granulocyte-macrophage CSF (GM-CSF), and monocyte macrophage CSF (M-CSF)), tumor necrosis factor (TNF) alpha and beta, and interferons such as interferon-a, P, or y (US Patent Nos. 4,925,793 and 4,929,554). Cytokine receptors typically consist of a ligand-specific alpha chain and a common beta chain. Exemplary cytokine receptors include those for GM-CSF, IL-3 20 (US Patent No. 5,639,605), IL-4 (US Patent No. 5,599,905), IL-5 (US Patent No. 5,453,491), IFNy (EP0240975), and the TNF family of receptors (e.g., TNFa (e.g. TNFR-I (EP 417, 563), TNFR-2 (EP 417,014) lymphotoxin beta receptor). In another embodiment, an scFv molecule of the invention may bind to a cytokine or a cytokine receptor. 25 Adhesion Proteins Adhesion molecules are membrane-bound proteins that allow cells to interact with one another. Various adhesion proteins, including leukocyte homing receptors and 30 cellular adhesion molecules, of receptor binding portions thereof, can be incorporated in a binding molecule of the invention. Leucocyte homing receptors are expressed on leucocyte cell surfaces during inflammation and include the P-I integrins (e.g. VLA-l, 2, 3, 4, 5, and 6) which mediate binding to extracellular matrix components, and the 02 134 WO 2007/109254 PCT/US2007/006883 integrins (e.g. LFA-1, LPAM-1, CR3, and CR4) which bind cellular adhesion molecules (CAMs) on vascular endothelium. Exemplary CAMs include ICAM-1, ICAM-2, VCAM-1, and MAdCAM-I. Other CAMs include those of the selectin family including E-selectin, L-selectin, and P-selectin. 5 In another embodiment, an scFv molecule of the invention may bind to an adhesion protein or an adhesion protein receptor. Chemokines 10 Chemokines, chemotactic proteins which stimulate the migration of leucocytes towards a site of infection or chemokine receptor binding portions thereof, can also be incorporated into a binding molecule of the invention. Exemplary chemokines include Macrophage inflammatory proteins (MIP-l-a and MIP-1-), neutrophil chemotactic factor, and RANTES (regulated on activation normally T-cell expressed and secreted). 15 In another embodiment, an scFv molecule of the invention may bind to a chemokine or a receptor. Growth Factors and Growth Factor Receptors Growth factors or their receptors (or receptor binding or ligand binding portions 20 thereof) or molecules which bind to them may be incorporated in the binding molecule of the invention. Exemplary growth factors include angiopoietin, Vascular Endothelial Growth Factor (VEGF) and its isoforms (U.S. Pat. No. 5,194,596); Epidermal Growth Factors (EGFs); Fibroblastic Growth Factors (FGF), including aFGF and bFGF; atrial natriuretic factor (ANF); hepatic growth factors (HGFs; US Patent Nos. 5,227,158 and 25 6,099,841), neurotrophic factors such as bone-derived neurotrophic factor (BDNF), neurotrophin-3, -4, -5, or -6 (NT-3, NT-4, NT-5, or NT-6), or a nerve growth factor such as NGF-13 platelet-derived growth factor (PDGF) (U.S. Pat. Nos. 4,889,919, 4,845,075, 5,910,574, and 5,877,016); transforming growth factors (TGF) such as TGF-alpha and TGF-beta (WO 90/14359), osteoinductive factors including bone morphogenetic protein 30 (BMP); insulin-like growth factors-I and -II (IGF-I and IGF-II; US Patent Nos. 6,403,764 and 6,506,874); Erythropoietin (EPO); stem-cell factor (SCF), thrombopoietin (c-Mpl ligand), and the Wnt polypeptides (US Patent No. 6,159,462). Exemplary growth factor receptors which may be used include EGF receptors 135 WO 2007/109254 PCT/US2007/006883 (EGFRs); VEGF receptors (e.g. FltI or FlkI/KDR), PDGF receptors (WO 90/14425); HGF receptors (US Patent Nos. 5,648,273, and 5,686,292); IGF receptors (e.g. IGFR1 and IGFR2) and neurotrophic receptors including the low affinity receptor (LNGFR), also termed as p75NTR or p75, which binds NGF, BDNF, and NT-3, and high affinity 5 receptors that are members of the trk family of the receptor tyrosine kinases (e.g. trkA, trkB (EP 455,460), trkC (EP 522,530)). In another embodiment, both IGFR1 and VEGF are targeted. In yet another embodiment, VLA4 and VEGF are targeted. In another embodiment, both LFAI and VLA4 are targeted. Other cell surface receptors and/or their ligands can also be targeted (e.g., the 10 TNF family receptors or their ligands (as described in more detail herein). In another embodiment, an scFv molecule of the invention may bind to a growth factor or growth factor receptor. 15 Hormones Exemplary growth hormones or molecules which bind to them for use as targeting agents in a binding molecule of the invention include renin, human growth hormone (HGH; US Patent No. 5,834,598), N-methionyl human growth hormone; bovine growth hormone; growth hormone releasing factor; parathyroid hormone (PTH); 20 thyroid stimulating hormone (TSH); thyroxine; proinsulin and insulin (US Patent Nos. 5,157,021 and 6,576,608); follicle stimulating hormone (FSH), calcitonin, luteinizing hormone (LH), leptin, glucagons; bombesin; somatropin; mullerian-inhibiting substance; relaxin and prorelaxin; gonadotropin-associated peptide; prolactin; placental lactogen; OB protein; or mullerian-inhibiting substance. 25 In another embodiment, an scFv molecule of the invention may bind to a hormone or a hormone receptor. Clotting Factors 30 Exemplary blood coagulation factors for use as targeting agents in the binding molecules of the invention include the clotting factors (e.g., factors V, VII, VIII, X, IX, XI, XII and XIII, von Willebrand factor); tissue factor (U.S. Pat. Nos. 5,346,991, 5,349,991, 5,726,147, and 6,596,84); thrombin and prothrombin; fibrin and fibrinogen; 136 WO 2007/109254 PCT/US2007/006883 plasmin and plasminogen; plasminogen activators, such as urokinase or human urine or tissue-type plasminogen activator (t-PA). In another embodiment, an scFv molecule of the invention may bind to a clotting factor. 5 In other embodiments, stabilized binding protein of the invention is a modified immunoadhesin. As is well-known in the art, an immunoadhesin is a fusion protein that combines the target-binding region of a receptor, an adhesion molecule, a ligand, or an enzyme, with the Fc region of an antibody. Exemplary immunoadhesins are described, 10 for example, in US Patent Nos. 5,116,964; 5,428,130; 5,714,147; and 6,406,697, each of which is incorporated by reference herein. In one embodiment, a modified immunoadhesin of the invention is formed by linking a stabilized scFv molecule to an immunoadhesin. Previously reported immunoadhesins may be stabilized as described herein to 15 provide the modified immunoadhesins of the present invention. Exemplary immunoadhesins that may be used to generate or derive the disclosed modified immunoadhesin include, but are not limited to, abatecept (Orencia@, Bristol-Meyers Squibb, Princeton, NJ), alefacept (Amevive@, Biogenldec, Cambridge, MA), etanercept (Enbrel@, Amgen, Thousand Oaks, CA), SMART TM Anti-Gamma Interferon (Protein 20 Design Labs, Fremont, CA), SMART TM Anti-L-Selectin (Protein Design Labs. Fremont, CA), rilonacept (Regeneron Pharmaceuticals Inc., Tarrytown, NY), regavirumab (TI-23, Teijin America, New York, NY), R24 (National Cancer Institute (Bethesda, MD), Oprelvekin (NEUMEGA@ , Genetics Institute, Cambridge, MA), ONCOLYSIN B, ONCOLYSIN CD6, ONCOLYSIN M, and ONCOLYSIN S (all of 25 ImmunoGen Inc., Cambridge, MA, USA), ONCOLYM TM 131 (Techniclone Corp., Tustin, CA), ImmuRAIT-LL2 (Immunomedics Inc., Morris Plains, NJ), IL-4 RA (BAY 16-9996; Bayer Corp., Berkeley, CA), IC 14 (ICOS Corporation, Bothell, WA), CYT 356-Y-90 (ONCORAD@PR, Cytogen Corp., Princeton, NJ), COTARA TM (Techniclone Corp., Tustin, CA), CMB-401 (Wyeth Pharmaceuticals, Madison, NJ), AVICIDIN@ 30 conjuguate (NeoRx Corp., Seattle, WA) and anti-CD 18 immunoadhesin (Genentech Inc., S. San Francisco, CA). Another exemplary molecule that may be included in a binding molecule of the invention is immunoglobulin super family member 9 (IGSF9; Genomics. 2002. 79:663 137 WO 2007/109254 PCT/US2007/006883 70). C. Binding Specificity In one embodiment, a binding molecule of the invention binds to a target 5 molecule that is present on the surface of a cell or that is soluble. In one embodiment, at least one binding specificity of a binding molecule of the invention is catalytic. Catalytic binding specificities can be made using art recognized techniques (see, e.g., U.S. Pat. No. 6,590,080, U.S. Pat. No. 5,658,753). Catalytic binding specificities can work by a number of basic mechanisms similar to those 10 identified for enzymes to stabilize the transition state, thereby reducing the free energy of activation. For example, general acid and base residues can be optimally positioned for participation in catalysis within catalytic active sites; covalent enzyme-substrate intermediates can be formed; catalytic antibodies can also be in proper orientation for reaction and increase the effective concentration of reactants by at least seven orders of 15 magnitude (Fersht, A. R., et al., Am. Chem. Soc. 90 (1968):5833) and thereby greatly reduce the entropy of a chemical reaction. Finally, catalytic antibodies can convert the energy obtained upon substrate binding to distort the reaction towards a structure resembling the transition state. In one embodiment, acid or base residues can be brought into the binding site by 20 using a complementary charged molecule as an immunogen. This technique proved successful for elicitation of antibodies with a hapten containing a positively-charged ammonium ion (Shokat, et al., Chem. Int. Ed. Engl. 27 (1988):269-271). In another approach, antibodies can be elicited to stable compounds that resemble the size, shape, and charge of the transition state of a desired reaction (i.e., 25 transition state analogs). See U.S. Pat. No. 4,792,446 and U.S. Pat. No. 4,963,355 which describe the use of transition state analogues to immunize animals and the production of catalytic antibodies. Both of these patents are hereby incorporated by reference. In one embodiment, such molecules can be administered as part of an immunoconjugate, e.g., with an immunogenic carrier molecule, such as KLH. 30 Exemplary catalytic binding specificities can have, e.g., esterase activity (involving a charged transition state whose electrostatic and shape characteristics can be mimicked by a phosphonate structure; Jacobs, et al., J. Am. Chem. Soc. 109 (1987):2174-2176; Durfor, et al., J. Am. Chem. Soc. 110 (1988):8713-8714; 138 WO 2007/109254 PCT/US2007/006883 Tramontano, et al., J. Am. Chem. Soc. 110 (1988):2282; Pollack, et al., J. Am. Chem. Soc. 11 (1989):5961-5962); peptidase or amidase activity (Janda, et al., Science 241 (1988):1188-1191; Iverson, et al., Science 243 (1989):1184-1188; Paul, et al., Science 244 (1989):1158-1162); Claisen rearrangement (Jackson, et al., J. Am. Chem. Soc. 110 5 (1988):4841-4842; Hilvert, et al., Proc. Natl. Acad. Sci. USA 85 (1988):4953-4955; Hilvert, et al., J. Am. Chem. Soc. 110 (1988):5593-5594); redox reactions (Shokat, et al., Angew. Chem. Int. Ed. Engl. 27 (1989):269-271); photochemical cleavage of a thymine dimer (Cochran, et al., J. Am. Chem. Soc. 110 (1988):7888-7890); stereospecific transesterification rearrangements (Napper, et al., Science 237 (1987):1041-1043); or a 10 bimolecular amide synthesis (Benkovic, et al., Proc. Natl. Acad. Sci. USA 85 (1988):5355-5358; Janda, et al., Science 241 (1988):1188-1191). In another approach, conventional binding specificities can be mutated to render them catalytic. Methods of screening for catalytic antibody activity are well known in the art 15 (e.g., Reymond, J.L. 2002. Journal of Immunological Methods 269:125; Mouratou et al. 2002. J. of Immunological Methods. 269:147. In yet another embodiment, catalytic B cells can be selected, e.g., as described in US patent 6,590,080 using a molecule can be constructed which facilitates selection of catalytic B cells. In another embodiment, catalytic binding specificities can be developed as part 20 of a two step process. Catalytic antibodies can be selected only if displaying the following binding features: binding both the substrate and a reactive group in such a way that the two groups are in a reactive position towards each other. Second, the selected antibodies can be chemically engineered by covalently binding a reactive group into the binding pocket of the antibody. J Immunol Methods. 2002. 269:81-98. 25 In one embodiment, a catalytic binding specificity is specific for a prodrug. Such a binding specificity can be used to catalyze the conversion of a prodrug into a drug which is effective in vivo. Preferably, the reaction catalyzed is one that cannot be accomplished by natural enzymes in vivo. Examples of prodrug activation by antibodies are known in the art (see, e.g., Miyashita et al. 1993. Proc. Natl. Acad. Sci. USA 30 90:5337). In one embodiment, a binding molecule of the invention comprises at least one binding specificity for a target cell and at least one binding specificity for a prodrug. For example, in a preferred embodiment, a stabilized binding molecule molecule of the 139 WO 2007/109254 PCT/US2007/006883 invention comprises at least one binding specificity for a tumor cell and at least one binding specificity for a prodrug which can be converted to cytotoxic drug. In one example, a stabilized binding molecule of the invention comprises a binding specificity for a carbamate prodrug 4-[N,N,-bis(2-chloroethyl)]aminophenyl-N-[(1S-(1,3 5 dicarboxy)propyl]carbamate and generates the corresponding cytotoxic nitrogen mustard (Wentworth et al. 1996. Proc Natl. Acad. Sci. USA. 93:799). -In one embodiment, binding molecule is administered prior to administration of the prodrug to allow accumulation at the site of the target cell. Exemplary prodrugs are known in the art. Prodrugs can also be synthesized by incorporating a portion designed 10 to be released by catalytic action, e.g., by sequential retro-aldol/retro-Michael reactions catalyzed by an antibody with aldolase activity. (Shabat et al. 2001. Proc. Natl. Acad. Sci. USA 98:7428). Such drug masking portions can be made, e.g., by modification of hydroxyl or thiol groups of drugs. In one embodiment, a binding molecule of the invention is multispecific, i.e., has 15 at least one binding site that binds to a first target molecule or epitope of the target molecule and at least one second binding site that binds to a second, different target molecule or to a second, different epitope of the first target molecule. In certain embodiments, multispecific binding molecules of the invention (e.g. bispecific binding molecules) comprise at least one binding site from any of the antibodies described in 20 herein, e.g., Section A, supra. In one embodiment, a binding molecule of the invention is bispecific. Bispecific molecules can bind to two different target sites, e.g., on the same target molecule or on different target molecules. For example, in the case of antibodies, bispecific molecules can bind to two different epitopes, e.g., on the same antigen or on two different antigens. 25 Bispecific molecules can be used, e.g., in diagnostic and therapeutic applications. For example, they can be used to immobilize enzymes for use in immunoassays. They can also be used in diagnosis and treatment of cancer, e.g., by binding both to a tumor associated molecule and a detectable marker (e.g., a chelator which tightly binds a radionuclide. Bispecific molecules can also be used for human therapy, e.g., by 30 directing cytotoxicity to a specific target (for example by binding to a pathogen or tumor cell and to a cytotoxic trigger molecule, such as the T cell receptor or the Fcy receptor. Bispecific antibodies can also be used, e.g., as fibrinolytic agents or vaccine adjuvants. In one embodiment, the multispecific binding molecules of the invention include 140 WO 2007/109254 PCT/US2007/006883 those with at least one arm (ie. binding site) directed against a cell-surface molecule, and at least one arm directed against a soluble molecule. In another embodiment, a multispecific antibody of the invention has two binding sites that bind to soluble molecules. In another embodiment, a multispecific antibody of the invention has two 5 binding sites that bind to cell surface molecules. The multispecific binding molecules of the invention may be monovalent for each specificity or multivalent for each specificity. In one embodiment, a bispecific binding molecule of the invention may comprise one binding site that reacts with a first target molecule and one binding site that reacts with a second target molecule (e.g. a 10 bispecific antibody molecule, fusion protein, or minibody). In another embodiment, a bispecific binding molecule of the invention may comprise two binding sites that react with a first target molecule and two binding sites that react with a second target molecule (e.g. a bispecific scFv2 tetravalent antibody, tetravalent minibody, or diabody). In certain embodiments, at least one binding site of a multispecific binding 15 molecule of the invention is an antigen binding region of an antibody or an antigen binding fragment thereof. In one embodiment, the multispecific binding molecules of the invention are bivalent antibodies or antibody variants with one arm containing at least one stabilized scFv directed to a first target molecule and a second arm containing at least one 20 stabilized scFv directed to a second target molecule. In one embodiment, the multispecific binding molecules of the invention comprise at least one stabilized scFv (e.g. 2, 3, or 4 scFvs) linked to the C-terminus of a heavy chain, wherein the scFvs have the same or different binding specificity. An exemplary binding molecule of this type (a "C-Hercules" antibody) is shown in Figure 25 13. In another embodiment, the multispecific binding molecules of the invention comprise at least one stabilized scFv (e.g. 2, 3, or 4 scFvs) linked to the N-terminus of a heavy chain, wherein the scFvs have the same or different binding specificity. An exemplary binding molecule of this type (a "NH-Hercules" antibody) is shown in Figure 13. In another embodiment, the multispecific binding molecules of the invention 30 comprise at least one stabilized scFv (e.g. 2, 3, or 4 scFvs) linked to the N-terminus of a light chain, wherein the scFvs have the same or different binding specificity. An exemplary binding molecule of this type (a "NL-Hercules" antibody) is shown in Figure 13. In another embodiment, the multispecific binding molecules of the invention 141 WO 2007/109254 PCT/US2007/006883 comprise at least one stabilized scFv (e.g., 2, 3, or 4 scFvs linked to the N-terminus of the heavy chain or light chain and at least one stabilized scFv (e.g., 2, 3, or 4 scFvs) linked to the C-terminus of the heavy chain, wherein the scFvs have the same or different binding specificity. 5 In one embodiment, the multispecific binding molecules of the invention are bivalent minibodies with one arm containing at least one scFv fragment directed to a first target molecule and a second arm containing at least one scFv directed to a second target molecule wherein at least one of the scFv molecules is stabilized. An exemplary bispecific bivalent minibody construct is shown in Figure 42. In Figure 42 a CH3 10 domain is fused at its N-terminus to a connecting peptide which is fused at its N terminus to a VH domain which is fused via its N-terminus to a (Gly4Ser)n flexible linker which is fused at its N-terminus to a VL domain. In another embodiment, the multispecific binding molecules of the invention are scFv tetravalent minibodies, with each heavy chain portion of the scFv tetravalent 15 minibody containing first and second scFv fragments wherein at elsat one of the scFv molecules is stabilized. Said second scFv fragment may be linked to the N-terminus of the first scFv fragment (e.g. bispecific NH scFv tetravalent minibodies or bispecific NL scFv tetravalent minibodies). An example of a bispecific N-scFv tetravalent minibody is shown in Figure 43. Alternatively, the second scFv fragment may be linked to the C 20 terminus of said heavy chain portion containing said first scFv fragment (e.g. bispecific C-scFv tetravalent minibodies). An example of a bispecific C-scFv tetravalent minibody is shown in Figure 44. In one embodiment, the first and second scFv fragments of may bind the same or different target molecule. Where the first and second scFv fragments of a first heavy chain portion of a bispecific tetravalent minibody bind the same target 25 molecule, at least one of the first and second scFv fragments of the second heavy chain portion of the bispecific tetravalent minibody binds a different target molecule. In another embodiment, the multispecific binding molecules of the invention are bispecific diabodies, with each arm of the diabody comprising tandem scFv fragments at least one of which is stabilized. In one embodiment, a bispecific diabody may comprise 30 a first arm with a first binding specificity and a second arm with a second binding specificity (see, for example, Figure 45). In another embodiment, each arm of the diabody may comprise a first scFv fragment with a first binding specificity and a second scFv fragment with a second binding specificity. 142 WO 2007/109254 PCT/US2007/006883 In another embodiment, the multispecific binding molecules of the invention are scFv2 tetravalent antibodies with each heavy chain portion of the scFv2 tetravalent antibody containingan scFv molecule, wherein at least one of the scFv molecules are stabilized. The scFv fragments may be linked to the N-termini of a variable region of the 5 heavy chain portions (e.g. bispecific NH scFv2 tetravalent antibodies or bispecific N 1 scFv2 tetravalent antibodies). Alternatively, the scFv fragments may be linked to the C termini of the heavy chain portions of the scFv2 tetravalent antibody (e.g. bispecific C scFv2 tetravalent antibodies, see for example Figure 46). Each heavy chain portion of the scFv2 tetravalent antibody may have variable regions and scFv fragments that bind 10 the same or different target molecules. Where the scFv fragment and variable region of a first heavy chain portion of a bispecific scFc2 tetravalent antibody bind the same target molecule, at least one of the first and second scFv fragments of the second heavy chain portion of the bispecific tetravalent minibody binds a different target molecule. In another embodiment, the multispecific binding molecules of the invention are 15 scFv2 tetravalent domain-deleted antibodies with each heavy chain portion of the scFv2 tetravalent antibody containingan scFv fragment at least one of which is stabilized. The scFv fragments may be linked to the N-termini of a variable region of the heavy chain portions (e.g. bispecific NH scFv2 tetravalent domain-deleted antibodies (see Figure 48) or bispecific NL scFv2 tetravalent antibodies (see Figure 49). Alternatively, the scFv 20 fragments may be linked to the C-termini of the heavy chain portions of the scFv2 tetravalent domain-deleted antibody (e.g. bispecific C-scFv2 tetravalent domain deleted antibodies, see for example Figure 47). Exemplary cell-surface molecules to which a binding molecule of the invention may bind include receptors or tumor cell antigens that are overexpressed on the surface 25 of a tumor or neoplastic cell, as well as any of the cytokine receptors, adhesion molecules, or growth factor receptors described in herein, e.g. section B supra. Exemplary soluble molecules include anti-tumor agents (e.g. toxins, chemotherapeutics, and prodrugs thereof), soluble enzymes (e.g. prodrug converting enzymes), cytokines, chemokines, hormones, growth factors, or clotting factors, e.g., as described herein, e.g. 30 in section A supra. Bispecific molecules which bind to both tumor cell antigens and anti-tumor agents or soluble enzymes can therefore localize the anti-cancer agent to a tumor cell expressing said tumor cell antigen, thereby maximizing the toxic effects of the anti 143 WO 2007/109254 PCT/US2007/006883 cancer agent on a tumor cell and minimizing a toxic effect of the anti-cancer agent on normal cells. Exemplary bispecific binding molecules with at least one binding site for a tumor antigen and at least one binding site for a toxin include anti-saporin/anti-[d-1, anti 5 CD22/anti-saporin, anti-CD7/anti-saporin, anti-CD38/anti-saporin, anti-CEA/anti-ricin A chain, anti-interferon-.alpha.(IFN-. alpha.)/anti-hybridoma idiotype, anti-CEA/anti vinca alkaloid). Exemplary bispecific binding molecules with at least one binding site for a cell-surface molecule and at least one binding site for a prodrug converting enzyme include for example, anti-CD3O/anti-alkaline phosphatase (which catalyzes conversion 10 of mitomycin phosphate prodrug to the chemotherapeutic mitomycin alcohol). In other embodiments, the bispecific binding molecules bind to both tumor cell antigens and diagnostic agents, thereby localizing said diagnostic agent to a tumor cell expressing said tumor cell antigen and facilitating tumor detection in vitro or in vivo. Exemplary bispecific binding molecules include anti-CEA/anti-EOTUBE, anti 15 CEA/anti-DPTA, anti-CEA/anti-.beta.-galactosidase, and anti-p185HER2/anti- -hapten). In other embodiments, bispecific binding molecules of the invention bind to both soluble molecules (e.g. soluble antigens) and cell surface molecules on non-tumor cells (e.g. immune cells). For example, can be used to target soluble immune complexes to cell surface receptors on immune cells, thereby facilitating their clearance from the body 20 by cell-mediated immune mechanisms. Exemplary bispecific molecules of this type include anti-low density lipoprotein (LDL)/anti-Fc receptor (e.g. Fc.gamma.RI, Fc.gamma.RII or Fc.gamma.RIII)) and bispecific binding molecules for use in therapy of infectious diseases (such as anti-CD3/anti-herpes simplex virus (HSV), anti-T-cell receptor:CD3 complex/anti-influenza, anti-Fc.gamma.R/anti-HIV. 25 In other embodiments, bispecific binding molecules of the invention are capable of binding to both cell surface receptors and soluble ligands thereof. In one embodiment, the ligand is the cognate ligand of a TNF family receptor. Exemplary cell surface receptors to which the bispecific binding molecules can bind are tumor cell antigens or immune cell receptors. Exemplary cell surface receptors 30 also include cytokine receptors, adhesion molecules, or growth factor receptors, e.g., as described in herein, e.g. in section B supra. Exemplary soluble ligands include cytokines, chemokines, hormones, growth factors, or clotting factors, e.g., as described in section B supra. 144 WO 2007/109254 PCT/US2007/006883 Exemplary bispecific binding molecules include anti-VLA4/anti-Mac-1, anti VLA4/anti-VEGF, anti-VLA4/anti-angiopoietin, anti-VLA4/anti-TNF., anti IGFR1/anti-VEGF, anti-IGFRI/anti-angiopoietin, anti-IGFRI/anti-EGFR, anti-HGF SF/anti-VEGF, anti-HGF-SF/anti-angiopoietin, and HGF-SF/any second antigen (See, 5 e.g., Cao et al. Proc. Natl. Acad. Sci 2001. 98:7443; Lu et al. 2004. J. Biol. Chem. 279:2856). In other embodiments, the bispecific binding molecules of the invention include those with at least one arm (ie. binding site) directed against a first soluble molecule (e.g. soluble ligand), and at least one arm directed against a second soluble molecule 10 (e.g. soluble ligand). Such bispecific binding molecules can be employed as diagnostic tools (e.g. anti-rabbit IgG/anti-ferritin, anti-horse radish peroxidase (HRP)/anti-hormone, anti-somatostatin/anti-substance P, anti-HRP/anti-FITC (see Nolan et al., supra)) or fibrinolytic agents (e.g. anti-fibrin/anti-tissue plasminogen activator (tPA), anti fibrin/anti-urokinase-type plasminogen activator (uPA)). 15 In a preferred embodiment, the soluble molecule to which a bispecific binding molecule of the invention binds is a soluble ligand of the TNF family. Examples of TNF family ligands include, but are not limited to, LTA (which binds TNFRI/TNFRSF1A), TNF (which binds CD120b/TNFRSFIB), LTB (which binds LTBR/TNFRSF3), OX40L (which binds OX40/TNFRSF4), CD40L (which binds CD40/TNFRSF5), (which binds 20 Fas/TNFRSF6 and DcR3/TNFRSF6B), CD27L (which binds CD27/TNFRSF7), CD30L (which binds CD30/TNFRSF8), 4-1-BB-L (which binds 4-1-BB/TNFRSF9), TRAIL (which binds TRAIL-R 1 /TNFRSF 10 A, TRAIL-R2/TNFRSF lOB, TRAIL R3/TNFRSF I 0C, and TRAIL-R4/TNFRSF I OD), RANKL (which binds RANK/TNFRSF1 IA and Osteoprotegrin/TNFRSF1 IB), APO-3L (which binds APO 25 3/TNFRSF12 and DR3L/TNFRSF12L), APRIL (which binds TACI/TNFRSF13B), BAFF (which binds BAFFR/TNFRSF13A), LIGHT (which binds HVEM/TNFRSF14), NGF ligands (which bind LNGFR, e.g. NGF-p, NGF-2/NTF3, NTF5, BDNF, IFRD 1), GITRL (which binds GITR/TNFRSF18), EDARI & XEDAR ligand, Fn14 ligand, andTroy/Trade ligand. 30 In other exemplary embodiments, the bispecific binding molecules of the invention have at least one binding site for a first cell-surface molecule and at least one binding site for a second cell-surface molecule. In one embodiment, the first and second cell-surface molecules are located on different cells (e.g. different cell types). For 145 WO 2007/109254 PCT/US2007/006883 example, bispecific molecules may have at least one arm directed against a tumor cell antigen and at least one arm directed against cell-surface receptor on a non-tumor cell (e.g. an immune cell). Exemplary bispecific binding molecules of this type include those having at least one binding site for a tumor cell antigen and at least one binding 5 site directed against a cytotoxic trigger molecule of an immune effector cell (such as anti-Fc.gamma.RI/anti-CD15, anti-p185.sup.HER2/Fc.gamma.RIII (CD16), anti pI85.sup.HER2/anti-VEGF, anti-CD3/anti-malignant B-cell (ID 10), anti-CD3/anti p185.sup.HER2, anti-CD3/anti-p97, anti-CD3/anti-renal cell carcinoma, anti-CD3/anti OVCAR-3, anti-CD3/L-D1 (anti-colon carcinoma), anti-CD3/anti-melanocyte 10 stimulating hormone analog, anti-EGF receptor/anti-CD3, anti-CD3/anti-CAMAl, anti CD3/anti-CD 19, anti-CD3/MoV18, anti-neural cell adhesion molecule (NCAM)/anti CD3, anti-folate binding protein (FBP)/anti-CD3, and anti-pan carcinoma associated antigen (AMOC-3 1)/anti-CD3)). Bispecific molecules which bind to both tumor cell antigens and cytotoxic trigger molecule are capable of effectively juxtaposing a tumor 15 cell with an immune effector cell, thereby activating the effector cell to destroy the tumor cell by cell-mediated immune mechanisms. In another embodiment, the first and second cell-surface molecules to which a bispecific binding molecule is capable of binding are located on the same cell or cell type. By crosslinking the first and second receptors on the same cell, the bispecific 20 binding molecules of the invention may inhibit or enhance an activity (e.g. signal transduction activity) associated with one or both of the first and second receptors. In one embodiment the first and second cell surface molecules are of the same type (e.g., are in the same family of molecules). In another embodiment said first and second cell surface molecules are distinct types (e.g., are in different families of molecules). 25 Exemplary cell surface receptors to which the bispecific binding molecules bind include tumor cell antigens or immune cell receptors. Exemplary cell surface receptors include any of the cytokine receptors, adhesion molecules, or growth factor receptors described in section B supra. In one embodiment, exemplary target molecules to which a binding molecule of 30 the invention binds include one or more epitopes of e.g., heparin sulfate, growth factors or their receptors (.e.g, epidermal growth factor receptor, insulin-like growth factor receptor, hepatocyte growth factor (HGF/SF) receptor (See, e.g., Cao et al. Proc. Natl. 146 WO 2007/109254 PCT/US2007/006883 Acad. Sci. 2001. 98:7443; Lu et al. 2004. J. Biol. Chem. 279:2856). In an exemplary embodiment, at least one of the molecules to which a bispecific binding molecule of the invention binds is a member of the TNF receptor (TNFR) family. In another exemplary embodiment, the first and second target molecules to 5 which a bispecific binding molecule of the invention binds are both TNFR family members. In another embodiment, a binding molecule of the invention binds to a TNFR family ligand.- In yet another embodiment, a binding molecule of the invention binds to one TNFR family member and an antigen expressed on the surface of a tumor cell, e.g., preferentially expressed on a tumor cell. The limiting factor in the treatment of tumors 10 with monospecific TNFR binding molecules is that often only a subset of tumors appears to be sensitive to such therapies. Multispecific TNFR binding molecules can specifically activate TNFRs, and enhance receptor signaling by, for example, bringing the TNFRs into close proximity. The invention provides improved bispecific TNFR binding molecules which can target more than one TNFR or TNFR type and enhance 15 signaling, thus providing an improved method of treating cancer. In one embodiment, the bispecific TNFR binding molecule increases the signal strength by binding to two or more TNFRs of the same type increasing the number of TNFRs being brought together. In another more preferred embodiment, the bispecific TNFR binding molecule is capable of binding to two different receptors of the TNF family. 20 In one embodiment, at least one of the TNFRs to which a bispecific TNFR binding molecule binds contains a death domain. The term "death domain" refers to a cytoplasmic region of a TNF family receptor which is involved TNF-mediated cell death or apoptotic signaling and cell-cytotoxicity induction mediated by these receptors. This region couples the receptor to caspase activation via adaptor proteins resulting in 25 activation of the extrinsic death pathway. Examples of TNF receptors which contain death domains include, but are not limited to, TNFR1 (TNFRSF1A), Fas (TNFRSF6), DR-3 (TNFRSF6B), LNGFR (TNFRSF16) TRAIL-RI (TNFRSFIOA), TRAIL-R2 (TNFRSF1OB) and DR6 (TNFRSF2 1). The apoptotic signaling of these receptors is modulated upon binding of a 30 cognate ligand and formation of any of the following receptor-ligand pairs: TNFRI/TNFa, Fas/FasL, DR-3/DR-3LG, TRAIL-RI/TRAIL, or TRAIL-R2/TRAIL. Bispecific TNFR binding molecules that target TNF family receptors containing 147 WO 2007/109254 PCT/US2007/006883 death domains are useful for the treatment of cancer since the TNFRs of this type are often overexpressed on tumor cells and stimulating of the receptor can activate tumor cell apoptosis. In preferred embodiments, the death-domain containing TNFR to which the bispecific TNFR binding molecule of the invention binds is TRAIL-R2. TRAIL-R2 5 is preferred for human tumor therapy since its activation does not trigger hepatocyte apoptosis and hence should have reduced toxicity. While the activation of some of death domain containing receptors, e.g. TNFRI or Fas, has been toxic in in vivo applications, it is likely that tethering these receptors to other TNF receptors may diminish toxicity and thus render a toxic antibody less toxic. 10 In one embodiment, a bispecific TNFR binding molecule of the invention comprises at least one binding site directed to a TNFR containing a death domain and at least one binding site directed to a TNFR lacking a death domain. In certain exemplary embodiments, TNFRs lacking a death domain include TNFRs involved in tissue differentiation. Examples of TNFR receptors involved in 15 tissue differentiation include LTpR, RANK, EDAR1, XEDAR, Fn14, Troy/Trade, and NGFR. TNFRs involved in tissue differentiation may influence tissue differentiation following binding of a cognate ligand. TNFR binding molecules that target TNFRs involved in tissue differentiation can affect tumors in several ways. First, they have the potential to directly slow tumor growth by altering cell cycle progression. Second, 20 tissue differentiation in the context of tumor cell transformation may lead to cell cycle conflict and default apoptosis. Third, such conflicting input may render a cell more sensitive to chemotherapy. In certain preferred embodiments, TNFR involved in tissue differentiation is lymphotoxin P receptor (LTpR). LTpR is involved in the control of the maturation. 25 status of various specialized stromal cells in the immune system and plays a critical role during the development of the stromal elements of the lymph node anlagen. It has been proposed that activation of a developmental program in epithelial or fibroblastoid cells in the context of a transformed cell is detrimental to their survival and this action may account for some of the anti-tumor activity of LTPR activation. These receptors can 30 also initiate inflammatory programs that involve chemokine release or promote immunological anti-tumor responses. Such release could affect the inflammatory status 148 WO 2007/109254 PCT/US2007/006883 of the tumor and/or invoke infiltration of lymphoid elements promoting an immunological reaction to the tumor. Thus, bispecific TNFR binding molecules which bind LTpR, alone or in combination with TNF receptors containing death domains (e.g. TRAIL-R2), are encompassed by the invention. 5 In certain exemplary embodiments, the TNFRs lacking a death domain include TNFRs involved in immune regulation. Such receptors include TNFR2, HVEM, CD27, CD30, CD40, 4-1BB, OX40, GITR, TACI, BAFF-R, BCMA, and RELT. Additional TNF family receptors involved in immune regulation include TRAIL-R3 and TRAIL R4. 10 Other target TNF family receptors with a role in tumor formation can be identified using existing RNA databases of receptor expression in various cell types which allow one to define TNF family receptors that are present or ideally overexpressed on various tumors. Moreover, existing RNA databases provide an additional advantage in that the pair of TNF family receptors to which a bispecific 15 TNFR binding molecule of the invention binds could be optimized by identifying those receptor pairs that are more uniquely expressed on a tumor type or subset of tumors but are not abundant on normal tissues, especially liver and vasculature. In such a manner receptor pairs (or more) are identified that could deliver a potent signal to the tumor and spare normal tissues. 20 Methods of producing multispecific molecules are well known in the art. For example, recombinant technology can be used to produce multispecific molecules, e.g., diabodies, single-chain diabodies, tandem scFvs, etc. Exemplary techniques for producing multispecific molecules are known in the art (e.g., Kontermann et al. Methods in Molecular Biology Vol. 248: Antibody Engineering: Methods and 25 Protocols. Pp 227-242 US 2003/0207346 Al and the references cited therein). In one embodiment, a multimeric multispecific molecules are prepared using methods such as those described e.g., in US 2003/0207346 Al or US patent 5,821,333, or US2004/0058400. In another embodiment, a multispecific binding molecule of the invention is a 30 multispecifc fusion protein. As used herein the phrase "multispecific fusion protein" designates fusion proteins (as hereinabove defined) having at least two binding specificities (i.e. combining a binding domains of a ligand or receptor). Multispecific fusion proteins can be assembled, e.g., as heterodimers, heterotrimers or heterotetramers, 149 WO 2007/109254 PCT/US2007/006883 essentially as disclosed in WO 89/02922 (published Apr. 6, 1989), in EP 314, 317 (published May 3, 1989), and in U.S. Pat. No. 5,116,964 issued May 2, 1992. Preferred multispecific fusion proteins are bispecific. Examples of bispecific fusion proteins include CD4-scFv/TNFreceptor-IgG and CD4-scFv/L-selectin-IgG. The last mentioned 5 molecule combines the lymph node binding function of the lymphocyte homing receptor (LHR, L-selectin), and the HIV binding function of CD4, and finds potential application in the prevention or treatment of HIV infection, related conditions, or as a diagnostic. In another embodiment, the invention pertains to multispecific stabilized binding molecules, e.g., bispecific binding molecules, e.g., antibodies, which incorporate at least 10 one binding site that binds to a known target and at least one binding site which recognizes an unknown target (for example, in one embodiment, the bispecific molecule incorporates binding sites selected from a semi-synthetic antibody phage display library) and a stabilized scFv of the invention. In one embodiment of the invention, one of ordinary skill in the art could start 15 with a single chain antibody of known specificity and build a Fab library using techniques known in the art or, alternatively, the skilled artisan could start with an Fab fragment of known specificity and build a stabilized single chain library using techniques known in the art. It is known in the art that libraries from nonimmunized sources and prepared by synthetic recombination of V-gene sequences (preferably 20 recombination of VH with, DH and JH, and VL with JL sequences) can be used to isolate antibodies to any antigen. For example, patent application W092/01047 teaches that antibody fragments can be displayed on the surface of bacteriophage and that they will bind antigen. Antibody fragments (e.g., Fab, Fv, scFv and VH) can be directly selected using this characteristic. Other methods known in the art include those taught, 25 e.g., in U.S. patent 5,698,426; 6,291,159; 5,658,727; 5,667,988; and 5,969,108. In another embodiment, scFv which recognize a known target can be dimerized with scFv isolated from a semi-synthetic human phage antibody display library. (see, e.g., Kruif and Logtenberg 1996. J. Biol. Chem. 271:7630). In one embodiment, the subject multispecific molecule is expressed in an 30 expression system used to express antibody molecules, for example mammalian cells, yeast such as Picchia, E. coli, Bacculovirus, etc. In one embodiment, the subject bispecific molecule is expressed in the NEOSPLA vector system (see, e.g., U.S. patent 6,159,730). This vector contains the cytomegalovirus promoter/enhancer, the mouse 150 WO 2007/109254 PCT/US2007/006883 beta globin major promoter, the SV40 origin of replication, the bovine growth hormone polyadenylation sequence, neomycin phosphotransferase exon I and exon 2, the dihydrofolate reductase gene and leader sequence. In one embodiment, the subject multispecific molecules comprise a synthetic 5 connecting peptide. These multispecific molecules have one or more binding sites for a known target and express a library at one or more binding sites. Such multispecific molecules can be used, e.g., to identify molecules in close proximity to or associated with the known target. For example, the skilled artisan could use the subject multispecific molecules in 10 an assay to select for those that induce a particular response, e.g., apoptosis or cellular activation, using screening methods well known in the art. The bispecific molecule identified as producing the response screened for can then be identified and its specificity determined. Using such methods it is possible to identify molecules in close association with particular targets of interest, e.g., T cell markers or other signaling 15 molecules (such as CRIPTO-I, death domain molecules, or molecules involved in apoptosis). The proximity of the known target and the molecule newly identified as a "nearest neighbor" can be confirmed using immunoprecipitation or other techniques known to those of skill in the art. Using these methods it is possible to identify molecules as targets for modulating a particular cellular response. 20 Binding specificities comprising antigen recognition sites or entire variable regions of multispecific binding molecule, in particular multispecific antibodies or antibody variants of the invention may be derived from one or more parental antibodies. The parental antibodies can include naturally occurring antibodies or antibody fragments, antibodies or antibody fragments adapted from naturally occurring 25 antibodies, antibodies constructed de novo using sequences of antibodies or antibody fragments known to be specific a target molecule. Sequences that may be derived from parental antibodies include heavy and/or light chain variable regions and/or CDRs, framework regions or other portions thereof. In one exemplary embodiment of the invention, the parental antibodies used to 30 construct a multispecific TNFR binding molecule are an anti-TRAIL-R2 antibody, for example 14A2, and an anti- LTpR antibody, for example CBE1 1 or BHA10. Multivalent, multispecific antibodies may contain a heavy chain comprising two or more variable regions and/or a light chain comprising one or more variable regions wherein at 151 WO 2007/109254 PCT/US2007/006883 least two of the variable regions recognize different epitopes of LTpR. Multispecific, e.g., bispecific TNFR binding molecules may be constructed in a variety different ways using a variety of different sequences derived from parental anti LTPR antibodies, including murine or humanized BHA10 (Browning et al., J. Immunol. 5 154: 33 (1995); Browning et al. J. Exp. Med. 183:867 (1996)), murine or humanized CBE I1 (U.S. Patent 6,312,691 and WO 02/30986, respectively), and/or parental anti TRAIL-R2 murine or chimeric 14A2. Examples of anti-LTOR antibodies which can be used for the bispecific TNFR binding molecules of the invention include consisting: BKAI 1, CDH10, BCG6, AGH1, BDA8, CBE I1 and BHA10 or BHA10. The following 10 hybridoma cell lines producing monoclonal anti- LT-p-R antibodies may be used to produce anti-LT R antibodies from which to derive antibody construct sequences, which have been previously deposited with the American Type Culture Collection (ATCC) according to the provisions of the Budapest Treaty and have been assigned the indicated ATCC accession numbers: 15 Cell Line mAb Name Accession No. a) AG.Hl.5.1 AGHI HB 11796 b) BD.A8.AB9 BDA8 HB 11798 c) BC.G6.AF5 BCG6 B 11794 20 d) BH.A10 BHAIO B 11795 e) BK.All.ACI0 BKAI I B 11799 f) CB.E11.1 CBE1I B 11793 g) CD.H 10.1 CDHIO B 11797 25 Other examples of anti-TNF receptor antibodies which can be used in the multispecific TNFR binding molecules of the invention include antibodies directed to TNF receptors containing a death domain. A number of antibodies have been generated to death domain containing TNF receptors and are well known in the art. Such antibodies include anti-TNF-R1 monoclonal antibodies (R&D systems anti-TNF-RI; 30 Tularik mAb #985, US Patent Nos. 6,110,690; 6,437,113), anti-Fas receptor mAb CH 11 (US Patent No. 6,312,691; WO 95/10540), anti-DR3 antibodies (US Patent No. 5,985,547; Johnson, et al. (1984) ImmunoBiology of HLA, ed. Dupont, B.O., Springer, 152 WO 2007/109254 PCT/US2007/006883 New York; US Patent Nos. 6,462,176; 6,469,166), and anti-TRAIL-R antibodies (US Patent Nos. 5,763,223; 6,072,047; 6,284,236; 6,521,228; 6,569,642; 6,642,358; and US Patent No 6,417,328). A number of antibodies have been also raised to TNF receptors involved in tissue 5 differentiation and are known in the art. Examples of anti-TNF receptor antibodies specific to TNF receptors involved in tissue differentiation include: anti-RANK monoclonal antibodies (Immunex - US Patent Nos. 6-,562,948; 6,537,763; 6,528,482; 6,479,635; 6,271,349; 6,017,729; Komed - WO 03/080671), anti-EDAR polyclonal (anti-human) and monoclonal (anti-mouse) antibodies (R&D Systems - MAB745, 10 BAF157; Elomaa et al. (2001) Human Molecular Genetics. 10:953), anti-XEDAR monoclonal and polyclonal antibodies (R&D Systems - MAB1093 and AF1093), anti Fn14 monoclonal antibodies (Nakayama et al. (2003) J. Immunology 170:341; ITEM-1, ITEM-2, and ITEM-4 clones available from eBioscience), anti-TROY antibody (T3323 from Sigma-Aldrich), and anti-NGFR (anti-rodent) antibodies (Chemicon USA). 15 A number of antibodies have been also raised to TNF receptors involved in immune regulation and are known in the art. Examples of anti-TNF receptor antibodies specific to TNF receptors involved in immune regulation include: anti-HVEM antibodies (HGSI - WO 03/086301), anti-CD40 antibodies (Biogen - WO 97/20063; Chiron - US Patent Nos. 5,677,165; 5,874,082; 6,004,552; 6,056,959; 6,315,998; US 20 Application Publication No. 2002/0106371; US Application Publication Nos. 2003/0059427; US20030118588A; 2003/0211 100A; US2002020142358A1; US Patent Nos. US6312693; US6051228; Fanslow et al. - US5801227), anti-4-1BB (PCT Publication No. WO 03/084999; EP 0948353; US Patent No. 6210669; Genecraft - WO 03/083069), and anti-BAFF-R antibodies (rabbit polyclonal - ProSci catalog #3097), 25 among many other antibodies raised to immune regulation receptors. A variety of other multivalent antibody constructs may be developed by one of skill in the art using routine recombinant DNA techniques, for example as described in PCT International Application No. PCT/US86/02269; European Patent Application No. 184,187; European Patent Application No. 171,496; European Patent Application No. 30 173,494; PCT International Publication No. WO 86/01533; U.S. Pat. No. 4,816,567; European Patent Application No. 125,023; Better et al. (1988) Science 240:1041-1043; Liu et al. (1987) Proc. Natl. Acad. Sci. USA 84:3439-3443; Liu et al. (1987) J Immunol. 139:3521-3526; Sun et al. (1987) Proc. Natl. Acad. Sci. USA 84:214-218; Nishimura et 153 WO 2007/109254 PCT/US2007/006883 al. (1987) Cancer Res. 47:999-1005; Wood et al. (1985) Nature 314:446-449; Shaw et al. (1988) J. Natl. Cancer Inst. 80:1553-1559); Morrison (1985) Science 229:1202-1207; Oi et al. (1986) BioTechniques 4:214; U.S. Pat. No. 5,225,539; Jones et al. (1986) Nature 321:552-525; Verhoeyan et al. (1988) Science 239:1534; Beidler et al. (1988) J. 5 Immunol. 141:4053-4060; and Winter and Milstein, Nature, 349, pp. 293-99 (1991)). Preferably non-human antibodies are "humanized" by linking the non-human antigen binding domain with a human constant domain (e.g. Cabilly et al., U.S. Pat. No. 4,816,567; Morrison et al., Proc. Natl. Acad. Sci. U.S.A., 81, pp. 6851-55 (1984)). Other methods which may be used to prepare multivalent antibody constructs are 10 described in the following publications: Ghetie, Maria-Ana et al. (2001) Blood 97:1392 1398; Wolff, Edith A. et al. (1993) Cancer Research 53:2560-2565; Ghetie, Maria-Ana et al. (1997) Proc. Natl. Acad. Sci. 94:7509-7514; Kim, JC. et al. (2002) Int. J. Cancer 97(4):542-547; Todorovska, Aneta et al. (2001) Journal ofImmunological Methods 248:47-66; Coloma M.J. et al. (1997) Nature Biotechnology 15:159-163; Zuo, Zhuang et 15 al. (2000) Protein Engineering (Suppl.) 13(5):361-367; Santos A.D., et al. (1999) Clinical Cancer Research 5:3118s-3123s; Presta, Leonard G. (2002) Current Pharmaceutical Biotechnology 3:237-256; van Spriel, Annemiek et al., (2000) Review Immunology Today 21(8) 391-397. 20 XI. Expression of Binding Molecules Following manipulation of the isolated genetic material to provide polypeptides of the invention as set forth above, the genes are typically inserted in an expression vector for introduction into host cells that may be used to produce the desired quantity of polypeptide that, in turn, provides the claimed binding molecules. 25 The term "vector" or "expression vector" is used herein for the purposes of the specification and claims, to mean vectors used in accordance with the present invention as a vehicle for introducing into and expressing a desired gene in a cell. As known to those skilled in the art, such vectors may easily be selected from the group consisting of plasmids, phages, viruses and retroviruses. In general, vectors compatible with the instant 30 invention will comprise a selection marker, appropriate restriction sites to facilitate cloning of the desired gene and the ability to enter and/or replicate in eukaryotic or prokaryotic cells. For the purposes of this invention, numerous expression vector systems may be 154 WO 2007/109254 PCT/US2007/006883 employed. For example, one class of vector utilizes DNA elements which are derived from animal viruses such as bovine papilloma virus, polyoma virus, adenovirus, vaccinia virus, baculovirus, retroviruses (RSV, MMTV or MOMLV) or SV40 virus. Others involve the use of polycistronic systems with internal ribosome binding sites. 5 Additionally, cells which have integrated the DNA into their chromosomes may be selected by introducing one or more markers which allow selection of transfected host cells. The marker may provide for prototrophy to an auxotrophic host, biocide resistance (e.g., antibiotics) or resistance to heavy metals such as copper. The selectable marker gene can either be directly linked to the DNA sequences to be expressed, or 10 introduced into the same cell by cotransformation. Additional elements may also be needed for optimal synthesis of mRNA. These elements may include signal sequences, splice signals, as well as transcriptional promoters, enhancers, and termination signals. In particularly preferred embodiments the cloned variable region genes are inserted into an expression vector along with the heavy and light chain constant region genes 15 (preferably human) synthetic as discussed above. Preferably, this is effected using a proprietary expression vector of IDEC, Inc., referred to as NEOSPLA (U.S. patent 6,159,730). This vector contains the cytomegalovirus promoter/enhancer, the mouse beta globin major promoter, the SV40 origin of replication, the bovine growth hormone polyadenylation sequence, neomycin phosphotransferase exon 1 and exon 2, the 20 dihydrofolate reductase gene and leader sequence. As seen in the examples below, this vector has been found to result in very high level expression of antibodies upon incorporation of variable and constant region genes, transfection in CHO cells, followed by selection in G418 containing medium and methotrexate amplification. Vector systems are also taught in U.S. Pat. Nos. 5,736,137 and 5,658,570, each of which is 25 incorporated by reference in its entirety herein. This system provides for high expression levels, e.g., > 30 pg/cell/day. Other exemplary vector systems are disclosed e.g., in U.S. patent 6,413,777. In other preferred embodiments the polypeptides of the invention of the instant invention may be expressed using polycistronic constructs such as those disclosed in 30 copending United States provisional application No. 60/331,481 filed November 16, 2001 and incorporated herein in its entirety. In these novel expression systems, multiple gene products of interest such as heavy and light chains of antibodies may be produced from a single polycistronic construct. These systems advantageously use an internal 155 WO 2007/109254 PCT/US2007/006883 ribosome entry site (IRES) to provide relatively high levels of polypeptides of the invention in eukaryotic host cells. Compatible IRES sequences are disclosed in U.S. Pat. No. 6,193,980 which is also incorporated herein. Those skilled in the art will appreciate that such expression systems may be used to effectively produce the full 5 range of polypeptides disclosed in the instant application. More generally, once the vector or DNA sequence encoding a monomeric subunit of the binding molecule (e.g. a modified antibody) has been prepared, the expression vector may be introduced into an appropriate host cell. That is, the host cells may be transformed. Introduction of the plasmid into the host cell can be accomplished 10 by various techniques well known to those of skill in the art. These include, but are not limited to, transfection (including electrophoresis and electroporation), protoplast fusion, calcium phosphate precipitation, cell fusion with enveloped DNA, microinjection, and infection with intact virus. See, Ridgway, A. A. G. "Mammalian Expression Vectors" Chapter 24.2, pp. 470-472 Vectors, Rodriguez and Denhardt, Eds. (Butterworths, 15 Boston, Mass. 1988). Most preferably, plasmid introduction into the host is via electroporation. The transformed cells are grown under conditions appropriate to the production of the light chains and heavy chains, and assayed for heavy and/or light chain protein synthesis. Exemplary assay techniques include enzyme-linked immunosorbent assay (ELISA), radioimmunoassay (RIA), or flourescence-activated cell sorter analysis 20 faces) , immunohistochemistry and the like. As used herein, the term "transformation" shall be used in a broad sense to refer to the introduction of DNA into a recipient host cell that changes the genotype and consequently results in a change in the recipient cell. Along those same lines, "host cells" refers to cells that have been transformed 25 with vectors constructed using recombinant DNA techniques and encoding at least one heterologous gene. In descriptions of processes for isolation of polypepties from recombinant hosts, the terms "cell" and "cell culture" are used interchangeably to denote the source of antibody unless it is clearly specified otherwise. In other words, recovery of polypeptide from the "cells" may mean either from spun down whole cells, or from 30 the cell culture containing both the medium and the suspended cells. In one embodiment, the host cell line used for protein expression (e.g., of multivalent binding molecules) is of mammalian origin; those skilled in the art are credited with ability to preferentially determine particular host cell lines which are best 156 WO 2007/109254 PCT/US2007/006883 suited for the desired gene product to be expressed therein. Exemplary host cell lines include, but are not limited to, DG44 and DUXBI 1 (Chinese Hamster Ovary lines, DHFR minus), HELA (human cervical carcinoma), CVI (monkey kidney line), COS (a derivative of CVI with SV40 T antigen), R1610 (Chinese hamster fibroblast) 5 BALBC/3T3 (mouse fibroblast), HAK (hamster kidney line), SP2/O (mouse myeloma), P3.times.63-Ag3.653 (mouse myeloma), BFA-1c1BPT (bovine endothelial cells), RAJI (human lymphocyte) and 293 (human kidney). In one -embodiment NSO cells may be used. CHO cells are particularly preferred. Host cell lines are typically available from commercial services, the American Tissue Culture Collection or from published 10 literature. In vitro production allows scale-up to give large amounts of the desired polypeptides. Techniques for mammalian cell cultivation under tissue culture conditions are known in the art and include homogeneous suspension culture, e.g. in an airlift reactor or in a continuous stirrer reactor, or immobilized or entrapped cell culture, e.g. in 15 hollow fibers, microcapsules, on agarose microbeads or ceramic cartridges. If necessary and/or desired, the solutions of polypeptides can be purified by the customary chromatography methods, for example gel filtration, ion-exchange chromatography, chromatography over DEAE-cellulose or (immuno-)affinity chromatography, e.g., after preferential biosynthesis of a synthetic hinge region polypeptide or prior to or 20 subsequent to the HIC chromatography step described herein. Genes encoding the polypeptide of the invention can also be expressed non mammalian cells such as bacteria or yeast or plant cells. In this regard it will be appreciated that various unicellular non-mammalian microorganisms such as bacteria can also be transformed; i.e. those capable of being grown in cultures or fermentation. 25 Bacteria, which are susceptible to transformation, include members of the enterobacteriaceae, such as strains of Escherichia coli or Salmonella; Bacillaceae, such as Bacillus subtilis; Pneumococcus; Streptococcus, and Haemophilus influenzae. It will further be appreciated that, when expressed in bacteria, the polypeptides typically become part of inclusion bodies. The polypeptides must be isolated, purified and then 30 assembled into functional molecules. Where tetravalent forms of antibodies are desired, the subunits will then self-assemble into tetravalent antibodies (WO02/096948A2). In addition to prokaryates, eukaryotic microbes may also be used. Saccharomyces cerevisiae, or common baker's yeast, is the most commonly used among 157 WO 2007/109254 PCT/US2007/006883 eukaryotic microorganisms although a number of other strains are commonly available. For expression in Saccharomyces, the plasmid YRp7, for example, (Stinchcomb et al., Nature, 282:39 (1979); Kingsman et al., Gene, 7:141 (1979); Tschemper et al., Gene, 10:157 (1980)) is commonly used. This plasmid already contains the TRP1 gene which 5 provides a selection marker for a mutant strain of yeast lacking the ability to grow in tryptophan, for example ATCC No. 44076 or PEP4-1 (Jones, Genetics, 85:12 (1977)). The presence of the trpl lesion as a characteristic of the yeast host cell genome then provides an effective environment for detecting transformation by growth in the absence of tryptophan. 10 XII. Labeling or Con jugation of Bindini Molecules The binding molecules of the present invention may be used in non-conjugated form or may be conjugated to at least one of a variety of effector, i.e., functional, molecules, e.g., to facilitate target detection or for imaging or therapy of the patient. The polypeptides of the invention can be labeled or conjugated either before or after 15 purification, when purification is performed. In particular, the polypeptides of the present invention may be conjugated to cytotoxins (such as radioisotopes, cytotoxic drugs, or toxins) therapeutic agents, cytostatic agents, biological toxins, prodrugs, peptides, proteins, enzymes, viruses, lipids, biological response modifiers, pharmaceutical agents, immunologically active ligands (e.g., lymphokines or other 20 antibodies wherein the resulting molecule binds to both the neoplastic cell and an effector cell such as a T cell), PEG, or detectable molecules useful in imaging. In another embodiment, a polypeptide of the invention can be conjugated to a molecule that decreases vascularization of tumors. In other embodiments, the disclosed compositions may comprise polypeptides of the invention coupled to drugs or prodrugs. Still other 25 embodiments of the present invention comprise the use of polypeptides of the invention conjugated to specific biotoxins or their cytotoxic fragments such as ricin, gelonin, pseudomonas exotoxin or diphtheria toxin. The selection of which conjugated or unconjugated polypeptide to use will depend on the type and stage of cancer, use of adjunct treatment (e.g., chemotherapy or external radiation) and patient condition. It 30 will be appreciated that one skilled in the art could readily make such a selection in view of the teachings herein. It will be appreciated that, in previous studies, anti-tumor antibodies labeled with isotopes have been used successfully to destroy cells in solid tumors as well as 158 WO 2007/109254 PCT/US2007/006883 lymphomas/leukemias in animal models, and in some cases in humans. Exemplary radioisotopes include: 0 Y, 1251 1 1 In , osh 's3Sm, 67 Cu, 67 Ga, 1 66 Ho, '"Lu, 1 86 Re and '"Re. The radionuclides act by producing ionizing radiation which causes multiple strand breaks in nuclear DNA, leading to cell death. The isotopes used to 5 produce therapeutic conjugates typically produce high energy X- or p-particles which have a short path length. Such radionuclides kill cells to which they are in close proximity, for example neoplastic cells to which the conjugate has attached or has entered. They have little or no effect on non-localized cells. Radionuclides are essentially non-immunogenic. 10 With respect to the use of radiolabeled conjugates in conjunction with the present invention, polypeptides of the invention may be directly labeled (such as through iodination) or may be labeled indirectly through the use of a chelating agent. As used herein, the phrases "indirect labeling" and "indirect labeling approach" both mean that a chelating agent is covalently attached to a binding molecule and at least one radionuclide 15 is associated with the chelating agent. Such chelating agents are typically referred to as bifunctional chelating agents as they bind both the polypeptide and the radioisotope. Particularly preferred chelating agents comprise I -isothiocycmatobenzyl-3 methyldiothelene triaminepentaacetic acid ("MX-DTPA") and cyclohexyl diethylenetriamine pentaacetic acid ("CHX-DTPA") derivatives. Other chelating agents 20 comprise P-DOTA and EDTA derivatives. Particularly preferred radionuclides for indirect labeling include 1 "In and 90 Y. As used herein, the phrases "direct labeling" and "direct labeling approach" both mean that a radionuclide is covalently attached directly to a polypeptide (typically via an amino acid residue). More specifically, these linking technologies include random 25 labeling and site-directed labeling. In the latter case, the labeling is directed at specific sites on the polypeptide, such as the N-linked sugar residues present only on the Fc portion of the conjugates. Further, various direct labeling techniques and protocols are compatible with the instant invention. For example, Technetium-99m labeled polypeptides may be prepared by ligand exchange processes, by reducing pertechnate 30 (TcO 4 ) with stannous ion solution, chelating the reduced technetium onto a Sephadex column and applying the polypeptides to this column, or by batch labeling techniques, e.g. by incubating pertechnate, a reducing agent such as SnCl 2 , a buffer solution such as a sodium-potassium phthalate-solution, and the antibodies. In any event, preferred 159 WO 2007/109254 PCT/US2007/006883 radionuclides for directly labeling antibodies are well known in the art and a particularly preferred radionuclide for direct labeling is 131 covalently attached via tyrosine residues. Polypeptides according to the invention may be derived, for example, with radioactive sodium or potassium iodide and a chemical oxidizing agent, such as sodium 5 hypochlorite, chloramine T or the like, or an enzymatic oxidizing agent, such as lactoperoxidase, glucose oxidase and glucose. However, for the purposes of the present invention, the indirect labeling approach is particularly preferred. Patents relating to chelators and chelator conjugates are known in the art. For instance, U.S. Patent No. 4,831,175 of Gansow is directed to polysubstituted 10 diethylenetriaminepentaacetic acid chelates and protein conjugates containing the same, and methods for their preparation. U.S. Patent Nos. 5,099,069, 5,246,692, 5,286,850, 5,434,287 and 5,124,471 of Gansow also relate to polysubstituted DTPA chelates. These patents are incorporated herein in their entirety. Other examples of compatible metal chelators are ethylenediaminetetraacetic acid (EDTA), 15 diethylenetriaminepentaacetic acid (DPTA), 1,4,8,11-tetraazatetradecane, 1,4,8,11 tetraazatetradecane- 1,4,8,11 -tetraacetic acid, 1 -oxa-4,7,12,15-tetraazaheptadecane 4,7,12,15-tetraacetic acid, or the like. Cyclohexyl-DTPA or CHX-DTPA is particularly preferred and is exemplified extensively below. Still other compatible chelators, including those yet to be discovered, may easily be discerned by a skilled artisan and are 20 clearly within the scope of the present invention. Compatible chelators, including the specific bifunctional chelator used to facilitate chelation in co-pending application Serial Nos. 08/475,813, 08/475,815 and 08/478,967, are preferably selected to provide high affinity for trivalent metals, exhibit increased tumor-to-non-tumor ratios and decreased bone uptake as well as greater in 25 vivo retention of radionuclide at target sites, i.e., B-cell lymphoma tumor sites. However, other bifunctional chelators that may or may not possess all of these characteristics are known in the art and may also be beneficial in tumor therapy. It will also be appreciated that, in accordance with the teachings herein, polypeptides may be conjugated to different radiolabels for diagnostic and therapeutic 30 purposes. To this end the aforementioned co-pending applications, herein incorporated by reference in their entirety, disclose radiolabeled therapeutic conjugates for diagnostic "imaging" of tumors before administration of therapeutic antibody. "In2B8" conjugate comprises a murine monoclonal antibody, 2B8, specific to human CD20 antigen, that is 160 WO 2007/109254 PCT/US2007/006883 attached to "'In via a bifunctional chelator, i.e., MX-DTPA (diethylene triaminepentaacetic acid), which comprises a 1:1 mixture of 1 -isothiocyanatobenzyl-3 methyl-DTPA and 1 -methyl-3-isothiocyanatobenzyl-DTPA. 1'In is particularly preferred as a diagnostic radionuclide because between about 1 to about 10 mCi can be 5 safely administered without detectable toxicity; and the imaging data is generally predictive of subsequent 90 Y-labeled antibody.distribution. Most imaging studies utilize 5 mCi "'In-labeled antibody, because this dose is both safe and has increased imaging efficiency compared with lower doses, with optimal imaging occurring at three to six days after antibody administration. See, for example, Murray, J. Nuc. Med. 26: 3328 10 (1985) and Carraguillo et aL., J. Nuc. Med. 26: 67 (1985). As indicated above, a variety of radionuclides are applicable to the present invention and those skilled in the art can readily determine which radionuclide is most appropriate under various circumstances. For example, 13I is a well known radionuclide used for targeted immunotherapy. However, the clinical usefulness of ' 3 1 I can be limited 15 by several factors including: eight-day physical half-life; dehalogenation of iodinated antibody both in the blood and at tumor sites; and emission characteristics (e.g., large gamma component) which can be suboptimal for localized dose deposition in tumor. With the advent of superior chelating agents, the opportunity for attaching metal chelating groups to proteins has increased the opportunities to utilize other radionuclides such as 'In and 20 90Y. 90 Y provides several benefits for utilization in radioimmunotherapeutic applications: the 64 hour half-life of 90Y is long enough to allow antibody accumulation by tumor and, unlike e.g., 1311, 90 Y is a pure beta emitter of high energy with no accompanying gamma irradiation in its decay, with a range in tissue of 100 to 1,000 cell diameters. Furthermore, the minimal amount of penetrating radiation allows for outpatient administration of 25 90Y-labeled antibodies. Additionally, internalization of labeled antibody is not required for cell killing, and the local emission of ionizing radiation should be lethal for adjacent tumor cells lacking the target molecule. Those skilled in the art will appreciate that these non-radioactive conjugates may also be assembled using a variety of techniques depending on the selected agent to be 30 conjugated. For example, conjugates with biotin are prepared e.g. by reacting the polypeptides with an activated ester of biotin such as the biotin N-hydroxysuccinimide ester. Similarly, conjugates with a fluorescent marker may be prepared in the presence of a coupling agent, e.g. those listed above, or by reaction with an isothiocyanate, 161 WO 2007/109254 PCT/US2007/006883 preferably fluorescein-isothiocyanate. Conjugates of the polypeptides of the invention with cytostatic/cytotoxic substances and metal chelates are prepared in an analogous manner. Many effector molecules lack suitable functional groups to which antibodies can 5 be linked. In one embodiment, an effector molecule, e.g., a drug or prodrug is attached to the antibody through a linking molecule. In one embodiment, the linking molecule contains a chemical bond that allows for the activation of cytotoxicity at a particular site. Suitable chemical bonds are well known in the art and include disulfide bonds, acid labile bonds, photolabile bonds, peptidase labile bonds, thioether bonds formed between 10 sulfhydryl and maleimide groups, and esterase labile bonds. Most preferably, the linking molecule comprises a disulfide bond or a thioether bond. In accordance with the invention, the linking molecule preferably comprises a reactive chemical group. Particularly preferred reactive chemical groups are N-succinimidyl esters and N sulfosuccinimidyl esters. In a preferred embodiment, the reactive chemical group can be 15 covalently bound to the effector via disulfide bonding between thiol groups. In one embodiment an effector molecule is modified to comprise a thiol group. One of ordinary skill in the art will appreciate that a thiol group contains a sulfur atom bonded to a hydrogen atom and is typically also referred to in the art as a sulfhydryl group, which can be denoted as "--SH" or "RSH." 20 In one embodiment, a linking molecule may be used to join the effector molecule with the binding molecule. The linking molecule of the invention may be cleavable or non-cleavable. In one embodiment, the cleavable linking molecule is a redox cleavablelinking molecule, such that the linking molecule is cleavable in environments with a lower redox potential, such as the cytoplasm and other regions with higher 25 concentrations of molecules with free sulfhydryl groups. Examples of linking molecules that may be cleaved due to a change in redox potential include those containing disulfides. The cleaving stimulus can be provided upon intracellular uptake of the binding protein of the invention where the lower redox potential of the cytoplasm facilitates cleavage of the linking molecule. In another embodiment, a decrease in pH 30 triggers the release of the maytansinoid cargo into the target cell. The decrease in pH is implicated in many physiological and pathological processes, such as endosome trafficking, tumor growth, inflammation, and myocardial ischemia. The pH drops from a physiological 7.4 to 5-6 in endosomes or 4-5 in lysosomes. Examples of acid sensitive 162 WO 2007/109254 PCT/US2007/006883 linking molecules which may be used to target lysosomes or endosomes of cancer cells, include those with acid-cleavable bonds such as those found in acetals, ketals, orthoesters, hydrazones, trityls, cis-aconityls, or thiocarbamoyls (see for example, Willner et al., (1993), Bioconj. Chem., 4: 521-7; US Pat. Nos. 4,569,789, 4,631,190, 5 5,306,809, and 5,665,358). Other exemplary acid-sensitive linking molecules comprise dipeptide sequences Phe-Lys and Val-Lys (King et al., (2002), J. Med. Chem., 45: 4336 43). The cleaving stimulus can be provided upon intracellular uptake trafficking to low pH endosomal compartments (e.g. lysosomes). Other exemplary acid-cleavable linking molecules are the molecules that contain two or more acid cleavable bonds for 10 attachment of two or more maytansinoids (King et aL., (1999), Bioconj. Chem., 10: 279 88; WO 98/19705). Cleavable linking molecules may be sensitive to biologically supplied cleaving agents that are associated with a particular target cell, for example, lysosomal or tumor associated enzymes. Examples of linking molecules that can be cleaved enzymatically 15 include, but are not limited to, peptides and esters. Exemplary enzyme cleavable linking molecules include those that are sensitive to tumor-associated proteases such as Cathepsin B or plasmin (Dubowchik et al., (1999), Pharm. Ther., 83: 67-123; Dubowchik et al., (1998), Bioorg. Med. Chem. Lett., 8: 3341-52; de Groot et al., (2000), J. Med. Chem., 43: 3093-102; de Groot et al., (1999)m 42: 5277-83). Cathepsin B 20 cleavable sites include the dipeptide sequences valine-citrulline and phenylalanine lysine (Doronina et aL., (2003), Nat. Biotech., 21(7): 778-84); Dubowchik et aL., (2002), Bioconjug. Chem., 13: 855-69). Other exemplary enzyme-cleavable sites include those formed by oligopeptide sequences of 4 to 16 amino acids (e.g., Suc-P-Ala-Leu-Ala-Leu) which recognized by trouse proteases such as Thimet Oliogopeptidase (TOP), an 25 enzyme that is preferentially released by neutrophils, macrophages, and other granulocytes. In a further embodiment, the linking molecule is formed by reacting a binding molecule of the invention with a linking molecule of the formula: X-Y-Z 30 wherein: X is an attachment molecule; Y is a spacer molecule; and Z is a effector attachment moeity. 163 WO 2007/109254 PCT/US2007/006883 The term "binding molecule attachment molecule" includes molecules which allow for the covalent attachment of the connecting peptide to a binding molecule of the invention. 5 The attachment molecule may comprise, for example, a covalent chain of 1-60 carbon, oxygen, nitrogen, sulfur atoms, optionally substituted with hydrogen atoms and other substituents which allow the binding molecule to perform its intended function. The attachment molecule may comprise peptide, ester, alkyl, alkenyl, alkynyl, aryl, ether, thioether, etc. functional groups. Preferably, the attachment molecule is selected 10 such that it is capable of reacting with a reactive functional group on a polypeptide comprising at least one antigen binding site, to form a binding molecule of the invention. Examples of attachment molecules include, for example, amino, carboxylate, and thiol attachment molecules. Amino attachment molecules include molecules which react with amino groups 15 on a polypeptide, such that a binding molecule of the invention is formed. Amino attachment molecules are known in the art. Examples of amino attachment molecules include, activated carbamides (e.g., which may react with an amino group on a binding molecule to form a linking molecule which comprises urea group), aldehydes (e.g., which may react with amino groups on a binding molecule), and activated isocyanates 20 (which may react with an amino group on a binding molecule to from a linking molecule which comprises a urea group). Examples of amino attachment molecules include, but are not limited to, N-succinimidyl, N-sulfosuccinimidyl, N-phthalimidyl, N sul fophthalimidyl, 2-nitrophenyl, 4-nitrophenyl, 2,4-dinitrophenyl, 3-sulfonyl-4 nitrophenyl, or 3-carboxy-4-nitrophenyl molecule. 25 Carboxylate attachment molecules include molecules which react with carboxylate groups on a polypeptide, such that a binding molecule of the invention is formed. Carboxylate attachment molecules are known in the art. Examples of carboxylate attachment molecules include, but are not limited to activated ester intermediates and activated carbonyl intermediates, which may react with a COOH 30 group on a binding moleculeto form a linking molecule which comprises a ester, thioester, or amide group. Thiol attachment molecules include molecules which react with thiol groups present on a polypeptide, such that a binding molecule of the invention is formed. Thiol 164 WO 2007/109254 PCT/US2007/006883 attachment molecules are knovyn in the art. Examples of thiol attachment molecules include activated acyl groups (which may react with a sulfhydryl on a binding molecule to form a linking molecule which comprises a thioester), activated alkyl groups (which may react with a sulfhydryl on a binding molecule to form a linking molecule which 5 comprises a thioester molecule), Michael acceptors such as maleimide or acrylic groups (which may react with a sulfhydryl on a binding molecule to form a Michael-type addition product), groups which react with sulfhydryl groups via redox reactions, activated di-sulfide groups (which may react with a sulfhydryl group on a binding molecule to form, for example, a linking molecule which comprises a disulfide 10 molecule). Other thiol attachment molecules include acrylamides, alpha iodoacetamides, and cyclopropan- 1,1 -dicarbonyl compounds. In addition, the thiol attachment molecule may comprise a molecule which modifies a thiol on the binding molecule to form another reactive species to which the linking molecule can be attached to form a binding molecule of the invention. 15 The spacer molecule, Y, is a covalent bond or a covalent chain of atoms which may contain one or more aminoacid residues. It may also comprise 0-60 carbon, oxygen, sulfur or nitrogen atoms optionally substituted with hydrogen or other substituents which allow the resulting binding molecule to perform its intended function. In one embodiment, Y comprises an alkyl, alkenyl, alkynyl, ester, ether, carbonyl, or 20 amide molecule. In another embodiment, a thiol group on the binding molecule is converted into a reactive group, such as a reactive carbonyl group, such as a ketone or aldehyde. The attachment molecule is then reacted with the ketone or aldehyde to form the desired compound of the invention. Examples of carbonyl reactive attachment molecules 25 include, but are not limited to, hydrazines, hydrazides, O-substituted hydroxylamines, alpha-beta-unsaturated ketones, and H 2
C=CH-CO-NH-NH
2 . Other examples of attachment molecules and methods for modifying thiol molecules which can be used to form binding molecules of the invention are described Pratt, M. L. et al. J Am Chem Soc. 2003 May 21;125(20):6149-59; and Saxon, E. Science. 2000 Mar 30 17;287(5460):2007-10. The linking molecule may be a molecule which is capable of reacting with an effector molecule or a derivative thereof to form a binding molecule of the invention. For example, the effector molecule may be linked to the remaining portions of the 165 WO 2007/109254 PCT/US2007/006883 molecule through a disulfide bond. In such cases, the linking molecule is selected such that it is capable of reacting with an appropriate effector moeity derivative such that the effector molecule is attached to the binding molecule of the invention. As described above, the linking molecule and/or the connecting peptide as a whole may be selected 5 that the connecting peptide is cleaved in an appropriate environment. Particularly preferred connecting peptide molecules include, for example, N succinimidyl 3-(2-pyridyldithio)propionate (SPDP) (see, e.g., Carlsson et al., Biochem. J., 173, 723-737 (1978)), N-succinimidyl 4-(2-pyridyldithio)butanoate (SPDB) (see, e.g., U.S. Pat. No. 4,563,304), N-succinimidyl 4-(2-pyridyldithio)pentanoate (SPP) (see, e.g., 10 CAS Registry number 341498-08-6), N-succinimidyl 4-(N-maleimidomethyl)cyclohe xane-l -carboxylate (SMCC) (see, e.g., Yoshitake et al., Eur. J. Biochem., 101, 395-399 (1979)), and N-succinimidyl 4-methyl-4-[2-(5-nitro-pyridyl)-- dithiolpentanoate (SMNP) (see, e.g., U.S. Pat. No. 4,563,304) The most preferred connecting peptide molecules for use in the inventive composition are SPP, SMCC, and SPDB. In a 15 preferred embodiment, SPDB is used to link an effector molecule to a binding molecule of the invention. Preferred cytotoxic effector molecules for use in the present invention are cytotoxic drugs, particularly those which are used for cancer therapy. As used herein, "a cytotoxin or cytotoxic agent" means any agent that is detrimental to the growth and 20 proliferation of cells and may act to reduce, inhibit or destroy a cell or malignancy. Exemplary cytotoxins include, but are not limited to, radionuclides, biotoxins, enzymatically active toxins, cytostatic or cytotoxic therapeutic agents, prodrugs, immunologically active ligands and biological response modifiers such as cytokines. Any cytotoxin that acts to retard or slow the growth of immunoreactive cells or 25 malignant cells is within the scope of the present invention. Exemplary cytotoxins include, in general, cytostatic agents, alkylating agents, antimetabolites, anti-proliferative agents, tubulin binding agents, hormones and hormone antagonists, and the like. Exemplary cytostatics that are compatible with the present invention include alkylating substances, such as mechlorethamine, 30 triethylenephosphoramide, cyclophosphamide, ifosfamide, chlorambucil, busulfan, melphalan or triaziquone, also nitrosourea compounds, such as carmustine, lomustine, or semustine. Particularly preferred molecules for conjugation are maytansinoids. 166 WO 2007/109254 PCT/US2007/006883 Maytansinoids were originally isolated from the east African shrub belonging to the genus Maytenus, but were subsequently also discovered to be metabolites of soil bacteria, such as Actinosynnema pretiosum (see, e.g., U.S. Pat. No. 3,896,111). Maytansinoids are known in the art to include maytansine, maytansinol, C-3 esters of 5 maytansinol, and other maytansinol analogues and derivatives (see, e.g., U.S. Pat. Nos. 5,208,020 and 6,441,163). C-3 esters of maytansinol can be naturally occurring or synthetically derived. Moreover, both naturally occurring and synthetic C-3 maytansinol esters can be classified as a C-3 ester with simple carboxylic acids, or a C-3 ester with derivatives of N-methyl-L-alanine, the latter being more cytotoxic than the former. 10 Synthetic maytansinoid analogues also are known in the art and described in, for example, Kupchan et al., J. Med. Chem., 21, 31-37 (1978). Methods for generating maytansinol and analogues and derivatives thereof are described in, for example, U.S. Pat. No. 4,151,042. Suitable maytansinoids for use as antibody conjugates can be isolated from 15 natural sources, synthetically produced, or semi-synthetically produced using methods known in the art. Moreover, the maytansinoid can be modified in any suitable manner, so long as sufficient cytotoxicity is preserved in the ultimate conjugate molecule. Particularly preferred maytansinoids comprising a linking molecule that contains a reactive chemical group are C-3 esters of maytansinol and its analogs where the 20 linking molecule contains a disulfide bond and the attachment molecule comprises a N succinimidyl or N-sulfosuccinimidyl ester. Many positions on maytansinoids can serve as the position to chemically link the linking molecule, e.g., through an effector attachment molecule. For example, the C-3 position having a hydroxyl group, the C-14 position modified with hydroxymethyl, the C-15 position modified with hydroxy and the 25 C-20 position having a hydroxy group are all useful. The linking molecule most preferably is linked to the C-3 position of maytansinol. Most preferably, the maytansinoid used in connection with the inventive composition is N.sup.2'-deacetyl N.sup.2'-(- 3-mercapto- I -oxopropyl)-maytansine (DM1) or N.sup.2'-deacetyl-N.sup.2' (4-- mercapto-4-m ethyl-I -oxopentyl)-maytansine (DM4). 30 Linking molecules with other chemical bonds also can be used in the context of the invention, as can other maytansinoids. Specific examples of other chemical bonds which may be incorportated in the linking molecules include those described above, such as, for example acid labile bonds, thioether bonds, photolabile bonds, peptidase 167 WO 2007/109254 PCT/US2007/006883 labile bonds and esterase labile bonds. Methods for producing maytansinoids with linking molecules and/or effector attachment molecules are described in, for example, U.S. Pat. Nos. 5,208,020, 5,416,064, and 6,333,410. The linking molecule (and/or the effector attachment molecule) of a 5 maytansinoid typically and preferably is part of a larger connecting peptide molecule that is used to join the antibody to the maytansinoid. Any suitable connecting peptide molecule can be used in connection with the invention, so long as-the linking molecule provides for retention of the cytotoxicity and targeting characteristics of the maytansinoid and the antibody, respectively. The linking molecule joins the 10 maytansinoid to the antibody through chemical bonds (as described above), such that the maytansinoid and the antibody are chemically coupled (e.g., covalently bonded) to each other. Desirably, the linking molecule chemically couples the maytansinoid to the antibody through disulfide bonds or thioether bonds. Most preferably, the antibody is chemically coupled to the maytansinoid via disulfide bonds. 15 Other preferred classes of cytotoxic agents include, for example, the anthracycline family of drugs, the vinca drugs, the mitomycins, the bleomycins, the cytotoxic nucleosides, the pteridine family of drugs, diynenes, and the podophyllotoxins. Particularly useful members of those classes include, for example, adriamycin, carminomycin, daunorubicin (daunomycin), doxorubicin, aminopterin, methotrexate, 20 methopterin, mithramycin, streptonigrin, dichloromethotrexate, mitomycin C, actinomycin-D, porfiromycin, 5-fluorouracil, floxuridine, ftorafur, 6-mercaptopurine, cytarabine, cytosine arabinoside, podophyllotoxin, or podophyllotoxin derivatives such as etoposide or etoposide phosphate, melphalan, vinblastine, vincristine, leurosidine, vindesine,'leurosine and the like. Still other cytotoxins that are compatible with the 25 teachings herein include taxol, taxane, cytochalasin B, gramicidin D, ethidium bromide, emetine, tenoposide, colchicin, dihydroxy anthracin dione, mitoxantrone, procaine, tetracaine, lidocaine, propranolol, and puromycin and analogs or homologs thereof. Hormones and hormone antagonists, such as corticosteroids, e.g. prednisone, progestins, e.g. hydroxyprogesterone or medroprogesterone, estrogens, e.g. diethylstilbestrol, 30 antiestrogens, e.g. tamoxifen, androgens, e.g. testosterone, and aromatase inhibitors, e.g. aminogluthetimide are also compatible with the teachings herein. As noted previously, one skilled in the art may make chemical modifications to the desired compound in order to make reactions of that compound more convenient for purposes of preparing 168 WO 2007/109254 PCT/US2007/006883 conjugates of the invention. One example of particularly preferred cytotoxins comprise members or derivatives of the enediyne family of anti-tumor antibiotics, including calicheamicin, esperamicins or dynemicins. These toxins are extremely potent and act by cleaving 5 nuclear DNA, leading to cell death. Unlike protein toxins which can be cleaved in vivo to give many inactive but immunogenic polypeptide fragments, toxins such as calicheamicin, esperamicins and other enediynes are small molecules which are essentially non-immunogenic. These non-peptide toxins are chemically-linked to the dimers or tetramers by techniques which have been previously used to label monoclonal 10 antibodies and other molecules. These linking technologies include site-specific linkage via the N-linked sugar residues present only on the Fc portion of the constructs. Such site-directed linking methods have the advantage of reducing the possible effects of linkage on the binding properties of the constructs. Among other cytotoxins, it will be appreciated that polypeptides can also be 15 associated with a biotoxin such as ricin subunit A, abrin, diptheria toxin, botulinum, cyanginosins, saxitoxin, shigatoxin, tetanus, tetrodotoxin, trichothecene, verrucologen or a toxic enzyme. Preferably, such constructs will be made using genetic engineering techniques that allow for direct expression of the binding molecule-toxin construct. Other biological response modifiers that may be associated with the polypeptides of the invention 20 of the present invention comprise cytokines such as lymphokines and interferons. In view of the instant disclosure it is submitted that one skilled in the art could readily form such constructs using conventional techniques. Another class of compatible cytotoxins that may be used in conjunction with the disclosed polypeptides are radiosensitizing drugs that may be effectively directed to tumor 25 or immunoreactive cells. Such drugs enhance the sensitivity to ionizing radiation, thereby increasing the efficacy of radiotherapy. A conjugate internalized by the tumor cell would deliver the radiosensitizer nearer the nucleus where radiosensitization would be maximal. The unbound radiosensitizer linked polypeptides of the invention would be cleared quickly from the blood, localizing the remaining radiosensitization agent in the target tumor and 30 providing minimal uptake in normal tissues. After rapid clearance from the blood, adjunct radiotherapy would be administered in one of three ways: 1.) external beam radiation directed specifically to the tumor, 2.) radioactivity directly implanted in the tumor or 3.) systemic radioimmunotherapy with the same binding molecule. A potentially attractive 169 WO 2007/109254 PCT/US2007/006883 variation of this approach would be the attachment of a therapeutic radioisotope to the radiosensitized immunoconjugate, thereby providing the convenience of administering to the patient a single drug. In one embodiment, a molecule that enhances the stability or efficacy of the 5 polypeptide can be conjugated. For example, in one embodiment, PEG can be conjugated to the polypeptides of the invention to increase their half-life in vivo. Leong, S.R., et al. 2001. Cytokine 16:106; 2002; Adv. in Drug Deliv. Rev. 54:531; or Weir et al. 2002. Biochem. Soc. Transactions 30:512. As previously alluded to, compatible cytotoxins may comprise a prodrug. As used 10 herein, the term "prodrug" refers to a precursor or derivative form of a pharmaceutically active substance that is less cytotoxic to tumor cells compared to the parent drug and is capable of being enzymatically activated or converted into the more active parent form. Prodrugs compatible with the invention include, but are not limited to, phosphate containing prodrugs, thiophosphate-containing prodrugs, sulfate containing prodrugs, 15 peptide containing prodrugs, p-lactam-containing prodrugs, optionally substituted phenoxyacetamide-containing prodrugs or optionally substituted phenylacetamide containing prodrugs, 5-fluorocytosine and other 5-fluorouridine prodrugs that can be converted to the more active cytotoxic free drug. In one embodiment, a cytotoxic agent, such as a maytansinoid, is administered as a prodrug which is released by the hydrolysis of 20 disulfide bonds. Further examples of cytotoxic drugs that can be derivatized into a prodrug form for use in the present invention comprise those chemotherapeutic agents described above. XIII. Administration of Binding Molecules 25 Methods of preparing and administering binding molecules of the invention to a subject are well known to or are readily determined by those skilled in the art. The route of administration of the binding molecules of the invention may be oral, parenteral, by inhalation or topical. The term parenteral as used herein includes intravenous, intraarterial, intraperitoneal, intramuscular, subcutaneous, rectal or vaginal 30 administration. The intravenous, intraarterial, subcutaneous and intramuscular forms of parenteral administration are generally preferred. While all these forms of administration are clearly contemplated as being within the scope of the invention, a form for administration would be a solution for injection, in particular for intravenous or 170 WO 2007/109254 PCT/US2007/006883 intraarterial injection or drip. Usually, a suitable pharmaceutical composition for injection may comprise a buffer (e.g. acetate, phosphate or citrate buffer), a surfactant (e.g. polysorbate), optionally a stabilizer agent (e.g. human albumin), etc. However, in other methods compatible with the teachings herein, the polypeptides can be delivered 5 directly to the site of the adverse cellular population thereby increasing the exposure of the diseased tissue to the therapeutic agent. Preparations for parenteral administration include sterile aqueous-or non-aqueous solutions, suspensions, and emulsions. Examples of non-aqueous solvents are propylene glycol, polyethylene glycol, vegetable oils such as olive oil, and injectable organic esters 10 such as ethyl oleate. Aqueous carriers include water, alcoholic/aqueous solutions, emulsions or suspensions, including saline and buffered media. In the subject invention, pharmaceutically acceptable carriers include, but are not limited to, 0.01-0.1M and preferably 0.05M phosphate buffer or 0.8% saline. Other common parenteral vehicles include sodium phosphate solutions, Ringer's dextrose, dextrose and sodium chloride, 15 lactated Ringer's, or fixed oils. Intravenous vehicles include fluid and nutrient replenishers, electrolyte replenishers, such as those based on Ringer's dextrose, and the like. Preservatives and other additives may also be present such as for example, antimicrobials, antioxidants, chelating agents, and inert gases and the like. More particularly, pharmaceutical compositions suitable for injectable use include sterile 20 aqueous solutions (where water soluble) or dispersions and sterile powders for the extemporaneous preparation of sterile injectable solutions or dispersions. In such cases, the composition must be sterile and should be fluid to the extent that easy syringability exists. It should be stable under the conditions of manufacture and storage and will preferably be preserved against the contaminating action of microorganisms, such as 25 bacteria and fungi. The carrier can be a solvent or dispersion medium containing, for example, water, ethanol, polyol (e.g., glycerol, propylene glycol, and liquid polyethylene glycol, and the like), and suitable mixtures thereof. The proper fluidity can be maintained, for example, by the use of a coating such as lecithin, by the maintenance of the required particle size in the case of dispersion and by the use of surfactants. 30 Prevention of the action of microorganisms can be achieved by various antibacterial and antifungal agents, for example, parabens, chlorobutanol, phenol, ascorbic acid, thimerosal and the like. In many cases, it will be preferable to include isotonic agents, for example, sugars, polyalcohols, such as mannitol, sorbitol, or sodium chloride in the 171 WO 2007/109254 PCT/US2007/006883 composition. Prolonged absorption of the injectable compositions can be brought about by including in the composition an agent which delays absorption, for example, aluminum monostearate and gelatin. In any case, sterile injectable solutions can be prepared by incorporating an 5 active compound (e.g., a polypeptide by itself or in combination with other active agents) in the required amount in an appropriate solvent with one or a combination of ingredients enumerated herein, as required, followed by filtered sterilization. Generally, dispersions are prepared by incorporating the active compound into a sterile vehicle, which contains a basic dispersion medium and the required other ingredients from those 10 enumerated above. In the case of sterile powders for the preparation of sterile injectable solutions, the preferred methods of preparation are vacuum drying and freeze-drying, which yields a powder of an active ingredient plus any additional desired ingredient from a previously sterile-filtered solution thereof. The preparations for injections are processed, filled into containers such as ampoules, bags, bottles, syringes or vials, and 15 sealed under aseptic conditions according to methods known in the art. Further, the preparations may be packaged and sold in the form of a kit such as those described in co-pending U.S.S.N. 09/259,337 and U.S.S.N. 09/259,338 each of which is incorporated herein by reference. Such articles of manufacture will preferably have labels or package inserts indicating that the associated compositions are useful for treating a subject 20 suffering from, or predisposed to autoimmune or neoplastic disorders. Effective doses of the stabilized binding molecules of the present invention, for the treatment of the above described conditions vary depending upon many different factors, including means of administration, target site, physiological state of the patient, whether the patient is human or an animal, other medications administered, and whether 25 treatment is prophylactic or therapeutic. Usually, the patient is a human, but non-human mammals including transgenic mammals can also be treated. Treatment dosages may be titrated using routine methods known to those of skill in the art to optimize safety and efficacy. For passive immunization with a binding molecule of the invention, the dosage 30 may range, e.g., from about 0.0001 to 100 mg/kg, and more usually 0.01 to 5 mg/kg (e.g., 0.02 mg/kg, 0.25 mg/kg, 0.5 mg/kg, 0.75 mg/kg, 1mg/kg, 2 mg/kg, etc.), of the host body weight. For example dosages can be 1 mg/kg body weight or 10 mg/kg body weight or within the range of 1-10 mg/kg, preferably at least 1 mg/kg. Doses 172 WO 2007/109254 PCT/US2007/006883 intermediate in the above ranges are also intended to be within the scope of the invention. Subjects can be administered such doses daily, on alternative days, weekly or according to any other schedule determined by empirical analysis. An exemplary 5 treatment entails administration in multiple dosages over a prolonged period, for example, of at least six months. Additional exemplary treatment regimes entail administration once per every two weeks or once a month or once every 3 to 6 months. Exemplary dosage schedules include 1-10 mg/kg or 15 mg/kg on consecutive days, 30 mg/kg on alternate days or 60 mg/kg weekly. In some methods, two or more 10 monoclonal antibodies with different binding specificities are administered simultaneously, in which case the dosage of each antibody administered may fall within the ranges indicated. Binding molecules of the invention can be administered on multiple occasions. Intervals between single dosages can be, e.g., daily,weekly, monthly or yearly. Intervals 15 can also be irregular as indicated by measuring blood levels of polypeptide or target molecule in the patient. In some methods, dosage is adjusted to achieve a certain plasma binding molecule or toxin concentration, e.g., 1-1000 pg/ml or 25-300 pg/ml. Alternatively, binding molecules can be administered as a sustained release formulation, in which case less frequent administration is required. Dosage and frequency vary 20 depending on the half-life of the antibody in the patient. In general, humanized antibodies show the longest half-life, followed by chimeric antibodies and nonhuman antibodies. In one embodiment, the binding molecules of the invention can be administered in unconjugated form, In another embodiment, the polypeptides of the invention can be administered multiple times in conjugated form. In still another 25 embodiment, the binding molecules of the invention can be administered in unconjugated form, then in conjugated form, or vise versa. The dosage and frequency of administration can vary depending on whether the treatment is prophylactic or therapeutic. In prophylactic applications, compositions containing the present antibodies or a cocktail thereof are administered to a patient not 30 already in the disease state to enhance the patient's resistance. Such an amount is defined to be a "prophylactic effective dose." In this use, the precise amounts again depend upon the patient's state of health and general immunity, but generally range from 0.1 to 25 mg per dose, especially 0.5 to 2.5 mg per dose. A relatively low dosage is 173 WO 2007/109254 PCT/US2007/006883 administered at relatively infrequent intervals over a long period of time. Some patients continue to receive treatment for the rest of their lives. In therapeutic applications, a relatively high dosage (e.g., from about 1 to 400 mg/kg of binding molecule, e.g., antibody per dose, with dosages of from 5 to 25 mg 5 being more commonly used for radioimmunoconjugates and higher doses for cytotoxin drug conjugated molecules) at relatively short intervals is sometimes required until progression of the disease is reduced or terminated, and preferably until the patient shows partial or complete amelioration of symptoms of disease. Thereafter, the patent can be administered a prophylactic regime. 10 In one embodiment, a subject can be treated with a nucleic acid molecule encoding a polypeptide of the invention (e.g., in a vector). Doses for nucleic acids encoding polypeptides range from about 10 ng to 1 g, 100 ng to 100 mg, 1 pg to 10 mg, or 30-300 pg DNA per patient. Doses for infectious viral vectors vary from 10-100, or more, virions per dose. 15 Therapeutic agents can be administered by parenteral, topical, intravenous, oral, subcutaneous, intraarterial, intracranial, intraperitoneal, intranasal or intramuscular means for prophylactic and/or therapeutic treatment. Intramuscular injection or intravenous infusion are preferred for administration of a binding molecule of the invention. In some methods, particular therapeutic binding molecules are injected 20 directly into the cranium. In some methods, binding molecules are administered as a sustained release composition or device, such as a Medipad T M device. Agents of the invention can optionally be administered in combination with other agents that are effective in treating the disorder or condition in need of treatment (e.g., prophylactic or therapeutic). Preferred additional agents are those which are art 25 recognized and are standardly administered for a particular disorder. Effective single treatment dosages (i.e., therapeutically effective amounts) of 9 0 Y-labeled polypeptides of the invention range from between about 5 and about 75 mCi, more preferably between about 10 and about 40 mCi. Effective single treatment non-marrow ablative dosages of 1 31 I-labeled antibodies range from between about 5 and 30 about 70 mCi, more preferably between about 5 and about 40 mCi. Effective single treatment ablative dosages (i.e., may require autologous bone marrow transplantation) of 1 3 1 I-labeled antibodies range from between about 30 and about 600 mCi, more preferably between about 50 and less than about 500 mCi. In conjunction with a 174 WO 2007/109254 PCT/US2007/006883 chimeric modified antibody, owing to the longer circulating half life vis-a-vis murine antibodies, an effective single treatment non-marrow ablative dosages of iodine-131 labeled chimeric antibodies range from between about 5 and about 40 mCi, more preferably less than about 30 mCi. Imaging criteria for, e.g., the '"In label, are typically 5 less than about 5 mCi. While a great deal of clinical experience has been gained with 1311 and 90 Y, other radiolabels are known in the art and have been used for similar purposes: Still other radioisotopes are used for imaging. For example, additional radioisotopes which are compatible with the scope of the instant invention include, but are not limited to, 1231, 10 12 32 p, s7Co, "Cu, 67Cu, 77Br, 81Rb, 8-Kr, 87Sr, "3In, 1"Cs, 12Cs, 321 197Hg, 203Pb, 206Bi, 177 186 212 212 i,47SC 1055p8 1 25 21 1 Lu, Re, Pb, Bi, Sc, ' 05 Rh, ' 09 Pd, mSm, '8'Re, 199Au, 2Ac, 21 At, and mBi. In this respect alpha, gamma and beta emitters are all compatible with in the instant invention. Further, in view of the instant disclosure it is submitted that one skilled in the art could readily determine which radionuclides are compatible with a selected course of 15 treatment without undue experimentation. To this end, additional radionuclides which have already been used in clinical diagnosis include 1251, 1231, 99 Tc, 43 K, 2Fe, 67 Ga, 6 Ga, as well as "'In. Antibodies have also been labeled with a variety of radionuclides for potential use in targeted immunotherapy (Peirersz et al. Immunol. Cell Biol. 65: 111-125 (1987)). These radionuclides include ' 88 Re and '" 6 Re as well as 199Au and 67 Cu to a 20 lesser extent. U.S. Patent No. 5,460,785 provides additional data regarding such radioisotopes and is incorporated herein by reference. Whether or not the binding molecules of the invention are used in a conjugated or unconjugated form, it will be appreciated that a major advantage of the present invention is the ability to use these polypeptides in myelosuppressed patients, especially those who are 25 undergoing, or have undergone, adjunct therapies such as radiotherapy or chemotherapy. In other preferred embodiments, the polypeptides (again in a conjugated or unconjugated form) may be used in a combined therapeutic regimen with chemotherapeutic agents. Those skilled in the art will appreciate that such therapeutic regimens may comprise the sequential, simultaneous, concurrent or coextensive administration of the disclosed 30 antibodies and one or more chemotherapeutic agents. Particularly preferred embodiments of this aspect of the invention will comprise the administration of a toxin conjugated binding molecule, e.g., conjugated to a maytansinoid such as a D4 maytansinoid. While the binding molecules may be administered as described immediately above, 175 WO 2007/109254 PCT/US2007/006883 it must be emphasized that in other embodiments conjugated and unconjugated polypeptides may be administered to otherwise healthy patients as a first line therapeutic agent. In such embodiments the polypeptides may be administered to patients having normal or average red marrow reserves and/or to patients that have not, and are not, 5 undergoing adjunct therapies such as external beam radiation or chemotherapy. However, as discussed above, selected embodiments of the invention comprise the administration of binding molecules to myelosuppressed patients or in combination or conjunction with one or more adjunct therapies such as radiotherapy or chemotherapy (i.e. a combined therapeutic regimen). As used herein, the administration of polypeptides in 10 conjunction or combination with an adjunct therapy means the sequential, simultaneous, coextensive, concurrent, concomitant or contemporaneous administration or application of the therapy and the disclosed binding molecules. Those skilled in the art will appreciate that the administration or application of the various components of the combined therapeutic regimen may be timed to enhance the overall effectiveness of the treatment. 15 For example, chemotherapeutic agents could be administered in standard, well known courses of treatment followed within a few weeks by radioimmunoconjugates of the present invention. Conversely, cytotoxin associated polypeptides could be administered intravenously followed by tumor localized external beam radiation. In yet other embodiments, the polypeptide may be administered concurrently with one or more selected 20 chemotherapeutic agents in a single office visit. A skilled artisan (e.g. an experienced oncologist) would readily be able to discern effective combined therapeutic regimens without undue experimentation based on the selected adjunct therapy and the teachings of the instant specification. In this regard it will be appreciated that the combination of the binding molecules 25 (either conjugated or unconjugated) and the chemotherapeutic agent may be administered in any order and within any time frame that provides a therapeutic benefit to the patient. That is, the chemotherapeutic agent and polypeptide may be administered in any order or concurrently. Binding molecules and chemotherapeutic agents may be administered separately or may be administered in the form of one composition. In selected 30 embodiments the polypeptides of the present invention will be administered to patients that have previously undergone chemotherapy. In yet other embodiments, the polypeptides and the chemotherapeutic treatment will be administered substantially simultaneously or concurrently. For example, the patient may be given the binding molecule while 176 WO 2007/109254 PCT/US2007/006883 undergoing a course of chemotherapy. In preferred embodiments the binding molecule will be administered within 1 year of any chemotherapeutic agent or treatment. In other preferred embodiments the polypeptide will be administered within 10, 8, 6, 4, or 2 months of any chemotherapeutic agent or treatment. In still other preferred embodiments the 5 polypeptide will be administered within 4, 3, 2 or 1 week of any chemotherapeutic agent or treatment. In yet other embodiments the polypeptide will be administered within 5, 4, 3, 2 or 1 days of the selected chemotherapeutic agent or treatment.- It will further be appreciated that the two agents or treatments may be administered to the patient within a matter of hours or minutes (i.e. substantially simultaneously). 10 Moreover, in accordance with the present invention a myelosuppressed patient shall be held to mean any patient exhibiting lowered blood counts. Those skilled in the art will appreciate that there are several blood count parameters conventionally used as clinical indicators of myelosuppresion and one can easily measure the extent to which myelosuppresion is occurring in a patient. Examples of art accepted myelosuppression 15 measurements are the Absolute Neutrophil Count (ANC) or platelet count. Such myelosuppression or partial myeloablation may be a result of various biochemical disorders or diseases or, more likely, as the result of prior chemotherapy or radiotherapy. In this respect, those skilled in the art will appreciate that patients who have undergone traditional chemotherapy typically exhibit reduced red marrow reserves. As discussed above, such 20 subjects often cannot be treated using optimal levels of cytotoxin (i.e. radionuclides) due to unacceptable side effects such as anemia or immunosuppression that result in increased mortality or morbidity. More specifically conjugated or unconjugated polypeptides of the present invention may be used to effectively treat patients having ANCs lower than about 2000/mm 3 or 25 platelet counts lower than about 150,000/ mm 3 . More preferably the polypeptides of the present invention may be used to treat patients having ANCs of less than about 1500/ mm 3 , less than about 1000/mm 3 or even more preferably less than about 500/ mm 3 . Similarly, the polypeptides of the present invention may be used to treat patients having a platelet count of less than about 75,000/mm 3 , less than about 50,000/mm 3 or even less than about 30 10,000/mm 3 . In a more general sense, those skilled in the art will easily be able to determine when a patient is myelosuppressed using government implemented guidelines and procedures. As indicated above, many myelosuppressed patients have undergone courses of 177 WO 2007/109254 PCT/US2007/006883 treatment including chemotherapy, implant radiotherapy or external beam radiotherapy. In the case of the latter, an external radiation source is for local irradiation of a malignancy. For radiotherapy implantation methods, radioactive reagents are surgically located within the malignancy, thereby selectively irradiating the site of the disease. In any event, the 5 disclosed polypeptides may be used to treat disorders in patients exhibiting myelosuppression regardless of the cause. In this regard it will further be appreciated that the polypeptides of the instant invention may be used in conjunction or combination with any chemotherapeutic agent or agents (e.g. to provide a combined therapeutic regimen) that eliminates, reduces, 10 inhibits or controls the growth of neoplastic cells in vivo. As discussed, such agents often result in the reduction of red marrow reserves. This reduction may be offset, in whole or in part, by the diminished myelotoxicity of the compounds of the present invention that advantageously allow for the aggressive treatment of neoplasias in such patients. In other preferred embodiments the radiolabeled immunoconjugates disclosed 15 herein may be effectively used with radiosensitizers that increase the susceptibility of the neoplastic cells to radionuclides. For example, radiosensitizing compounds may be administered after the radiolabeled binding molecule has been largely cleared from the bloodstream but still remains at therapeutically effective levels at the site of the tumor or tumors. 20 With respect to these aspects of the invention, exemplary chemotherapeutic agents that are compatible with the instant invention include alkylating agents, vinca alkaloids (e.g., vincristine and vinblastine), procarbazine, methotrexate and prednisone. The four-drug combination MOPP (mechlethamine (nitrogen mustard), vincristine (Oncovin), procarbazine and prednisone) is very effective in treating various types of 25 lymphoma and comprises a preferred embodiment of the present invention. In MOPP resistant patients, ABVD (e.g., adriamycin, bleomycin, vinblastine and dacarbazine), ChIVPP (chlorambucil, vinblastine, procarbazine and prednisone), CABS (lomustine, doxorubicin, bleomycin and streptozotocin), MOPP plus ABVD, MOPP plus ABV (doxorubicin, bleomycin and vinblastine) or BCVPP (carmustine, cyclophosphamide, 30 vinblastine, procarbazine and prednisone) combinations can be used. Arnold S. Freedman and Lee M. Nadler, Malignant Lymphomas, in HARRISON'S PRINCIPLES OF INTERNAL MEDICINE 1774-1788 (Kurt J. Isselbacher et al., eds., 13' ed. 1994) and V. T. DeVita et al., (1997) and the references cited therein for standard dosing and scheduling. 178 WO 2007/109254 PCT/US2007/006883 These therapies can be used unchanged, or altered as needed for a particular patient, in combination with one or more polypeptides of the invention as described herein. Additional regimens that are useful in the context of the present invention include use of single alkylating agents such as cyclophosphamide or chlorambucil, or combinations 5 such as CVP (cyclophosphamide, vincristine and prednisone), CHOP (CVP and doxorubicin), C-MOPP (cyclophosphamide, vincristine, prednisone and procarbazine), CAP-BOP (CHOP plus procarbazine and bleomycin), m-BACOD (CHOP plus methotrexate, bleomycin and leucovorin), ProMACE-MOPP (prednisone, methotrexate, doxorubicin, cyclophosphamide, etoposide and leucovorin plus standard MOPP), 10 ProMACE-CytaBOM (prednisone, doxorubicin, cyclophosphamide, etoposide, cytarabine, bleomycin, vincristine, methotrexate and leucovorin) and MACOP-B (methotrexate, doxorubicin, cyclophosphamide, vincristine, fixed dose prednisone, bleomycin and leucovorin). Those skilled in the art will readily be able to determine standard dosages and scheduling for each of these regimens. CHOP has also been combined with bleomycin, 15 methotrexate, procarbazine, nitrogen mustard, cytosine arabinoside and etoposide. Other compatible chemotherapeutic agents include, but are not limited to, 2 chlorodeoxyadenosine (2-CDA), 2'-deoxycoformycin and fludarabine. For patients with intermediate- and high-grade NHL, who fail to achieve remission or relapse, salvage therapy is used. Salvage therapies employ drugs such as cytosine 20 arabinoside, cisplatin, etoposide and ifosfamide given alone or in combination. In relapsed or aggressive forms of certain neoplastic disorders the following protocols are often used: IMVP- 16 (ifosfamide, methotrexate and etoposide), MIME (methyl-gag, ifosfamide, methotrexate and etoposide), DHAP (dexamethasone, high dose cytarabine and cisplatin), ESHAP (etoposide, methylpredisolone, HD cytarabine, cisplatin), CEPP(B) 25 (cyclophosphamide, etoposide, procarbazine, prednisone and bleomycin) and CAMP (lomustine, mitoxantrone, cytarabine and prednisone) each with well known dosing rates and schedules. The amount of chemotherapeutic agent to be used in combination with the polypeptides of the instant invention may vary by subject or may be administered 30 according to what is known in the art. See for example, Bruce A Chabner et aL., Antineoplastic Agents, in GOODMAN & GILMAN'S THE PHARMACOLOGICAL BASIS OF THERAPEUTICS 1233-1287 ((Joel G. Hardman et aL., eds., 9h ed. 1996). In one embodiment, a binding molecule of the invention may be administered to 179 WO 2007/109254 PCT/US2007/006883 a subject who has undergone, is undergoing, or will undergo a surgical procedure, e.g., to remove a primary tumor, a metastasis or precancerous growth or tissue as a preventative therapy. In another embodiment, a binding molecule of the invention is administered in 5 conjunction with a biologic. Biologics useful in the treatment of cancers are known in the art and a binding molecule of the invention may be administered, for example, in conjunction with such known biologics. For example, the FDA has approved the following biologics for the treatment of breast cancer: Herceptin* (trastuzumab, Genentech Inc., South San Francisco, CA; a 10 humanized monoclonal antibody that has antitumor activity in HER2-positive breast cancer); Faslodex* (fulvestrant, AstraZeneca Pharmaceuticals, LP, Wilmington, DE; an estrogen-receptor antagonist used to treat breast cancer); Arimidex* (anastrozole, AstraZeneca Pharmaceuticals, LP; a nonsteroidal aromatase inhibitor which blocks aromatase, an enzyme needed to make estrogen); Aromasin* (exemestane, Pfizer Inc., 15 New York, NY; an irreversible, steroidal aromatase inactivator used in the treatment of breast cancer); Femara* (letrozole, Novartis Pharmaceuticals, East Hanover, NJ; a nonsteroidal aromatase inhibitor approved by the FDA to treat breast cancer); and Nolvadex* (tamoxifen, AstraZeneca Pharmaceuticals, LP; a nonsteroidal antiestrogen approved by the FDA to treat breast cancer). Other biologics with which the binding 20 molecules of the invention may be combined include: AvastinTM (bevacizumab, Genentech Inc.; the first FDA-approved therapy designed to inhibit angiogenesis); and Zevalin* (ibritumomab tiuxetan, Biogen Idec, Cambridge, MA; a radiolabeled monoclonal antibody currently approved for the treatment of B-cell lymphomas). In addition, the FDA has approved the following biologics for the treatment of 25 colorectal cancer: AvastinTM ;ErbituxTM (cetuximab, ImClone Systems Inc., New York, NY, and Bristol-Myers Squibb, New York, NY; is a monoclonal antibody directed against the epidermal growth factor receptor (EGFR)); Gleevec* (imatinib mesylate; a protein kinase inhibitor); and Ergamisol® (levamisole hydrochloride, Janssen Pharmaceutica Products, LP, Titusville, NJ; an immunomodulator approved by the FDA 30 in 1990 as an adjuvant treatment in combination with 5-fluorouracil after surgical resection in patients with Dukes' Stage C colon cancer). For use in treatment of Non-Hodgkin's Lymphomas currently approved therapies include: Bexxar* (tositumomab and iodine 1-131 tositumomab, GlaxoSmithKline, 180 WO 2007/109254 PCT/US2007/006883 Research Triangle Park, NC; a multi-step treatment involving a mouse monoclonal antibody (tositumomab) linked to a radioactive molecule (iodine 1-131)); Intron* A (interferon alfa-2b, Schering Corporation, Kenilworth, NJ; a type of interferon approved for the treatment of follicular non-Hodgkin's lymphoma in conjunction with 5 anthracycline-containing combination chemotherapy (e.g., cyclophosphamide, doxorubicin, vincristine, and prednisone [CHOP])); Rituxan* (rituximab, Genentech Inc., South San Francisco, CA, and Biogen Idec, Cambridge, MA; a monoclonal antibody approved for the treatment of non-Hodgkin's lymphoma; Ontak* (denileukin diftitox, Ligand Pharmaceuticals Inc., San Diego, CA; a fusion protein consisting of a 10 fragment of diphtheria toxin genetically fused to interleukin-2); and Zevalin* (ibritumomab tiuxetan, Biogen Idec; a radiolaebeled monoclonal antibody approved by the FDA for the treatment of B-cell non-Hodgkin's lymphomas). For treatment of Leukemia, exemplary biologics which may be used in combination with the binding molecules of the invention include Gleevec*; Campath* 15 1 H (alemtuzumab, Berlex Laboratories, Richmond, CA; a type of monoclonal antibody used in the treatment of chronic Lymphocytic leukemia). In addition, Genasense (oblimersen, Genta Corporation, Berkley Heights, NJ; a BCL-2 antisense therapy under development to treat leukemia may be used (e.g., alone or in combination with one or more chemotherapy drugs, such as fludarabine and cyclophosphamide) may be 20 administered with the claimed binding molecules. For the treatment of lung cancer, exemplary biologics include TarcevaTM (erlotinib HCL, OSI Pharmaceuticals Inc., Melville, NY; a small molecule designed to target the human epidermal growth factor receptor I (HERl) pathway). For the treatment of multiple myeloma, exemplary biologics include Velcade* 25 Velcade (bortezomib, Millennium Pharmaceuticals, Cambridge MA; a proteasome inhibitor). Additional biologics include Thalidomid* (thalidomide, Clegene Corporation, Warren, NJ; an immunomodulatory agent and appears to have multiple actions, including the ability to inhibit the growth and survival of myeloma cells and antiangiogenesis). 30 Other exemplary biologics include the MOAB IMC-C225, developed by ImClone Systems, Inc., New York, NY. In addition, the claimed binding molecules may be administered in conjunction with vaccines or other agents (e.g., cytokines) to modulate anti-cancer immune 181 WO 2007/109254 PCT/US2007/006883 responses. For example, Melacine* (Corixa Corporation, Seattle, WA) is an allogeneic tumor vaccine that has been reported to have promising results in the treatment of T3NOMO resected melanoma. GMK* (Progenics Pharmaceutical, Inc., Tarrytown, NY) is a ganglioside antigen administered as an adjuvant phase III agent in patients who are 5 at high risk for melanoma recurrence. Anti-gastrin therapeutic vaccine* (Aphton Corporation, Miami, FL) neutralizes hormones G 17 and glyextened and is in phase III clinical trials for patients with colorectal, pancreatic, and stomach cancers. CeaVac* (Titan Pharmaceuticals, Inc., South San Francisco, CA) is an anti-idiotype antibody vaccine being studied in colorectal cancer. Finally, Theratope* (Biomira Inc., 10 Edmonton, Alberta, Canada) is a synthetic carbohydrate therapeutic vaccine being investigated as a phase III agent in patients with metastatic breast cancer (Pharmaceutical Research and Manufacturers of America, 2000). In another embodiment, a binding molecule of the invention may be administered in conjunction with an anti-angiogenic agent, e.g., Endostatin (an 15 endogenous, tumor-derived, endothelial-specific inhibitor that halts microvascular endothelial cell production); anti-VEGF antibody; thalidomide; or matrix metalloproteinase inhibitors inhibit the synthesis and degradation of the basement membrane of blood vessels). 20 As previously discussed, the binding molecules of the present invention, immunoreactive. fragments or recombinants thereof may be administered in a pharmaceutically effective amount for the in vivo treatment of mammalian disorders. In this regard, it will be appreciated that the disclosed binding molecules will be formulated so as to facilitate administration and promote stability of the active agent. Preferably, 25 pharmaceutical compositions in accordance with the present invention comprise a pharmaceutically acceptable, non-toxic, sterile carrier such as physiological saline, non toxic buffers, preservatives and the like. For the purposes of the instant application, a pharmaceutically effective amount of a binding molecule of the invention, conjugated or unconjugated to a therapeutic agent, shall be held to mean an amount sufficient to 30 achieve effective binding to a target and to achieve a benefit, e.g., to ameliorate symptoms of a disease or disorder or to detect a substance or a cell. In the case of tumor cells, the polypeptide will be preferably be capable of interacting with selected immunoreactive antigens on neoplastic or immunoreactive cells and provide for an increase in the death of those cells. Of course, the pharmaceutical compositions of the 182 WO 2007/109254 PCT/US2007/006883 present invention may be administered in single or multiple doses to provide for a pharmaceutically effective amount of the polypeptide. In keeping with the scope of the present disclosure, the polypeptides of the invention may be administered to a human or other animal in accordance with the 5 aforementioned methods of treatment in an amount sufficient to produce a therapeutic or prophylactic effect. The polypeptides of the invention can be administered to such human or other animal in a conventional dosage form prepared by combining the binding molecule of the invention with a conventional pharmaceutically acceptable carrier or diluent according to known techniques. It will be recognized by one of skill in 10 the art that the form and character of the pharmaceutically acceptable carrier or diluent is dictated by the amount of active ingredient with which it is to be combined, the route of administration and other well-known variables. Those skilled in the art will further appreciate that a cocktail comprising one or more species of polypeptides according to the present invention may prove to be particularly effective. 15 XIV. Methods of Use The molecules of the invention can be used in circumstances where it is desirable to use stabilized scFv molecules or compositions comprising such scFv molecules, e.g., for diagnostic or therapeutic purposes. Preferred embodiments of the present invention 20 provide compounds, compositions, kits and methods for the diagnosis and/or treatment of disorders that would benefit from administration of a binding molecule of the invention, e.g., neoplastic disorders in a mammalian subject in need of such treatment. Preferably, the subject is a human. In one embodiment, the subject binding molecules may be used in an assay to 25 detect a tumor antigen in vitro, e.g., using an ELISA assay. Exemplary assays are known in the art, see, e.g., United States Application Number 20040077025. In another embodiment, the subject binding molecules are useful for detecting the presence of tumor antigen bearing cells using imaging technology. For such applications, it may be desirable to conjugate the binding molecule to a detectable 30 molecule, e.g., a radiolabel, as described further below. In another embodiment, the subject binding molecules are useful for reducing or eliminating cells (e.g. by apoptosis) bearing an epitope (e.g., an epitope of Cripto or an epitope of a TNF receptor family member, eg. TRAIL-R2 or LTQR) recognized by a 183 WO 2007/109254 PCT/US2007/006883 binding molecule of the invention. In another embodiment, the subject binding molecules are effective in reducing the concentration of or eliminating soluble target molecules in the circulation. In another embodiment, a binding molecule of the invention reduces tumor size, 5 inhibits tumor growth and/or prolongs the survival time of a tumor-bearing subject. Accordingly, this invention also relates to a method of treating tumors in a human or other animal by administering to such human or animal an effective, non-toxic amount of polypeptide. One skilled in the art would be able, by routine experimentation, to determine what an effective, non-toxic amount of polypeptide would be for the purpose 10 of treating malignancies. For example, a therapeutically active amount of a polypeptide may vary according to factors such as the disease stage (e.g., stage I versus stage IV), age, sex, medical complications (e.g., immunosuppressed conditions or diseases) and weight of the subject, and the ability of the binding molecule to elicit a desired response in the subject. The dosage regimen may be adjusted to provide the optimum therapeutic 15 response. For example, several divided doses may be administered daily, or the dose may be proportionally reduced as indicated by the exigencies of the therapeutic situation. Generally, however, an effective dosage is expected to be in the range of about 0.05 to 100 milligrams per kilogram body weight per day and more preferably from about 0.5 to 10, milligrams per kilogram body weight per day. 20 For purposes of clarification "mammal" refers to any animal classified as a mammal, including humans, domestic and farm animals, and zoo, sports, or pet animals, such as dogs, horses, cats, cows, etc. Preferably, the mammal is human. "Treatment" refers to both therapeutic treatment and prophylactic or preventative measures. Those in need of treatment include those already with the disease or disorder as well as those in which the 25 disease or disorder is to be prevented. Hence, the mammal may have been diagnosed as having the disease or disorder or may be predisposed or susceptible to the disease. In general, the disclosed compositions may be used to prophylactically or therapeutically. For example, a neoplasm comprising a marker that allows for the targeting of the cancerous cells by the binding molecule may be detected or inhibited 30 (e.g., killed) using a binding molecule of the invention. In a preferred embodiment, the binding molecules of the invention are used to treat solid tumors. Exemplary cancers that may be treated include, but are not limited to, prostate, gastric carcinomas such as colon, skin, breast, ovarian, lung and pancreatic cancer. In another embodiment, the 184 WO 2007/109254 PCT/US2007/006883 antibodies of the instant invention may be used to treat Kaposi's sarcoma, CNS neoplasias (capillary hemangioblastomas, meningiomas and cerebral metastases), melanoma, gastrointestinal and renal sarcomas, rhabdomyosarcoma, glioblastoma (preferably glioblastoma multiforme), leiomyosarcoma, retinoblastoma, papillary 5 cystadenocarcinoma of the ovary, Wilm's tumor or small cell lung carcinoma. It will be appreciated that appropriate polypeptides may be derived for tumor associated molecules related to each of the forgoing neoplasias without undue experimentation in view of the instant disclosure. Exemplary hematologic malignancies that are amenable to treatment with the 10 disclosed invention include Hodgkins and non-Hodgkins lymphoma as well as leukemias, including ALL-L3 (Burkitt's type leukemia), chronic lymphocytic leukemia (CLL) and monocytic cell leukemias. It will be appreciated that the compounds and methods of the present invention are particularly effective in treating a variety of B-cell lymphomas, including low grade/ follicular non-Hodgkin's lymphoma (NHL), cell 15 lymphoma (FCC), mantle cell lymphoma (MCL), diffuse large cell lymphoma (DLCL), small lymphocytic (SL) NHL, intermediate grade/ follicular NHL, intermediate grade diffuse NHL, high grade immunoblastic NHL, high grade lymphoblastic NHL, high grade small non-cleaved cell NHL, bulky disease NHL and Waldenstrom's Macroglobulinemia. It should be clear to those of skill in the art that these lymphomas 20 will often have different names due to changing systems of classification, and that patients having lymphomas classified under different names may also benefit from the combined therapeutic regimens of the present invention. In addition to the aforementioned neoplastic disorders, it will be appreciated that the disclosed invention may advantageously be used to treat additional malignancies bearing compatible tumor 25 associated molecules. In one embodiment, a binding molecule of the invention is capable of binding specifically to a tumor cell antigen and inhibiting growth of tumor cells in a patient. In certain embodiments, the tumor cells are brain, head, neck, prostate, breast, testicular, colon, lung, ovary, bladder, uterine, cervical, pancreatic and stomach tumor cells. In 30 other embodiments, a binding molecule of the invention binds specifically to the tumor cell antigen and inhibits growth of tumor cells which overexpress the antigen. In one embodiment, the tumor cells are cell lines which overexpress the antigen, such as cell lines derived from brain, breast, testicular, colon, lung, ovary, bladder, uterine, cervical, 185 WO 2007/109254 PCT/US2007/006883 pancreatic and stomach cancers. In yet other embodiments the binding molecules of the present invention may be used to treat immune disorders that include, but are not limited to, allergic bronchopulmonary aspergillosis; Allergic rhinitis Autoimmune hemolytic anemia; 5 Acanthosis nigricans; Allergic contact dermatitis; Addison's disease; Atopic dermatitis; Alopecia areata; Alopecia universalis; Amyloidosis; Anaphylactoid purpura; Anaphylactoid reaction; Aplastic anemia; Angioedema,- hereditary; Angioedema, idiopathic; Ankylosing spondylitis; Arteritis, cranial; Arteritis, giant cell; Arteritis, Takayasu's; Arteritis, temporal; Asthma; Ataxia-telangiectasia; Autoimmune oophoritis; 10 Autoimmune orchitis; Autoimmune polyendocrine failure; Behcet's disease; Berger's disease; Buerger's disease; bronchitis; Bullous pemphigus; Candidiasis, chronic mucocutaneous; Caplan's syndrome; Post-myocardial infarction syndrome; Post pericardiotomy syndrome; Carditis; Celiac sprue; Chagas's disease; Chediak-Higashi syndrome; Churg-Strauss disease; Cogan's syndrome; Cold agglutinin disease; CREST 15 syndrome; Crohn's disease; Cryoglobulinemia; Cryptogenic fibrosing alveolitis; Dermatitis herpetifomis; Dermatomyositis; Diabetes mellitus; Diamond-Blackfan syndrome; DiGeorge syndrome; Discoid lupus erythematosus; Eosinophilic fasciitis; Episcleritis; Drythema elevatum diutinum; Erythema marginatum; Erythema multiforme; Erythema nodosum; Familial Mediterranean fever; Felty's syndrome; 20 Fibrosis pulmonary; Glomerulonephritis, anaphylactoid; Glomerulonephritis, autoimmune; Glomerulonephritis, post-streptococcal; Glomerulonephritis, post transplantation; Glomerulopathy, membranous; Goodpasture's syndrome; Granulocytopenia, immune-mediated; Granuloma annulare; Granulomatosis, allergic; Granulomatous myositis; Grave's disease; Hashimoto's thyroiditis; Hemolytic disease of 25 the newborn; Hemochromatosis, idiopathic; Henoch-Schoenlein purpura; Hepatitis, chronic active and chronic progressive; Histiocytosis X; Hypereosinophilic syndrome; Idiopathic thrombocytopenic purpura; Job's syndrome; Juvenile dermatomyositis; Juvenile rheumatoid arthritis (Juvenile chronic arthritis); Kawasaki's disease; Keratitis; Keratoconjunctivitis sicca; Landry-Guillain-Barre-Strohl syndrome; Leprosy, 30 lepromatous; Loeffler's syndrome; lupus; Lyell's syndrome; Lyme disease; Lymphomatoid granulomatosis; Mastocytosis, systemic; Mixed connective tissue disease; Mononeuritis multiplex; Muckle-Wells syndrome; Mucocutaneous lymph node syndrome; Mucocutaneous lymph node syndrome; Multicentric reticulohistiocytosis; 186 WO 2007/109254 PCT/US2007/006883 Multiple sclerosis; Myasthenia gravis; Mycosis fungoides; Necrotizing vasculitis, systemic; Nephrotic syndrome; Overlap syndrome; Panniculitis; Paroxysmal cold hemoglobinuria; Paroxysmal nocturnal hemoglobinuria; Pemphigoid; Pemphigus; Pemphigus erythematosus; Pemphigus foliaceus; Pemphigus vulgaris; Pigeon breeder's 5 disease; Pneumonitis, hypersensitivity; Polyarteritis nodosa; Polymyalgia rheumatic; Polymyositis; Polyneuritis, idiopathic; Portuguese familial polyneuropathies; Pre eclampsia/eclampsia; Primary biliary cirrhosis; Progressive systemic sclerosis (Scleroderma); Psoriasis; Psoriatic arthritis; Pulmonary alveolar proteinosis; Pulmonary fibrosis, Raynaud's phenomenon/syndrome; Reidel's thyroiditis; Reiter's syndrome, 10 Relapsing polychrondritis; Rheumatic fever; Rheumatoid arthritis; Sarcoidosis; Scleritis; Sclerosing cholangitis; Serum sickness; Sezary syndrome; Sjogren's syndrome; Stevens Johnson syndrome; Still's disease; Subacute sclerosing panencephalitis; Sympathetic ophthalmia; Systemic lupus erythematosus; Transplant rejection; Ulcerative colitis; Undifferentiated connective tissue disease; Urticaria, chronic; Urticaria, cold; Uveitis; 15 Vitiligo; Weber-Christian disease; Wegener's granulomatosis and Wiskott-Aldrich syndrome. In another embodiment, the binding molecules of the invention can be used for pretargeting applications. For example, the same advantages will be apparent in pretargeting applications for chemotherapeutic drug delivery. 20 For example, in pretargeting a tumor is pretargeted with a binding construct that has affinity for the tumor-associated antigen on the one hand and for, e.g., a radiolabeled hapten on the other. The radiolabeled hapten is administered later, preferably after the binding construct that has affinity for the tumor-associated antigen has cleared (see, e.g., Boerman et al. 2003. J. Nuclear Med. 44:400). In another example, an antibody which 25 is non-toxic, but has been derivitized to react with a drug or prodrug that is toxic only when bound by the binding molecule. Given the biodistribution data in the instant examples, the binding molecules of the invention are well suited to use in pretargeting applications. In one embodiment, a clearing agent could be eliminated from the pretargeting methodology by using the instant binding molecules. 30 This invention is further illustrated by the following examples which should not be construed as limiting. The contents of all references, patents and published patent applications cited throughout this application are incorporated herein by reference. 187 WO 2007/109254 PCT/US2007/006883 EXAMPLES Throughout the examples, the following materials and methods were used unless otherwise stated. 5 General Materials and Methods In general, the practice of the present invention employs, unless otherwise indicated, conventional techniques of chemistry, biophysics, molecular biology, recombinant DNA technology, immunology (especially, e.g., antibody technology), and 10 standard techniques in electrophoresis. See, e.g., Sambrook, Fritsch and Maniatis, Molecular Cloning: Cold Spring Harbor Laboratory Press (1989); Antibody Engineering Protocols (Methods in Molecular Biology), 510, Paul, S., Humana Pr (1996); Antibody Engineering: A Practical Approach (Practical Approach Series, 169), McCafferty, Ed., Irl Pr (1996); Antibodies: A Laboratory Manual, Harlow et al., C.S.H.L. Press, Pub. 15 (1999); and Current Protocols in Molecular Biology, eds. Ausubel et al., John Wiley & Sons (1992). Expression Constructs In general, unless otherwise indicated, the expression constructs for scFvs in the 20 following Examples included an N-terminal Gene III signal peptide as well as a C terminal purification peptide comprising Myc and His tags and an Enterokinase cleavage site. DNA sequence for each peptide are set forth below: N-terminal Gene III signal peptide DNA sequence 25 ATGAAAAAACTGCTGTTCGCGATTCCGCTGGTGGTGCCGTTCTATAGCCATA GT (SEQ ID NO:25) N-terminal Gene III signal peptide DNA sequence MKKLLFAIPLVVPFYSHS (SEQ ID NO:26) 30 C-terminal purification peptide DNA Sequence GACGACGACGACAAAAGCTTTCTAGAACAAAAACTCATCTCAGAAGAGGAT CTGAATAGCGCCGTCGACCATCATCATCATCATCATTGA (SEQ ID NO:27) 188 WO 2007/109254 PCT/US2007/006883 C-terminal purification peptide DNA Sequence DDDDKSFLEQKLISEEDLNSAVDHHHHHH (SEQ ID NO:58) 5 Antibodies BIIBAntibodies used in certain Examples (BIIB1-BIIB18) are a collection of therapeutic antibodies of various specificities. Seventeen of the 18 antibodies were expressed from either stable bulk or clonal CHO cell lines and 1 of the antibodies (BIIB 13) was expressed using transient transfection in HEK293E cells. All 18 10 antibodies contain kappa light chains. The majority of the antibodies were human IgG1; however, BIIB2, BIIB5 and BIIB 11 were human IgG4. Seventeen of the 18 antibodies were human or humanized. BIIB 15 was PRIMATIZED* (Nakamura et al., 2000). The identity of the human germline for each antibody was assessed by ClustalW alignment (ClustalW WWW Service at the European Bioinformatics Institute, Thompson et al., 15 1994) of the BIIB VH or V, sequences against the publicly available human germlines (Lefranc et al., 1999). Fifteen of the 18 total antibodies were IgGI subclass and the remaining 3 were IgG4 (BIIB2, BIIB5 and BIIB 11). IgGI and IgG4 CHI sequences have 10 amino acid differences (6 are conservative) and an alternate disulfide-bonding pattern with the light 20 chain. An IgGI and an IgG4 construct with a duplicate Fv region was available to investigate the effect of IgG subclass on Fab stability. Two constructs were created which contained the VH region of BIIB7 grafted to either an IgG I or an IgG4 heavy chain constant region. 25 Example 1. Preparation of Conventional BHA10 scFv and Fab proteins BHA10 scFv was subcloned from plasmid pXWU034, using the Polymerase Chain Reaction (PCR) with oligonucleotide primers shown in Table 2 below. The forward primer BHA 10-01 F contains a unique Sph I restriction endonuclease site 30 (underlined sequence) followed by 18 bases of sequence complementary to the BHA 10 N-terminal heavy variable domain gene. The reverse primer, BHA10-0IR, contains 24 bases of sequence complementary to the BHAI0 C-terminal light variable domain gene, 15 bases of sequence complementary to an Enterokinase Site, and adjacent Hind III and 189 WO 2007/109254 PCT/US2007/006883 Xba I restriction endonuclease sites (endonuclease sites are underlined). Following PCR amplification, a PCR product corresponding to the expected size was resolved by agarose gel electrophoresis, excised, and purified using the Millipore Ultrafree-DA extraction kit according to manufacturer's instructions (Millipore; Bedford, MA). The 5 purified PCR product was digested with Sph I, made blunt-end by digesting with DNA Polymerase I in the presence of dNTPs, and then digested with Hind III. The blunt ended/ Hind III digested PCR product was ligated to Sca I/ Hind III digested pKJS216. pKJS216 is an E. coli vector that drives recombinant protein expression under the control of an inducible ara C promoter. A portion of the ligation mixture was used to 10 transformed E. coli strain XL 1-Blue. Ampicillin drug resistant colonies were screened and DNA sequence analysis confirmed the correct sequence of the final pIEHO03 construct. DNA and amino acid sequences of BHAIO scFv are shown in Figures IA and I B, respectively. 15 Table 2. Oligonucleotides for PCR amplification of a conventional BHA10 scFv. Primers Sequence BHA1O-01F 5'- CAGTAGCATGCAGGTCCAACTGGTGCAG -3' (SEQ ID NO:1) BHA10-02R 5'-GTTCTAGAAAGCTTTTGTCGTCGTCGTCTTTGATCTC (SEQ ID NO:2) CACCTTGGTACCCTG -3' For expression of BHA10 scFv, freshly isolated colonies of E. coli strain W31 10 (ATCC, Manassas, Va. Cat. # 27325) transformed with plasmid pIEH003 were grown in 4 x 250 ml SB media (Teknova, Half Moon Bay, Ca. Cat. # S0140) containing 50 pLg/ml 20 carbenicillin in I L baffled flasks to OD 6 0 0 ~ 0.8, induced by adding to 0.02% arabinose, and cultured overnight. Bacteria were collected by centrifugation. The pellets were solubilized and lysed using 40 mL B-PER protein extraction reagent (Cat#78243, Pierce). Solubilized scFv was applied to a 5 mL Ni-NTA-Superflow column (Cat# 30410, Qiagen). Bound scFv was washed with 60 mM imidazole, pH 8.0 and eluted 25 with 300 mM imidazole, pH 8.0. Eluted scFv was loaded onto a 6 mL Protein L agarose column (Cat#20510, Pierce). Bound protein was washed with phosphate buffered saline (PBS) and eluted with 0.1 M glycine, pH 3.0. Purified scFvs were dialyzed against PBS 190 WO 2007/109254 PCT/US2007/006883 and stored at -20 *C. Protein concentrations were determined using an E280 nm = 2.1 ml in -I c -I. mg~ cm~ For enzymatic preparation of BHA10 Fab, BHA10 IgG was mixed with 4 pLl of a concentrated papain stock (0.3 mg/mL, 30 Units/mg - Cat#108014, Roche) in 8.3 ml 5 solution containing 2.4 mg/mL BHA 10 IgG1, 100 mM Tris-HCI, 20 mM EDTA at pH 7.0. The reaction was allowed to proceed for 90 minutes at 25 "C. The digest solution was diluted 1:5 with 20 mM acetate, pH 5.0 and loaded onto a 6 ml SP-Sepharose FF column equilibrated with dilution buffer. The column was washed with 2 column volumes of dilution buffer. Crude Fab fragments were eluted with 30 column volumes 10 of a 0-200 mM NaCl linear gradient. A broad peak centering at 140 mM NaCl (-24 mL total) was collected and contained the majority of the digested IgG material. The eluted volume was reduced to 2 mL by concentration and loaded onto a preparative G300 SW Tosohaas SEC column (109 mL) equilibrated with PBS. Fab was eluted at 0.8 column volumes over 15 mL. Purified Fab was concentrated to between 2-11 mg/mL. Fab 15 concentrations were determined using an e 280 nm = 1.5 mL mg' cm~ . Example 2. Thermal Stability of Conventional BHA10 scFv molecules Differential Scanning Calorimetry (DSC) was used to test whether an isolated 20 BHA10 scFv is intrinsically less stable than its Fab counterpart. Scans were performed using an automated capillary differential scanning calorimeter (capDSC, MicroCal, LLC). Protein and reference solutions were sampled automatically from 96-well plates using the robotic attachment. Prior to each protein scan, 2 scans were performed with buffer in the sample cell and used for background subtraction. A single cleaning scan 25 was performed using 5% Liquinox after every protein scan. After every scan, the instrument automatically rinsed both the reference and sample cells three times with 2 ml distilled deionized H 2 0 containing 0.01% sodium azide. Scans were performed at 1 *C/min using the medium feedback mode for enhanced peak resolution. The scan range was 20-95 *C. All 96-well plates containing protein were stored within the instrument at 30 6 *C. Figure 2 shows DSC measurements with purified BHA1O Fab and scFv antibody fragments. Within the calorimeter, the temperature is raised until the Fab or scFv unfolds. The temperature at which each protein unfolds (i.e., the Tm value) can be 191 WO 2007/109254 PCT/US2007/006883 indicative of the overall stability. All four domains of the BHA10 Fab (VH, VL, CH I and CL) unfold cooperatively at 78 *C (Figure 2A). The scFv domain lacks the CHI and CL domains. Without this scaffolding, the domains of the scFv unfold at much lower temperatures than the Fab and there is a significant decrease in the calorimetric enthalpy 5 of the domains indicating the loss of stabilizing interactions. The VL domain unfolds with a TM of 68 *C, while the VH domain unfolds at 58 "C, 20 *C lower than what was observed for the unfolding transition of the BHA 10 Fab (Figure 2B). Additionally, there is a scan rate dependence of the TM, suggesting that protein aggregation is occurring during the heating phase which artificially lowers the TM of the scFv as the scan rate is 10 slowed (SAnchez-Ruiz et al., Biochemistry, 27: 1648-1652, 1988). The low apparent stability and the propensity to aggregate may be determining factors for the inability of CHO cells to produce significant quantities of stable, non-aggregated material which contain scFvs such as the conventional Hercules bispecific antibody molecules. 15 Example 3. Construction of BHA10 scFv Molecules with Improved Thermal Stability Knowing that the BHA10 scFv domain, as evidenced in Example 2, is intrinsically unstable, it was hypothesized that engineering the scFv through the use of recombinant DNA technology to produce a modified scFv that is thermodynamically or 20 functionally equivalent to a Fab under thermal challenge conditions should result inan scFv domain that is useful for constructing a bispecific antibody. Moreover, it was also hypothesized that engineering of the isolated scFv domain by itself should impart whatever beneficial biophysical properties are gained when re-introduced as a component of a full bispecific molecule. Towards that end, an effort to improve the 25 biophysical stability of the BHA 10 scFv domain using an E. coli expression system and monitored improvements in stability by measuring binding of thermally resistant scFv domains to ligand in a thermal challenge assay was begun. To stabilize the scFv domains two methods were applied: 1) introducing a disulfide bond between the VH and VL domain of the BHA1O scFv; and 2) optimizing 30 the length of the (Gly 4 Ser)n linker that connects the VH and VL domains of the BH A10 scFv. A. Construction of disulfide-stabilized BHA10 scFvs The BHA 10 scFv producing bacterial expression vector, pIEHOO3, was utilized 192 WO 2007/109254 PCT/US2007/006883 as the parental vector. The QuickChange Site-Directed Mutagenesis Kit (Stratagene; La Jolla, CA) was used, according to the manufacturer's instructions, to introduce two cysteine residues, one in VH and a second in VL that could participate in forming a stabilizing disulfide bond. Primer pairs VH44-F and VH44-R (Table 3) were used to 5 mutagenize the Gly residue (GGA) at position 44 (Kabat numbering system) of BHA10 variable heavy chain to a Cys residue (TGC). The mutagenesis product was digested with methylation sensitive enzyme Dpn I according to the kit protocol and transformed into the E. coli strain XL1 0-GOLD* (Stratagene; La Jolla, CA). E. coli colonies transformed to ampicillin drug resistance were screened for the correct sequence 10 mutation by DNA sequence analysis. The resulting plasmid, pIEH004, was.utilized for a subsequent reaction to mutate the Gln residue (CAG) at position 100 (Kabat numbering system) of BHA10 variable light chain to a Cys residue (TGC) using primer pairs VL100-F and VLOO-R (Table 3). Forward (VH44-F) and reverse (VH44-R) primers mutate Gly (GGA) to a Cys (TGC) at VH position 44 (TGC indicated by underlined 15 sequence). Forward (VL 100-F) and reverse (VL100-R) primers mutate Gln (CAG) to a Cys (TGC) at VL position 100 (TGC indicated by underlined sequence). XL 10-GOLD* E. coli colonies transformed to ampicillin drug resistance were screened for the correct sequence mutation by DNA sequence analysis and plasmid pIEH006 was identified as containing the double cysteine mutations at positions VH44 20 and VL100. DNA and amino acid sequences of VH44/VLIOO disulfide-stabilized BHA 10 scFv are shown in Figures 3A and 3B, respectively. 193 WO 2007/109254 PCT/US2007/006883 Table 3. Oligonucleotides for construction of VH44/VL100 disulfide-stabilized BHAI0scFv. Primers Sequence VH44-F 5'- GCAGGCCCCTGGACAGTGCCTTGAGTGGATGGGATG -3' (SEQ ID NO: 5). VH44-R 5'- CATCCCATCCACTCAAGGCACTGTCCAGGGGCCTGC -3' (SEQ ID NO:6) VL100-F 5'- CCTATCCATTCACGTTCGGCTGCGGTACCAAGGTGG (SEQ ID NO: AGATC -3' 7) VLI 00-R 5'- GATCTCCACCTTGGTACCGCAGCCGAACGTGAATGG (SEQ ID NO: ATAGG -3' 18) 5 B. Construction of BHA1O scFv with alternative (Gly 4 Ser). linkers Plasmid pIEH003 encoding huBHA10 scFv with the oconventional (Gly 4 Ser) 3 linker was modified to contain a (Gly 4 Ser) 4 or (Gly 4 Ser)s linker by PCR amplification using the oligonucleotide primers described in Table 4. 10 BHA10 scFv (Gly 4 Ser) 4 was assembled using a forward 5' PCR primer designated pXWU002-F1 and a reverse 3' PCR primer designated XWU002-R. The 5' VH PCR primer XWO02-FI included a Btg I restriction endonuclease site (underlined sequence) located at the carboxyl terminus of BHA10 VH followed by sequence encoding a (Gly 4 Ser) 4 linker. The 3' VL PCR primer XWO02-R included an Xba I site 15 and a partial Enterokinase site. The partial BHA1O VH + (Gly 4 Ser) 4 linker + the BHA1O VL regions were amplified in a PCR reaction using the XWO02-F1/ XWO02-R PCR primer set from plasmid DNA plEH003 (described in Example 1). The partial BHA10 scFv- (Gly 4 Ser) 4 linker gene fragment corresponding to the expected size was resolved by agarose gel electrophoresis, excised, and purified using the Millipore Ultrafree-DA 20 extraction kit according to manufacturer's instructions (Millipore; Bedford, MA). The purified PCR product was digested and cloned into the Btg I/Xba I digested pIEH003 vector resulting in plasmid pXWU002 encoding BHA 10 scFv containing a (Gly 4 Ser) 4 194 WO 2007/109254 PCT/US2007/006883 linker. BHA 10 scFv containing the (Gly 4 Ser)s linker was constructed in similar fashion using PCR primers XWO03-F and XWO02-R to produce plasmid pXWU003. Forward 5' PCR primer (XWU003-F) contained a Btg I site (underlined sequence) followed by sequence encoding a few amino acid of the carboxyl terminus of BHA 10 VH and 5 sequence encoding a partial (Gly 4 Ser)s linker. Correct sequences were confirmed by DNA sequence analysis. DNA and amino acid sequences of BHA 10 scFv containing the (Gly 4 Ser) 4 linker are shown in Figures 4A and 4B, respectively. DNA and amino acid sequences of BHA1O scFv containing the (Gly4Ser)5 linker are shown in Figures 5A and 5B, respectively. 10 Table 4. Oligonucleotides for construction of BHA10 scFv with (Gly 4 Ser) 4 or (Gly 4 Ser)5 linkers. Primers Sequence XWO02-F1 5'- AAGGGACCACGGTCACCGTCTCCTCAGGCGGTGGAGGGTCCG (SEQ ID GTGGGGGCGGATCTGGGGGCGGCGGATCCGGTGGTGGTGGTAG-3' NO:11) XWO02-R 5'- TTTTGTTCTAGAAAACTTTTGTCGTCG-3' (SEQ ID NO:12) XWO03-F 5'- AAGGGACCACGGTCACCGTCTCCTCAGGAGGGGGCGGTTCAG (SEQ ID GCGGTGGAGGGTCCGGTGGGGGCGGATCTGGGGGCGGCGGATC-3' NO:13) 15 Example 4. Characterization of BHA10 scFv Molecules with Improved Thermal Stability A. Expression and Western blot analysis of engineered BHA1O scFvs 20 For expression of engineered BHAI 0 scFvs, E. coli strain W3 110 (ATCC, Manassas, Va. Cat. # 27325) was transformed with plasmids pIEH003, pXWU002, pXWU003 and pIEH006 and ampicillin resistant colonies selected and grown in 10 ml SB media (Teknova, Half Moon Bay, Ca. Cat. # S0140) containing 50 jig/ml carbenicillin in a 50 ml conical centrifuge tube to.D 600 = 0.8, induced by adding to 25 0.02% arabinose, and cultured overnight. Bacteria were collected by centrifugation and 195 WO 2007/109254 PCT/US2007/006883 the pellets resuspended in 1/20 volume of an ice-cold iso-osmotic solution of 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 20% sucrose (w/v) and chilled on ice. Equal volume of 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 20% sucrose (w/v) containing 2 mg/ml Lysozyme (Sigma) was added to the bacterial suspension and incubated on ice 5 with occasional mixing for 10 minutes. The bacterial suspension was centrifuged for 10 minutes at 8000 x g, 4 "C and the periplasmic fraction retained. Samples were mixed with native sample buffer or sample buffer containing the reducing agent dithiothreitol and heated at 90 0 C for 3 minutes. Reduced and non reduced samples were electrophoresed on an SDS-PAGE Tris-glycine polyacrylamide 10 gel and electrophoretically transferred onto a nitrocellulose membrane (Invitrogen, Life Technologies, Carlsbad, CA). The membrane was blocked with PBS containing 5% (w/v) non-fat milk and 0.1% Triton X-100 and incubated with an anti-human kappa antibody (Roche Applied Science, Indianapolis, IN). The membrane was washed and then incubated with an anti-rabbit HRP antibody (Amersham Biosciences, Piscataway, 15 NJ). Immune complexes were detected using the ECL Western Blotting Analysis System according to the manufacturer (Amersham Biosciences, Piscataway, NJ). It was found that one of three disulfide pairs tested, namely VH44:VLIOO, produced suitable amounts of protein when expressed in E. coli (Figure 6, lane 2). (The 20 VH 44:VL 105 and VH 106:VL 4 3 disulfides tested did not produce as much intact scFv). Similarly, extending the length of the (Gly 4 Ser)n linker to n=4 (lane 3) or n=5 (not shown) also produced suitable amounts of protein in E. coli. Combining both the VH44:VLOO and (Gly 4 Ser) 4 linker modifications into BHA1O scFv using the methods decribed in Example 3 also led to suitable amounts of expressed protein (Figure 6, lane 25 4). DNA and amino acid sequences of BHA1O scFv containing the combination of VH44:VL100 and (Gly 4 Ser) 4 linker modifications are shown in Figures 7A and 7B, respectively. In both cases where the BHA10 scFv contains the VH44:VLl00 mutations, the scFvs were found to migrate with increased mobility in the denaturing, non-reducing polyacrylamide gel, yet migrated similarly to conventional BHA10 scFv under 30 denaturing, reducing conditions (Figure 6, lanes 2 and 4). This analysis suggests that the BHA 10 scFv variants containing the VH44:VLI00 mutations are likely forming intact disulfide bonds and may be attaining a more compact structure. 196 WO 2007/109254 PCT/US2007/006883 B. Thermal denaturation assay. The activities of the conventional and engineered BHA 10 scFvs were then compared in a thermal challenge assay which can be used to determine the temperature at which 50% of scFv molecules retain their antigen binding activity following a thermal 5 challenge event. The numerical value corresponding to this temperature is referred to as the"T 50 " value and the units are in *C. In this assay, the scFvs were subjected to a range of temperatures that encompass the thermal transition temperature of conventional BHA10 scFv. E. coli strain W3 110 (ATCC, Manassas, Va. Cat. # 27325) was transformed with 10 plasmids encoding the conventional and engineered BHA10 scFvs under the control of an inducible ara C promoter. Transformants were grown overnight in expression media consisting of SB (Teknova, Half Moon Bay, Ca. Cat. # SO 140) supplemented with 1% glycine, 1% Triton X100, 0.02% arabinose, and 50 pg/ml carbenicillin at either 37"C or 32"C. Bacteria was pelleted by centrifugation and supernatants harvested for further 15 treatment. Including glycine and Triton X- 100 in the media results in the release of periplasmic contents (native E. coli protein and scFv) into the media (Yang et al. Applied and Enviromental Microbiology. (1998) 64:2869-2874). The presence of E. coli proteins in the supernatant is essential to the performance of this assay because the thermally denatured proteins act as a "sink", trapping transiently unfolded scFv 20 molecules into irreversible inactive aggregates. Each library was screened in duplicate using a thermal challenge assay with supernatant from one replicate subjected to treatment conditions and the second supernatant serving as untreated reference. Thermal denaturation assays can be run at a single or range of temperatures for measuring stability. Thermocycler machines capable 25 of generating stable thermal gradients were used for treating sample supernatants (iCycler, Bio-Rad, Gaithersburg, Md). The challenge temperature varied depending on the properties of the parental BHA 10 scFv variant and was generally two to three degrees Celsius higher than the experimentally determined T 50 value. Frozen Master plates were thawed and used to 30 inoculate deep-well microtiter plates containing 250 pL of expression media per well, and cultures grown overnight at 32 0 C. As a control, cultures containing the parental plasmid were grown under the same conditions and processed simultaneously as the library. Bacteria were pelleted and 50-100 pl aliquots of test supernatant were placed in 197 WO 2007/109254 PCT/US2007/006883 either PCR strip tubes (Applied Biosystems, Foster City, Ca, Cat. # N801-535) or 96 well plates (Applied Biosystems, Foster City, Ca, Cat. # N801-560) and the samples were heated for 60-90 minutes. Samples were transferred to 96-well v-bottomed plates (Coming, Coming, NY, Cat. # 3357) and centrifuged in a refrigerated clinical centrifuge 5 (IEC model 8R, Thermo Electron, Waltham, Ma) for 30 minutes, and 100 Jl of the supernatant was transferred to standard micotiter plates (Coming, Coming, NY, Cat. # 3357). An aliquot of the supernatant was reserved for the reference DELFIA. For most libraries, the plates containing the remainder of the supernatants were sealed (Nalge Nunc, Rochester, NY, Cat.# 235205) and placed in an incubator set to the appropriate 10 challenge temperature (Echo Therm, Torrey Pines Scientific, San Marcos, Ca) for 90 minutes. For screens requiring multiple challenge temperatures (or for temperatures greater than 75*C) the supernatants were transferred to 96-well PCR plates (Applied Biosystems, Foster City, Ca, Cat. # N801-560) and incubated for 90 minutes at the desired temperature. 15 After thermal challenge, the samples were centrifuged at 2,000 RPM to remove aggregated material. Soluble BHA10 scFv samples remaining in the treated, cleared supernatant were assayed for binding to cognate LTQR Ig antigen by DELFIA assay. 96-well plates (MaxiSorp, Nalge Nunc, Rochester, NY, Cat. # 437111) were coated with fusion protein consisting of the ectodomain of the LT p receptor (LTpR) fused to a 20 human Fc region at I ig/ml in 0.1 M sodium carbonate buffer, pH 9.5. Plates were coated overnight at 4*C, and blocked with DELFIA assay buffer (DAB, 10 mM Tris HCI, 150 mM NaCl, 20 pM EDTA, 0.5 % BSA, 0.02% Tween 20, 0.01% NaN 3 , pH 7.4) for one hour with shaking at room temperature. Plates were washed 3 times with DAB without BSA (Wash buffer), and test samples diluted in DAB were added to the plates in 25 a final volume of 100 p1. The plates were incubated for one hour with shaking at room temperature, and then washed 3 times with Wash buffer to remove unbound and functionally inactivated scFv molecules. Bound BHA1O scFv was detected by addition of 100 p1 per well of DAB containing 40 ng/ml of Eu-labeled anti-His 6 antibody (Perkin Elmer, Boston, MA, Cat. # ADO109) and incubated at room temperature with shaking 30 for one hour. The plates were washed 3 times with Wash buffer, and 100 pI of DELFIA enhancement solution (Perkin Elmer, Boston, MA, Cat. # 4001-0010) was added per well. Following incubation for 15 minutes, the plates were read using the Europium method on a Victor 2 (Perkin Elmer, Boston, MA). 198 WO 2007/109254 PCT/US2007/006883 Assay data was processed using Spotfire DecisionSite software (Spotfire, Somerville, Ma.) and expressed as the ratio of the DELFIA counts observed at challenge temperature to the reference temperature for each clone. Clones that reproducibly gave ratios greater than or equal to twice what was observed for the parental plasmid were 5 considered hits. Plasmid DNAs from these positive clones were isolated by mini-prep (Wizard Plus, Promega, Madison, WI) and retransformed back into E. coli W3 110 for confirmation secondary thermal challenge assays. For thermal gradients, the data was analyzed using Prism 4 software (GraphPad Software, San Diego, Ca) using a sigmoidal dose response with variable slope as the 10 model. The values obtained for the mid-point of the thermal denaturation curves are referred to as T 5 0 values, and are not construed as being equivalent to biophysically derived Tm values. The results of the thermal challenge assay are depicted in Figure 8. As depicted in Figure 8, all of the stabilized scFv molecules of the invention resulted in 15 improvements in binding activity (Tso >49*C) as compared with the conventional scFv. In particular, the T 50 values of BHA10 library position VL46 scFv (S46L), library position V H16 scFv (S16E and S16Q), and library positions VL49: VL50 scFv exhibited increases in thermal stability ranging from + 3*C to + 12*C relative to the conventional BHA10 scFv. In addition, the T 50 values of BHA 10 library position VL3 scFv (Q3A, 20 Q3G, Q3S, Q3V, and Q3D), library position VH67 scFv (V671 and V67L), library position VH48 scFv (M481 and M48G), library position VH20 scFv (V20I) and library position VH 101 scFv (P101 D) exhibited increases in thermal stability ranging from + 4* C to + 18*C relative to the conventional BHA10 scFv. One of the stabilizing mutations (VL K13E) serendipitously resulted from a PCR error. Incorporation of one of these 25 stabilizing mutations into pIEHOO9 was found to further improve thermal stability and even exceed that of BHA10 Fab under these conditions. Importantly, the non-covalent VL46, VH101 mutation, and VH55 mutations derived from one or more of the four design methods (VLI/VH interface homology modeling, consensus scoring, computational modeling, and covariation analysis) resulted in an improvement in scFv thermal stability 30 nearly approaching that observed with the disulfide mutations and validating the utility and novelty of these design tools. In particular, the T 50 of the pIEH006 construct, BHAIO Vs44:VLl00 scFv (59*C), differed from BHA10 Fab (62*C) by only 3 *C, while the pIEH009 construct, BHA10 scFv VH44:VLOO/(Gly4Ser)4 linker, was found to be 199 WO 2007/109254 PCT/US2007/006883 functionally equivalent to BHA 10 Fab under these conditions. These results demonstate that the stabilized scFvs of the invention have improved activity following a thermal challenge event. 5 c. Affinity Measurements Isothermal titration calorimetry (ITC) was used to measure the affinity of sLTpR to the BHA10 Fab derived using enzymatic cleavage of the BHA10 IgGI molecule. sLTsR was prepared as described previously (Eldredge et al., Biochemistry, (2006)). The Fab and sLTpR were concentrated to 6.0 and 2.0 mg/mL, respectively, using 10 Amicon ultracentrifugal filter devices (MWCO 10,000). The concentrated stock solutions were simultaneously dialyzed against PBS prior to the ITC measurements. ITC was performed on a VP-ITC unit (MicroCal LLC, Northampton, MA) set to 30 "C. Approximately 500 pL of a 7 pM sLTpR solution was placed into the sample cell and PBS dialysate was placed in the reference cell. A total of 234 pL of 70 pM BHA10 Fab 15 was titrated into the sample cell in 7 x 10 piL, 12 x 7 pL, followed by 8 x 10 pL injections. The reaction stoichiometry was 1:1. ITC curves were analysed using the Origin Software supplied by the manufacturer. KD measurements of conventional BHAI0 scFv, BHA10 (Gly 4 Ser)4 scFv, and BHA 10 VH44:VLI00/(Gly 4 Ser) 4 scFv preparations as described in Examples 3 and 20 expressed and purified using the methods described in Example 1 were performed using surface plasmon resonance (SPR) on a Biacore 3000 instrument (Biacore Inc., Piscataway, NJ). All experiments were performed in HBS-EP buffer, pH 7.4. Biotinylated PENTA-His antibody (20 pg/mL, Cat#34440, Qiagen) was immobilized onto a streptavidin-coated CM5 chip at a flow rate of 10 pl/min for approximately 1 25 minute. 0.1 pM solutions of test scFvs were injected over the chip at a flow rate of 5 pl/min for 10 minutes and captured onto the surface via the immobilized PENTA-His antibody. To investigate binding between sLTpR and captured scFv, a concentration series of sLTsR (1, 2, 5, 10, 25, 50, 100, and 200 nM) was double injected at a flow rate of 30 pl/min onto the scFv-coated surface. Background was subtracted from test sample 30 sensorgrams using sensorgram data from a flow cell containing buffer only and sensorgram data from the PENTA-His surface where scFv was injected followed by an injection of buffer instead of sLTsR. The curves were analyzed using the BiaEval 3.0 200 WO 2007/109254 PCT/US2007/006883 manufacturer's software. KD values were calculated by fitting kinetic association and dissociation curves to a 1:1 Langmuir binding model. Chip surfaces were regenerated by two consecutive 10 pL injections of 0.1 M glycine, pH 3.0. Table 5 shows the results of the Biacore affinity assay. The binding affinity of 5 recombinantly produced BHA 10 scFv is essentially the same as the BHA 10 Fab prepared as described in Example 1. Additionally, introduction of the (G 4
S)
4 linker alone or in the presence of the stabilizing disulfide (VH44:VL100), or the combination of the two did not result in any significant loss of affinity towards antigen. 10 Table 5. Biacore affinity assay-binding to LTpR. Kd (nM) Construct BHA10 scFv 3.7 BHA 10 (Gly 4 Ser) 4 scFv 2.5 BHA 10 VH44:VLl00/(Gly4Ser)4 scFv 4.3 BHA 10 Fab (CHO) 1.9±1.0* *Measured by isothermal titration calorimetry D. Differential Scanning Calorimetry Studies Differential Scanning Calorimetry (DSC) analyses were performed with BHA10 15 scFv, BHA10 (Gly 4 Ser) 4 scFv, and BHA10 VH44:VLI00/(Gly4Ser)4 scFv preparations. Experiments were performed as described above for the initial comparisons of the conventional scFv against the enzymatically produced Fab except they were performed at a scan rate of 4 "C/min. BHA10 VH44:VL100/(Gly4Ser)4 scFv demonstrated superior thermostability properties compared to the wild-type scFv. In particular, the less stable 20 of the two domains, the VH domain of BHA10 VH44:VL100/(Gly4Ser) 4 scFv, denatured at approximately 5 *C higher than that of conventional BHA 10 scFv (Figure 9). Increases in the "melting temperature" or Tm could be due to two separate factors, (i) an increase in the thermostability of the equilibrium folded state or (ii) a decreased tendency to aggregate. The ~ 5 *C temperature difference between the conventional 25 BHA10 scFv and BHA10 VH44:VL100/(Gly4Ser)4 scFv was relatively independent of 201 WO 2007/109254 PCT/US2007/006883 scan rate (between 0.2 and 4.0 *C/min) suggesting the observed Tm change was due to stabilization of the equilibrium folded state. E. ANS Binding Studies 5 The ability to bind the hydrophobic fluorescent dye 1 -anilino-8-naphthaline sulfonate (ANS) is often a hallmark of interactions with significantly large hydrophobic regions either in a protein's native state or partially unfolded state that can occur upon increased treatment with temperature. The intrinsic fluorescence of the dye is quenched in solvent and significantly increases when exposed to large hydrophobic surface areas. 10 The conventional BHA10 scFv intrinsically bound ANS at a much stronger level than the BHA10 (Gly 4 Ser) 4 scFv, or BHA10 VH44:VLl00/(Gly4Ser)4 scFv indicating the presence of hydrophobic exposure forced upon the scFv by an inadequately sized linker. Addition of the longer linker, specifically (Gly 4 Ser)4, appeared to mediate this effect. The apparent exposure of a hydrophobic surface by the conventional BHA 10 scFv may 15 lead to an increased level of aggregation in the presence of other proteins or molecules, or even in the presence of itself in isolation. Heating BHA1O scFv above its TM appeared to induce ANS binding for the conventional and engineered BHA 10 scFvs. Interestingly, the conventional BHA 10 scFv showed a gradual increase in ANS binding, as a function of increasing temperature, suggesting possible increases in hydrophobic 20 exposure at temperatures below the TM - such as temperatures used in bacterial and mammalian cell culture (i.e. 37 *C) (Figure 10). In contrast, both BHA1O (Gly 4 Ser) 4 and BHA 10 VH44:VL100/(Gly 4 Ser) 4 scFvs did not exhibit this property suggesting that the stabilizing mutations described in this invention decrease the tendency to self-associate or associate with other cell-culture protein components based on reduced exposure of 25 hydrophobic surface areas. Example 5. Identification of Stabilizing Mutations which confer Improved Intrinsic Protein Stability 30 (i) Identification of Stabilizing Mutations A variety of sequence-based methods (e.g., consensus scoring, covariation analysis, VH/VL interface homology modeling) were used to identify stabilizing mutations which conferred improved protein stability to binding molecules of interest. 202 WO 2007/109254 PCT/US2007/006883 These stabilizing mutations were used in the design and construction of antibody variable region expression libraries. A. Covariation Analysis A number of designs were developed for stabilizing the BHA 10 scFv VH and VL 5 domains based on the strong covariation exhibited by two or more residues within a single sequence. Covariation analysis was performed on the BHA1O scFv VH and VL domains using methods similar to those described in Example 17 infra in order to identify missing or violative covariations such that stabilizing mutations could be predicted and included in an antibody variable region expression library for 10 experimental screening. In a first example, mutation from Ser to Leu at position 46 (S46L; Kabat numbering system) within the BHA10 VL exhibited a positive connection to an existing Tyr at residue 36 which the Covariation Analysis Tool showed to covary strongly with Leu 46 (see Figure 78). This mutation was also predicted to be stabilizing by residue 15 frequency analysis. In addition to S46L, a second mutation predicted by covariation to be stabilizing was a V55G mutation within the BHAlO VH domain. While residue frequency indicated that Val is infrequently observed at this position, this position is embedded within CDR2 and is variable. Therefore, without additional information, no changes at this position 20 were previously attempted. Upon inspection by covariation analysis, however, the covariation data suggested that this position was strongly correlated with at least 10 other amino acids that already exist within the BHA10 VH domain. Mutation to Gly at position 55 satisfied all 10 covariations. Another example of the utility of the Covariation Analysis Tool was the 25 predicted negative effect of a BHA 10 Vi Q6E mutation. Single residue frequency analysis suggested that mutation to the much more commonly observed Glu at this position would lead to an increase in stability. However, the Covariation Analysis Tool indicated that single mutation to Glu violates several existing covariations present within the BHAlO VH sequence. To obtain an improvement in stability, one must replace 30 several amino acids that preferentially stabilize Glu at this position. Yet, another example of the predictive value of the Covariation Analysis Tool is shown in Figure 79. Met 80 was mutated to Leu as part of a single residue library design. It was thought that this single mutation would be highly stabilizing, as Leu is 203 WO 2007/109254 PCT/US2007/006883 the most frequent amino acid observed at this position within the sequence database. However, covariation analyses indicate that two other amino acids must be mutated (V67F and T70S) in order to achieve covariation harmony. 5 B. Consensus Scoring Consensus scoring was utilized as a method for identifying amino acid residues within the scFv VH and VL regions for mutagenesis to improve the intrinsic stability of the scFv. The scoring assesses the relative drift from consensus VH and VK sequences due to hypersomatic mutations and evolutionary germline variations. Information 10 derived from this analysis was then used to design a library to screen for scFv variants with improved stability. The reference set of mammalian VH and V, kappa sequences used to derive the consensus sequence for scoring and the individual amino acid frequencies at each residue position were collected, sorted, and culled as described previously (Demarest, et 15 al., J. MoL. Bio. 335, 41-48 (2004); Demarest, et aL., Protein Eng. Des. SeL. In Press, (2006). The mammalian reference sets were naively constructed to include V-genes from various mammals in order to obtain diversity via the evolutionary drift between species. The VH mammalian reference set contains 61 VH sequences primarily from NCBI and TIGR representing a total of 17 different mammalian species. The V, 20 mammalian reference set contains 53 VK sequences from 13 different mammalian species. Statistical analysis of the BHA 10 VH and VL were performed using custom designed IgG databases and a modified PERL script (Demarest et aL., 2004; Demarest et aL., 2006). The amino acid frequency of every residue within the BHA10 VH and VL was 25 calculated from the custom database. The residue frequency of each amino acid within the BH A10 VH and VL, S;(r), for each position, i, in an individual sequence was calculated by the number of times that particular residue-type (r = A,C,D... V,W,Y) is observed within the data set divided by the total number of sequences. Figure 11 shows the BHA10 VH and VL residue frequencies divided by the residue frequency of the 30 database consensus residues. By dividing by the consensus residue frequency, a stringent cutoff for the creation of libraries at residue positions where the BHA 10 amino acid is infrequently observed among common VH or VL sequences is obtained. Library positions were determined by those whose residue frequency divided by the most 204 WO 2007/109254 PCT/US2007/006883 frequent residue frequency (S;(r)/MFRi(r)) was < 0.3. The residues listed to the right of the residue frequency calculations are those commonly found in human sequences as published (Chothia, el aL., J. Mol. Bio. 278, 457-479, (1998)) and can be specifically targeted in a library format to screen for the most stabilizing amino acid. The CDRs of 5 the BHA 10 VH and VL were not considered for stability optimization due to potential interruption of the interaction with LTpR, but could be considered for second-generation designs. C. VH/VL Interface Homology Modeling As described in Example 15 infra, differential scanning calorimetry analyses 10 were performed on 17 human antibodies. The top candidates, BIIB 1-4 all had exquisite (i.e., very high and desired) stability properties. These highly stable antibodies may be used as a platform for improving the stability of scFvs or antibody domains with low intrinsic stability. In particular, emphasis was placed on the interface between VH and VL to provide a potentially greater level of stability properties. 15 Converting the BHAIO Fab toan scFv format resulted in a change in the DSC thermogram indicative of a fully cooperative unfolding event, as observed by a single, higher temperature unfolding transition (Fab), to a largely non-cooperative unfolding event, as observed by two individual, lower temperature unfolding transitions (scFv). In other words, in the Fab format, the interdomain contacts were strong enough to lock all 20 the domains together into a single, higher temperature unfolding transition. Once the CH I and CL domains were removed, the VH and VL domains of the scFv no longer had the ability to drive the unfolding transition into a cooperative event. It is postulated that by stabilizing the interface, on may not only stabilize both the VH and VL domains simultaneously, but also promote the adherence of the VH/VL interface and preclude 25 aggregation. The top 4 most stable BIIB antibodies (BIIB1-4) were used to design stabilizing mutations into the BHA 10 scFv via a two-step procedure. First, a crystal structure of the humanized BHA10 Fab was utilized for structural analysis. Specifically, the crystal structure was employed to identify all the residues at the interface between VH and VL as 30 well as the amount of surface area that each residue contributes to the interface (using MOLMOL software). The residues that bury significantly more surface area than others at the interface are considered to be more crucial to the interface. As an arbitrary limit prioritization, those that bury between 30 and 40 A 2 are considered important. Those 205 WO 2007/109254 PCT/US2007/006883 that bury >40 A 2 are considered crucial. These two categories were given the highest priority in terms of homology modeling and mutagenesis. The amount of surface area each residue buries at the interface is listed in Table 6. Secondly, the amino acid types at the interface between VH and VL were compared with the amino acids that exist at the 5 same positions in BIIB1-4 (i.e. the extremely stable antibodies). This method allowed identification of a number of outstanding mutations for stabilizing the BHA 10 scFv. Two mutations, S46L in the VL and P1OlD in the VH were experimentally validated as described in Example 8. In fact, these two mutations were the single most stabilizing mutations tested in the thermal challenge assay (e.g., see Figure 50A). Both 10 mutations stabilize both VH and VL domains simultaneously, suggesting that they help to build cooperativity among the domains. Table 6. BHA10 VHVL Surface Area Analysis. Position of Region BHA10 AA Buried Residue Corresponding VHIVL Type Surface Area Frequency AA from Stable Interface (Kabat) (A 2 ) within Ab (BIIB) Residue mammalian (Kabat #) database VH 35 FR2 His 16.4 His VH 37 FR2 Val 6.5 Val VH 39 FR2 Gin 37.4 98% Gin VH 44 FR2 Gly 13.1 VH 45 FR2 LCu 97.8 97% LCu VH 47 CDR2 TIp 78.5 97% TIp VH 50 CDR2 Trp 24.1 VH 59 CDR2 Gin 28.8 VH 61 CDR2 Asn 26.4 VH 90 FR3 Tyr 37.3 Tyr VH 1001 CDR3 Gtu 101.8 Highly Variable Gly(BI1BI),Tyr (BIB2) VH I too CDR3 Gly 36.4 Ala. Gly (MCR); Phe (BIB1), Tyr (uncommon); Ala (B112) VH 1OOK CDR3 Phe 66.7 Met (MCR); Phe Phe (B1111); (SMCR) Met (BIB2) VH 101 CDR3 Pro 43.3 Asp(V. Cons); Asp*(BtlBI,BlB2) Ala (uncommon) 206 WO 2007/109254 PCT/US2007/006883 VH 103 CDR3 Trp, 78.9 98% Trp VH .104 FR4 Gly 6.1 ________ VH 108 Fr4 Gln 11.5 _ ______ VL 36 FR2 Tyr 44.2 86% Tyr (BIIBI, BIIB2) VL 38 FR2 Gin 34.5 92% Gin (BIIBI, BIIB2) VL 42 FR2 Lys___________ _____ VL 43 FR2 Ala 47.7 Set = 50%; Ala = Ala (BlIBI, B1132) 35%; Pro= 10% VL44 FR2 Pro 69.7 100% Pro (BIIBI. BIIB2) VL45 FR2 Lys 3.5 ________ VL46 FR2 Ser 29.7 Lcu =76%; Leu- (BIIBI. BIIB2) Arg = 8%. Scr = 2% VL49 FR2 Ser 14.8 Tyr =88%; Ser (BIlBlI) Glu =4%; ______________Ser = 2% VL 50 FR2 Ser 9.7 ________ Gly (BIIB I) VL 55 CDR2 Tyr 59.6 Ala = 34%; Glu* (B11B1) Glu = 28%; Tyr = 6% VL 87 FR3 Phe 35.1 Tyr =84%; Tyr-(BULBI) ____________Phe = 15% VL 89 CDR3 Gin 14.7 Gln (BIIB 1) VL 91 CDR3 Tyr 36.3 Phc (BIIB I) VL94 CDR3 Tyr 68.4 Thr =20%;, Ala (BUIB1); Tyr= 2% Leu (BIIB32) VL95 CDR3 Pro 38.8 Pro =90% Pro (BIIB 1); Leu (BI1B2) VL 96 FR4 Phe 76.6 Tyr= 12%. Tyr* (BIIBI);, __________Phe =6% Trp* (BIIB2) VL 98 FR4 Phe 102.5 100% Phe VL 99 FR4 Gly 5.4 _______ ________ VL 105 FR4 Glu 6.2 ___________________ (Residues positions labeled with ' indicate good opportunities for stability design - via mutation to residues found in BIIB 1,2.3. or 4). 207 WO 2007/109254 PCT/US2007/006883 D. Computational Analysis Computational methods were used to analyze BHA 10 for making recommendations on positions whereby amino acid substitutions might improve stability. These methods were composed of two steps: sequence-based analysis (step 1); 5 and structure-based analysis (step 2). During the first step, a database of sequences of variable domains of antibodies was used. Amino acids present in low frequencies in their respective positions (less than in 10% of database sequences) or those that did not match the corresponding consensus amino acid were selected for substitutions to high frequency or consensus amino acids (candidate mutations). During the second step a 10 three-dimensional structure or model of an Fab fragment of the antibody was used. The candidate mutations were evaluated in their structural context and were prioritized for experimental testing based on structural properties, compatibility with complementarity determining regions (CDRs) conformations, role in VL/VH interface packing, and folding of heavy and light chains. 15 BHA10 is unusual in having a serine in position 46 on the light chain. In the database, approximately I% of VL (kappa subtype) sequences possess S46, whereas approximately 79% of VL (kappa subtype) sequences have L46. Also, human consensus KVI has L46. Once identified as a candidate mutation, S46L substitution was evaluated in the context of a three-dimensional model of BHA10 Fab. Position 46 was found to 20 reside at the VL/VH interface making contacts with Y101 and W103 of VH and Y36 and Y55 of VL. Structure-base analysis revealed that S46L substitution was compatible with the integrity VifVH interface, CDR conformations, and VL folding. Serine is an infrequent amino acid at position 16 of the heavy chain. In the database, approximately 5% of VH sequences possess S16, 25% have A16, 21% have 25 G16, 11% have E16, and approximately 10% are Q16. Also human consensus HV1 has A 16. Once identified as candidate mutations, S 16A, G, E, and Q substitutions were evaluated in the context of three-dimensional model of BHA10 Fab. Position 16 was found to reside in the loop proximal to the VH/CHl interface making contacts with K13 and S16 of VH. Structure-base analysis revealed that S16A, G, E, and Q substitutions 30 were compatible with the integrity of the VL/VH interface, CDR conformations, and V folding. Also it was concluded that long hydrophilic sidechains of E16 and Q16 would be preferred. Accordingly, positions VL 46 and VH 16 were represented in the library design. 208 WO 2007/109254 PCT/US2007/006883 A complementary approach to computational analysis made use of structure based protein engineering techniques, such as Rosetta (Kuhlman B. et al., PNAS 200198(19):10687-91) and DEEK (Hanf K.J., Ph.D. Thesis, MIT, 2002). Given a three dimensional structure of a protein interface, these methods can yield mutations in amino 5 acid sequence that lead to improved AAG for VH toVL binding (difference between bound and unbound states) and/or AAG for VH and VL folding (difference between folded and reference unfolded states). Use of these methods revealed that S46L is a mutation that improves VH toVL binding, whereas S16E is a mutation that improves folding of the VH. 10 E. Combined Interface and Covariation Design Approaches Covariation analysis was used to determine residue networks important for providing and supporting the interface between VH and VL and maintaining a strong affinity between /VHNL, thereby resulting in increased scFv stability. Residues directly 15 involved in the interface between VH and VL of the BHAIO scFv were identified as described in Example SC and are listed in Table 6 supra. Residues that covary strongly with the most highly buried residues listed in Table 6 supra were calculated using the covariation methodology described in Section III (a) supra. The HMM used for covariation analysis eliminates the ability to investigate residues in CDR2 and CDR3 of 20 both the VH and VL domains that are buried at the interface. Nevertheless, it is believed that no strong covariations would arise from these residue positions as they are highly variable due to intense selective pressures that exist for antibodies to be able to recognize antigens with high affinity. Therefore, the residues that were used for determining whether a covariation network exists for supporting the interface between 25 both VH and VL were: VH (kabat#) VL (kabat#) V37/137 Y36 Q39 Q38 30 G44 A43 L45 P44 W47 L46 W50 Y49 Y91 Y95 209 WO 2007/109254 PCT/US2007/006883 W103 F98 Residues whose appearance in the V-class sequence alignment correlated with the interface residues listed above were determined using a phi-value cutoff> 0.25. No 5 residue was considered as important for maintaining a strong interface between VH and VL if it was not correlated with at least 2 of the interface residues listed directly above. Tables 7 and 8 list those residues within the VH and VL domains, respectively, that demonstrate 2 or more correlations with interface residues. The larger the number of links, the greater the impact each of these amino acid positions is conceived to have for 10 stabilizing the interface between VH and VL. Table 7 - Residues with VH with multiple strong correlations to structural residues observed at the interface of the BHA 10 xray-crystal structure. Covar# Amino Acid Bha10Xray# Kabat# #Links 67 W 47 47 6 48 V(I**) 37 37 5 61 L 45 45 5 66 E 46 46 5 71 1 51 51 5 85 Y 60 59 5 142 L 112 109 5 14 P 14 14 4 27 G 26 26 4 29 F 27 27 4 52 A 40 40 4 61 L 45 45 4 102 I 70 69 4 135 W 106 103 4 31 F 29 29 3 49 R 38 38 3 60 G 44 44 3 69 G 49 49 3 78 G 56 55 3 93 L 64 63 3 97 G 55 54 3 210 WO 2007/109254 PCT/US2007/006883 101 T 69 68 3 106 D 73 72 3 109 S 75 74 3 115 Y 80 79 3 120 T(S*) 85 82b 3 6 E 8 8 2 43 Y 32 32 2 50 Q 39 39 2 53 P 41 41 2 92 S 63 62 2 124 A 89 85 2 131 Y 95 91 2 24 A,V,G 24 24 3* 89 A,N,P 61 60 3* Residues in bold bury surface area at the interface. Residues in italics are those directly adjacent in primary sequence to residues that bury surface at the interface. *Distinguishing subclass feature *Covaries with many of the same residues as the primary residue, but at a lower 5 correlation level due to weaker residue frequency. Table 8. Residues with VL with multiple strong correlations to structural residues observed at the interface of the BHA 10 xray-crystal structure. 10 Covar# Amino Acid Bha10Xray# Kabat# #Links 49 Q 37 37 7 48 Y 36 36 6 61 P 44 44 6 93 P 59 59 6 91 G 57 57 6 102 G 64 64 6 67 L 46 46 4 69 1 48 48 4 101 S 63 63 4 107 S 67 67 4 108 G 68 68 4 52 K 39 39 3 211 WO 2007/109254 PCT/US2007/006883 66 K 45 45 3 85 R 54 54 3 90 S 56 56 3 122 Q 79 79 3 129 D 85 85 3 135 F 98 98 3 68 L 47 47 2 92 V 58 58 2 117 T 74 74 2 120 G 77 77 2 126 E 83 83 2 133 Q 89 89 2 143 L 104 104 2 Residues in bold bury surface area at the interface. Residues in italics are those directly adjacent in primary sequence to residues that bury surface at the interface. *Q50,A89,Y135 are conserved in both heavy and light chains and covariations do not appear to correlate with light chain positions. 5 The residues from Tables 7 and 8 have been mapped to the surfaces of the BHA10 VH and VL, respectively, and compared with the actual residues that make direct contact at the interface between the two domains (see Figures 87 and 88). Two supported from this analysis. First, covariations do not provide information (at this 10 stage) about networks involving CDR2 and CDR3 residues at the interface. Second, covariations suggest that not only are residues that make direct contacts within the interface important for its maintenance, but that many other residues outside the direct interface residues are important for scaffolding and supporting the positions of the interface residues. This point is illustrated by those residues that are illustrated on the 15 surface of the covariation network but are absent on the surface obtained simply using the structures to calculate those residues that directly bury surface at the interface. Violations of a few of these interface covariation networks were found to exist within the BHA10 or p5E8 (described below in Example 21) scFvs: VH or VL sequences. Mutation of the native/non-ideal amino acids at these positions to the ideally 20 supportive interface amino acid residues described in Tables 7 and 8 were shown to be highly stabilizing to either the BHAIO or p5E8 scFvs: (a) BHA1O VL S46L, and (b) Idec152 VH S49G (kabat and Xray#) 212 WO 2007/109254 PCT/US2007/006883 E72D(73 for Xray). One of these residues (S46LVL) was directly at the interface, one (S49GvH) was directly adjacent to the interface, and one was distal to the interface (E72DvH). However, all three were predicted based on Covariation Analysis to be stabilizing to the interface. Stabilization of the interface can be observed using DSC 5 measurements as an increase in the thermal stability of both the VH and the VL. ii) Construction and Screening of BHA10 scFv Libraries with Improved Thermal Stability A) Construction of scFv Libraries 10 Libraries designed to contain the desired amino acid replacements in the conventional BHA10 scFv (pXWU002) using the methods described in Examples 4, 5, and 6 were created using the QuikChange II Site-Directed Mutagenesis Kit following instructions provided by the manufacturer (Stratagene, La Jolla, Ca.) using oligonucleotides listed in Table 9. The DNA sequence of the conventional scFv is 15 depicted in Figure 1. Individual transformed colonies were picked into deep-well 96 well dishes (Corning, Coming, NY, Cat.# 3960) containing 400 l/well LB plus 50 mg/ml carbenicillin and grown overnight at 37*C. Master plates were created by adding an equal volume of LB containing 20% glycerol to each well of the deep-well 96 well 20 dishes and transferring 50 pl aliquots of the bacterial suspension to sterile microtiter plates (Corning, Corning, NY, Cat. # 3359) and freezing for storage at -80 0 C. Table 9. Oligonucleotides for stability engineering of BHA10 scFv. BHA10 Library Position Sequeneet Design Method(s) VL 49 and 50 5'-GTACCGGTAGGAGGCMNNR Consensus (SEQ ID NO:123) KAAATCAGTGATTTAGG -3' VL 46 5'- Consensus, (SEQ ID NO: 124) GGGAAGGCTCCTAAATTACTG Computation, ATTTCCTCGGCC - 3' VL/VH Interface, Covariation VH 16 5'-. Consensus, (SEQ ID NO: 125) GGACACCTTCACTGACBNCCC Computation AGGCTTCTTCAC - 3' 213 WO 2007/109254 PCT/US2007/006883 VH 101 5'- V/VH Interface, (SEQ ID NO:126) GATCCTGGGAAGGTTTTGACT ACTGGGGCCAAGGGAC -3' VH20 5'- Consensus (SEQ ID NO:127) GGGTCCTCAGTGAAGWTRTCC TGCAAGGCTTCTG -3' VH 48 5'- Consensus (SEQ ID NO: 128) CAGGGACTTGAGTGGVKKGG A TGGATTTATCCTG -3' VH 67 5'- Consensus (SEQ ID NO: 129) GAAGTTCAAGGGCAGGNYCA CAATCACTGCAGAC -3' VH 55 5'- Covariation (SEQ ID NO:130) GGATGGATTTATCCTGGAAA TGGTCATGCTCAGTACAATGA G -3' VL3 5'- Consensus (SEQ ID NO:131) GGTGGTAGTGACATTVNSATG ACCCAGTCTCCTAGC -3' tPositions targeted for mutagenesis are indicated by underline. Ambiguous bases are abbreviated as follows: W = A or T, V = A or C or G, Y = C or T, S = C or G, M = A or C, N = A or C or G or T, R = A or G, K = G or T, B = C or G or T (JBiol Chem. 261(l):13-7 (1986)). 5 B. Thermal Challenge Assay The activities of the conventional and engineered BHA10 scFvs were then compared in a thermal challenge assay which can be used to determine the temperature at which 50% of scFv molecules retain their antigen binding activity following a thermal 10 challenge event. The numerical value corresponding to this temperature is referred to as theT 50 value and the units are in *C. In this assay, the scFvs were subjected to a range of temperatures that encompass the thermal transition temperature of conventional BHA10 scFv. E. coli strain W3 110 (ATCC, Manassas, Va. Cat. # 27325) was transformed with 15 plasmids encoding conventional and engineered BHA10 scFvs under the control of an inducible ara C promoter. Transformants were grown overnight in expression media consisting of SB (Teknova, Half Moon Bay, Ca. Cat. # SO140) supplemented with 1% 214 WO 2007/109254 PCT/US2007/006883 glycine, 1% Triton X 100, 0.02% arabinose, and 50 jig/ml carbenicillin at either 37"C or 32 0 C. Bacteria was pelleted by centrifugation and supernatants harvested for further treatment. Including glycine and Triton X- 100 in the media resulted in the release of periplasmic contents (native E. coli protein and scFv) into the media. The presence of E. 5 coli proteins in the supernatant is essential to the performance of this assay because the thermally denatured proteins act as a "sink", trapping transiently unfolded scFv molecules into irreversible inactive aggregates. Each library was screened in duplicate using a thermal challenge assay with supernatant from one replicate subjected to treatment conditions and the second 10 supernatant serving as untreated reference. Thermal challenge assays can be run at a single or range of temperatures for measuring stability. Thermocycler machines capable of generating stable thermal gradients are used for treating sample supernatants (iCycler, Bio-Rad, Gaithersburg, Md). The challenge temperature varied depending on the properties of the parental 15 BHA10 scFv variant and was generally two to three degrees Celsius higher than the experimentally determined T 50 value. Frozen Master plates were thawed and used to inoculate deep-well microtiter plates containing 250 pl of expression media per well, and cultures grown overnight at 32 0 C. As a control, cultures containing the parental plasmid were grown under the same conditions and processed simultaneously as the 20 library. Bacteria were pelleted in the deep-well plates by centrifugation at 2000 rpm (IEC model 8R, Thermo Electron, Waltham, Ma) for 30 minutes, and 100 pl of the supernatant was transferred to standard micotiter plates (Corning, Coming, NY, Cat. # 3357). An aliquot of the supernatant was reserved for the reference DELFIA. For most libraries, the plates containing the remainder of the supernatants were sealed (Nalge 25 Nunc, Rochester, NY, Cat.# 235205) and placed in an incubator set to the appropriate challenge temperature (Echo Therm, Torrey Pines Scientific, San Marcos, Ca) for 90 minutes. For screens requiring multiple challenge temperatures (or for temperatures greater than 75*C) the supernatants were transferred to 96-well PCR plates (Applied Biosystems, Foster City, Ca, Cat. # N801-560) and incubated for 90 minutes at the 30 desired temperature. After thermal challenge, the aggregated material was removed by centrifugation and soluble BHA10 scFv samples remaining in the treated, cleared supernatant were assayed for binding to cognate LTPR Ig antigen by DELFIA assay. 96-well plates 215 WO 2007/109254 PCT/US2007/006883 (MaxiSorp, Nalge Nunc, Rochester, NY, Cat. # 437111) were coated with fusion protein consisting of the ectodomain of the LTp receptor (LTpR) fused to a human Fc region at I .ig/ml in 0.1M sodium carbonate buffer, pH 9.5. Plates were coated overnight at 4"C, and blocked with DELFIA assay buffer (DAB, 10 mM Tris HCI, 150 mM NaCl, 20 pM 5 EDTA, 0.5 % BSA, 0.02% Tween 20, 0.01% NaN 3 , pH 7.4) for one hour with shaking at room temperature. Plates were washed 3 times with DAB without BSA (Wash buffer), and test samples diluted in DAB were added to the plates in a final volume of 100 pl. The plates were incubated for one hour with shaking at room temperature, and then washed 3 times with Wash buffer to remove unbound and functionally inactivated 10 scFv molecules. Bound BHA10 scFv was detected by addition of 100 pl per well of DAB containing 40 ng/ml of Eu-labeled anti-His 6 antibody (Perkin Elmer, Boston, MA, Cat. # ADO 109) and incubated at room temperature with shaking for one hour. The plates were washed 3 times with Wash buffer, and 100 pl of DELFIA enhancement solution (Perkin Elmer, Boston, MA, Cat. # 4001-0010) was added per well. Following 15 incubation for 15 minutes, the plates were read using the Europium method on a Victor 2 (Perkin Elmer, Boston, MA). Assay data was processed using Spotfire DecisionSite software (Spotfire, Somerville, Ma.) and expressed as the ratio of the DELFIA counts observed at challenge temperature to the reference temperature for each clone. Clones that reproducibly gave 20 ratios greater than or equal to twice what was observed for the parental plasmid were considered hits. Plasmid DNAs from these positive clones were isolated by mini-prep (Wizard Plus, Promega, Madison, WI) and retransformed back into E. coli W31 10 for confirmation secondary thermal challenge assays. For thermal gradients, the data was analyzed using Prism 4 software (GraphPad 25 Software, San Diego, Ca) using a sigmoidal dose response with variable slope as the model. The values obtained for the mid-point of the thermal denaturation curves are referred to as Tso values, and are not construed as being equivalent to biophysically derived Tm values. Primary and confirmatory results from these assays are shown in Table 10. 30 Several of the stabilized scFv molecules of the invention resulted in improvements in binding activity (T 50 >49*C) as compared with the conventional scFv. In particular, the
T
5 0 values of BHA 10 library position VL46 scFv (S46L), library position VH 16 scFv (S16E and SI6Q), and library positions VL 4 9 : VL50 scFv exhibited increases in thermal 216 WO 2007/109254 PCT/US2007/006883 stability ranging from + 3*C to + 12*C relative to the conventional BHA10 scFv. stability ranging from + 3*C to + 12"C relative to the conventional BHAIO scFv. In addition, the T 50 values of BHA10 library position VL3 scFv (Q3A, Q3G, Q3S, Q3V, and Q3D), library position VH67 scFv (V671 and V67L), library position VH48 scFv 5 (M481 and M48G), library position VH20 scFv (V201) and library position VH101 scFv (P101D) exhibited increases in thermal stability ranging from + 4*C to + 18"C relative to the conventional BHA10 scFv. One of the stabilizing mutations (VL K13E) serendipitously resulted from a PCR error. Incorporation of one of these stabilizing mutations into pIEH009 was found to further improve thermal stability and even exceed 10 that of BHA10 Fab under these conditions (Figure 12). Importantly, the non-covalent VL46 mutation and VH55 mutations derived from one or more of the four design methods (VL/VH interface homology modeling, consensus scoring, computational modeling, and covariation analysis) resulted in an improvement in scFv thermal stability nearly approaching that observed with the disulfide mutations and validating the utility 15 and novelty of these design tools. Figure 1B shows the amino acid sequences of conventional BH A10 scFv (SEQ ID NO: 4) and Figures 94A and B shows the amino acid sequence of stabilized BHAIO scFVs containing the S46L(VL) stabilizing mutation (SEQ ID NO: 137), and the V55G (VH) stabilizing mutation (SEQ ID NO: 138), respectively. The DNA sequence of wild 20 type BHA1O scFv (SEQ ID NO: 3) is depicted in Figure IA. The stabilizing mutation is indicated by the boxed residue. The leader sequence, gly/ser connecting peptide, and CH1 domain are indicating by the underlined, bolded, and italicized residues, respectively. Table 10. BHA 10 VH and VL library positions, library composition, and screening 25 results. Position Library Hit Seq. ATso "C Observed VL49 and 49 (Y or S) YT, YS, +3-4 50 50 (all AA's) YR, YG, SR, SK VH16 Q, K, E, R, W, G, P, S16E +8 S, T, A S16Q +4 VL46 L S46L +10 VH13 na K13E +3 217 WO 2007/109254 PCT/US2007/006883 VH101 na PIOD +18 VH20 F, M, L, I V201 +7 VH48 R, S, M, I, L, V, G M481 +3-4 M48G VL3 N, Q, K, H, E, D, R, Q3A, Q3S, +3-5 W, G, P, S, T, A Q3V, Q3D, Q3G, VH55 na VH V55G +12 VH67 T, P, I, L, V, A V671 +5 V67L +8 na = not applicable Table 11 shows the results of a comprehensive thermal stability analysis of the various individual and combined stabilizing mutations introduced into a conventional 5 scFv. These results demonstrate that the improvements in activity are additive and that the methods described in this invention are capable of improving the thermal stability properties of scFvs even beyond that of native Fabs. Even in the absence of a covalent disulfide bond at positions VH44-VL100, stabilizing mutations were identified that upon combination exhibited increases in thermal stability ranging from + 19*C to + 33*C 10 relative to the conventional BHA 10 scFv. Table 11. Characteristics of BHA10 constructs used to produce variant proteins and T 50 results from thermal challenge assay. Plasmid Disulfide Linker Other Mutation T 5 o "C Length (aa) pIEH003 no 15 na 49 pIEH006 VH44-VL100 15 na 59 pXWUOOI Fab na Fab 61 pXWU002 no 20 na 51 pXWU003 no 25 na 51 pIEH009 VH44-VL100 20 na 61 plEH029 no 20 VL S46L 61 plEHO31 VH44-VLOO 20 VL S46L 71 pIEH032 no 20 VH S16E 60 pIEHO34 no 20 VH S16Q 56 pIEH058 no 20 VH P1O1D 67 218 WO 2007/109254 PCT/US2007/006883 pIEHO59 no 20 VH V20I 56 pIEHO60 no 20 VH M48G 54 pIEHO61 no 20 VH M48I 53 plEH062 no 20 VL Q3A 53 pIEHO63 no 20 VL Q3S 53 plEH065 no 20 VL Q3V 53 pIEH066 no 20 VL Q3D 54 plEH067 no 20 VL Q3G 54 pIEH068 no 20 VH V671 55 pIEHO69 no 20 VH V67L 58 pIEHO70 no 20 VH V55G 64 p1EH076 no 20 VH Sl6E + VL S46L 71 plEH078 no 20 VH S16Q +VL S46L68 pIEH080 VH44-VL100 20 VH Sl6E + VL S46L 75 pIEH081 VH44-VL100 20 VH S16Q + VL S46172 pIEH087 no 20 VH S16E, V55G + 75 VL S46L pIEH094 no 20 VH S16E, V55G, 82 1__ PIOID +VL S46L na = not applicable aa = amino acids 5 Summary Three stabilizing mutations (VS46L, VHV55G, and VH_PIOD) were experimentally validated by thermal challenge (T 5 o) assay. Both of the V 1 _S46L and VH_V55G mutations were predicted to be stabilizing to the individual VH and VL domains based on the covariation analyses described in 10 Example 3 supra. As depicted in Figure 50A and B, both of these mutations led to significant increases in the T 50 of the BHAI0 scFv. In particular, the VLS46L mutation stabilizes the scFv by -7-8 *C while the VHV55G mutation stabilizes the scFv by -12 "C. Both of these mutations also significantly shrink the Tm gap between VH and V 1 as determined by DSC (see Figure 51). The DSC data shows that both 15 mutations lead to significant increases in both the TM (midpoint of the thermal unfolding transitions) of both of the VH and VL domains and in the calorimetric enthalpy (i.e. area 219 WO 2007/109254 PCT/US2007/006883 under the curve) for the scFv. Increases in these values provide experimental validation that these predicted stabilizing mutations are in fact stabilizing to the scFv. Furthermore, since these two mutations stabilize both VH and VL domains simultaneously, they are likely important in building cooperativity between the VH and 5 VL domains. Accordingly, it is expected that strengthening the interface between V" and VL may not only help increase stability, but reduce the tendency to form aggregates as well. The VL_S46L stabilizing mutation and a third stabilizing mutation (VHPIOID) were also predicted to be stabilizing based on VH/VL interface homology modeling 10 described in Example 4 supra. The VH__P101D stabilizing mutation stabilized the scFv by -15 "C as measured by thermal challenge assay (Figure 50C). In addition, DSC (see Figure 52) shows that this stabilizing mutation leads to significant increases in Tm both of the VH and VL domains and in the calorimetric enthalpy, indicating that this mutation is also likely cooperatively stabilize the scFv as a whole via the VH/VL interface. 15 Example 6. Biophysical Characterization of Stabilized BHA10 scFvs comprising Stabilizing Mutations V region gene sequences from various plasmids listed in Table 11 were subcloned into a modified E. coli expression vector to drive recombinant protein 20 expression under the control of an inducible ara C promoter. Variant BHA 10 scFvs were expressed and purified using methods described above. An apparent cleavage site at the N-terminus of the VL domain led to a low level of VL in most scFv preparations as judged by SDS-Page analysis (Figure 53). The hydrodynamic properties of each scFv were investigated by size exclusion 25 HPLC (Agilent Technologies) with static light scattering and refractive index detectors (MiniDAWN/ReX, Wyatt Technology). Each scFv was found to be predominantly monomeric; however, a few of the scFvs, including the wild-type (unstabilized) scFv, exhibited a low level of dimeric material (Table 12). None of the purified scFvs had detectable levels of oligomers larger than dimer. 30 Table 12. SEC/Light Scattering Results of scFvs Sample Expressed %Monomer %Dimer %Larger Yield ___ _ Aggregates 220 WO 2007/109254 PCT/US2007/006883 WT n.d.b 87 13 . 0 GS4 0.6±0.1 97 3 0 009 (ss) n.d. 99 i' 0 072 V.S46L 2.0 85 15 0 073 VS46L + ss *0.15 78 22 0 074 VH_S16E. 2.4*1.7 .94 6 0 075 VHS16Q 1.1±0.5 96 4 0 076 V_.S46L + 1.8±0.4 80 20 0 VH_S16E 077 VHP1OID 2.5 93 7 0 078 VS46L + 1.0 85 15 0 VH_SI6Q 079 VHV55G 2.0 92 8 0 080 VS46L + *0.21 76 24 0 VH_SI6E + ss 087 VS46L + 2.8 76 24 0 VH_S16E, V55G 094 VLS46L + 2.5 79 21 0 VH SI6E,V55G,PIO1D aProteins without standard error were expressed one time. bn.d., not determined. 'These values were determined AFTER preparative SEC was performed to reduce aggregates 5 The ultimate goal of the experiment was to investigate whether the designed mutations enhance stability and lead to a lower propensity for aggregation. The thermostability of each scFv was assessed using two separate methods. First, thermal unfolding of the scFvs was analyzed using DSC (Figure 54A). Secondly, the thermostability of each scFv was investigated by heating the proteins in the presence of 10 a fluorescent hydrophobic dye, 1-anilino-8-sulfonate (ANS, Figure 54B). ANS generally binds to partially unfolded proteins or heat-denatured proteins in compact unfolded states. Upon association with exposed hydrophobic surface areas, the fluorescence of ANS increases significantly presumably due to sequestration of the quenching affects of solvent. 15 The temperature dependent ANS curves were fitted to a two-state protein unfolding transition as described below. The buffer was PBS and the concentration of 221 WO 2007/109254 PCT/US2007/006883 scFv used in every experiment was 750 nM. Fluorecence measurements were performed on a JASCO model 812 Circular Dichroism spectropolarimeter equipped with a peltier heating device and external water bath. Fluorescence was collected using an accessory containing a photomultiplier tube perpendicular to the light path. The accessory is 5 equipped with an adjustable monochrometer set to 480 nm. The sensitivity was set to 600 V. Heating was performed at a continuous rate of 120 *C/min. The excitation monochrometer was set to 370 tim. Thermal unfolding of every scFv in this study led to an increase in ANS fluorescence. The midpoint of thermal unfolding (i.e. the TM) of each scFv protein 10 domain was determined using DSC by fitting each unfolding peak to the Gibbs Helmholtz equation using the Origin 7.0 software provided by the manufacturer (MicroCal, Inc). TMs were also derived using ANS fluorescence by incorporation of the Gibbs-Helmholtz equation into the non-linear curve fitting routine in KaleighdaGraphTm: 15 AGU-(T) = AHU*(T) TASU*(T), (1) where AGu*(T) is the temperature-dependent change in Gibbs free energy upon unfolding, AH*(T) is the enthalpy change associated with unfolding, and ASu*(T) is 20 the entropy associated with unfolded. The equation can be expanded to: T (T)1 AGu (T) = AH*(1 )-TM AC,* Tl -T)+Tln - , (2) TM T where TM is the midpoint of the unfolding curve and AC,* is the change in the heat 25 capacity between the folded and unfolded states. This equation can be used to derive the temperature-dependence of the ANS-fluorescence intensity by making the assumption that the fluorescence intensity is the sum of the intrinsic ANS fluorescence in the presence of folded and unfolded scFv, 30 iT(T)=iFF +UfU (3) 222 WO 2007/109254 PCT/US2007/006883 where i. (T) is the temperature-dependent total fluorescence signal, iF and iU are the fluorescence intensities of the folded and unfolded states, respectively, and fF and fu are the fractions of folded and unfolded scFv, respectively at the given temperature. The fractions folded are related to the equilibrium unfolding constant and free energy of 5 unfolding: Ku (T)= = exp( AGu (T) (4) fF RT By factoring Ku (T) and AGu *(T) into equation (3) and assuming that the fluorescence 10 intensities of ANS in the presence of folded and unfolded scFv are linearly dependent on the temperature (i.e. iF= I +i 2 T and iU = i3 +i 4 T) the following equation was obtained: F AGU" 0 (T)1 iK + i 2 (T)+(i +i 4 T)expL u RT iT (T ) - IF+ iUKU RT .(5) 1+Ku x AGu O(T) RT 15 To obtain the final equation utilized to fit the data, the relationship for AGu *(T) in equation (2) is substituted into equation (5) assuming AC,* is independent of temperature and proportional to the difference in solvent exposed surface area between the folded and unfolded states (Haynie & Friere, Proteins: Struct. Funct. Genet. 16: 115-140, (1993); Myers et al., Protein Sci. 4: 2138-2148 (1995)). 20 The thermostability measurements for all the scFvs derived from DSC and ANS binding experiments were comparable. Unfolding of each scFv was irreversible; therefore, aggregation had an affect on the absolute TM measured by DSC or ANS binding. Since the two techniques heated the samples differently and at different rates, the Tms were not expected to be identical across the experimental formats (Sinchez-Ruiz 25 et al., Biochemistry, 27:1648-1652 (1988)). However, the trend observed for each experimental technique was identical (see Table 13). The DSC experiments readily discriminated between the VH and VL unfolding transitions. The ANS-binding experiments were not capable of accurately discriminating 2 transitions (i.e. VH vs. VL unfolding); thus, only the apparent TM was provided for the ANS-binding experiments. 223 WO 2007/109254 PCT/US2007/006883 The apparent Tm observed by ANS-binding appeared to correlate well with the Tm of the last domain to unfold, either VH or VL depending on the mutation, as determined by DSC. Due to aggregation of unfolded material in both assay formats, the TMs were used as a guide to rank order the stability enhancements afforded by each mutation without 5 additional interpretation of scFv free-energies of unfolding, etc. Table 13. Thermostability measurements of each scFv. Construct VH Tm VL TM VH TM scFv Tm scFv KD (M) *10'9 ("C DSC) (*C DSC) (*C ANS) (*C DSC)" (Biacore) Wild-type, (G 4 S)*3 55.4 68.7 65.6 - 2.7 linkerb Wild-type, (G4S)*4 linker 57.7 67.4 67.0 - 2.1 V_SI6E 60.7 68.1 66, 76 - 2.0 VHS16Q 59.4 68.4 69c - 1.9 V,_S46L 65.6 74.2 71.0 - 4.0 VsV55G - - 75.4 68.4 2.2 V_PIOD - - - 71.9 5.2 VH_S16E, VjS46L 71 74.5 - 72.2 3.4 VtS16Q, VS46L 67.9d 74.9' 75.4 - 4.2 Vn_S16E, VH_V55G, - - 79.3 77.7 3.8 VeS46L Vs _SI6E, Vt_V55G, 84.1 77 81.3 - n.d. VH_P101D, V,_S46L Wild-type, (G 4 S)*4 linker 61 68 n.d. - 3.3 w/ VH 4 4 NLIOO disulfide VH_SI6E, VeS46L 69 77 n.d. - 4.0 VH_S16Q, VLS46L n.d. n.d. n.d. n.d. 3.3 aThese particular mutations led to single, cooperative VH/VL unfolding events. bAll subsequent scFvs were constructed with the (G 4 S)*4 20 amino acid linker. 10 'Multiple transitions complicated analysis. dUnable to discriminate VH vs. VL. Summary All designed single mutations picked from the library screens: VH SI 6E, S16Q, 15 V55G, PIOID, and VL S46L significantly stabilized the VH domain and in some 224 WO 2007/109254 PCT/US2007/006883 instances, the VL domain as well (Figure 53). The rationale behind testing these positions for stability enhancements often came from multiple forms of analysis. Consensus methods predicted that all of the mutations VH Sl6E, SI6Q, V55G, P101D and VL S46L would stabilize the BHAIO scFv (Example 3). However, VH V55G and 5 P101D (incidentally, the two most stabilizing mutations) are positioned within CDR2 and CDR3 of VH, respectively, and were not considered for mutagenesis until further predictive evidence was compiled that suggested that mutation at these two positions could lead to stabilizing events. Both VI V55G and VL S46L were also predicted to be stabilizing based on Covariation Analyses. Finally, VH P1OID and VL S46L were 10 predicted to be potentially stabilizing to the VH/VL interface based on the interface composition of highly stable human antibodies (Example 3). Single mutations at the VH/VL interface increased the stability of both domains, while VH mutations outside the interface only stabilized the VH domain itself. VH S16E, S I 6Q, and V55G were outside the interface and increased the apparent TM of the 15 BHA10 VH by 3, 2, and I I "C, respectively. The VL stability was relatively unaffected by these mutations. VH V55G mutation appeared to increase the stability of the VH to match that of the VL domain, without leading to detectable increases in VL stability. Both mutations at the interface, VH P101D and VL S46L significantly increased the stability of both domains. The PIOD mutation in particular led to the fully cooperative 20 unfolding of the VH and VL domains suggesting that the interface and cooperativity of folding between VH and VL was significantly strengthened. Thus, mutations at the VH/VL interface may provide the most effective means of forming a stabilized Fv region and provide a rationale for prioritizing stability designs to the VH/VL interface of scFvs over designs directed towards at residues outside the Fv interface. 25 Combining VH mutations eventually led to a disconnect between the cooperativity of VH/VL unfolding (Table 13). As the VH TM climbed above 75 *C (the result of cQmbinations of VH mutations), the TM of the VL began to move towards lower temperatures. This suggests hyperstabilization of the V domain without building in compensating mutations within the VL domain may lead to decreased folding 30 cooperativity and a weakened interaction between the two domains. While the TMs of the VH and VL domains (including the TM of the wild-type BHA 10 scFv) were often quite different, evidence that mutations at the interface between the two domains could stabilize both domains simultaneously suggests that the two domains are interacting at 225 WO 2007/109254 PCT/US2007/006883 some level with one another - although the apparent affinity between the isolated domains is not expected to be extremely high, except perhaps in the VH P101D stabilized variant (Brandts & Lin, Biochemistry, 29:6927-6940, (1990)). Increasing the stability of the scFv led to decreased intrinsic binding of the 5 hydrophobic, fluorescent dye ANS at ambient temperature (15 "C, Figure 55). The wild-type scFv binds weakly to ANS suggesting that the protein may permanently or transiently expose hydrophobic-surface area to solvent. Stabilized scFvs appeared to completely lose the ability to bind ANS under ambient conditions - the fluorescence of ANS in the presence of the most stabilized scFvs was no greater than that of ANS alone 10 in solvent. Plots of VHTM (measured by DSC or temperature-dependent ANS-binding experiments) vs. ANS-fluorescence in the presence of folded protein at 15 *C demonstrate that scFv-dependent ANS fluorescence decreases as scFv stability increases (Figure 55). Reduction of ANS-binding by scFv stabilization may indicate that less hydrophobic surface area gets exposed to solvent and that stabilized scFvs may have a 15 lower intrinsic propensity for aggregation. KD measurements of various individual and combined stabilizing BHAl0 scFv mutations were performed using surface plasmon resonance (SPR) on a Biacore 3000 instrument. Table 13 shows the results of the Biacore affinity assay. The binding affinity of the variant BHA10 scFvs ranged from IX-2.5X of the parent BHA10 scFv. 20 Example 7. Production of Stabilized Bispecific "Hercules" Antibodies Both conventional BHA1O scFv and the stabilized BHA10 scFvs of the invention were used to construct a panel of bispecific antibodies referred to as "Hercules". 25 Hercules bispecific antibodies comprise a fusion of chimeric 14A2 IgG antibody that binds to TRAIL R2 receptor with a BHA10 scFv that binds to LTpR. The Hercules antibodies were constructed both as N- terminal and C- terminal BHA10 scFv fusions (see Figure 13). N- terminal scFv fusions can be engineered as light chain and/or heavy chain fusions ("NL-Hercules" or "NH-Hercules", respectively). The decision as to which 30 amino terminal V region (heavy or light) is selected for adjoining the scFv is primarily driven by which chain is thought to tolerate a fused scFv which is capable of recognizing the first target antigen while not appreciably interfering with the binding of the Fab domain to the second target antigen. 226 WO 2007/109254 PCT/US2007/006883 It is also possible using the methods described in this invention to engineer tetravalent or bispecific antibodies consisting solely of stabilized scFvs fused directly to antibody hinge regions or to CH2 or CH3 domains as depicted, for example, in Figures 42-49. Said antibodies may comprise full-length Fc regions (see, for example, Figure 5 46) or CH2-domain deleted Fc regions (see, for example, Figure 42). In other exemplary embodiments, two or more stabilized scFv domains may be fused to the same terminus of a heavy or light chain (see, for example, Figure 43). A. Construction of "NH- Hercules" with BHA10 conventional, (G4S)4, 10 VH44:V100, and VH44:V100/(Gly4Ser) 4 scFvs Four anti-LTpR (BHA1O) x anti-TRAIL R2 (chil4A2) bispecific antibody designs were based on appending the conventional and variant BHA 10 scFvs to the amino terminus of the anti-TRAIL R2 antibody heavy chain. The BHA 10 scFvs DNAs described in Example 3 were used to construct a panel of NH- Hercules bispecific 15 antibodies by PCR amplification using the oligonucleotide primers described in Table 14. A (Gly 4 Ser)s linker was used to connect the BHAlO scFvs to the mature amino terminus of chil4A2 heavy chain. The forward 5' VH PCR primer (scFvBHAIO-FI) includes an Mlu I restriction endonuclease site (underlined sequence) for cloning followed by sequence encoding the last three amino acids of the heavy chain signal 20 peptide and the amino terminus of BHAlO VH. Two reverse 3' PCR primers were used to generate the PCR products with internal reverse primer XWU005-R encoding the carboxyl terminus of BHA10 VL followed by the (Gly 4 Ser)s linker, and reverse primer scFvBHA10-Ri encoding a partial anti-TRAIL R2 VH region and a Bgl II site (underlined sequence) for cloning. 25 227 WO 2007/109254 PCT/US2007/006883 Table 14. Oligonucleotides for PCR amplification of N- Hercules with conventional, (GIy 4 Ser) 4 , VH 4 4 :VL100 and VH44:VLI00/(Gly 4 Ser) 4 BHA10 scFvs scFvBHA10-Fl 5' (SEQ ID NO: 134) AGAGAGACGCGTGTCCTGTCCCAGGTCCAACTGGTGCA G-3' XWU005-R 5' (SEQ ID NO: 135) AGCCACCTCCCCCCGATCCACCGCCCCCTGAACCGCCCC CTCCAGAGCCCCCTCCACCGGACCCTCCACCGCCTTTGA TCTCCACCTTG -3' scFvBHA10-Rl 5' (SEQ ID NO: 136) AGAGAGAGATCTTGACTGTCTCTCCAGGCTTCTTCAGCT CAGGTCCAGACTGCACCAACTGGATCTGGGAGCCACCT CCCCCCGATCCAC-3' The BHA 10 scFv + (Gly 4 Ser) 5 linker + partial anti-TRAIL R2 VH gene 5 sequences were amplified in two sequential PCR reactions through the common overlapping sequences encoding the (Gly 4 Ser) 5 linker as represented in Figure 14A to prepare conventional BHA10 scFv from plasmid DNA pXWU034, BHA 10 scFv (Gly 4 Ser) 4 from plasmid pXWU002, BH A10 scFv VH44:VLI00 from plasmid pIEH006, and BHA10 scFv VH44:VLI00/(Gly4Ser)4 from plasmid pIEH009. The PCR products 10 from the panel of amplified BHA10 scFvs were purified by agarose gel electrophoresis using the Millipore Ultrafree-DA extraction kit according to manufacturer's instructions (Millipore; Bedford, MA). The purified PCR products were digested with Mlu I/Bgl II restriction endonculeases and ligated into an Mlu I/Bgl II digested, pN5KG1 vector containing the chi 14A2 IgG I previously modified to remove an internal Bgl II site 15 present in the chil4A2 IgGI coding sequence. The mammalian expression vector pN5KG I contains a translation-impaired, modified (intron-containing) neomycin phosphotransferase gene to select for transcriptionally active integration events, and a murine dihydrofolate reductase gene to permit amplification with methotrexate (Barnett, et al., Antibody Expression and Engineering. (Imanaka, H. Y. W. a. T., ed), pp. 27-40, 20 Oxford University Press, New York, NY, (1995)). The resulting panel of constructs form fusion proteins of the variant BHA 10 scFvs to the amino terminus of the anti-TRAIL R2 antibody VH domain through the 25 amino acid (Gly 4 Ser) 5 linker. Fusion of the conventional BHA10 scFv gene sequence to 228 WO 2007/109254 PCT/US2007/006883 the amino terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWU005. Fusion of BHAIO scFv (Gly 4 Ser) 4 gene sequence to the amino terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWUO26. Fusion of BHAl0 scFv V.44:VLOO gene sequence to the amino terminus 5 of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWU027. Fusion of BHA10 scFv VH44:VLI00/(Gly4Ser)4 gene sequence to the amino -terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWUO28. The ligation mixtures were used to transform E. coli strain TOP 10 competent cells (Invitrogen Corporation, Carlsbad, CA). E. coli colonies transformed to 10 ampicillin drug resistance were screened for presence of inserts. DNA sequence analysis confirmed the correct sequence of the final constructs. The chimeric 14A2 light chain used is common among all the N- and C-Hercules bispecific antibodies and the DNA (SEQ ID NO:28) and amino acid sequences (SEQ ID NO:29) are shown in Figures 15A and 15B. The chimeric 14A2 light chain is 15 expressed with a signal peptide at the N-terminus having the following DNA and amino acid sequences: ATGGCCTGGACTCCTCTCTTCTTCTTCTTTGTTCTTCATTGCTCAGGGT CTTTCTCC (SEQ ID NO:59) 20 MAWTPLFFFFVLHCSGSFS (SEQ ID NO:60) The heavy chain DNA (SEQ ID NO:30)_and amino acid sequences (SEQ ID NO:3 1) for conventional BHA10 scFv NH- Hercules are shown in Figures 16 and 17, respectively. The heavy chain DNA (SEQ ID NO:32) and amino acid (SEQ ID NO:33) 25 sequence for BHA 10 scFv (Gly 4 Ser) 4 NH- Hercules are shown in Figures 18 and 19, respectively. Heavy chain DNA (SEQ ID NO:34) and amino acid (SEQ ID NO:35) sequences for BHA10 scFv VH44:VLOO N- Hercules are shown in Figures 20 and 21, respectively. Heavy chain DNA (SEQ ID NO:36) and amino acid (SEQ ID NO:37) sequences for BHA10 scFv VH44:VL100/(Gly4Ser)4 N- Hercules are shown in Figures 30 22 and 23, respectively. Each of the heavy chains was expressed with a signal peptide at the N-terminus having the following DNA and amino acid sequences: ATGGGTTGGAGCCTCATCTTGCTCTTCCTTGTCGCTGTTGCTACGCGT 229 WO 2007/109254 PCT/US2007/006883 GTCCTGTCC (SEQ ID NO:61) MGWSLILLFLVAVATRVLS (SEQ ID NO:62) 5 B. Construction of C-scFv "Hercules" with BHA10 conventional, (G4S) 4 . VH 4 4 V100. and VH44:VL100/(Gly 4 Ser) 4 scFvs Four anti-LTPR (BHA10) x anti-TRAIL R2 (chil4A2) bispecific antibody designs were based on appending the conventional and variant BHA10 scFvs to the 10 carboxyl terminus of the anti-TRAIL R2 antibody heavy chain. The BHA 10 scFvs DNAs described in Example 3 were used to construct a panel of C- Hercules bispecific antibodies by PCR amplification using the oligonucleotide primers described in Table 15. A Ser(Gly 4 Ser) 3 linker was used to connect the BHAIO scFvs to the carboxyl terminus of chil4A2 heavy chain. The forward 5' VH PCR primer (XWU006-FI) 15 includes a BamH I restriction endonuclease site (underlined sequence) for cloning followed by sequence encoding a portion of the Ser(Gly 4 Ser) 3 linker peptide and the amino terminus of BHA1O VH. The reverse 3' VL PCR primer (XWU006-R1) primes BHA 10 scFv light chain and includes a stop codon followed by a BamH I site (underlined sequence) for cloning. Forward 5' internal overlapping PCR primer 20 (XWU006-F2) includes sequence encoding the (Gly 4 Ser) 3 linker and contains a silent mutation indicated in bold type. Reverse 3' internal overlapping PCR primer (XWU006 R2) includes sequence encoding the (Gly 4 Ser) 3 linker and contains a silent mutation indicated in bold type to remove a BamHI site located at BHA10 scFv (Gly 4 Ser) 3 linker region. 25 Table 15. Oligonucleotides for PCR amplification of C- Hercules with BHA 10 conventional, (G 4
S)
4 , VH 4 4 :VL100, and VH44:VLl00/(Gly4Ser)4 scFvs. 230 WO 2007/109254 PCT/US2007/006883 XWU006-Fl 5' (SEQ ID NO:38) GGGGGTGGATCCGGTGGAGGGGGCTCCGGCGGTGGCGG GTCCCAGGTCCAACTGGTGCAGTCTG -3' XWU006-F2 5'- TGGGGGCGGCGGGTCCGGTGGTGGTGGTAG -3' (SEQ ID NO:39) XWU006-R2 5'- TACCACCACCACCGGACCCGCCGCCCCCAG -3' (SEQ ID NO:40) XWU006-RI 5'- GTTAACGGATCCTCATTTGATCTCCACCTTGG -3' (SEQ ID NO:41) As represented in Figure 14B, BHA 10 scFv gene sequences were amplified in a two-step PCR reaction using the 5' VH XWU006-F1 + 3' VL XWU006-R1 PCR primer set and a set of internal overlapping PCR primers (XWU006-F2 and XWU006-R2) 5 designed to eliminate a BamH I site within the BHA10 scFv (Gly 4 Ser) 3 linker region. These PCR conditions were used to prepare conventional BHA10 scFv from plasmid pXWU034, BHA1O scFv (Gly 4 Ser) 4 from plasmid pXWU002, BHA1O scFv VH44:VLIOO from plasmid pIEH006, and BHA1O scFv VH44:VLOO/(Gly4Ser)4 from plasmid pIEHOO9. The PCR products from the panel of amplified BHAlO scFvs were 10 purified by agarose gel electrophoresis using the Millipore Ultrafree-DA extraction kit according to manufacturer's instructions (Millipore; Bedford, MA). The purified PCR products were digested with BamH I restriction endonuclease and ligated into a single BamH I site of the pN5KG I vector previously engineered to contain several modifications. Briefly, the pN5KG 1 vector containing chi 14A2 IgG I was modified to 15 remove the stop codon at the 3' end of the heavy chain gene and introduce nucleotides coding for the amino acid sequence Ser-Gly-Gly-Gly immediately followed by a BamH I restriction endonuclease site (coding for Gly-Ser) for cloning. Lastly, an internal unwanted BamH .1 restriction endonuclease site in chil4A2 VL region was also eliminated. 20 The resulting panel of constructs form fusion proteins of the variant BHAlO scFvs to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain through the 16 amino acid Ser(Gly 4 Ser) 3 linker. Fusion of the conventional BHA1O scFv gene sequence to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWU006. Fusion of BHA10 scFv (Gly 4 Ser) 4 gene 25 sequence to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain gene 231 WO 2007/109254 PCT/US2007/006883 sequence produced plasmid pXWU034. Fusion of BHA10 scFv VH44:VL100 gene sequence to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWU035. Fusion of BHA10 scFv VH44:VLI00/(Gly4Ser)4 gene sequence to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain 5 gene sequence produced plasmid pXWU036. The ligation mixtures were used to transform E. coli strain TOP 10 competent cells (Invitrogen Corporation, Carlsbad, CA). E. coli colonies transformed to ampicillin drug resistance were screened for presence of inserts. DNA sequence analysis confirmed the correct sequence of the final constructs. The chimeric 14A2 light chain 10 used is common among all the N- and C-Hercules bispecific antibodies and DNA and amino acid sequences are shown in Figures 15A and 15B. Heavy chain DNA and amino acid sequences for conventional BHA10 scFv C- Hercules are shown in Figures 24 and 25, respectively. Heavy chain DNA and amino acid sequences for BHA 10 scFv (Gly 4 Ser) 4 C- Hercules are shown in Figures 26 and 27, respectively. Heavy chain 15 DNA and amino acid sequences for BHA10 scFv VH44:VLIOO C- Hercules are shown in Figures 28 and 29, respectively. Heavy chain DNA and amino acid sequences for BHAI0 scFv VH44:VLI00/(Gly4Ser)4 C- Hercules are shown in Figures 30 and 31, respectively. A summary of the C- terminal bispecific Hercules constructs is found in Table 16. 20 Table 16. Intermediate and expression plasmids encoding 'Hercules' Vector Composition Antibody pXWU002 pIEH003 + BHA10 scFv (G 4
S)
4 BHA10 scFv with (G 4
S)
4 linker linker pXWU005 pN5KG1 + BHA1O scFv N- Hercules BHA10 scFv N- Hercules pXWU006 pN5KGI + BHA10 scFv C- Hercules BHA10 scFv C- Hercules pXWU026 pXWUOO5 + (G 4
S)
4 linker BHA10 scFv (G 4
S)
4 N- Hercules pXWUO27 pXWU005 + VH44:VL100 BHA1O scFv VH44:VL1OO N- Hercules pXWUO28 pXWU005 + VH44:VLIOO/(Gly 4 Ser)4 BHA10 scFv
VH
4 4 :VLI 00/(Gly 4 Ser) 4 N- Hercules pXWUO33 Modified pXWU006 Bam HI site removed in 14A2 VL of C- Hercules 232 WO 2007/109254 PCT/US2007/006883 pXWUO34 pXWU033 + (G 4
S)
4 linker BHA1OscFv (G 4
S)
4 linker C Hercules pXWU035 pXWU033 + VH44:VL100 BHA10 scFv VH 4 4 :VLOO C Hercules pXWU036 pXWUO33 + VH44:VL100/(Gly 4 Ser) 4 BHA10 scFv VH44:VLI00/(Gly 4 Ser) 4 C- Hercules C. Transient Expression of Bispecific Antibodies in CHO cells Plasmid DNAs pXWU005, pXWUO26, pXWU027, pXWUO28; and pXWU006, pXWU034, pXWU035, and pXWU036 (Table 16) were used to transform CHO DG44 5 cells for transient production of antibody protein. Each 20ug of plasmid DNA was combined with 4 x 106 cells in a volume of 0.4 mIs of 1XPBS. The mixture was added to a 0.4 cm cuvette (BioRad) and placed on ice for 15 min. The cells were electroporated at 600uF and 350volts with a Gene Pulser electroporator (BioRad). The cells were placed in the CHO-SSFM II media containing 1OuM Hypoxanthine and l6uM 10 Thymidine into a T-25 flask and incubated at 370 for 4 days. Supernatants containing Hercules proteins produced by this transient CHO expression system were collected and evaluated by Western Blot. Figure 32-shows that the Hercules antibodies containing either the VH44:VL1 00 (ds=disulfide) or VH44:VLOO/(Gly4Ser)4 linker stabilized BHA10 scFvs dramatically improved 15 expression yields independent of whether the scFv was fused to the NH- or C- terminus (lanes 3, 4, 7, 8). Surprisingly, no secreted protein was detected with wild type BHA10 scFv nor (Gly 4 Ser) 4 linker BHAI0 scFv N-terminal fusions indicating the benefit of scFv stabilization on expression (lanes 6 and 7). In addition, conventional BHA10 scFv and (Gly 4 Ser) 4 linker BHA10 scFv C-terminal fusions displayed a significant amount of 20 a -55-60 molecular weight byproduct which is substantially reduced in the stabilized constructs suggesting that scFv stabilization may improve product quality. D. Bispecific Binding ELISA Assay Supernatants were tested for individual binding activity to recombinantly 25 produced TRAIL R2 and LTpR receptors in ELISA assays. In both assays, receptor was immobilized onto plates and test samples incubated to permit binding to receptor. Bound samples were detected with labeled antibody. 96-well microtiter Immulon II plates (Fisher, Cat# 14245-61) were coated with 233 WO 2007/109254 PCT/US2007/006883 100 1/well of 2pg/ml LT3R-Ig in Na 2
CO
3 / NaHCO 3 buffer pH 9.5, overnight at 4 *C and blocked with 200pl dilution Buffer (0.5% Nonfat Dry Milk in PBS plus 0.01% Thimerosal) for 1 h at 37 *C. In the next step, 100lp individual 'Hercules' supernatant or purified protein in dilution buffer was added to duplicate wells and incubated for 1 h at 5 37 "C. After washing with tap water, l00pl of 1OOng/ml TRAIL-R2 Fc 6X-His tagged fusion protein (R&D Systems, Minneapolis, MN) was added to the wells and incubated for 1 h at 37 "C. After washing, I100pl of a 1/2000 dilution of Penta-His HRP Conjugate (QIAGEN, Cat# 34460) was added to each well and incubated for 1 h at 37 *C. After washing, 100pil/well of a HRPO Substrate combined TMB Peroxidase Substrate 10 [Peroxidase Solution B (Kirdgaard and Perry Labs, Cat. 50-76-00) was added. The reaction was stopped with 100 p of 2M H 2
SO
4 after 5 to 10 min. The OD was measured at 450nm and 540nm using a Molecular Devices plate reader and binding curves were generated. Figures 33 and 34 show that both the N- terminal (Figure 33A, 34A) and the C 15 terminal (Figure 33B, 34B) Hercules antibodies containing either the VH 4 4 :VL100 (ds=disulfide) or VH44:VLI00/(Gly4Ser)4 linker stabilized BHA 10 scFv show improved binding to LTpR (Figure 33) and TRAIL R2 (Figure 34), respectively, compared to Hercules containing a conventional BHA 10 scFv (wt). 20 E. Stable Expression of Bispecific Antibodies in CHO cells, Antibody Purification, and Characterization. Plasmid DNAs pXWUO27, pXWUO28, and pXWU006, pXWU035, and pXWU036 (Table 16) were used to transform DHFR-deficient CHO DG44 cells for stable production of antibody protein. Transfected cells were grown in alpha minus 25 MEM medium containing 2 mM glutamine supplemented with 10% dialyzed fetal bovine serum (Invitrogen Corporation) and enriched as a stable bulk culture pool using fluorescently labeled antibodies and reiterative fluorescent-activated cell sorting (FACS) (Brezinsky, et al. JImmunol Methods. 277(1-2):141-55 (2003)). FACS was also used to generate individual cell lines. Cell pools or cell lines were adapted to serum-free 30 conditions and scaled for antibody production. Supernatants from transfected CHO cell pools or cells lines expressing 1) C terminal Hercules with the VH 4 4 :VLI00 stabilized BHA10 scFv and 2) C- terminal Hercules with the VH44:VLl00/(Gly 4 Ser)4 linker stabilized BHA10 scFv, as well as 234 WO 2007/109254 PCT/US2007/006883 clonal CHO cell lines expressing 3) N- terminal Hercules with the VH 44 :VLl00 stabilized BHA1O scFv and 4) N- terminal Hercules with the VH44:VLI0O/(Gly4Ser)4 linker stabilized BHA10 scFv were collected and purified using Protein A Sepharose FF (6 mL) column using a PBS, 10mM EDTA running buffer. Hercules proteins were 5 eluted using 0.1 M glycine, pH 3.0 and neutralized immediately to pH 7.5-8.5 using Tris base. In addition, C-Hercules containing the conventional BHA 10 scFv was expressed and purified by Protein A chromatography for comparison. Protein A eluates from C Hercules containing the conventional BHA 10 scFv, C-terminal Hercules with the VH44:VLl00 BHAl0 scFv, and C-terminal Hercules with the VH44:VLl00/(Gly4Ser)4 10 BHA 10 scFv were examined for the presence of aggregates by analytical size exclusion chromatography (Figure 35). The chromatogram profile of C-Hercules containing the conventional BHA 10 scFv showed - 40% aggregates. In contrast, C-terminal Hercules with the VH44:VL100 BHA 10 scFv reduced the aggregate level to 20% and further stabilization achieved through C-terminal Hercules with the VH44:VL1 00/(Gly 4 Ser) 4 15 BHA 10 scFv reduced the level of aggregates to 10%. The level of aggregation appeared to depend upon the properties of the scFv and not whether it was appended to the NH- or C-terminus of the 14A2 IgG (data not shown). The protein A Hercules eluants were further purified by dialysis into 0.1 M acetate, pH 5.0, and purification by MonoS (GE Healthcare) cation exchange chromatography using an identical running buffer as the 20 dialysate. Hercules proteins were eluted using a step gradient to 0.1 M acetate, pH 5.0, 0.5 M NaCl. The MonoS eluants were collected and passed through a TosoHaas preparative SEC column to remove aggregates. Figures 36A and 36B show SDS-PAGE gels of purified N- terminal Hercules with the VH44:VL100 stabilized BHA 10 scFv and NH- terminal Hercules with the 25 VH44:VL100/(Gly4Ser)4 linker stabilized BHA10 scFv and C- terminal Hercules with the VH44:VL100 stabilized BHA10 scFv and C- terminal Hercules with the VH44:VLOO/(Gly4Ser)4 linker stabilized BHA10 scFv, respectively. The reduced lanes show the expected sizes of the heavy and light chain proteins. Importantly, there is no significant level of degraded or unwanted lower molecular weight byproducts that has 30 often been observed with Hercules containing wild type scFv domains. Figure 37 (Panels A and C) shows the analytical SEC elution profiles of NH terminal Hercules with the VH44:VL100 stabilized BHA10 scFv and NH- terminal Hercules with the VH44:V L00/(Gly4Ser)4 linker stabilized BHA 10 scFv, respectively, 235 WO 2007/109254 PCT/US2007/006883 subsequent to the initial protein A purification step. Figure 37 (Panels B and D) also shows the analytical SEC elution profiles of NH-terminal Hercules with the VH44:VL100 stabilized BHA10 scFv and NH- terminal Hercules with the VH44:VLOO/(Gly4Ser)4 linker stabilized BHA1O scFv, respectively, following preparative SEC removal of 5 residual non-monomeric protein contaminants. Similarly, Figure 38 (Panels A and C) shows the analytical SEC elution profiles of C- terminal Hercules with the Va44:VL1OO stabilized BHAI0 scFv and C- terminal Hercules with the VH44:VL100/(Gly4Ser)4 linker stabilized BHA1O scFv, respectively, subsequent to the initial protein A purification step. Figure 38 (Panels B and D) shows the analytical SEC elution profiles of C 10 terminal Hercules with the VH44:VLl00 stabilized BHA 10 scFv and C- terminal Hercules with the VH44:VLIOO/(Gly4Ser)4 linker stabilized BHAIO scFv, respectively, following preparative SEC removal of residual non-monomeric protein contaminants. These studies demonstrate that stabilization of the BHA 10 scFv by addition of either the
VH
4 4 :VLl00 disulfide or the VH44:VL100 disulfide and (Gly 4 Ser) 4 linker results in 15 preparative quantities of >98% pure, monomeric Hercules bispecific antibody that is essentially free of higher order molecular weight species. DSC studies performed under identical conditions demonstrate that C-Hercules VH44:VL100/(Gly4Ser)4 BHA 10 scFv is more thermostable than C-Hercules containing conventional BHA10 scFv (Figure 39). The denaturation profile for C-Hercules 20 containing conventional BHA10 scFv begins -5 "C lower temperature than the profile observed for the C-Hercules VH44:VLIOO/(Gly4Ser)4 BHA 10 scFv. These results suggest that the scFv may limit the overall thermostability of the Hercules molecules. This correlates with a 5 *C Tm observed increase for the VH44:VL OO/(Gly4Ser)4 BHAI 0 scFv over that of the conventional BHAI0 scFv. 25 Purified versions of the stabilized Hercules constructs were then tested for bispecific binding activity to recombinantly produced TRAIL-R2 and LTpR receptors in an ELISA assay. In this assay LTpR Ig receptor was immobilized onto plates and test samples were then incubated to permit binding to receptor. Unbound samples were removed by washing followed by a second incubation step with TRAIL R2- (His) 6 . 30 Following a wash step, doubly-bound complexes were detected with a labeled anti (His) 6 antibody. The results of this study and the effective concentration (EC50, in ptg/ml) at which each construct are depicted in Figure 40. Figure 40 shows that both the N- terminal and C- terminal Hercules antibodies with either the VH 4 4 :VL 00 236 WO 2007/109254 PCT/US2007/006883 (ds=disulfide) or VH44:VL100/(Gly 4 Ser) 4 linker stabilized BHA 10 scFv demonstrate clear bispecific binding to both LTPR and TRAIL R2. A multivalent (pentameric) version of an LTPR-binding antibody (CBE 11) was used as a control to demonstrate that single recognition of LTpR alone does not induce signal in the binding assay. 5 Example 8. Scale-up Production of Stabilized Bispecific "Hercules" Antibodies in CHO Cells DHFR-deficient CHO cell lines stably transfected with plasmid DNAs 10 pXWUO28 and pXWU036 (Table 16) were screened to select for single cell isolates that are capable of expressing high levels of solubilized and properly-folded Hercules molecules that were stabilized using the methods of the invention. The cell screening methods employed the fluorescence-activated cell sorting (FACS) analysis of Brezinky et al. (Brezinky et al., J Immunol Meth (2003). 277:141-155). Briefly, fluorescently 15 tagged anti-Hercules antibodies were used to label CHO cells exhibiting transient expression of Hercules antibodies on their surface. Cells exhibiting a signature fluorescence intensity were then selected by tailoring the gating of the cell sorter apparatus to that signature. Single cells exhibiting high levels of productivity were thereby selected and adapted to serum-free conditions to establish stable producer cell 20 lines. The producer cell lines were subsequently scaled-up for production and purification of the bispecific antibody protein. 80 L of N- terminal Hercules containing VH44:VLOO/(Gly4Ser)4 BH A10 scFv (XWUO28) supernatant from an 11 day bioreactor run was harvested and precleared by ultrafiltration. The bispecific antibody was captured by Protein A chromatography and 25 eluted in 1218 mis volume. The Protein A fraction was further purified in two separate batches by anion exchange chromatography followed by preparative size exclusion chromatography. The first batch yielded 825.5 mg at a concentration of 4.85 mg/ml in PBS with 98.9% purity and an endotoxin load of 0.13 EU/mg. The second batch yielded 318 mg at a concentration of 10.3 mg/ml with a purity of 99.1% and an endotoxin load 30 of 2.37 EU/mg. A total of 1143.5 mg highly purified, monomeric N- terminal Hercules containing VH44:VLIOO/(Gly4Ser)4 BHAl0 scFv was recovered. 24 L of C- terminal Hercules containing VH44:VLIOO/(Gly 4 Ser) 4 BHA10 scFv (XWU036) supernatant from an 11 day bioreactor run was harvested and precleared by ultrafiltration. The bispecific antibody was purified in two separate batches as described 237 WO 2007/109254 PCT/US2007/006883 above using Protein A chromatography to capture crude product followed by anion exchange chromatography and preparative size exclusion chromatography. The first batch yielded 985.5 mg at a concentration of 13 mg/ml with 98.5% purity and an endotoxin load of 0.01 EU/mg. The second batch yielded 570 mg at a concentration of 5 11.42 mg/mI with a purity of 99 % and an endotoxin load of 1.91 EU/mg. A total of 1555.5 mg highly purified, monomeric C- terminal Hercules containing VH44:VL100/(Gly4Ser)4 BHA1O scFv was recovered. A CHO cell line producing the XWU054 bispecific antibody was isolated and found to produce 21.5 mg/L. This is within the expected range for research grade 10 unamplified CHO cell lines and we would anticipate much higher productivity if we were to generate a Production Cell Line. These scale-up production studies were conducted with non-amplified CHO cell 15 lines and yet resulted in yields exceeding 1 gm of high quality, biophysically stable bispecific antibodies. Presumably a cell line amplification strategy would greatly enhance the cellular productivity of the transfected CHO cell lines enabling the development of processes suitable for commercial applications. These studies exemplify the utility of the methods described in this invention for enabling the scale-up production 20 of stable bispecific antibodies in a cell line (e.g. CHO) suitable for manufacturing. Example 9. Biological Activity of Bispecific "Hercules" Antibodies Tumor cell lines WiDr, (ATCC CCL-218) a human colon carcinoma cell line, 25 Mel 80, (ATCC HTB 33) a human cervical epithelial carcinoma cell line, and MDA23 1, (Dr. Dajun Yang, University of Michigan) a human breast carcinoma cell line were cultured in MEM-Earles with 10% FCS, 2 mM L-Glutamine,IX non-essential amino acids, 0.5 mM sodium pyruvate, and Penicillin/Streptomycin. Tumor cell lines were rinsed once with PBS and cells released by digestion with trypsin. Cells were collected 30 by centrifugation, resuspended in complete media, counted and 96-well tissue culture plates seeded at 5000 cells/well for WiDr and Me180 and 1500 cell/well for MDA231. Human IFNy (Biogen Idec, Corp) is added to the cell suspensions to result in a final cytokine concentration of 80 U/ml for WiDr and MDA231 and 50 U/ml for Mel 80. 50 pl of the tumor cell/ IFNy suspension were mixed with 50 pl of 2X concentrated 3-fold 238 .
WO 2007/109254 PCT/US2007/006883 serial dilutions of test antibodies prepared in complete media. The final concentrations of test antibodies typically ranged from 5000 pM to 0.07 pM. Cells were grown for 4 days (WiDr & Mel 80) or 3 days (MDA23 1) at 37 "C in a 5% CO 2 humidified chamber and cell killing assessed by the addition of 20 p!L/well Promega CellTiter 96 AQueous 5 One Solution Cell Proliferation Assay reagent (Promega Corporation, Madison, WI). Plates were read in a micotiter plate reader at 490 nM (Spectromax Plus, Molecular Devices, Sunnyvale CA). Data was graphed using Microsoft Excel (Microsoft Inc, WA). Results from these studies are shown in Figures 41A-41D. Figure 41A shows that 14A2 IgG antibody had modest activity in inhibiting growth of WiDr tumor cells 10 and in combination with BHA10 IgG exhibited a slight increase in anti-tumor cell activity compared to BHA10 IgG alone. In contrast, both of the bispecific Hercules antibodies showed enhanced tumor cell killing of the WiDr cells. Figure 41B shows that both the 14A2 IgG and BHA10 IgG antibodies had modest, if not negligible, activity in inhibiting growth of Mel 80 tumor cells as single agents or in combination. In 15 contrast, both of the bispecific Hercules antibodies showed tumor cell killing of the Me180 cells. Figure 41C shows that both the 14A2 IgG and BHA10 IgG antibodies had negligible activity in inhibiting growth of MDA231 tumor cells as single agents and perhaps some activity when used in combination. In contrast, the bispecific Hercules antibody XWU036 (C- terminal Hercules with the VH44:VLl0O/(Gly 4 Ser) 4 linker 20 stabilized BHA1O scFv) showed tumor cell killing of the MDA231 cells. Figure 41D shows that none of the antibodies demonstrate any activity towards control cultured human umbilical vein endothelial cells (HUVEC) demonstrating that the cell killing activity of the bispecific antibodies is not indescriminant. HUVEC shows positive staining for LTPR and TRAIL-R2 by FACS analysis (data not shown) indicating 25 presence of receptors and that the activity of the bispecific antibody may be dependent on triggering specific or unique pathway components in tumor cells. Example 10. Stability Studies of Bispecific "Hercules" Antibodies 30 Real-time stability studies of the N- and C-terminal bispecific antibody samples XWUO28 (BHA10 scFv VH44:VL100/(Gly4Ser)4N- Hercules) and XWU036 (BHA10 scFv VH4 4 :VLI 00/(Gly 4 Ser) 4 C-Hercules) stored at 2-8 "C for three months were conducted. Protein quality was assessed for (1) aggregation, (2) precipitation, (3) polypeptide cleavage or proteolysis, and (4) post-translational modifications such as 239 WO 2007/109254 PCT/US2007/006883 deamidation or oxidation. High and low concentration N- and C-terminal bispecific antibody samples used for the studies were XWUO28 at 1.8 mg/mL (Low) and 10.3 mg/mL (High) and XWU036 at 5.0 mg/mL (Low) and 11.4 mg/mL (High). Antibodies were formulated in 5 PBS. For the stability study, initial (T=0), intermediate (T= I wk, 2 wk, 1 mo, 2 mo) and final (T=3 mo) time point samples were analyzed immediately following sample collection. In addition, initial (T=0) samples were frozen and stored at -70 "C until thawed for secondary analyses at the end of the 3-month study. BHA10 IgG (8.7 mg/mL) was used as control and was similarly handled. 10 A. Protein agregation and precipitation analysis Protein aggregation or precipitation was monitored using analytical size exclusion chromatography (SEC) linked to inline light scattering and refractive index detectors (Wyatt Technologies, MiniDawn and rEX, respectively). SEC/Light 15 scattering analysis showed no evidence of aggregation or material loss that might occur through precipitation. The SEC elution profiles of the initial, T=0, and final, T= 3 mo, time points for XWUO28, XWU036, and the BHA 10 antibody were nearly identical (Figures 56A, 56B, and 56C, respectively). XWU028 and XWU036 samples both accumulated -1% aggregated material by the completion of the study; however 20 detection of these aggregates was near the lower limit of detection determined for this method (Table 17). Aggregate formation was independent of protein concentration. Based on the molecular masses determined using light scattering detection, it is possible that that the low levels of aggregates in the test samples are likely a product of monomeric species converting to dimers. Neither of the bispecific antibodies samples, 25 XWUO28 or XWUO36, accumulated detectable levels of high-order aggregates, even though this type of aggregate would have easily been observed using these methods. The BHA10 antibody demonstrated no increase in aggregates over the 3-month course of the study (Table 17). 30 Table 17. Percent monomer detected using analytical size exclusion chromatography. T=0 T=I week T=2 week T=I mo T=2 mo T=3 mo SEQID#49 Low 99.0 98.7 98.4 98.3 98.2 97.9 SEQID#49 High 99.1 99.0 98.9 98.6 98.2 98.1 240 WO 2007/109254 PCT/US2007/006883 SEQID#51 Low 98.8 97.5 97.5 97.4 97.2 96.7 SEQID#51 High 98.9 98.8 98.7 98.5 98.1 96.7 BHA10 IgG 97.4 97.5 97.5 97.5 97.3 97.5 B. Proteolysis and Post-translational modification analyses Proteolysis was monitored using SDS-PAGE and liquid chromatography/mass spectroscopy (LC/MS) intact mass analysis (Agilent LC/MSD TOF'coupled to an 5 Agilent 1100 LC system via an electrospray interface). For SDS-PAGE analysis, 5 ug of protein was loaded per lane. Reduced samples were prepared in IX Tris-glycine sample buffer containing 5% p-mercaptoethanol. XWUO28 and XWU036 showed no evidence of proteolysis over the 3-month storage period at 2-8 *C as determined by non reducing and reducing SDS-PAGE analysis (Figures 57A and B). Similarly, XWU028 10 and XWU036 showed no evidence of lower molecular weight proteolytic products over the 3-month storage period at 2-8 "C as determined by LC/MS analysis (Figure 58). Post-translational modifications were monitored using LC/MS. T=0, T=1 mo, and T=3 mo samples were analyzed in both the non-reduced and reduced form. Reduced samples were prepared by treating in 50mM DTT/4M guanidine hydrochloride 15 for 1 hour at 37"C. HPLC Buffer A consists of 0.03% TFA in water and HPLC buffer B contains 0.025% TFA in acetonitrile. Flow rate was kept constant at 100pl per minute. 7.5pg of each sample (reduced and non-reduced) was injected onto a 2.1 x 50mm C4 column and analyzed by Agilent ESI-TOF. A bind-and-elute method was used for non reduced samples while a gradient method was used for reduced samples. Spectra were 20 obtained using the Analyst and deconvoluted using the MaxEnt 1 software packages included with the instrumentation. Deconvoluted mass spectra of XWUO28 and XWUO36 (high concentrations) at T=0, T=1 mo, and T=3 mo are shown in Figure 58. Neither XWUO28, nor XWU036, demonstrated any detectable changes in their mass spectra, indicating absence of post-translational modification that could potentially 25 adversely affect bispecific antibody function or stability. Taken together, the data indicate that both the N- and C-terminal bispecific antibodies containing stabilized scFv domains are stable under extended storage conditions such as those required for biological drug products. These results are particularly encouraging because these stability studies were conducted with the N- and 30 C-terminal bispecific antibodies prepared in a simple buffer (PBS) and not with 241 WO 2007/109254 PCT/US2007/006883 solutions prepared with optimal formulations. Example 11. In Vivo Pharmacokinetic Activity and Serum Stability of Bispecific 5 "Hercules" Antibodies A single bolus injection of 10 mg/kg (1 mg/ml) of N- terminal Hercules (XWUO28) or C- terminal Hercules .(XWU036) diluted in phosphate-buffered saline (PBS) was administered intraperitoneally into male CB17- scid mice. Mice were 10 sacrificed at 0, 0.5, 2, 6, 24, 48, 72, 96, 168, 240, and 336 hours post-injection using three mice per timepoint for each bispecific antibody. Serum samples were prepared for analysis by ELISA assay to quantify levels of the bispecific antibodies. ELISA plates were coated with goat-anti-human IgG, blocked with PBS/1%BSA, and dilutions of serum containing the bispecific antibodies were serially diluted in PBS/i %BSA, added 15 to the plates and incubated. Captured antibodies were detected with a goat-anti-human kappa chain-HRP-linked antibody. Results of the pharmacokinetic study, are shown in Table 18. N- terminal Hercules (XWUO28) has an elimination half-life (tj) of 10.3 days with a peak serum concentration of 85.67 tg/ml. C- terminal Hercules (XWU036) has .a longer elimination half-life (t%) of 15.1 days with a peak serum concentration of 105.67 20 pg/ml. Both molecules have similar volumes of distribution (Vd) though these values have not been adjusted for bioavailability. Table 18. Pharmacokinetic Parameters N- terminal Hercules C- terminal Hercules (XWUO28) (XWU036) Parameter Units Value Value T1/2 hr 247.9386 362.3607 Cmax ug/ml 85.67 105.67 Vd * Lkg 0.1707 0.1748 Cl * ml/min/kg 0.008 0.0056 1 L not adjusted for bioavailability 25 Serum samples were also analyzed in the bispecific binding ELISA described in Example 7. In this assay, the serum samples containing N- terminal Hercules (XWU028) or C- terminal Hercules (XWUO36) described above were tested for bispecific binding activity to recombinantly produced TRAIL-R2 and LTpR receptors in an ELISA assay. Figures 59A and B show that serum samples from N- and C- terminal 242 WO 2007/109254 PCT/US2007/006883 Hercules treated mice contain antibodies that bispecifically bind to both TRAIL-R2 and LTsR and closely parallel the elimination profiles observed from the PK study indicating that the bispecific antibodies are remaining intact under physiological conditions for extended periods of time. 5 Example 12. In Vivo Biological Activity of Bispecific "Hercules" Antibodies (A) In Vivo Biological Activity of Bispecific "Hercules" Antibodies in a Colon Cancer Tumor Model 10 WiDr human colon carcinoma (2x1OE6 cells per mouse) cells were implanted subcutaneously into 125 athymic nude mice. The tumors were grown until they reached approximately 100 mg, at which point a total of 70 mice were selected for the study, divided into 7 groups of 10 mice. IP treatments were administered, beginning on day 13 post-implantation, as follows: Group 1 = pyrogen-free PBS; Group 2= CBEl 1, 2 15 mg/kg, Ix/wk; Group 3 = hBHA10, 2 mg/kg, 2x/wk; Group 4 = ch14A2, 2 mg/kg, 2x/wk; Group 5 = Hercules-II XWUO28, 2 mg/kg, 1x/wk; Group 6 = Hercules-II XWU036, 2 mg/kg, 1x/wk; Group 7 = hBHA10 + chl4A2, 1 mg/kg each, 2x/wk. Tumor sizes and body weights were recorded bi-weekly. Study was terminated when average tumor size of vehicle group reached approximately 2000 mg. Tumor volume 20 was calculated using the formula: (L x W 2 /2). Interim analyses of mice treated with the XWU028 and XWU036 showed significant anti-tumor activity (p<0.001) for both bispecific antibodies compared to PBS vehicle control (Figure 60). For XWUO28, significant anti-tumor responses (p<0.05) were observed compared to responses with single hBHAlO or chl4A2 mAb treatment. For XWU036, a 25 significant anti-tumor response (p<0.05) was observed compared to the response with ch 14A2 mAb treatment and, notably, a greater significant response (p<0.0 1) compared to hBHAIO mAb. Importantly, the bispecific antibodies showed superior in vivo anti tumor activity though administered only once per week suggesting that the stability enhancements described in this invention result in improved antibody properties and 30 physical stability under physiological conditions. (B) In Vivo Biological Activity of Bispecific "Hercules" Antibodies in a Breast Cancer Tumor Model 243 WO 2007/109254 PCT/US2007/006883 MDA-MB-231 human breast carcinoma cells were implanted subcutaneously into 135 athymic nude mice (2x 10E6 cells per mouse). The tumors were grown until day 13 at which point 75 tumor-bearing mice with an average size of approximately 168 mg were assigned to treatment (N=10) and vehicle control (N=15) groups. Mice 5 received antibodies and vehicle IP starting at day 13. Exemplary groups are shown as follows: Group I = pyrogen-free PBS; 1x/wk; Group 2 = hBHA10, 2 mg/kg, 2x/wk; Group 3 = chl4A2, 2 mg/kg, 2x/wk; Group 4 = Hercules-II XWUO36, 2 mg/kg, lx/wk; Group 5 = hBHA 10 + chl4A2, 1 mg/kg each, 2x/wk. Tumor sizes and body weights were recorded bi-weekly. Study was terminated when average tumor size of vehicle 10 group reached approximately 2800 mg. Tumor volume was calculated using the formula: (L x W 2 /2). XWUO36 demonstrated statistically significant (p<0.001) anti tumor activity compared to either single treatment with hBHA 10 or ch 1 4A2 mAbs or to treatment with a mixture of the two antibodies. Importantly, the bispecific antibody XWU036 demonstrated good anti-tumor activity in vivo administered on a once per 15 week dosing schedule suggesting that the stability enhancements described in this invention result in improved antibody properties and physical stability under physiological conditions (see Figure 86). Example 13. Production of Stabilized Bispecific "Hercules" Antibodies Containing 20 Non-covalent Stabilizing scFv Mutations BHA 10 scFvs of the invention containing non-covalent stabilizing mutations alone and in combination with a VH44-VL100 disulfide bond were used to construct stabilized bispecific C-Hercules antibodies comprising a fusion of chimeric 14A2 IgG 25 antibody that binds-to TRAIL R2 receptor with a BHA1O scFv that binds to LTPR. The bispecific antibodies were constructed as C- terminal BHA10 scFv fusions using methods essentially as described in Example 7. A. Construction of C-Hercules bispecific antibody with BHA10 VH S16E 30 + V 1 S46L and VH44-VL100/VH S16E + VL S46L scFvs PCR was used to amplify variant BHA10 scFv gene fragments from plasmid DNAs pIEH-050 (parent plasmid of the high expression plasmid pIEH076) and pIEH 244 WO 2007/109254 PCT/US2007/006883 052 (parent plasmid of the high expression plasmid pIEH080) containing the BHAI 0 VH S16E + VL S46L and VH 4 4-VL100NH S16E + VL S46L scFvs, respectively using the oligonucleotide primers described in Table 19. The variant BHA]10 scFv gene fragments were gel isolated. Due to a BamHI site in the linker region of plasmids pIEH-050 and 5 pIEH-052, the gene fragments were digested with pPuM I and Kpn I restriction endonucleases and ligated to the modified plasmid pN5KG 1 digested with the same restriction endonucleases resulting in a fusion product of the stabilized anti-LTBR (BHA 10) scFvs to the carboxyl terminus of the anti-TRAILR2 (14A2) antibody CH3 domain through a 16 amino acid Ser(Gly 4 Ser) 3 linker. Correct sequences were 10 confirmed by DNA sequence analysis. Table 19. Oligonucleotides for PCR amplification of BHA 10 VH S16E + VL S46L and VH44-V100/VH S16E + VL S46L scFvs. pXWU006-Fl 51 (SEQ ID NO:56) GGGGGTGGATCCGGTGGAGGGGGCTCCGGCGGTGGCGG GTCC CAGGTCCAACTGGTGCAGTCTG -3' XWU006-Rl 5'- GTTAACGGATCCTCATTTGATCTCCACCTTGG -3' (SEQ ID NO:57) 15 Fusion of BHAIO scFv VH S16E + VL S46L gene sequence to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWU054. Fusion of BHA10 scFv VH44-VL100/VH SI6E + VL S46L gene sequence to the carboxyl terminus of the anti-TRAIL R2 antibody heavy chain gene sequence produced plasmid pXWU055. 20 The ligation mixtures were used to transform E. coli strain TOP 10 competent cells (Invitrogen Corporation, Carlsbad, CA). E. coli colonies transformed to ampicillin drug resistance were screened for presence of inserts. DNA sequence analysis confirmed the correct sequence of the final constructs. The chimeric 14A2 light chain DNA (SEQ ID NO:28) and amino acid sequences (SEQ ID NO:29) are shown in Figures 25 15A and 15B. The heavy chain DNA (SEQ ID NO:52) and amino acid sequence (SEQ ID NO:53) for C-Hercules BHAI0 scFv VH S16E + VL S46L bispecific antibody are shown in Figures 61 and 62, respectively. The heavy chain DNA (SEQ ID NO: 54) and amino acid sequence (SEQ ID NO:55) for C-Hercules BHAl0 scFv VH 4 4 -VLI00/VH SI6E + VL S46L bispecific antibody are shown in Figures 63 and 64, respectively. 245 WO 2007/109254 PCT/US2007/006883 The heavy chains employed the same signal peptide as used previously. B. Stable Expression of Bispecific Antibodies in CHO cells, Antibody Purification, and Characterization. 5 DHFR-deficient CHO cell lines stably transfected with plasmid DNAs pXWU054 and pXWU055 and adapted to serum-free conditions were scaled for production of bispecific antibody protein and proteins purified as described in Example 8. Protein A eluates from supernatants containing C-Hercules with the stabilized VH 10 S16E + VL S46L BH A10 scFv and C- bispecific antibody with the stabilized VH44 VL100HV S16E + VL S46L BHA10 scFv were examined for the presence of aggregates by analytical size exclusion chromatography (Figure 65). The chromatogram profile of C-Hercules containing the conventional BHA 10 scFv showed ~ 40% aggregates. In contrast, C-terminal Hercules with the stabilized VH SI 6E + VL S46L BHA 10 scFv 15 significantly reduced aggregates to levels comparable to that observed with standard IgGs. Figures 66A and 66B show SDS-PAGE gels of purified C-terminal Hercules with the VH S16E + VL S46L BHA10 scFv and C-terminal Hercules with the Vt44:VL 00/VH S 16E + VL S46L BHA 10 scFv, respectively. The reduced lanes show 20 the expected sizes of the heavy and light chain proteins. Importantly, there is no significant level of degraded or unwanted lower molecular weight byproducts that has often been observed with Hercules containing wild type scFv domains. Figures 67A and 67B show the analytical SEC elution profiles of C-terminal Hercules with the VH S1 6E + VL S46L BHA 10 scFv and C-terminal Hercules with the 25 VH44:VL10 0 I VH S16E + VL S46L BHA10 scFv, respectively, subsequent to the initial protein A purification step. These studies demonstrate that stabilization of the BHA10 scFv by addition of VH S16E + VL S46L mutations alone and in combination with the VH44:VL100 disulfide results in preparative quantities of >98% pure, monomeric Hercules bispecific antibody that is essentially free of higher order molecular weight 30 species. Example 14. Biological Activity of Bispecific "Hercules" Antibodies Containing Non-covalent Stabilizing scFv Mutations 246 WO 2007/109254 PCT/US2007/006883 In vitro biological activity of C-terminal Hercules with the VH Si 6E + VL S46L BHA10 scFv (XWU054) and C-terminal Hercules with the VH 4 4 :VL1OO/VH SI6E + VL S46L BHA1O scFv (XWU055) were tested in the tumor cell proliferation assay as described in Example 7. 5 Results from these studies are shown in Figure 68. Figure 68B shows that 14A2 IgG antibody had moderate activity in inhibiting growth of WiDr tumor cells and in combination with BHA 10 IgG exhibited a slight increase in anti-tumor cell activity compared to BHA 10 IgG alone. In contrast, both of the bispecific Hercules antibodies (XWU054 and XWU055) showed enhanced tumor cell killing of the WiDr cells 10 comparable to that observed with the XWU036 molecule (C- terminal Hercules with the VH44:VLl00/(Gly4Ser)4 linker stabilized BHA10 scFv) described in Example 10. Figure 68A shows that both the 14A2 IgG and BHA10 IgG antibodies had negligible activity in inhibiting growth of MDA231 tumor cells as single agents as well as when used in combination. In contrast, the bispecific Hercules antibodies (XWU054 and 15 XWU055) showed enhanced tumor cell killing of the MDA231 cells comparable to that observed with the XWU036 molecule (C- terminal Hercules with the VH44:VLOO/(Gly4Ser)4 linker stabilized BHA 10 scFv) described in Example 10. Example 15: Utilization of Differential Scanning Calorimetry (DSC) to Measure 20 Thermal Stability of Human or Humanized Antibody Sequences To assess the range of apparent stabilities of antibodies, differential scanning calorimetry (DSC) was performed on a set of 17 human or humanized BIIB antibodies. The BIIB antibodies were dialyzed exhaustively against a 20 mM sodium citrate, 150 mM sodium chloride buffer at pH 6.0. Dialysates were used within the reference 25 cell of a calorimeter to define the baseline of each antibody scan. Subsequent to dialysis, the concentrations of BIIBI -17 were measured by 280 nm absorbance (Pace et al., Protein Sci. 4, 2411-2423, 1995). Antibody solutions used for DSC were universally prepared at I mg/mL by diluting the concentrated stocks with their dialysates. Scans were performed using an automated capillary differential scanning 30 calorimeter (capDSC, MicroCal, LLC). Protein and reference solutions were sampled automatically from 96-well plates using the robotic attachment. Prior to each protein scan, 2 scans were performed with buffer in the sample cell and used for background subtraction. A single cleaning scan was performed using 5% Liquinox after every 247 WO 2007/109254 PCT/US2007/006883 protein scan. After every scan, the instrument automatically rinsed both the reference and sample cells with 3X-2 mL distilled deionized H 2 0 containing 0.01% sodium azide. Scans were performed at 1 "C/min using the medium feedback mode for enhanced peak resolution. The scan range was 20-95 "C. All 96-well plates containing protein were 5 stored within the instrument at 6 *C. Scans were analyzed using the Origin software supplied by the manufacturer. Subsequent to background subtraction, non-zero baselines were corrected-using a third order polynomial. The unfolding transitions of each IgGI domain were deconvoluted by fitting the multi-peak curves to 3 separate transitions. Recombinant IgGI and IgG4 Fcy 10 domains were utilized to identify the CH 2 and CH 3 peaks within each transition. The remaining transition(s) was assumed to correspond to the unfolding of the Fab portion of the antibody. Most of the Fab portions of the 17 antibodies unfolded in apparent single transitions with midpoints of thermal unfolding (Tms) ranging from 57 to 82 *C 15 depending on the unique properties of each individual Fv. Only three antibodies, BIIB15, BIIB16, and BIIB17 exhibited 4 transitions. A typical IgGI DSC trace is shown in Figure 69A. The Cpma (i.e. the height of a DSC peak) for Fab portions is generally significantly greater than that of the CH 2 or CH 3 domains. This is reasonable considering the Fab protein mass within an antibody is 4-fold greater than that of either 20 the CH 2 or the CH 3 domains, and the Fab portion of the antibody also contains buried surface area between 4 domains as opposed to the CH2 and CH 3 domains whose transitions each involve the disruption of only a single homodimeric interface. In general, the transitions observed for the CH2 and CH 3 domains of each IgG 1 antibody superimposed well upon one another. The same is true for the CH2 and CH3 domains of 25 the 3 IgG4s. For these reasons, the Fab transitions of each IgG 1 or IgG4 molecule were readily identified. One antibody (BIIB18) was not studied by DSC. BIIB18 expressed poorly and never resulted in soluble/non-aggregated material. This antibody was included in the sequence analyses described in Example 16. One of the antibodies, BIIB7, was analyzed in both the human IgGI and IgG4 30 format. The apparent Fab thermostabilities of BIIB7 in the IgGI and IgG4 format as measured by the midpoints of their thermal unfolding transitions (TM) were very similar (see Figure 69B). Variation in the constant domain subclass did not affect the Fab TM or Fab calorimetric unfolding enthalpy. 248 WO 2007/109254 PCT/US2007/006883 The overlapping transitions corresponding to the CH2 and CH 3 domains of the IgG4 Fc occur at significantly lower temperatures than those of IgG I and artificially make the Fab transitions look -1-2 "C lower for IgG4. Once the IgG transitions are completely deconvoluted, the Fab TMs are much closer than they appear in the DSC 5 traces (ATM < 1 "C). The fact that the Fab DSC curves of BIIB7 in the IgG1 and IgG4 formats were highly similar suggests that the thermostability of the IgGI Fabs and IgG4 Fabs can be directly compared. DSC measurements with 17 of the humanized BIIB antibodies indicate that the Fv regions heavily influenced the apparent stability of their respective Fabs. For 10 example, Figure 70 demonstrates the very different DSC curves obtained for BIIBI, BIIB4, BIIB6 and BIIBl6. All four antibodies are IgGI, and the only sequence differences between them reside in the Fv region. The position and magnitude of the
CH
2 and CH3 DSC transitions of all 17 antibodies are independent of the Fab to which they are attached (see Figure 70). The range of Fab TMs for the top 17 antibodies in 15 Table 20 was between 57.2 and 81.6 "C (+24.4 0C), with BIIBI exhibiting the highest apparent Fab stability and BIIB 17 the lowest. The three antibodies with the lowest measureable Fab TM values, BIIB1 5, BIIB 16, and BIIB 17, demonstrate a disconnect between domain unfolding transitions within the Fab (See Figure 70 for BIIBI6's DSC curve). The expected Fab peak was 20 split into two separate transitions. The lower TM transitions were listed in Table 20, but a second transition, which may represent the unfolding of the CH /CL structural unit, occurred with Tms of 74, 72 and 70 0C for BIIB 15, BIIB16 and BIIB 17, respectively. BIIB18 was included in the study as an example of an antibody which would not express and whose VH sequence diverged significantly from consensus (see Example 16). 25 The DSC analyses clearly showed that the unique properties of individual Fv regions can highly attenuate the apparent TMs of an entire Fab portion of an antibody. In the extreme cases of BIIB 15, BIIB 16 and BIIB 17, the apparent Fv stabilities were low enough that the unfolding transitions of the domains within the Fab portion became uncoupled from one another. In order to maintain an apparent single transition for the 30 unfolding of all four domains of the Fab, there may be a minimum overall TM limit (>70 *C for IgG 1s under the scan conditions described here), otherwise multiple transitions appear. It would be convenient to assign the low temperature unfolding transitions of BIIB15, BIIB 16, and BIIB17 to the Fv and the second, higher temperature transitions, to 249 WO 2007/109254 PCT/US2007/006883 the CHl/CL domain. However, Ewert and coworkers have shown that VH/VL domain unfolding is not always coupled and can depend on subclass, VH/VL stability, and VH/VL complementarity (Ewert et al., .J. Mol. Biol. 325: 531-553, 2003). In fact, very few of their VHNL scFv constructs with consensus-derived frameworks exhibited cooperative 5 unfolding. Therefore, it is likely that, as the transitions uncouple, the unfolding scenario may be more complicated than a simple split into two transitions representing the Fv region and CH1/CL-regions. The apparent stabilities as measured by DSC for the BIIB7 IgGI and IgG4 Fabs were highly similar even though there are substantial inherent differences in CH 1 10 sequence and disulfide-bonding between the two subclasses. There are many possible explanations for this such as (1) the biophysical properties of the Fv set the thermostability ceiling of the BIIB7 Fab regardless of whether the CHI is IgGI or IgG4, (2) the apparent stability of the CHI/CL region is identical between IgGI and IgG4, or (3) the CH1/CL structural unit does not unfold within the timeframe of the DSC 15 experiment. The fact that the BIIB7 Fab unfolded at a single Tm and with a similar calorimetric enthalpy in both the IgGI and IgG4 formats suggests a potential stability dependence upon the Fv and not the constant regions 250 WO 2007/109254 PCT/US2007/006883 m cc l-4 r4 00 e en rJ r e 00 14 C > eAGn w n 00 w ~ e w n w N 0 n 00 N =n ON ON el e s N ( 19 0 > 0 00 14 " en en r-- en ON 00 ;'I- 'gr ml en '.0 -;r Gn r- w0 0w 00 - =0 r- 00 as 0'. ON a' , 0 00 00 00 O~~ n ell elc C4 I V* a, CD' CD 0'D C' 0 0 mien e >0 en en on r iU . ~ ~ . . .+ + + V) - - - -: -l " " 6 i (N -i CD 6 - =N 6 " - "N (Ni (Ni (N > M - w - - 0 '.'r- CN NO r. - r- al r- C7% C7' O a , 0 = 0 00 eq %N 0 0'- - e 00 t-' a' -, "- - Mr 0" %Q IT 0> > > G U, ( r- ON '.0 cr r- 4n' a' 110 r- No007 %.0 ' a O 00 00 > >> n > e > e en>> r-e m e 00 = W. 0 0. 0 4a'(a > 0 m ' > . > ' > en en> > en > a' > >' a' Z -) C) C) 0 sC Cl4 ON '0C.)' c> a = ~ ~ ~ ~ ~ 0 e en 'CS Ut 7 a 0 D 2, 270 6 00 > - xl en >_ U, >'<.0 12- 0 0------------------ C25 WO 2007/109254 PCT/US2007/006883 Example 16: Utilization of Consensus Scoring to identify Antibody variable region with sub-optimal stability Consensus scoring was utilized as a method for identifying which of the BIIB antibodies contained significantly large numbers of non-optimal amino acids within their 5 Fv regions. The scoring assesses the relative drift from consensus VH and VK sequences due to hypersomatic mutations and evolutionary germline variations. The DSC measurements for the 17 antibodies were used to qualify the ability of the consensus scoring approach for predicting poor antibody stability. The reference set of mammalian VH and V, kappa sequences used to derive the 10 consensus sequence for scoring and the individual amino acid frequencies at each residue position were collected, sorted, and culled as described previously (Demarest et al., J Mol. Biol. 335: 41-48, 2004). The mammalian reference sets were naively constructed to include V-genes from various mammals in order to obtain diversity via the evolutionary drift between species. The VH mammalian reference set contains 61 VH 15 sequences primarily from NCBI and TIGR representing a total of 17 different mammalian species. The V, mammalian reference set contains 53 V" sequences from 13 different mammalian species. Human VH and V, sequences for building a reference set (or for use as test sequences) were gathered from the NCBI database using the search criteria "homo 20 sapiens antibody variable heavy chain" and "homo sapiens antibody kappa variable light chain", respectively. A total of 182 human V-gene sequences, 114 VH and 68 VK, were semi-randomly selected from the NCBI database for comparison against mammalian VH and V, databases. The subclass distribution of human VH and V, sequences obtained by semi-random cherry-picking of the NCBI database yielded the expected subclass 25 representation based on natural germline VH or V, usage (Guigou et al., 1990) suggesting little bias in the way sequences were chosen from the NCBI. The sequences were then categorized into their individual variable domain subclasses (VH1-VH7 and VK 1- V, 4) by ClustalW alignment against the subclass consensus sequences obtained from Aho's Amazing Atlas of Antibody Anatomy 30 website. The subclasses were additionally confirmed by alignment against the publicly available human germline sequences (Lefranc et al., Nucleic Acids Res. 27, 209-212, 1999). No more than 5 human sequences from any one multi-sequence NCBI submission were included within the reference sets to reduce potential bias introduced - 252 - WO 2007/109254 PCT/US2007/006883 by primer sets or antigen-based V-gene usage. The sequence gathering was performed naively; therefore, there was a natural abundance of VH3 and VI sequences after subclass grouping. The average score and standard deviation (described below and in Table 20) were calculated for each subclass using 25 V 1 l, 3 VH 2 , 44 VH3, 34 VH 4 , 5 5 VH5, 3 VH7, 29 V, 1, 4 V. 2, 19 V. 3 and 16 V.4 human sequences. Only 2 of the 18 antibodies (BIIB5 and BIIB 14) contained VH domains with minimal subgroup sequence populations (i.e. VH2, VH5 or VH7). Four of the 18 antibodies (B133, BIIB4, BIIB17 and BIIB 18) contained V,,2 subclass variable domains whose subclass was underrepresented in sequence space and whose scoring accuracy was ill defined. No 10 natural VH6 sequences were found within the human Vi sequence reference set. This was not a concern as none of the BIIB antibody constructs contained a VH6 domain. The variable domain sequences of the 18 BIIB antibodies and the 182 human VH and V, sequences obtained from the NCBI were scored against the reference set of mammalian antibody sequences. All VH and VK sequences were truncated prior to 15 CDR3 (two residues after the consensus YYC residues) to reduce the unpredictability introduced by the hypervariable nature of the V-(D-)J-C joining region. Residue frequencies (p,(r)), at every position within the mammalian database were calculated by summing the number of times each amino acid (A,C,D... V,W,Y,-) occurs at each residue position within the database divided by the total number of sequences. Scores 20 were evaluated for each human VH and V, test sequence using the following formula: score = h (r) c, (r) wherein c,(r) equals the consensus residue frequency at each position of the mammalian variable domain databases and h,(r) equals the test residue frequency of each amino acid of the human variable domain test sequence. Perfect consenus scores 25 for the consensus sequences of the mammalian databases were 104 for VH and 100 for VI. A. Scoring Random VH and VK Sequences Using a Mammalian V-Gene Database VH and V, subclass consensus sequences were scored against the mammalian 30 databases consensus and plotted in Figures 71A and 71C, respectively. The VH 3 subclass consensus scored only 1 point lower than the hypothetical "perfect" score of the mammalian database consensus and differed from the mammalian database consensus - 253 - WO 2007/109254 PCT/US2007/006883 by a total of three residues. Thus, it is apparent that the mammalian database is biased towards VH3-like sequences. All other human VH subclass consensus sequences scored significantly lower. A bias towards V4 sequences was observed for the mammalian VK database (Figure 71C). The human or humanized.BIIB sequence scores were analyzed 5 based on the performance of human NCBI sequences of comparable subclass. Distributions of the individual VH sequence scores are shown in Figure 71B, and distributions of the individual V, sequence scores are shown in Figure 71D. B. Scoring Results for BIIB1 through BIIB18 VH and V, scores for all 18 BIIB antibodies were calculated and compared to the 10 relative numbers derived from the NCBI human V-gene sequence reference set (Table 20). Also included in Table 20 is the difference between the BIIB V-gene score and the average subclass score. It is the difference between these scores that was used for examining potential stability trends. BIIB 1 through 18 were labeled in descending order of their measured Fab stabilities. The V-gene subclass and the closest individual human 15 germline for the BIIB VH and V, genes were determined by ClustalW analysis. Consensus-based scores were derived for the 18 BIIB antibodies and compared to the experimentally determined Fab thermal unfolding profiles. BIIB antibodies with greatest difference between the BIIB V-gene score and their average subclass score (ie. lowest sequence scores) was correlated with the low Fab Tms measured in Example 15 20 (Table 20, Figure 72A). There was no apparent trend correlating V, scores with the Tms measured by DSC (Figure 72B). In fact, the lowest scoring V, sequence belonged to BIIB2 which had the second highest apparent Fab stability. Unusual CDR amino acid insertions and deletions were found in BIIB 15 and BIIB 16, as well as in BIIB 18, which had the worst VH score of all the BIIB antibodies 25 and never expressed at appreciable levels. VH scoring did not suggest potential stability problems for BIIB15 and BIIB 16; however, their Fab Tms were two of the lowest tested (Figure 72A). BIIB15 contained an unusually long CDR1 (13 amino acids in length and 2 longer than any VH3-like domain in our mammalian database or VH3 domain in our set of human sequences) and BIIB 16 contained an unusually short CDR2 (15 amino acids in 30 length and I amino acid shorter than the significant majority of the other VH3 CDR2s in both our mammalian VH database and our reference set of human VHs). C. Utility of Small Mammalian Sequence Databases for Residue -254- WO 2007/109254 PCT/US2007/006883 Frequency Analysis and Stability Prediction of the BHIB Fabs Scoring the BIIB VH domains against the mammalian database of VH Sequences proved to be a useful tool for determining the abundance of non-optimal amino acids within each VH domain, which may in turn limit the overall stability of a Fab. The 5 scores were useful for picking out Fabs with low stability and the potential for reduced in vitro longevity and increased aggregation rates. Based on the subclass biases observed for the mammalian V-gene databases, one might question whether it would be best to use strictly human V-gene databases for performing the consensus scoring as opposed to the mammalian database used here. However, based on the significant 10 number of stability determining variables that the scoring does not reflect, such as the exact residue positions of non-consensus amino acids, the strength of the VHNL interaction, the nature of the V-(D-)J-C joining and the incidence of hypersomatic insertions or deletions, it is difficult to believe that an all-human VH database would provide a significant improvement. A strictly human database used for scoring may also 15 require a significantly greater number of sequences to match the evolutionary diversity naively included within the mammalian database. One piece of data that indicates the mammalian database approach works well is that the human subclass consensus VH sequences scored in ascending order of their genomic subclass abundance and apparent subclass usage (Guigou et al., 1990; Teale and Medina, 1992; Tomlinson et al., 1992). 20 The mammalian VH database appears to weight the overall contributions of the human VH subclasses appropriately. The fact that each VH and VK subclass consensus sequence generally scored higher than the vast majority of the individual and unique human sequences picked from the NCBI suggests that the mammalian database is capturing consensus information for all subclasses, even though a bias exists towards VH3. 25 Therefore, due to the existence of other stability-influencing factors that the scoring does not incorporate, the trend observed in Figure 72A is unlikely to improve by fine-tuning of the VH sequence database used for scoring. Unusual insertions or deletions in VH domains appear to have significant effects on stability not reflected by consensus scoring. The unusually long (13 amino acids) 30 CDR1 of BIIB15 and the unusually short (15 amino acids) CDR2 of BIIB16 are likely carry-overs of humanization from the original mouse sequence. In fact, the BIIB 16 VH sequence aligns closely with two mouse VH sequences in our mammalian database that also contained a 15 amino acid CDR2. Forcing the unusually small CDR onto the - 255 - WO 2007/109254 PCT/US2007/006883 human framework may have adversely affected the stability of this antibody. BIIB 18 not only displayed the lowest VH score of all the BIIB sequences but also contained an extraordinarily long CDR2 (22 residues compared to the usual 19 amino acid maximum). Thus, it is not surprising considering the multiple possible problems 5 inherent within the sequence of BIIB 18 that it never reached expression levels allowing for anything other than extremely crude biological assays with unpurified supernatants. The results for BIIB1 5, BIIB 16 and BIIB 18 suggest that unusual VH insertions or deletions may have a greater effect on antibody thermostability than most single point mutations. The exceptionally poor behavior of BIIB18 was not predicted based on a 10 perfunctory glance over its sequence as its Fv contained all the "essential" amino acids strictly conserved within the V-gene folds. Having the ability to pick out troublesome antibodies using a somewhat finer comb is the real the utility of the scoring approach. No trends were apparent when comparing the measured Fab stabilities against the VK scores. Combining the V. scores with the VH scores simply reduced the trending 15 observed for the Vi scores alone. Unlike the VH scoring results, the human VK consensus scores were not in line with the natural abundance of V, germlines and gene usage (Guigou et al., 1990; Meindl et al., 1990; Cox et al., 1994); even though the consensus scores themselves all scored better than their individual human kappa subclass sequences (Figure 72B). It is only speculation that these deviations from the 20 expected order of V,, consensus scores may deter the mammalian databases' ability to accurately predict low stability V, domains. Also, there were not enough VK 2 sequences for defining the VK 2 subclass average score. After subclass distributions were determined for the 68 human V, sequences, only 4 were VK 2 (which reflects VK 2 usage pretty well). Even with these drawbacks, the V, results still suggest that the greater 25 sequence diversity of VH genes may, more often than not, make V" domains the stability-determining component of Fabs (Demarest et al., 2006); although examples of the opposite are available in the literature (R6thlisberger et al., 2006). D. Stability Dependence upon V. or Vi. Germline and V/VL pairing 30 Many of the 18 antibodies contained overlapping germlines V-genes whose CDRs and hypersomatic mutations varied and whose VH or V, pairing was the same or different. This allowed for a comparison of antibody Fab stabilities against the - 256 - WO 2007/109254 PCT/US2007/006883 individual germlines from which they were derived. The two most stable Fabs, BIIB 1 and BIIB2, have the same VHI and VK1 germline combinations. VH and V, pairings do not appear to have clear intrinsic biases based on a study by Winter and coworkers, but are believed to be "receptor driven" (de Wildt et al., 1999b). BIIBI and BIIB2 bind 5 relatively dissimilar antigens, CCL2 and VLA4; therefore, the combination either occurred randomly or was the result of humanization. There was a clear abundance of non-optimal amino-acids within the VH domains of the least stable Fabs, BIIB 17 and BII18 (i.e. low sequence scores and BIIB 18 contained an unusual insertion). Both of these antibodies also contained the same V,2 germline, potentially suggesting light chain 10 protein stability issues on top of the VH-based issues identified by the scoring. An interesting future study would be to determine whether this particular V.2 was a contributor to the poor biophysical behavior of these two antibodies. V, 2 subclass in general was not associated with poor Fab stabilities as BIIB3 and BIIB4 also contain a V,2 subclass gene. The ABO19438VH1 germline (Lefranc et al., 1999) cropped up four 15 times. The TM-values for these Fabs (BIIB12, BIIB1 1, BIIB7 and BIIB3) ranged from 71.2 to 78.2. Interestingly, their VH sequence scores correlated with their apparent Fab stabilities (BIIB12 < BIIB II < BIIB7 = BIIB3). A positive VH scoring correlation was also found in the BIIBI and BIIB2 antibodies described above with the same VHI germline. The apparent imprecision of the VH scoring, however, was emphasized by a 20 negative stability correlation observed for BIIB4 and BIIB6 which both contained the M99649VH3 germline. While usage of VH germline families in adult humans appears to be fairly random, the VH3 family contains the most members in humans. VH3 or VH3-like genes stochastically appeared more often in the human reference set and the mammalian 25 database. Ewert and Plickthun's results demonstrated that a consensus-derived human VH3 sequence was the most stable of the consensus-derived VH domains (Ewert and Pluckthun, 2003). This result may not be entirely unexpected considering the VH 3 subclass has more germline sequences to contribute to the creation of an optimal consensus compared to the other subclasses. Our stability data shows that IgGs 30 containing VH3 germlines do not in general exhibit higher Fab TM-values. There are
VH
3 and VHI subclass genes at both the top and bottom of the stability list. While many other factors contribute to the overall stability of the Fab, especially the properties of the V, counterpart within the Fv, one might expect a trend towards higher Fab stabilities for -257- WO 2007/109254 PCT/US2007/006883 those containing VH3 subclass germlines if the VH 3 domains in general were more stable than the other subclasses. This does not appear to be the case from the limited stability studies described here. 5 Example 17. Covariation Analysis of Ig-fold Polypeptides A. Collecting and Filtering Ig-Fold Structures Structures of Ig-Fold proteins -or Ig-Fold domains from multidomain proteins were gathered from the ASTRAL database, which contains domain structures matching the SCOP hierarchy. Immunoglobulin specific Ig-domains were found under the 10 following classifications within the SCOP hierarchy: "All beta proteins" --> "Immunoglobulin-like beta-sandwich" --> "Immunoglobulin". Under immunoglobulin, four sets of structures were available: "V set," "C1 set," "C2 set," and "I set." Four custom download scripts were separately run to obtain "V-class", "Cl-class", "I-class", and "C2-class" pdb files. 15 Once the structure files for each subfamily were downloaded, some filtering performed to remove erroneously categorized, incomplete, redundant, or domain swapped structures (Liu Y, et al., Protein Sci. (2002), 1285-1299). There were 4 main filtering steps, described below: Step #1. Removal sequences that have breaks 20 Structures from each subclass were visually inspected, using SwissPDB Viewer, for breaks in sequence (either unresolved densities or missing sections due to domain swapping). The PDB files of faulty structures were manually removed from the V-, Cl -, C2-, and I-class structure datasets. Step #2. Removal of 100% identical sequences 25 PDB format structures were converted to FASTA format amino acid sequences, and filtered to remove any sequences that were either 100% identical to, or perfect match substrings of, remaining sequences. Step #3. Removal of sequences of aberrant length PDB structures with aberrantly long or short amino acid sequences were 30 removed from the structure datasets. The length cutoff criteria were determined by examining the histogram of all sequence lengths (Figure 74). The histograms appeared -258- WO 2007/109254 PCT/US2007/006883 somewhat normally distributed. Sequences outside two standard deviations from the mean were removed from each subfamily dataset. The overall mean number of residues for the immunoglobulin superfamily was 106.10 and the standard deviation was 12.19. Consequently, the global cutoffs used were <= 81 and >=131 residues. A breakdown of 5 the mean lengths and standard deviations is shown in Table 21. Table^21. Sequence Length Criteria of Structure Datasets Subclass Average Length Standard Deviation Cl 99.44 8.79 C2 88.07 15.14 I 95.5 4.94 V 111.95 10.62 Overall 106.10 12.19 Step #4. Removal of misfolded sequences 10 The structures were visually inspected, using SwissPDB Viewer, for misfolding. Any structures that did not conform to the two beta-sheet sandwich topology were discarded. Since only five C2-class domains were obtained from SCOP, C2-subclass Ig Folds were not pursued further. Hereafter, Cl-class is referred to as C-class. 15 B. Obtaining Sequence Alignments from Structure Alignments For each separate class, the Ig-Fold structures were superimposed upon one another using Secondary Structure Matching (SSM) within the Schr6dinger structalign package (See Schr6dinger Prime program documentation for instructions on creating superpositions). Superpositions were performed on an 'all-to-all' or an 'all-to-one' 20 basis. Each algorithm led to similar quality alignments, so 'all-to-one' was chosen for the superpositions. Some superimpositions were corrected to insure that the core regions of each structure were superimposed instead of the loops. The superimpositions were then used to generate structure-based sequence alignments of all the V-class, I-class, and C-class sequences within each structural alignment. Alignment of each sequence onto 25 another was performed using the Schr6dinger package by matching amino acids from one sequence to that of second sequence based on the shortest distance between a carbons of the polypeptide backbones. -259- WO 2007/109254 PCT/US2007/006883 C. Construction of a Curated Ig-Fold Sequence Database A number of defined steps were created to generate a curated Ig-Fold dataset for calculating robust covariation statistics: 5 Step #1. Construction of HMM profiles Three Hidden Markov Model (HMM) profiles were built, each based upon the structure-based sequence alignments for one of the three Ig-Fold classes. The profiles were created with the HMMER software package (version 1.8), using the commands "hmmbuild" and "hmmcalibrate" with standard options. These HMMs were then used 10 to detect and align additional Ig-Fold sequences in the NR-database maintained at the NCBI. Step #2. Searching NR using the new V-, I-, and C-class HMM profiles. 15 The three class-specific HMMs were used to search for similar sequences in a local NR database. The NR database is a large file containing -3 million non-redundant protein sequences. For each of the V-, I-, and C-class HMMs, the HMMER command "hmmsearch" was used to search NR. Each output ranked NR sequences by their scores relative to the HMM used, and provided information about the number of regions hit by 20 the HMM and the positions of the hit regions within each NR sequence. For each Ig Fold class-specific HMM search, hit NR sequences above a recommended criterion score threshold were retained as candidate members of the Ig-Fold class whose HMM was used. For those hit regions that were subsequences of an NR sequence, the exact NR subsequence hit was extracted from the full NR sequence using a custom program. 25 Step #3. Validation of Ig-Fold class assignment using PFAM. PFAM is a protein family and domain classification tool, created and maintained at the Wellcome Trust Sanger Institute (Fin, et al., Nuc. Acids. Res., (2006), 34: D247 D251), that can be applied to individual protein sequences. A Pfan tool 'pfamverify' 30 was applied to each Ig-Fold candidate sequence obtained in step 2 above, to confirm that it was correctly classified by the Ig-Fold class-specific HMMs created from the - 260 - WO 2007/109254 PCT/US2007/006883 structure-based sequence alignments. PFAM's Ig-clan HMM profiles (including V-, I-, Cl-, C2-, and less specific Ig HMMs) were downloaded from the PFAM website. Each sequence from NR was scored by the Ig-clan HMMs, revealing how well each sequence conforms to each PFAM HMM. Sequences whose scores lay below recommended 5 cutoffs (TCl - defined at the PFAM website) for the V-, I-, or C-classes were removed from the respective sequence sets. Thus, the candidate class-specific Ig-Fold sequences found in step 2 were retained only if their PFAM scores validated their Ig-Fold class assignments. 10 Step #4. Aligning the new Ig-Fold sequences. The Ig-Fold sequences that were pulled from NR with HMM searches and that survived step 3 above, were next aligned to our custom HMMs of their assigned class. Since these HMMs had been based upon careful structural alignments, this process insured a structure-guided alignment of the new sequences. The HMMER package was 15 utilized to generate 'mapali' alignments in fasta output format, to be used in the Sequence Covariation Tool described below. The alignments were also output in 'Stockholm' output format, and inspected for aberrant or misaligned sequences, which were manually omitted. The resulting alignments, consisting of the original structurally aligned sequences and the added HMM-aligned sequences from NR, contained the 20 following numbers of sequences: -50,000 V-class; -10,000 C-class; -10,000 I-class. Step #5. Removing redundant or similar sequences. We expected these alignments to be biased towards frequently observed sequences, with under-representation of rare sequences and a consequent masking of 25 sequence diversity. For example, we expected a large bias in the V-class Ig-Fold sequence alignment towards murine and human immunoglobulin variable domains. Since over-representation of particular sequence types limits the usefulness of covariation analyses, a filtering tool was created and applied to reduce the alignment redundancy. The tool used a novel heuristic algorithm to find sequences with >80% 30 identity to one another, and removed them from the alignments. In brief, percent identities were calculated for all sequence pairs. The identity values were then grouped into bins of percent identity (i.e. 99% bin, 98% bin, 97% bin, etc.). During the reduction of each bin, the sequences of each bin were ranked by decreasing non-gap residue count, -261- WO 2007/109254 PCT/US2007/006883 and then by their Henikoff sequence weight (Henikoff S and Henikoff JG, J. Mol. Biol., (1994), 243: 184-199). After this step, the remaining number of sequences in each alignment were: 2,786 V-class; 1,587 I-class; and 518 C-class. 5 Step #6. Removal of gaps to create the final alignments In the final alignments, columns that were not match states in the HMM profile used to find these sequences (see HMMER manual) were removed. Therefore, while many of the sequences contain more amino acids than the final alignment length, the lengths were truncated to avoid calculations on the less informative gap regions of the 10 alignment. The final numbers of residues (including gaps) within the alignments were 144 for the V-class, 60 for the I-class, and 111 for the C-class. D. General description of the final alignments a) V-Class 15 After the final 80% filter was applied, the 2,786 V-class sequences could be divided into three categories: (1) Immunoglobulin variable gene class, 49%, which include both the VH and VL domains of diverse immunoglobulins; (2) T-Cell Receptor V-class genes, 16%, which contained the hypervariable domains; and (3) Other V-class genes, 35%. The Immunoglobulin V-class genes come from a huge variety of species 20 ranging from cartilaginous fish to primates. There was a bias towards human V-class sequences (537 of the 1272 sequences). However, other vertebrate species were only minimally represented: mouse (33), cow (5), camel (23), llama (31), macaque (17), chicken (6), etc. The T-cell receptor set of sequences was nearly as diverse ranging from fish to primate and did not contain a bias towards human sequences. The "Other" 25 category comprised a large number of diverse sequences with no real subcategories comprising over 1-2% of the remaining sequences. The "Other" category contained an even wider array of species including chordates (primarily), cartilaginous fish, cephalopods, and insects. b) C-Class 30 After the final 80% filter was applied, the 518 C-class sequences could be divided into three categories: (1) Immunoglobulin constant domains, 44%; (2) MHC - 262 - WO 2007/109254 PCT/US2007/006883 type Ig-Folds, 21%; and (3) Other C-class Ig-Folds, 35%. The Immunoglobulin constant domain category contained diverse sequences ranging from cartilaginous fish to primates. The C-class immunoglobulin subcategory was not biased by primate or even mammalian sequences (only 3 human sequences remained unfiltered: IgE-CH4, a kappa 5 CL, and a lambda CL). The MHC subcategory also ranged from cartilaginous fish to primates and also showed little bias towards one evolutionary group. The "Other" C class category contained many unknown proteins, a very small subcategory of various beta-2-microglobulins, and many other proteins that were not observed with a frequency deserved of a subcategory. 10 c) I-Class The I-class did not have any clear subcategories. Titan or titan-like molecules cropped up commonly as did cell-adhesion or cell-adhesion-like molecules. E. Calculation of Covariation Statistics 15 Correlation strength (in terms of a *-value) and significance (in terms of a X2_ value) of covariations between all possible pairs of amino acids in the alignments was calculated based on art-recognized methods, but included the following variations: 1) Gaps were included as a distinct residue type; (2) the Henikoff weighing scheme was not performed; (3) Sequence Average Identities (SAI) were not used to filter out covarying 20 pairs; (4) X2 values were calculated using event based counts (number of occurrences), rather than frequencies; and (5) covarying pairs were not reported by the program unless they were observed a minimum number of times. Two bodies of statistics were created, the first (and smaller set) with an event cutoff of 10 and a second (much larger set) with an event cutoff of 6 events. The ability to calculate these parameters on any given 25 sequence alignment was encoded into a Java executable and tested with Java Runtime Engine (JRE) version 1.4.2. Based on the number of positions within the alignment and a total of 23 possible residues at each position (including gaps, B, and Z ambiguous residues), there were 2.4 million correlations calculated for the C-class, 700 thousand calculated for the I-class, 30 and 4.1 million calculated for the V-class. Figure 75 shows the * values calculated for the V-class with high x 2 significance. values range from -i to +1. As positive * - 263 - WO 2007/109254 PCT/US2007/006883 values move away from 0 (i.e. toward +1), the strength of the correlation (i.e. the covariation) between two amino acids at defined positions within the alignments increases. Negative correlations (i.e. the presence of one amino acid at one site in the alignment forcing the absence of another specific amino acid at a second site) increase 5 as * values move towards -1. Cutoffs for what is a real covariation are somewhat arbitrary, but those with a * value > 0.5 are strongly correlated with one another. Thus, according to Figure 75, only 370 of the possible 2.4 million correlations within the C class are strong. However, correlations above 0.3 have been reported as important and would increase the number of positive correlations observed for all three classes by -10 10 fold. These correlations represent the very small, but important piece of the dataset that is used for further analyses, generating protein designs, or prediction of functional residue positions. F. Validation of Covariation Statistics There were several criteria used to validate that the strong covariations that were 15 derived from the Ig-Fold database were significant and meaningful apart from the statistical strengths based on the calculations. The first was to analyze whether there was a trend between residues that covaried with one another and their proximity inside Ig-Folds of known structure. Previous studies in the literature have demonstrated that a correlation between covarying positions and their proximity to one another do exist 20 although they are very weak - in agreement with our results. A second criterion looked at whether the Ig-Fold covariation calculations generated known connections between amino acid positions that have already been reported to exist within known Ig-Fold subsets. One clear test case was at N-terminus of human/murine IgG variable heavy chain folds (part of the V-class). Residues 6-10 are known to adopt very specific 25 conformations based on the conservation of covarying pairs of amino acids (Ewert S, et al., Methods, (2004), 24: 184-199). Two of the three reported pairs of covarying amino acids were found to have * values above 0.3 - well above the apparent background of the dataset. 30 Example 18. Preparation of a PRIMATIZED@ p5E8 Tetravalent Antibody Comprising a Conventional p5E8 scFv PRIMATIZED@ p5E8GI is a chimeric macaque/human (PRIMATIZED@) - 264 - WO 2007/109254 PCT/US2007/006883 monoclonal antibody containing macaque heavy and light variable regions fused to human gamma I and kappa constant regions, respectively. PRIMATIZED@ p5E8G1 binds to human CD23, the low affinity receptor for IgE (FccRII) (Mavromatis and Cheson. 2003. J. Clin. Oncol. 21:1874; US Patent Application 20030059424). A 5 tetravalent PRIMATIZED@ p5E8 antibody was constructed using a similar strategy as that described in Examples 1-8. The PRIMATIZED@ p5E8 scFv used for constructing the tetravalent antibody is comprised of p5E8 VL and VH region sequences tethered by a short linker in the VL -+ (Gly 4 Ser)3 linker -> VH orientation and is described in greater detail in Example 19 . Correct sequences were confirmed by DNA sequence 10 analysis. Plasmid DNA was used to transform CHO DG44 cells for stable production of antibody protein. Figure 80 shows the DNA sequence of heavy chain C-terminal tetravalent PRIMATIZED@ p5E8 antibody comprising a conventional scFv. Figure 81 shows the amino acid sequence of heavy chain C-terminal tetravalent PRIMATIZED@ p5E8 antibody comprising a conventional scFv. Figure 82A shows the DNA sequence 15 of PRIMATIZED@ p5E8 light chain. Figure 82B shows the amino acid sequence of PRIMATIZED@ p5E8 light chain. Example 19. Preparation of PRIMATIZED@ p5E8 scFv and Fab Proteins PRIMATIZED@ p5E8 scFvs in both orientations (VL -> (Gly 4 Ser) 3 linker -+ 20 VH (VL/VH) and VH -) (Gly 4 Ser) 3 linker -* VL (VH/VL)) were subcloned by PCR amplification from plasmids described in U.S. Patent Application 20050163782. Oligonucleotides used in the construction are shown in Table 22. PRIMATIZED p5E8 scFvs (VL/VH) was constructed by PCR using the forward primer P5E8-VLOIF which contains 29 bases encoding part of the gpIII leader sequence 25 followed by 15 bases of sequence complementary to the p5E8 N-terminal light variable domain gene and the reverse primer, P5E8-VHOlR, which contains 15 bases of sequence complementary to the p5E8 C-terminal heavy variable domain followed by a unique adjacent Sal I endonuclease site (endonuclease site is underlined). Similarly, PRIMATIZED p5E8 scFvs (VH/VL) was constructed by PCR using the forward primer 30. P5E8-VHO 1F which contains 29 bases encoding part of the gpIII leader sequence followed by 18 bases of sequence complementary to the p5E8 N-terminal variable heavy domain gene and the reverse primer, P5E8-VLO 1 R, which contains 12 bases of sequence complementary to the p5E8 C-terminal variable light domain gene followed by a unique - 265 - WO 2007/109254 PCT/US2007/006883 adjacent Sal I endonuclease site (endonuclease site is underlined). Table 22. Oligonucleotides for PCR amplification of a conventional PRIMATIZED@ p5E8 scFvs. Primers Sequence P5E8-VLOI F 5'- CGCTGGTGGTGCCGTTCTATAGCCATAGTGAC (SEQ ID NO:71) ATCCAGATGACC -3' P5E8-VLOIR 5'- GTGGTCGACTTTGATTTCCAC -3' (SEQ ID NO:72) P5E8-VHOlF 5'- CGCTGGTGGTGCCGTTCTATAGCCATAGTGAGG (SEQ ID NO:73) TGCAGCTGGTGGAG -3' P5E8-VHOIR 5'- GTGGTCGACTGAGGAGACGGTGAC -3' (SEQ ID NO:74) P5E8-LeaderO 1 5'- GGCATATGAAAAAACTGCTGTTCGCGATTCCGCTG (SEQ ID NO:75) GTGGTGCCGTTCTATAG -3' 5 Following PCR amplification, primer P5E8-Leader01 was added to both reactions for a second PCR reaction. Primer P5E8-Leader01 contains a unique Nde I endonuclease site followed by 2-5 bases encoding the N-terminal portion of the gpIII leader sequence, followed by 22 bases complementary to the 5' ends of P5E8-VLO1F 10 and P5E8-VHOIF and PCR amplified again. PCR products corresponding to the expected sizes were resolved by agarose gel electrophoresis, excised, and purified using the Millipore Ultrafree-DA extraction kit according to manufacturer's instructions (Millipore; Bedford, MA). The purified PCR products were subsequently digested with Nde I and Sal I and cloned into the Nde I/ Sal I sites of a modified E. coli expression 15 vector designed to drive recombinant protein expression under the control of an inducible ara C promoter. The expression vector contained a modification encoding a unique Nde I site overlapping the start codon of the BHA 10 scFv. Individual ligation reactions were performed with each of the gel purified PCR products and the digested expression vector and a portion of each of the ligation mixtures were used to transform 20 E. coli strain XL1-Blue. Ampicillin drug resistant colonies were screened and DNA sequence analysis confirmed the correct sequence of the final p5E8 (VH/VL) encoding - 266 - WO 2007/109254 PCT/US2007/006883 pIEH-162 and p5E8 (VL/VH) encoding pIEH-163 constructs. DNA and amino acid sequences of p5E8 (VL/VH) scFv are shown in Figures 83A and 83B, respectively. DNA and amino acid sequences of p5E8 (VH/VL) scFv are shown in Figures 84A and 84B, respectively. 5 For expression of conventional p5E8 scFvs, freshly isolated colonies of E. coli strain W3 110 (ATCC, Manassas, Va. Cat. # 27325) transformed with plasmids pIEH 162 and pIEH-163 were cultivated and either culture supernatants or periplasm extracts were prepared as described in Example 4. PRIMATIZED@ p5E8 Fab was prepared by enzymatic digestion of 10 PRIMATIZED@ p5E8 IgG following methods described in Example 1. Purified Fab was concentrated to between 2-11 mg/mL. Fab concentrations were determined using an E28 nm = 1.5 mL mg~ cm-. Example 20. Thermal Stability of Conventional p5E8 scFv Antibodies 15 Since an excellent correlation has been observed between T50 values obtained using the thermal challenge assay described in Examples 4 and 5 and Tm values calculated by DSC analysis (r 2 = 0.93), a thermal challenge assay was employed to evaluate the relative thermal stabilities of the p5E8 (VLNH) and p5E8 (VH/VL) scFvs. The thermal challenge assay was utilized as described in Examples 4 and 5 except that 20 soluble CD23 antigen at lug/ml was used to coat the plates and the concentration of the Eu-labelled anti-6 His antibody (Perkin Elmer, Boston, MA, Cat. # ADO 109) was increased to 250 ng/ml. Results from this assay determined the T50 value of p5E8 (VL/VH) to be 38 *C and p5E8 (VH/VL) to be slightly lower at 34 *C (Figure 85). Given the remarkably low T50 values of both scFvs and the observation that of the two p5E8 25 (VL/VH) was slightly more thermally stabile, p5E8 (VL/VH) was selected for further stability engineering. Example 21. Construction of p5E8 scFv Molecules with Improved Thermal Stability 30 Individual variants and libraries designed to contain the desired amino acid replacements in the conventional p5E8 (VL/VH) scFv using the methods described in Examples 3 and 5 were created as previously described using oligonucleotides listed in Table 23. - 267 - WO 2007/109254 PCT/US2007/006883 In Table 23, each oligonucleotide name gives reference to amino acid substitutions at position(s) in VH or VL according to Kabat numbering system. "Rationale" refers to the design method employed. Said methods are described in detail in Example 5 supra. Mutagenic residues are shown as capital letters. Oligonucleotide 5 pairs for introducing VH/VL disulfides are boxed. Abbreviations are: "COMP" Computational Analysis, "COVAR"-Covariation Analysis, "CONS" -Consensus Scoring, "INTER-VH/VL" Interface Design, "SS"-. VH/VL disulfide bond, and "TBD" to be determined. 10 Table 23. Oligonucleotides and rationale for construction of variant p5E8 (VL/VH) scFvs. Oligo_name Rationale Sequencet SEQ ID NO: VH E6Q COMP gaggtgcagctggtgCagtctgggggcggcttg 76 VH L11SDG COMP gagtctgggggCggcRRCgcaaagcctgggggg 77 VHA12VKK1 COVAR gtctgggggcggcttgRHGMAGcctggggggtccctg 78 3QER VH N32S CONS gttcaggttcaccttcAGCaactactacatggac 79 VH D35bHSN INTER caataactactacatgMRCtgggtccgccaggctc 80 VH Q43KR COVAR cgccaggctccagggARGgggctggagtgggtc 81 VH S49GA COVAR gggctggagtgggtcGSCcgtattagtagtagtg 82 VH 151M COMP gagtgggtctcacgtatGagtagtagtggtgatc 83 VL Q37A COMP ggtattatttaaattggtatGCCcagaaaccaggaaaag 84 VH D55GS CONS gtattagtagtagtggtRGCcccacatggtacgcag 85 VH P56STD CONS gtagtagtggtgatRVCacatggtacgcagac 86 VH W58YN INTER gtggtgatcccacaWACtacgcagactccgtg 87 VH E72DN COVAR gattcaccatctccagaRACaacgccaagaacacac 88 VH A74S COVAR catctccagagagaacAGCaagaacacactgtttc 89 VH F79YSV COVAR gccaagaacacactgKHCcttcaaatgaacagc 90 VH Q81E COMP gaacacactgtttcttGaaatgaacagcctgagag 91 VH A84G COMP caaatgaacagcctgagaGGCgaggacacggctgtc 92 VH V89AGT COMP gagctgaggacacggctRSCtattactgtgcg 93 VH S94RK CONS gtctattactgtgcgARGttgactacagggtctg 94 VH V107T COVAR TBD VH L108AGS COMP TBD - 268 - WO 2007/109254 PCT/US2007/006883 VH T110VS COVAR TBD VL L11G COVAR cagtctccatcttccGGCtctgcatctgtaggg 95 VL V15AGS COMP ccctgtctgcatctRSCggggacagagtcacc 96 VL V19L COVAR ctgtaggggacagaCTGaccatcacttgcagg 97 VL 121L COVAR ggggacagagtcaccCTGacttgcagggcaag 98 VL T22S COMP. gacagagtcaccatcAGCtgcagggcaagtcag. 99 VL D28SG CONS gcagggcaagtcagRGCattaggtattatttaaattg 100 VL R30SL CONS gcaagtcaggacattMKCtattatttaaattgg 101 VL K39A COVAR aattggtatcagcagGCCccaggaaaagctcc 102 VL K45ER INTER ccaggaaaagctcctSRCctcctgatctatgttg 103 VL V50n INTER ctaagctcctgatctatNNKgcatccagtttgcaaag 104 VL L54R CONS ctatgttgcatccagtCGCcaaagtggggtccc 105 VL V58S COVAR cagtttgcaaagtgggTCCccatcaaggttcagc 106 VL E70D COVAR cagtggatctgggacaGACttcactctcaccgtc 107 VL V751 COMP gagttcactctcaccATCagcagcctgcagcc 108 VL P80AG COMP cagcagcctgcagRGCgaagattttgcgac 109 VL F83AGST COMP ctgcagcctgaagatRSCgcgacttattactg 110 VL T85D COVAR cctgaagattttgcgGACtattactgtctacag 111 VL L89AQ INTER gcgacttattactgtSMRcaggtttatagtacc 112 VL R96LY INTER gtttatagtacccctMwAacgttcggccaaggg 113 VL F98W INTER gtacccctcggacgTGGggccaagggaccaag 114 VL 1106AGS COMP TBD VH L45C SS gctccagggcaggggTGCgagtgggtctcacg 115 VL F98C SS gtacccctcggacgTGCggccaagggaccaag 116 VH D101C SS cttgactacagggtctTGCtcctggggccagggag 117 VL L46C SS ggaaaagctcctaagTGCctgatctatgttgc 118 VH S102C SS gactacagggtctgacTGCtggggccagggagtc 119 VL L46C SS ggaaaagctcctaagTGCctgatctatgttgc 120 VH G44C SS caggctccagggcagTGCctggagtgggtctcac 121 VL Q100C SS cctcggacgttcggcTGCgggaccaaggtggaaatc 122 tPositions targeted for mutagenesis are indicated by underline. Ambiguous bases are abbreviated as follows: W = A or T, V = A or C or G, Y = C or T, S = C or G, M = A or C, N = A or C or G or T, R = A or G', K= G or T, B= C or G or T (J Biol Chem. 261(1):13-7 (1986)). 5 - 269 - WO 2007/109254 PCT/US2007/006883 Individual transformed colonies were picked into deep-well 96 well dishes, processed, and screened according to the methods detailed in Example 5. Transformants were grown overnight in expression media consisting of SB (Teknova, Half Moon Bay, Ca. Cat. # SO140) supplemented with 0.6 % glycine, 0.6 % Triton X100, 0.02% 5 arabinose, and 50 pg/ml carbenicillin at either 37*C or 32*C. After thermal challenge, the aggregated material was removed by centrifugation and assayed in the soluble CD23 DELFIA as described in Example 3. Assay data was processed using Spotfire DecisionSite software (Spotfire, Somerville, Ma.) and expressed as the ratio of the DELFIA counts observed at challenge temperature to the 10 reference temperature for each clone. Clones that reproducibly gave ratios greater than or equal to twice what was observed for the parental plasmid were considered hits. Plasmid DNAs from these positive clones were isolated by mini-prep (Wizard Plus, Promega, Madison, WI) and retransformed back into E. coli W3 110 for confirmation secondary thermal challenge assays as well as for DNA sequence determination. 15 Primary and confirmatory results from these assays are shown in Table 24. Several of the stabilized scFv molecules of the invention resulted in improvements in binding activity (T 50 >38*C) as compared with the conventional p5E8 scFv. In particular, the T 5 0 values of p5E8 library position VH6 (E6Q), library position VH49 (S49G and S49A), library position VH43 (Q43K), library positions VH7 2 (E72D and 20 E72N), and library position VH79 (F79S), exhibited increases in thermal stability ranging from + 4*C to + 5'C relative to the conventional p5E8 scFv. The T 50 values of p5E8 library position VL50 (V50D and V50S), library position VL75 (V751), library position VL80 (P80S), and library positions VL83 (F83A, F83G, F83S and F83T), exhibited increases in thermal stability ranging from + 3*C to + 7*C relative to the 25 conventional p5E8 scFv. In addition, the T 50 values of p5E8 library position VH32 (N32S) and library position VH79 (F79Y) exhibited increases in thermal stability of + 2*C relative to the conventional p5E8 scFv. Stabilizing mutations were identified using an assortment of the design methods 30 described in Example 5- the VH6 (E6Q), VL 7 5 (V751), VL80 (P80S) and VL 83 (F83A, F83G, F83S and F83T) mutations were derived using Computational Analysis, the VH 32 (N32S) mutation was derived using Consensus Scoring, and the four mutations VH49 (S49G and S49A), VH43 (Q43K), VH 72 (E72D and E72N), the VH 7 9 (F79S and F79Y) mutations were derived using Covariation Analysis, and the VL50 (V5OD and V50S) - 270 - WO 2007/109254 PCT/US2007/006883 mutations were derived using VH/VL interface analysis, validating the utility and novelty of these design tools. Combining VH43Q and VH32S mutations with Vt49G, VH72D, or VH49A stabilizing mutations enhanced the thermal stability (T 50 ) of p5E8 up to 53'C, an increase of 15*C relative to the conventional p5E8 scFv. 5 Table 24. p5E8 VH and VL library positions, library composition, and screening results. Position Library Hit Seq. ATso "C Observed VH6 Q E6Q +4 VH32 S N32S +2 VH49 S, A S49G, S49A +5, +5 VH43 K, R Q43K +4 VH72 D, N E72D, E72N +5, +4 VH79 S, V, Y F79S,F79Y +4,+2 VL50 all amino acids V50D, V50S +4, +3 VL75 I V751 +5 VL80 SG P80S +4 VL83 S, A, G, T F83S, F83A, +4, +6, F83G,F83T +7,+6 10 Table 25 shows the results of a comprehensive thermal stability analysis of the various individual and combined stabilizing mutations introduced into a conventional scFv. Stabilizing mutations were identified that upon combination exhibited increases in thermal stability ranging from + 10*C to + 15*C relative to the conventional p5E8 scFv. These results demonstrate, as similarly shown in Example 5, that improvements in 15 activity are additive and that the methods described in this invention are capable of improving the thermal stability properties of a second test scFv demonstrating the general usefulness of the methods. Table 25. Characteristics of p5E8 constructs used to produce variant 20 proteins and T 50 results from thermal challenge assay. Plasmid Disulfide Linker Other Mutation T 5 o *C Length (aa) pIEH162 no 15 na 34 p1EH163 no 15 na 38 pIEH 164 no 20 na 37 -271- WO 2007/109254 PCT/US2007/006883 pIEH165 no 20 na pIEH171 no 15 VH E6Q 42 pIEH172 no 15 VH N32S 40 pIEH173 no 15 VH S49G 43 pIEH174 no 15 VH E72D 43 pIEH175 no 15 VH Q43K 42 pIEH 176 no 15 VH S49A 45 pIEH177 no 15 VH E72N 42 pIEH178 no 15 VH F79S 2 pIEH 179 no 15 VH F79Y 0 pIEH 187 no 15 VL V75I 43 pIEH 188 no 15 VL P80S 42 pIEH 189 no 15 VL F83S 42 pIEH 190 no 15 VL F83A 44 pIEH 191 no 15 VL F83G 45 pIEH192 no 15 VL F83T 44 pIEH182 VH45-VL98 15 na TBD pIEH 183 ~VH1OI-VL46 15 na TBD pIEH184 VH102-VL46 15 na TBD pIEH 193 VH44-VL100 15 na TBD pIEH195 no 15 VL V50D 43 pIEH 196 no 15 VL V50S 41 pIEH 197 no 15 VH E6Q, S49G 51 pIEH 198 no 15 VH E6Q, N32S, 53 1 S49G pIEH 199 no 15 VH E6Q, E72N 48 pIEH-200 no 15 VH E6Q, N32S, 50 E72N pIEH-201 no 15 VH E6Q, N32S, 50 E72D plEH-202 no 15 VH E6Q, E72D 48 pIEH-203 no 15 VH E6Q, N32S, 51 S49A p1EH-204 no 15 VH E6Q, S49A 48 na = not applicable aa = amino acids TBD = to be determined -272- WO 2007/109254 PCT/US2007/006883 Example 22. Production of Stabilized p5E8 Tetravalent Antibodies Stabilized p5E8 scFvs of the invention are used to construct tetravalent antibodies as both N- terminal and C- terminal scFv fusions similar in configuration as 5 shown in Figure 13. A. Construction of NH- p5E8 Tetravalent Antibody The p5E8 scFvs DNAs described in Example 21 are used to construct an NH- p5E8 tetravalent antibody similar to that described in Example 7 supra. A (Gly 4 Ser) 5 linker is 10 used to connect the p5E8 scFv to the mature amino terminus of PRIMATIZED@ p5E8 IgG heavy chain. The organization of the molecule is a p5E8 scFv- (Gly4Ser)5 linker PRIMATIZED@ p5E8 IgG heavy chain. This molecule assembles with the light chain to form the tetravalent antibody. The correct sequence would be confirmed by DNA sequence analysis. The construct could be used for transfecting CHO cells as described 15 in Example 8 for scaled-up production of a NH- p5E8 Tetravalent Antibody. Transfected CHO-cells can be screened for binding to immobilized CD23 antigen by DELFIA assay. 96-well plates. For example, (MaxiSorp, Nalge Nunc, Rochester, NY, Cat. # 437111) can be coated with soluble CD23 antigen at I ug/ml in 0. 1M sodium carbonate buffer, pH 9.5. Plates are coated overnight at 4oC, and blocked with DELFIA assay buffer 20 (DAB, 10 mM Tris HCl, 150 mM NaCl, 20 mM EDTA, 0.5 % BSA, 0.02% Tween 20, 0.0 1% NaN3, pH 7.4) for one hour with shaking at room temperature. Plates are washed 3 times with DAB without BSA (Wash buffer), and test samples diluted in DAB added to the plates in a final volume of 100 ul. The plates are incubated for one hour with shaking at room temperature, and then washed 3 times with Wash buffer to 25 remove unbound material. Bound antibody is detected by addition of 100 ul per well of DAB containing 250 ng/ml of Eu-labeled anti-human antibody (Perkin Elmer, Boston, MA, Cat. # 1244-330) and incubated at room temperature with shaking for one hour. The plates are washed 3 times with Wash buffer, and 100 ul of DELFIA enhancement solution (Perkin Elmer, Boston, MA, Cat. # 4001-0010) is added per well. Following 30 incubation for 15 minutes, the plates are read using the Europium method on a Victor 2 plate reader (Perkin Elmer, Boston, MA). B. Construction of C- p5E8 Tetravalent Antibody The p5E8 scFvs DNAs described in Example 21 supra are used to construct a 35 C- p5E8 tetravalent antibody as described in Example 7 supra. A Ser(Gly4Ser) 3 linker peptide is used to connect the p5E8 scFv to the carboxy terminus of PRIMATIZED@ p5E8 IgG heavy chain. The organization of the molecule is a PRIMATIZED@ p5E8 - 273 - WO 2007/109254 PCT/US2007/006883 IgG heavy chain - Ser(Gly 4 Ser) 3 linker- p5E8 scFv. This molecule assembles with the light chain to form the tetravalent antibody. The correct sequence would be confirmed by DNA sequence analysis. The construct could be used for transfecting CHO cells as described in Example 8 for scaled-up production of a C- p5E8 Tetravalent Antibody. 5 Transfected CHO cells can be screened as described above. Equivalents Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments of the 10 invention described herein. Such equivalents are intended to be encompassed by the following claims. - 274 -

权利要求:
Claims (60)
[1] 1. A population of stabilized scFv molecules, wherein the stabilized scFv molecules comprise 5 i) a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH domain and an amino acid in the VL domain, and wherein the Vi-I and VL domains of the scFv molecules have Tm values of greater than 55"C; or 10 ii) a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain, and wherein scFv molecules have a T 50 of greater than 49 0 C. 15
[2] 2. A stabilized scFv molecule wherein the stabilized scFv molecule comprises i) a scFv linker having the amino acid sequence (Gly 4 Ser)n interposed between a VH 20 domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between VH amino acid 44 and VL amino acid 100; ii) a VH domain and a VL domain wherein the molecule has at least one substitution selected from a group consisting of: 25 a) substitution of an amino acid at Kabat position 13 of VH; b) substitution of an amino acid at Kabat position 16 of VH; c) substitution of an amino acid at Kabat position 46 of VL; d) substitution of an amino acid at Kabat position 49 of VL; 30 e) substitution of an amino acid at Kabat position 50 of VL; f) substitution of amino acids at Kabat positions 49 and 50 of VL; g) substitution of amino acid at Kabat position 101 of VH; h) substitution of amino acid at Kabat position 20 of VH; - 275 - WO 2007/109254 PCT/US2007/006883 i) substitution of amino acid at Kabat position 48 of VH; j) substitution of amino acid at Kabat position 3 of VL; k) substitution of amino acid at Kabat position 55; 1) substitution of amino acid at Kabat position 67 of VH; 5 m) substitution of amino acid at Kabat position 6 of VH; n) substitution of amino acid at Kabat position 32 of VH; o) substitution of amino acid at Kabat position 49 of VH; p) substitution of amino acid at Kabat position 43 of VH; q) substitution of amino acid at Kabat position 72 of VH; 10 r) substitution of amino acid at Kabat position 79 of VH; s) substitution of amino acid at Kabat position 50 of VL; t) substitution of amino acid at Kabat position 75 of VL; u) substitution of amino acid at Kabat position 80 of VH; v) substitution of amino acid at Kabat position 83 of VH ; or 15 iii) a V" domain and a VL domain with a stabilized interface between VH and VL of the scFv wherein the molecule has at least one substitution at an amino acid position directly within the interface and/or at an amino acid position that scaffolds the interaction between VH and VL, wherein the at least one amino acid position is selected from a 20 group consisting of: 47H, 37H, 45H, 46H, 51H, 59H, 109H, 14H, 26H, 27H, 40H, 69H, 103H, 29H, 38H, 44H, 49H, 55H, 63H, 54H, 68H, 72H, 74H, 79H, 82bH, 8H, 32H, 39H, 41H, 62H, 85H, 91H, 24H, 60H, 37L, 36L, 44L, 59L, 57L, 64L, 46L, 48L, 63L, 67L, 68L, 39L, 45L, 47L, 54L, 56L, 79L, 85L, 98L, 58L, 74L, 77L, 83L, 89L, and 104L, according to Kabat numbering, 25 and wherein the stabilized scFv molecule has a T 50 greater than that of a scFv molecule lacking the substitution.
[3] 3. The stabilized scFv molecule of claim 2, wherein the stabilized scFv molecule 30 has a stability equivalent to that of a conventional Fab fragment under conditions of thermal challenge.
[4] 4. A fusion protein comprising a stabilized scFv molecule of claim 2. - 276 - WO 2007/109254 PCT/US2007/006883
[5] 5. A population of stabilized multivalent antigen binding molecules, wherein the stabilized binding molecules comprise at least one stabilized scFv molecule comprising a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL 5 domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain, and wherein the population of stabilized binding molecules comprise monomeric,. soluble proteins of which not more than 10% is present in aggregated form. 10
[6] 6. The population of claim 5, wherein the stabilized multivalent antigen binding molecules are expressed in CHO or NSO cells.
[7] 7. A population of stabilized multivalent antigen binding molecules, wherein each multivalent antigen binding molecule comprises 15 i) at least one scFv molecule comprising a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain, and wherein the VH and VL domains of the at least one scFv molecule have a Tm of greater than 55*C; or 20 ii) at least one scFv molecule comprising a scFv linker interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between an amino acid in the VH and an amino acid in the VL domain, and wherein the at least one scFv molecule of the multivalent antigen binding molecule has a T 50 of greater 25 than 49*C.
[8] 8. A stabilized multivalent antigen binding molecule comprising i) at least one stabilized scFv molecule, wherein the stabilized scFv molecule comprises 30 a (Gly 4 Ser)n scFv linker interposed between a VH domain and a VL domain and wherein the VH and VL domains are linked by a disulfide bond; ii) at least one stabilized scFv molecule comprising a scFv linker having the amino acid - 277 - WO 2007/109254 PCT/US2007/006883 sequence (Gly 4 Ser), interposed between a VH domain and a VL domain, wherein the VH and VL domains are linked by a disulfide bond between VH amino acid 44 and VL amino acid 100; 5 iii) at least one stabilized scFv molecule, wherein the stabilized scFv molecule comprises at least one substitution selected from a group consisting of: a) substitution of an amino acid at Kabat position 13 of VH; b) substitution of an amino acid at Kabat position 16 of VH; c) substitution of an amino acid at Kabat position 46 of VL; 10 d) substitution of an amino acid at Kabat position 49 of VL; e) substitution of an amino acid at Kabat position 50 of VL; f) substitution of amino acids at Kabat positions 49 and 50 of VL; g) substitution of amino acid at Kabat position 101 of VH; h) substitution of amino acid at Kabat position 20 of VH; 15 i) substitution of amino acid at Kabat position 48 of VH4; j) substitution of amino acid at Kabat position 3 of VL; k) substitution of amino acid at Kabat position 55 of VH; 1) substitution of amino acid at Kabat position 67 of VH; m) substitution of amino acid at Kabat position 6 of VH; 20 n) substitution of amino acid at Kabat position 32 of VH; o) substitution of amino acid at Kabat position 49 of VH; p) substitution of amino acid at Kabat position 43 of VH; q) substitution of amino acid at Kabat position 72 of VH; r) substitution of amino acid at Kabat position 79 of VH; 25 s) substitution of amino acid at Kabat position 50 of VL; t) substitution of amino acid at Kabat position 75 of VL; u) substitution of amino acid at Kabat position 80 of VH; v) substitution of amino acid at Kabat position 83 of VH; or 30 iv) a VH domain and a VL domain with a stabilized interface between VH and VL of the scFv wherein the molecule has at least one substitution at an amino acid position directly within the interface and/or at an amino acid position that scaffolds the interaction between VH and VL, wherein the at least one amino acid position is selected from a - 278 - WO 2007/109254 PCT/US2007/006883 group consisting of: 47H, 37H, 45H, 46H, 51H, 59H, 109H, 14H, 26H, 27H, 40H, 69H, 103H, 29H, 38H, 44H, 49H, 55H, 63H, 54H, 68H, 72H, 74H, 79H, 82bH, 8H, 32H, 39H, 41H, 62H, 85H, 91H, 24H, 60H, 37L, 36L, 44L, 59L, 57L, 64L, 46L, 48L, 63L, 67L, 68L, 39L, 45L, 47L, 54L, 56L, 79L, 85L, 98L, 58L, 74L, 77L, 83L, 89L, and 5 104L, according to Kabat numbering.
[9] 9. The multivalent antigen binding molecule of claim 8, wherein the at least one stabilized scFv molecule is genetically fused to an antibody.
[10] 10 10. The multivalent antigen binding molecule of claim 8, wherein two stabilized scFv molecules are genetically fused to an antibody.
[11] 11. The multivalent antigen binding molecule of claim 9, wherein the at least one stabilized scFv molecule is genetically fused to the carboxy terminus of the light or 15 heavy chain of the antibody.
[12] 12. The multivalent antigen binding molecule of claim 9, wherein the at least one stabilized scFv molecule is genetically fused to the amino terminus of a light chain or heavy chain of the antibody. 20
[13] 13. The multivalent antigen binding molecule of any of claims 8-12, which comprises at least one binding site that binds to a molecule preferentially expressed on cancer cells 25
[14] 14. The multivalent antigen binding molecule of any of claims 8-12, wherein the binding molecule binds to an antigen which is selected from the group consisting of HERI, HER3, CD80, CD86, PD-1, CTLA4, B7-H4, RON, CD200, CD4, BAF R, EGFR, IGFR, VEGFR, a member of the TNF family of receptors, a Tie receptor, MET, IGF1, IGF2, TNF, a TNF ligand, IL-6, TWEAK, Fn14, CD20, CD23, CRIPTO, HGF, 30 alpha4betal integrin, alpha5betal integrin, alpha6beta4 integrin,, and alphaVbeta6 integrin.
[15] 15. The multivalent antigen binding molecule of any of claims 8-12, which - 279 - WO 2007/109254 PCT/US2007/006883 comprises at least one binding site that binds to a molecule involved in modulating immune responses.
[16] 16. The multivalent antigen binding molecule of any of claims 8-12, which 5 comprises at least one binding site that binds to a molecule involved in modulating angiogenesis.
[17] 17. The multivalent antigen binding molecule of any of claims 8-12, which comprises at least one binding site that binds to a neurological target. 10
[18] 18. The multivalent antigen binding molecule of any-one of claims 8-12, which is multispecific.
[19] 19. The multivalent antigen binding molecule of any one of claims 8-12, which is 15 bispecific.
[20] 20. The multispecific antigen binding molecule of claim 19, which binds to two members of the TNF receptor family. 20
[21] 21. A nucleic acid molecule comprising a nucleotide sequence which encodes the stabilized scFv molecule of any of claims 2-3 or the complement thereof.
[22] 22. The nucleic acid molecule of claim 21 which is in a vector. 25
[23] 23. A host cell comprising the vector of claim 22.
[24] 24. The host cell of claim 23, wherein the cell is a CHO cell or an NSO cell.
[25] 25. The host cell of claim 23, which expresses a population of binding molecules 30 comprising scFv molecules, the population having at least 10% fewer aggregates relative to a host cell which expresses a population of binding molecules comprising conventional scFv molecules. - 280 - WO 2007/109254 PCT/US2007/006883
[26] 26. A method of producing a stabilized binding molecule, comprising culturing the host cell of claim 23 under conditions such that the stabilized binding molecule is produced. 5
[27] 27. The method of claim 26, wherein at least 10 mg of the stabilized binding molecule is produced for every liter of the host cell culture medium and wherein not more than 10% of the binding molecule is present in aggregate form.
[28] 28. A method of treating a subject that would benefit from treatment with a binding 10 molecule of any of claims 8-12 comprising administering the binding molecule to the subject such that treatment occurs.
[29] 29. The method of 28, wherein the subject is suffering from a disease or disorder selected from the group consisting of cancer, an autoimmune disease or disorder, and a 15 neurological disease or disorder.
[30] 30. A method of stabilizing an scFv molecule comprising genetically fusing a VH domain and a VL domain using a (Gly 4 Ser), scFv linker and engineering the VH domain to comprise a cysteine at amino acid 44 and engineering the VL domain to comprise a 20 cysteine at amino acid 100 to form a stabilized scFv molecule.
[31] 31. A method of making a stabilized multivalent antibody comprising a stabilized scFv molecule, the method comprising genetically fusing a stabilized scFv molecule to the amino terminus or the carboxy terminus of a light or heavy chain of an antibody 25 molecule.
[32] 32. A method of making a stabilized scFv molecule comprising substituting at least one amino acid in the VH/VL interface of the scFv molecule and/or at least one amino acid that scaffolds the VH/VL interface of the scFv molecule substitution(s) wherein 30 said substitutions simultaneously improve the thermal stability of both the VH and VL domains of the scFv molecule as compared to a conventional scFv molecule.
[33] 33. A method of making a stabilized scFv molecule comprising substituting at least -281 - WO 2007/109254 PCT/US2007/006883 one amino acid in the VH domain or VL domain that covaries with two or more amino acids at the interface between the VH and VL domains.
[34] 34. A method of making a stabilized fusion protein comprising an scFv molecule 5 comprising substituting at least one amino acid in the.VH/VL interface of the scFv molecule and/or at least one amino acid in the VH/VL scaffold of the scFv molecule and genetically fusing the scFv molecule to a polypeptide to thereby make a stabilized fusion protein. 10
[35] 35. A method of improving the stability of a multivalent molecule comprising at least one scFv molecule, the method comprising introducing at least one stabilizing mutation into the at least one scFv molecule to thereby improve the stability of the multivalent molecule. 15
[36] 36. A method for large scale manufacture of a stabilized fusion protein, the method compnsing: (a) providing at least one candidate scFv molecule; (b) comparing the thermal stability of the at least one candidate scFv molecule 20 with a suitable control to identify at least one stabilized scFv molecules having increased thermal stability relative to the control, (c) selecting the at least one stabilized scFv molecules identified in step (b); (d) genetically fusing the at least one stabilized scFv molecule to a protein to form a stabilized fusion protein; 25 (e) transfecting a mammalian host cell with a nucleic acid molecule encoding the stabilized fusion protein, (f) culturing the host cell of step (e) under conditions such that the stabilized fusion protein is expressed; wherein the stabilized fusion protein is expressed with not more than 10% 30 protein aggregation when grown in 1OL or more of culture medium.
[37] 37. A graphical user interface comprising a graphical display of: (iii) a grid layout comprising a collection of residue usage frequency data - 282 - WO 2007/109254 PCT/US2007/006883 for a reference set of polypeptide sequences; and (iv) a co variation overlay.
[38] 38. A method for producing an immunoglobulin (Ig) superfamily polypeptide with 5 an improved biophysical property, said method comprising the steps of: a) providing an alignment of a curated reference set of sequences corresponding to an Ig fold of the polypeptide; b) calculating covariation between residues of the sequences of the alignment to identify covarying residues; 10 c) substituting a residue of the polypeptide that fails to satisfy a covariation with the covarying residue found at a corresponding position in the alignment.
[39] 39. The method of claim 38, wherein the biophysical property is selected from the group consisting of thermal stability, pH unfolding profile, stable removal of 15 glycosylation, solubility, biochemical function, and combinations thereof.
[40] 40. The method of claim 39, wherein the biochemical function is selected from the group consisting of recognition of a protein target or a chemical moiety, which chemical moiety is selected from the group consisting of phosphate, methyl, acetyl, lipid, 20 detergent, metal ion, halogen, RNA, DNA, oligonucleotide, and oligonucleoside.
[41] 41. The method of claim 38, wherein the Ig superfamily polypeptide is derived from a protein selected from the group consisting of an antibody or antigen-binding fragment thereof, an immunoglobulin receptor, a cell adhesion protein, an integrin, an allergen, a 25 T-cell receptor, and a major histocompatibility complex (MHC) molecule.
[42] 42. The method of claim 38, wherein the Ig fold is selected from the group consisting of a V-class fold, an I-class fold, a Cl-class fold, and a C2-class fold. 30
[43] 43. The method of claim 38, wherein the Ig superfamily polypeptide is a modified antibody selected from the group consisting of a domain antibody, a humanized antibody, a human antibody, a non-human monoclonal antibody, a chimeric antibody, a bispecific antibody, a scFv-containing antibody, and a domain-deleted antibody. - 283 - WO 2007/109254 PCT/US2007/006883
[44] 44. The method of claim 38, wherein the covarying residues are part of a structural feature selected from the group consisting of a disulfide bond, a salt bridge, a portion of a ligand binding pocket or surface, and a network of van Der Waals, hydrogen bond, and/or charge-charge interactions. 5
[45] 45. A method for producing an antibody or modified antibody with an improved stability, said method comprising the steps of: a) providing a high-resolution structural model of a template antibody that is experimentally validated as stable; 10 b) using the structural model to identify a VH/VL interface and/or interface scaffolding residue in the template antibody; c) using a homology model to identify a corresponding VH/VL interface residue of the antibody or modified antibody that is important for stabilizing the interface; and d) substituting the corresponding VH/VL interface residue in the antibody or modified 15 antibody with the interface residue from the template protein.
[46] 46. The method of claim 45, wherein the interface residue of the template protein buries at least 10 A 2 of surface area in the interface. 20
[47] 47. The method of claim 45, wherein the modified antibody is selected from the group consisting of a single domain antibody, a humanized antibody, a human antibody, a non-human monoclonal antibody, a chimeric antibody, a bispecific antibody, a scFv containing antibody, and a domain-deleted antibody. 25
[48] 48. A polypeptide produced by the method of any of claims 38-47.
[49] 49. The polypeptide of claim 48, wherein the polypeptide exhibits an improvement in a biophysical property selected from the group consisting of thermal stability, pH unfolding profile, stable removal of glycosylation, solubility, biochemical function, and 30 combinations thereof.
[50] 50. The polypeptide of claim 49, wherein biochemical function is ligand binding affinity or specificity. - 284 - WO 2007/109254 PCT/US2007/006883
[51] 51. The polypeptide of claim 48, which is derived from a protein selected from the group consisting of an antibody or antigen-binding fragment thereof, an immunoglobulin receptor, a cell adhesion protein, an integrin, an allergen, a T-cell receptor, and a major histocompatibility complex (MHC) molecule. 5
[52] 52. The polypeptide of claim 48, which is selected from the group consisting of a heavy chain variable region (VH), a light chain variable region (VL), and a single chain antibody (sFv). 10
[53] 53. The polypeptide of claim 48, wherein the Ig fold is selected from the group consisting of a V-class fold, an I-class fold, a Cl-class fold, and a C2-class fold.
[54] 54. The polypeptide of claim 48, which is a modified antibody selected from the group consisting of a single domain antibody, a humanized antibody, a human antibody, 15 a non-human monoclonal antibody, a chimeric antibody, a bispecific antibody, a scFv containing antibody, and a domain-deleted antibody.
[55] 55. A method for producing a library comprising stabilized scFv molecules, the method comprising: 20 (a) providing a reference set of sequences corresponding to a variable domain sequence ofan scFv molecule; (b) identifying an amino acid within the variable domain which is absent or found at low frequencies at a corresponding position in the reference set; (c) combining this information with covariation data suggesting that modifying 25 the amino acid identified with low frequency in the reference set may satisfy existing covariation restraints within the variable domain; and (d) substituting the amino acid of step (b) with a candidate stabilizing amino acid to thereby produce a library comprising stabilized scFv molecules. 30
[56] 56. The method of claim 55, wherein the amino acid of step (b) has a consensus score of less than 0.5.
[57] 57. A method for producing a stabilized scFv molecule, the method comprising: 35 (a) providing a scFv library designed according to the method of claim 56, -285- WO 2007/109254 PCT/US2007/006883 (b) screening the scFv library with a thermal challenge assay to identify candidate scFv molecules, (c) comparing the thermal stability of a candidate scFv molecule of the scFv library with a suitable control, 5 whereby an increase in thermal stability of a candidate scFv molecule relative to the control identifies the candidate scFv molecule as a stabilized scFv molecule.
[58] 58. A stabilized scFv molecule identified using the method of claim 57. 10
[59] 59. A fusion protein comprising the stabilized scFv molecule of claim 58.
[60] 60. A method for predicting the stability of a candidate protein comprising an amino acid sequence having candidate domain sequences, the method comprising: (a) providing a reference set of amino acid sequences corresponding to a 15 candidate domain sequence of the candidate protein; (b) determining residue frequencies at individual amino acid positions within the test domain sequence to obtain a consensus score; and (c) using the consensus score to predict the stability of the candidate protein, wherein the consensus score is correlated with the stability of the candidate 20 protein. - 286 -

类似技术:

---

公开号 | 公开日 | 专利标题

---

AU2007227292B2|2012-04-12|Stabilized polypeptide compositions

---

US9738707B2|2017-08-22|Heterodimeric Fc regions, binding molecules comprising same, and methods relating thereto

---

JP5398703B2|2014-01-29|Single-chain FC | region, binding polypeptide comprising the same, and methods related thereto

---

US8618252B2|2013-12-31|Neonatal Fc receptor |-binding polypeptide variants, dimeric Fc binding proteins and methods related thereto

---

US20090258420A1|2009-10-15|Altered polypeptides, immunoconjugates thereof, and methods related thereto

---

US20100203046A1|2010-08-12|Fc gamma receptor-binding polypeptide variants and methods related thereto

---

JP2008531060A|2008-08-14|A method for humanizing immunoglobulin variable regions via rational modification of complementarity determining residues

---

CN101448853A|2009-06-03|Stabilized polypeptide compositions

---

AU2012203476A1|2012-07-05|Neonatal Fc Receptor |-Binding Polypeptide Variants, Dimeric Fc Binding Proteins and Methods Related Thereto

同族专利:

---

公开号 | 公开日

---

CN103073639A|2013-05-01|

---

US20080050370A1|2008-02-28|

---

JP5374359B2|2013-12-25|

---

US7960142B2|2011-06-14|

---

EA015992B1|2012-01-30|

---

US20090048122A1|2009-02-19|

---

IL194120D0|2011-08-01|

---

WO2007109254A2|2007-09-27|

---

KR20080113241A|2008-12-29|

---

JP2009531028A|2009-09-03|

---

CA2646508A1|2007-09-27|

---

JP2013176394A|2013-09-09|

---

EP2024392A2|2009-02-18|

---

US8892364B2|2014-11-18|

---

NO20084339L|2008-12-16|

---

NZ591252A|2012-06-29|

---

AU2007227292B2|2012-04-12|

---

BRPI0709598A2|2011-07-19|

---

KR101516823B1|2015-05-07|

---

US20110301331A1|2011-12-08|

---

SG170750A1|2011-05-30|

---

US7951918B2|2011-05-31|

---

IL194120A|2015-03-31|

---

MX2008011874A|2009-01-07|

---

EA200802005A1|2009-06-30|

---

WO2007109254A3|2008-10-23|

---

BRPI0709598A8|2019-01-08|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

EP0281604B1|1986-09-02|1993-03-31|Enzon Labs Inc.|Single polypeptide chain binding molecules|

---

US6482919B2|1993-02-01|2002-11-19|Bristol-Myers Squibb Company|Expression vectors encoding bispecific fusion proteins and methods of producing biologically active bispecific fusion proteins in a mammalian cell|

---

AT207080T|1991-11-25|2001-11-15|Enzon Inc|MULTIVALENT ANTI-BINDING PROTEINS|

---

US6377893B1|1992-03-25|2002-04-23|Steven Albert Benner|Application of protein structure predictions|

---

US20050038609A1|1992-03-25|2005-02-17|Benner Steven Albert|Evolution-based functional genomics|

---

US5958784A|1992-03-25|1999-09-28|Benner; Steven Albert|Predicting folded structures of proteins|

---

WO1994012520A1|1992-11-20|1994-06-09|Enzon, Inc.|Linker for linked fusion polypeptides|

---

US5637481A|1993-02-01|1997-06-10|Bristol-Myers Squibb Company|Expression vectors encoding bispecific fusion proteins and methods of producing biologically active bispecific fusion proteins in a mammalian cell|

---

US5747654A|1993-06-14|1998-05-05|The United States Of America As Represented By The Department Of Health And Human Services|Recombinant disulfide-stabilized polypeptide fragments having binding specificity|

---

DE4425115A1|1994-07-15|1996-01-18|Boehringer Mannheim Gmbh|Method for modifying the stability of antibodies|

---

US6262238B1|1997-01-14|2001-07-17|Roche Diagnostic, Gmbh|Process for modifying the stability of antibodies|

---

US5656730A|1995-04-07|1997-08-12|Enzon, Inc.|Stabilized monomeric protein compositions|

---

AU740904B2|1997-03-20|2001-11-15|Government Of The United States Of America, As Represented By The Secretary Of The Department Of Health And Human Services, The|Recombinant antibodies and immunoconjugates targeted to CD-22 bearing cells and tumors|

---

US20020129889A1|2001-03-15|2002-09-19|Anderson Norman G.|Method and apparatus for making fibers for sectioned arrays|

---

WO2003099999A2|2002-05-20|2003-12-04|Abmaxis, Inc.|Generation and selection of protein library in silico|

---

ES2276735T3|2001-09-14|2007-07-01|Affimed Therapeutics Ag|SINGLE CHAIN MULTIMERIC FV ANTIBODIES IN TANDEM.|

---

ES2564570T3|2002-03-01|2016-03-23|Codexis Mayflower Holdings, Llc|Methods, systems and software for the identification of functional biomolecules|

---

CA2511013A1|2002-12-20|2004-07-15|Biogen Idec Ma Inc.|Multivalent lymphotoxin beta receptor agonists and therapeutic uses thereof|

---

NZ544923A|2003-06-27|2009-02-28|Biogen Idec Inc|Use of hydrophobic-interaction-chromatography or hinge-region modifications for the production of homogeneous anti-body solutions|

---

SG151294A1|2004-03-23|2009-04-30|Biogen Idec Inc|Receptor coupling agents and therapeutic uses thereof|

---

AR049390A1|2004-06-09|2006-07-26|Wyeth Corp|ANTIBODIES AGAINST HUMAN INTERLEUQUINE-13 AND USES OF THE SAME|

---

WO2006074399A2|2005-01-05|2006-07-13|Biogen Idec Ma Inc.|Multispecific binding molecules comprising connecting peptides|

---

CN103073639A|2006-03-17|2013-05-01|比奥根艾迪克Ma公司|Stabilized polypeptide compositions|

---

CA2697922A1|2007-08-28|2009-03-12|Biogen Idec Ma Inc.|Compositions that bind multiple epitopes of igf-1r|

---

US20090155255A1|2007-09-27|2009-06-18|Biogen Idec Ma Inc.|Cd23 binding molecules and methods of use thereof|TWI353991B|2003-05-06|2011-12-11|Syntonix Pharmaceuticals Inc|Immunoglobulin chimeric monomer-dimer hybrids|

---

WO2006068953A2|2004-12-21|2006-06-29|Astrazeneca Ab|Antibodies directed to angiopoietin-2 and uses thereof|

---

US8609816B2|2005-12-08|2013-12-17|Medarex, L.L.C.|Human monoclonal antibodies to O8E|

---

CN103073639A|2006-03-17|2013-05-01|比奥根艾迪克Ma公司|Stabilized polypeptide compositions|

---

EP2059533B1|2006-08-30|2012-11-14|Genentech, Inc.|Multispecific antibodies|

---

AT525392T|2007-01-30|2011-10-15|Epivax Inc|REGULATORY T-CELL EPITOPES AND COMPOSITIONS AND APPLICATIONS THEREOF|

---

EP2190861A4|2007-08-22|2011-03-30|Univ California|Activatable binding polypeptides and methods of identification and use thereof|

---

CA2697922A1|2007-08-28|2009-03-12|Biogen Idec Ma Inc.|Compositions that bind multiple epitopes of igf-1r|

---

ES2595638T3|2007-09-26|2017-01-02|Chugai Seiyaku Kabushiki Kaisha|Method to modify the isoelectric point of an antibody by replacing amino acids in a CDR|

---

EP2195341B1|2007-09-26|2017-03-22|UCB Biopharma SPRL|Dual specificity antibody fusions|

---

US20090155255A1|2007-09-27|2009-06-18|Biogen Idec Ma Inc.|Cd23 binding molecules and methods of use thereof|

---

US9266967B2|2007-12-21|2016-02-23|Hoffmann-La Roche, Inc.|Bivalent, bispecific antibodies|

---

US20090162359A1|2007-12-21|2009-06-25|Christian Klein|Bivalent, bispecific antibodies|

---

AU2009206724A1|2008-01-22|2009-07-30|Biogen Idec Ma Inc.|RON antibodies and uses thereof|

---

RU2509085C2|2008-01-28|2014-03-10|Медиммун Лимитед|Stabilised antibodies against angiopoietin-2 and their use|

---

WO2009114591A1|2008-03-11|2009-09-17|Smithkline Beecham Corporation|Method and apparatus for screening drugs for predictors of quantitatively measured events|

---

US20100260668A1|2008-04-29|2010-10-14|Abbott Laboratories|Dual Variable Domain Immunoglobulins and Uses Thereof|

---

AU2009241589B2|2008-04-29|2013-10-10|Abbvie Inc.|Dual variable domain immunoglobulins and uses thereof|

---

JP5723769B2|2008-06-03|2015-05-27|アッヴィ・インコーポレイテッド|Dual variable domain immunoglobulins and uses thereof|

---

US9035027B2|2008-06-03|2015-05-19|Abbvie Inc.|Dual variable domain immunoglobulins and uses thereof|

---

RU2011104348A|2008-07-08|2012-08-20|Эбботт Лэборетриз |IMMUNOGLOBULINS WITH DOUBLE VARIABLE DOMAIN AGAINST PROSTAGLANDINE E2 AND THEIR APPLICATION|

---

JP2011528035A|2008-07-14|2011-11-10|ユーシカゴアルゴン，エルエルシー|Method for systematically controlling protein stability|

---

AU2009288167B2|2008-09-03|2015-10-22|Genentech, Inc.|Multispecific antibodies|

---

PT2334705T|2008-09-26|2017-03-22|Ucb Biopharma Sprl|Biological products|

---

BRPI0919426A2|2008-09-30|2015-12-15|Abbott Lab|Enhanced RNA Exposure Methods|

---

PT2344540T|2008-10-02|2018-02-02|Aptevo Res & Development Llc|Cd86 antagonist multi-target binding proteins|

---

US8268314B2|2008-10-08|2012-09-18|Hoffmann-La Roche Inc.|Bispecific anti-VEGF/anti-ANG-2 antibodies|

---

EP3543256A1|2009-01-12|2019-09-25|Cytomx Therapeutics Inc.|Modified antibody compositions, methods of making and using thereof|

---

AU2010215761B2|2009-02-23|2017-04-06|Cytomx Therapeutics, Inc|Proproteins and methods of use thereof|

---

JP2012521360A|2009-03-20|2012-09-13|アムジエン・インコーポレーテツド|Selective and potent peptide inhibitors of Kv1.3|

---

EP2411050A4|2009-03-25|2013-07-17|Vet Therapeutics Inc|Antibody constant domain regions and uses thereof|

---

JP6072401B2|2009-03-30|2017-02-01|アバイア インク．|A system and method for managing a contact center with a graphical call connection display.|

---

SG175004A1|2009-04-02|2011-11-28|Roche Glycart Ag|Multispecific antibodies comprising full length antibodies and single chain fab fragments|

---

WO2010112194A1|2009-04-02|2010-10-07|F. Hoffmann-La Roche Ag|Antigen-binding polypeptides and multispecific antibodies comprising them|

---

SG175078A1|2009-04-07|2011-11-28|Roche Glycart Ag|Bispecific anti-erbb-1/anti-c-met antibodies|

---

KR20110124369A|2009-04-07|2011-11-16|로슈 글리카트 아게|Bispecific anti-erbb-3/anti-c-met antibodies|

---

MX2011010168A|2009-04-07|2011-10-11|Roche Glycart Ag|Trivalent, bispecific antibodies.|

---

RU2570633C2|2009-05-27|2015-12-10|Ф.Хоффманн-Ля Рош Аг|Tri- or tetraspecific antibodies|

---

CA2763488A1|2009-05-28|2010-12-02|Neil James Clarke|Antigen-binding proteins|

---

US9676845B2|2009-06-16|2017-06-13|Hoffmann-La Roche, Inc.|Bispecific antigen binding proteins|

---

TW201109438A|2009-07-29|2011-03-16|Abbott Lab|Dual variable domain immunoglobulins and uses thereof|

---

TW201119673A|2009-09-01|2011-06-16|Abbott Lab|Dual variable domain immunoglobulins and uses thereof|

---

US9493578B2|2009-09-02|2016-11-15|Xencor, Inc.|Compositions and methods for simultaneous bivalent and monovalent co-engagement of antigens|

---

US20120302737A1|2009-09-16|2012-11-29|Genentech, Inc.|Coiled coil and/or tether containing protein complexes and uses thereof|

---

JP2013507926A|2009-10-14|2013-03-07|メリマックファーマシューティカルズインコーポレーティッド|Bispecific binding agents targeting IGF-1R and ErbB3 signaling and uses thereof|

---

AR078651A1|2009-10-15|2011-11-23|Abbott Lab|IMMUNOGLOBULINS WITH DUAL VARIABLE DOMAIN AND USES OF THE SAME|

---

UY32979A|2009-10-28|2011-02-28|Abbott Lab|IMMUNOGLOBULINS WITH DUAL VARIABLE DOMAIN AND USES OF THE SAME|

---

JP2013511279A|2009-11-20|2013-04-04|アムジエン・インコーポレーテツド|Anti-Orai1 antigen binding protein and use thereof|

---

ES2592385T3|2009-12-29|2016-11-29|Aptevo Research And Development Llc|Heterodimer binding proteins and uses thereof|

---

MX341687B|2010-02-10|2016-08-30|Immunogen Inc|Cd20 antibodies and uses thereof.|

---

EP2542258A4|2010-03-04|2013-08-21|Vet Therapeutics Inc|Monoclonal antibodies directed to cd52|

---

US9616120B2|2010-03-04|2017-04-11|Vet Therapeutics, Inc.|Monoclonal antibodies directed to CD20|

---

GB201005064D0|2010-03-25|2010-05-12|Ucb Pharma Sa|Biological products|

---

GB201005063D0|2010-03-25|2010-05-12|Ucb Pharma Sa|Biological products|

---

AR080794A1|2010-03-26|2012-05-09|Hoffmann La Roche|BIVING SPECIFIC ANTIBODIES ANTI-VEGF / ANTI-ANG-2|

---

TW201138821A|2010-03-26|2011-11-16|Roche Glycart Ag|Bispecific antibodies|

---

US20120010388A1|2010-04-16|2012-01-12|Gottfried Himmler|LeY SPECIFIC BIOTHERAPEUTIC|

---

JP2013528362A|2010-04-21|2013-07-11|グラクソグループリミテッド|Binding domain|

---

AU2011249783B9|2010-05-06|2014-11-06|Novartis Ag|Compositions and methods of use for therapeutic low density lipoprotein -related protein 6antibodies|

---

AU2011249782B2|2010-05-06|2014-10-02|Novartis Ag|Compositions and methods of use for therapeutic low density lipoprotein - related protein 6multivalent antibodies|

---

CN102946858B|2010-05-10|2015-09-30|英塔斯制药有限公司|The liquid preparation of the polypeptide containing immunoglobulin Fc|

---

EP2571992B1|2010-05-21|2018-04-25|Merrimack Pharmaceuticals, Inc.|Bi-specific fusion proteins|

---

TW201217527A|2010-07-09|2012-05-01|Biogen Idec Hemophilia Inc|Processable single chain molecules and polypeptides made using same|

---

AU2011283694B2|2010-07-29|2017-04-13|Xencor, Inc.|Antibodies with modified isoelectric points|

---

EP2601218A4|2010-08-03|2015-02-18|Abbvie Inc|Dual variable domain immunoglobulins and uses thereof|

---

WO2012019061A2|2010-08-05|2012-02-09|Stem Centrx, Inc.|Novel effectors and methods of use|

---

WO2012021475A2|2010-08-12|2012-02-16|Eli Lilly And Company|ANTI-N3pGlu AMYLOID BETA PEPTIDE ANTIBODIES AND USES THEREOF|

---

RU2013110876A|2010-08-24|2014-09-27|Рош Гликарт Аг|ACTIVATED SPECIFIC ANTIBODIES|

---

EP2609111B1|2010-08-24|2017-11-01|F.Hoffmann-La Roche Ag|Bispecific antibodies comprising a disulfide stabilized-fv fragment|

---

US9046513B2|2010-08-26|2015-06-02|Abbvie Inc.|Dual variable domain immunoglobulins and uses thereof|

---

CN103260646B|2010-08-27|2016-01-13|施特姆森特克斯股份有限公司|Notum protein modulators and using method|

---

TW201302793A|2010-09-03|2013-01-16|Glaxo Group Ltd|Novel antigen binding proteins|

---

AU2011295715B9|2010-09-03|2017-02-23|Abbvie Stemcentrx Llc|Novel modulators and methods of use|

---

SG190990A1|2010-12-08|2013-07-31|Stem Centrx Inc|Novel modulators and methods of use|

---

SA4398B1|2011-02-18|2015-10-09|ستيم سينتركس، انك.|Novel modulators and methods of use|

---

US20130058947A1|2011-09-02|2013-03-07|Stem Centrx, Inc|Novel Modulators and Methods of Use|

---

BR112013006679A2|2010-09-22|2016-06-07|Amgen Inc|carrier immunoglobulins and their uses|

---

AR083495A1|2010-10-22|2013-02-27|Esbatech Alcon Biomed Res Unit|STABLE AND SOLUBLE ANTIBODIES|

---

KR20130135866A|2010-11-05|2013-12-11|자임워크스 인코포레이티드|Stable heterodimeric antibody design with mutations in the fc domain|

---

MX341921B|2011-02-28|2016-09-07|F Hoffmann-La Roche Ag|Antigen binding proteins.|

---

AR085403A1|2011-02-28|2013-09-25|Hoffmann La Roche|MONOVALENT PROTEINS THAT JOIN ANTIGENS|

---

JP2014517683A|2011-03-11|2014-07-24|アムジェンインコーポレイテッド|Methods of correlated mutation analysis to improve therapeutic antibodies|

---

EP2502934B1|2011-03-24|2018-01-17|Universitätsmedizin der Johannes Gutenberg-Universität Mainz|Single chain antigen recognizing constructsstabilized by the introduction of novel disulfide bonds|

---

TWI631136B|2011-04-19|2018-08-01|莫瑞麥克製藥公司|Monospecific and bispecific anti-igf-1r and anti-erbb3 antibodies|

---

CA2838833A1|2011-06-10|2012-12-13|Biogen Idec Ma Inc.|Pro-coagulant compounds and methods of use thereof|

---

EP3632462A1|2012-07-06|2020-04-08|Genmab B.V.|Dimeric protein with triple mutations|

---

SG10201605572TA|2011-07-06|2016-09-29|Genmab Bv|Antibody variants and uses thereof|

---

US11180572B2|2012-07-06|2021-11-23|Genmab B.V.|Dimeric protein with triple mutations|

---

WO2013025779A1|2011-08-15|2013-02-21|Amplimmune, Inc.|Anti-b7-h4 antibodies and their uses|

---

CA2843158A1|2011-08-26|2013-03-07|Merrimack Pharmaceuticals, Inc.|Tandem fc bispecific antibodies|

---

KR101684750B1|2011-09-23|2016-12-08|로슈 글리카트 아게|Bispecific anti-egfr/anti-igf-1r antibodies|

---

US10851178B2|2011-10-10|2020-12-01|Xencor, Inc.|Heterodimeric human IgG1 polypeptides with isoelectric point modifications|

---

CA2851642A1|2011-10-18|2013-04-25|Coherus Biosciences, Inc.|Etanercept formulations stabilized with sodium chloride|

---

US10493151B2|2011-10-18|2019-12-03|Coherus Biosciences, Inc.|Etanercept formulations stabilized with sodium chloride|

---

AU2012325232B2|2011-10-20|2017-08-31|Esbatech - A Novartis Company Llc|Stable multiple antigen-binding antibody|

---

AR088633A1|2011-11-04|2014-06-25|Novartis Ag|LRP6 MEDIUM LIFE EXTENSION CONSTRUCTS |

---

CN104080811B|2011-11-04|2019-09-27|酵活有限公司|There is the antibody design of the stabilization heterodimeric of mutation in Fc structural domain|

---

SI2794658T1|2011-12-19|2017-05-31|Synimmune Gmbh|Bispecific antibody molecule|

---

MX2014008101A|2011-12-30|2014-09-25|Abbvie Inc|Dual variable domain immunoglobulins against il-13 and/or il-17.|

---

CA2861124A1|2012-02-10|2013-08-15|Genentech, Inc.|Single-chain antibodies and other heteromultimers|

---

CA2868404A1|2012-04-05|2013-10-10|F. Hoffmann-La Roche Ag|Bispecific antibodies against human tweak and human il17 and uses thereof|

---

WO2014004586A1|2012-06-25|2014-01-03|Zymeworks Inc.|Process and methods for efficient manufacturing of highly pure asymmetric antibodies in mammalian cells|

---

MX354862B|2012-06-27|2018-03-23|Hoffmann La Roche|Method for selection and production of tailor-made highly selective and multi-specific targeting entities containing at least two different binding entities and uses thereof.|

---

CN104411725B|2012-07-04|2018-09-28|弗·哈夫曼-拉罗切有限公司|Anti-biotin antibodies and application method|

---

SG11201407420RA|2012-07-04|2015-02-27|Hoffmann La Roche|Covalently linked antigen-antibody conjugates|

---

SI2869837T1|2012-07-04|2016-12-30|F. Hoffmann-La Roche Ag|Anti-theophylline antibodies and methods of use|

---

EP3016681B1|2013-07-05|2019-12-18|Genmab A/S|Humanized or chimeric cd3 antibodies|

---

EA029193B1|2012-07-09|2018-02-28|Кохерус Байосайенсис, Инк.|Etanercept formulations exhibiting marked reduction in sub-visible particles|

---

PL3495387T3|2012-07-13|2022-01-10|Roche Glycart Ag|Bispecific anti-vegf/anti-ang-2 antibodies and their use in the treatment of ocular vascular diseases|

---

JP6774164B2|2012-08-24|2020-10-21|中外製薬株式会社|Mouse FcγRII specific Fc antibody|

---

BR112015005161A2|2012-09-11|2017-07-04|Coherus Biosciences Inc|correctly folded etanercept in high purity and excellent yield|

---

JP6581505B2|2012-10-03|2019-09-25|ザイムワークス，インコーポレイテッド|Methods for quantifying heavy and light chain polypeptide pairs|

---

MX2015005593A|2012-11-01|2016-02-05|Abbvie Inc|Anti-vegf/dll4 dual variable domain immunoglobulins and uses thereof.|

---

MX2015006758A|2012-11-28|2016-06-10|Zymeworks Inc|Engineered immunoglobulin heavy chain-light chain pairs and uses thereof.|

---

US9914785B2|2012-11-28|2018-03-13|Zymeworks Inc.|Engineered immunoglobulin heavy chain-light chain pairs and uses thereof|

---

CA2894689A1|2012-12-19|2014-06-26|Amplimmune, Inc.|Anti-human b7-h4 antibodies and their uses|

---

JP6377635B2|2013-01-10|2018-08-22|ゲンマブ ビー．ブイ．|Human IgG1 Fc region variants and uses thereof|

---

US10487155B2|2013-01-14|2019-11-26|Xencor, Inc.|Heterodimeric proteins|

---

US10858417B2|2013-03-15|2020-12-08|Xencor, Inc.|Heterodimeric proteins|

---

SI2943511T1|2013-01-14|2020-01-31|Xencor, Inc.|Novel heterodimeric proteins|

---

US9701759B2|2013-01-14|2017-07-11|Xencor, Inc.|Heterodimeric proteins|

---

US10131710B2|2013-01-14|2018-11-20|Xencor, Inc.|Optimized antibody variable regions|

---

US11053316B2|2013-01-14|2021-07-06|Xencor, Inc.|Optimized antibody variable regions|

---

AU2014207549B2|2013-01-15|2018-12-06|Xencor, Inc.|Rapid clearance of antigen complexes using novel antibodies|

---

US10047163B2|2013-02-08|2018-08-14|Abbvie Stemcentrx Llc|Multispecific constructs|

---

JP2016510755A|2013-03-06|2016-04-11|メリマック ファーマシューティカルズ インコーポレーティッド|Anti-C-MET Tandem Fc Bispecific Antibody|

---

SG10201707021RA|2013-03-11|2017-10-30|Genzyme Corp|Engineered anti-tgf-beta antibodies and antigen-binding fragments|

---

US10968276B2|2013-03-12|2021-04-06|Xencor, Inc.|Optimized anti-CD3 variable regions|

---

US9605084B2|2013-03-15|2017-03-28|Xencor, Inc.|Heterodimeric proteins|

---

US10047167B2|2013-03-15|2018-08-14|Eli Lilly And Company|Methods for producing fabs and bi-specific antibodies|

---

US10106624B2|2013-03-15|2018-10-23|Xencor, Inc.|Heterodimeric proteins|

---

CA2904448A1|2013-03-15|2014-09-18|Tariq Ghayur|Dual specific binding proteins directed against il-1.beta. and/or il-17|

---

CA3093606A1|2013-03-15|2014-09-18|Xencor, Inc.|Heterodimeric proteins for induction of t cells|

---

US10519242B2|2013-03-15|2019-12-31|Xencor, Inc.|Targeting regulatory T cells with heterodimeric proteins|

---

KR102049991B1|2013-03-28|2019-12-02|삼성전자주식회사|Bispecific anti-cMet/anti-Her2 antibodies|

---

KR102074421B1|2013-03-29|2020-02-10|삼성전자주식회사|Bispecific anti-cMet/anti-EGFR antibodies|

---

KR101641206B1|2013-06-24|2016-07-22|에이비엘바이오 주식회사|An antibody-drug conjugate having improved stability, and use thereof|

---

DK3024851T3|2013-07-25|2018-08-06|Cytomx Therapeutics Inc|MULTI-SPECIFIC ANTIBODIES, MULTI-SPECIFICALLY ACTIVABLE ANTIBODIES, AND METHODS FOR USING IT|

---

CN104341504B|2013-08-06|2017-10-24|百奥泰生物科技有限公司|Bispecific antibody|

---

MX2016002720A|2013-09-06|2016-06-08|Hoffmann La Roche|Method for improving antibody stability.|

---

JP6422956B2|2013-10-11|2018-11-14|エフ．ホフマン−ラ ロシュ アーゲーＦ． Ｈｏｆｆｍａｎｎ−Ｌａ Ｒｏｃｈｅ Ａｋｔｉｅｎｇｅｓｅｌｌｓｃｈａｆｔ|Multispecific domain exchange common variable light chain antibody|

---

EP3065769A4|2013-11-08|2017-05-31|Biogen MA Inc.|Procoagulant fusion compound|

---

MX2016006572A|2013-11-27|2016-12-09|Zymeworks Inc|Bispecific antigen-binding constructs targeting her2.|

---

WO2015085179A1|2013-12-06|2015-06-11|The Regents Of The University Of California|Alpha-v beta-6 integrin-binding antibody fragments|

---

JP6608827B2|2013-12-20|2019-11-20|エフ．ホフマン−ラロシュアーゲー|Humanized anti-tauantibodies and methods of use|

---

MX2016008187A|2014-01-03|2016-09-29|F Hoffmann-La Roche Ag|Bispecific anti-hapten/anti-blood brain barrier receptor antibodies, complexes thereof and their use as blood brain barrier shuttles.|

---

RU2682754C2|2014-01-03|2019-03-21|Ф. Хоффманн-Ля Рош Аг|Covalent bonded polypeptide toxin and antibody conjugates|

---

WO2015101587A1|2014-01-03|2015-07-09|F. Hoffmann-La Roche Ag|Covalently linked helicar-anti-helicar antibody conjugates and uses thereof|

---

JOP20200094A1|2014-01-24|2017-06-16|Dana Farber Cancer Inst Inc|Antibody molecules to pd-1 and uses thereof|

---

CR20160425A|2014-03-14|2017-05-26|Novartis Ag|ANTIBODY MOLECULES THAT JOIN LAG-3 AND USES OF THE SAME|

---

EP3119423A2|2014-03-15|2017-01-25|Novartis AG|Treatment of cancer using chimeric antigen receptor|

---

HUE048734T2|2014-03-28|2020-08-28|Xencor Inc|Bispecific antibodies that bind to cd38 and cd3|

---

GB201411320D0|2014-06-25|2014-08-06|Ucb Biopharma Sprl|Antibody construct|

---

AR100978A1|2014-06-26|2016-11-16|F Hoffmann-La Roche Ag|ANTI-Tau HUMANIZED ANTIBODY BRAIN LAUNCHERSAND USES OF THE SAME|

---

DK3160513T3|2014-06-30|2020-04-06|Glykos Finland Oy|SACCHARIDE DERIVATIVES OF A TOXIC PAYLOAD AND ANTIBODY CONJUGATES THEREOF|

---

EP3164417A1|2014-07-01|2017-05-10|Pfizer Inc|Bispecific heterodimeric diabodies and uses thereof|

---

US20180326084A1|2015-07-10|2018-11-15|Genmab A/S|Axl-specific antibody-drug conjugates for cancer treatment|

---

LT3169706T|2014-07-11|2020-03-25|Genmab A/S|Antibodies binding axl|

---

JP2017528433A|2014-07-21|2017-09-28|ノバルティス アーゲー|Low immunoenhancing dose of mTOR inhibitor and CAR combination|

---

RU2751660C2|2014-07-21|2021-07-15|Новартис Аг|Treatment of malignant neoplasm using humanized chimeric antigen receptor against bcma|

---

SG10201913765YA|2014-07-21|2020-03-30|Novartis Ag|Treatment of cancer using a cd33 chimeric antigen receptor|

---

US20170211055A1|2014-07-21|2017-07-27|Novartis Ag|Sortase synthesized chimeric antigen receptors|

---

AU2015292406B2|2014-07-25|2021-03-11|Cytomx Therapeutics, Inc|Anti-CD3 antibodies, activatable anti-CD3 antibodies, multispecific anti-CD3 antibodies, multispecific activatable anti-CD3 antibodies, and methods of using the same|

---

US20170209492A1|2014-07-31|2017-07-27|Novartis Ag|Subset-optimized chimeric antigen receptor-containing t-cells|

---

US10851149B2|2014-08-14|2020-12-01|The Trustees Of The University Of Pennsylvania|Treatment of cancer using GFR α-4 chimeric antigen receptor|

---

LT3183268T|2014-08-19|2020-06-10|Novartis Ag|Anti-cd123 chimeric antigen receptorfor use in cancer treatment|

---

EP3194435A1|2014-09-15|2017-07-26|Amgen Inc.|Bi-specific anti-cgrp receptor/pac1 receptor antigen binding proteins and uses thereof|

---

WO2017049004A1|2015-09-15|2017-03-23|Amgen Inc.|Tetravalent bispecific and tetraspecific antigen binding proteins and uses thereof|

---

KR20210149228A|2014-09-17|2021-12-08|노파르티스 아게|Targeting cytotoxic cells with chimeric receptors for adoptive immunotherapy|

---

MA41044A|2014-10-08|2017-08-15|Novartis Ag|COMPOSITIONS AND METHODS OF USE FOR INCREASED IMMUNE RESPONSE AND CANCER TREATMENT|

---

CR20170143A|2014-10-14|2017-06-19|Dana-Farber Cancer Inst Inc|ANTIBODY MOLECULES THAT JOIN PD-L1 AND USES OF THE SAME|

---

EP3220900B1|2014-11-21|2020-09-23|University of Maryland, Baltimore|Targeted structure-specific particulate delivery systems|

---

EP3223845B1|2014-11-26|2021-05-19|Xencor, Inc.|Heterodimeric antibodies that bind cd3 and cd20|

---

MX2017006918A|2014-11-26|2018-01-25|Xencor Inc|Heterodimeric antibodies that bind cd3 and cd38.|

---

US10259887B2|2014-11-26|2019-04-16|Xencor, Inc.|Heterodimeric antibodies that bind CD3 and tumor antigens|

---

JP6721590B2|2014-12-03|2020-07-15|エフ．ホフマン−ラ ロシュ アーゲーＦ． Ｈｏｆｆｍａｎｎ−Ｌａ Ｒｏｃｈｅ Ａｋｔｉｅｎｇｅｓｅｌｌｓｃｈａｆｔ|Multispecific antibody|

---

US20180334490A1|2014-12-03|2018-11-22|Qilong H. Wu|Methods for b cell preconditioning in car therapy|

---

WO2016094881A2|2014-12-11|2016-06-16|Abbvie Inc.|Lrp-8 binding proteins|

---

US10428155B2|2014-12-22|2019-10-01|Xencor, Inc.|Trispecific antibodies|

---

DK3242890T3|2015-01-08|2019-10-21|BioNTech SE|AGONISTIC TNF RECEPTOR BINDING AGENTS|

---

MA41375A|2015-01-22|2017-11-28|Lilly Co Eli|BISPECIFIC IGG ANTIBODIES AND THEIR PREPARATION PROCESSES|

---

TW201643190A|2015-02-05|2016-12-16|中外製藥股份有限公司|Antibodies comprising an ion concentration dependent antigen-binding domain, Fc region variants, IL-8-binding antibodies, and uses therof|

---

TWI733661B|2015-03-04|2021-07-21|美商健臻公司|MODIFIED-IgG ANTIBODIES THAT BIND TRANSFORMING GROWTH FACTOR-β1 WITH HIGH AFFINITY, AVIDITY AND SPECIFICITY|

---

TWI726870B|2015-03-04|2021-05-11|美商健臻公司|scFv-Fc DIMERS THAT BIND TRANSFORMING GROWTH FACTOR-β1 WITH HIGH AFFINITY, AVIDITY AND SPECIFICITY|

---

WO2016141387A1|2015-03-05|2016-09-09|Xencor, Inc.|Modulation of t cells with bispecific antibodies and fc fusions|

---

CA2981103A1|2015-03-31|2016-10-06|Vhsquared Limited|Polypeptides|

---

WO2016164731A2|2015-04-08|2016-10-13|Novartis Ag|Cd20 therapies, cd22 therapies, and combination therapies with a cd19 chimeric antigen receptor- expressing cell|

---

US20180298068A1|2015-04-23|2018-10-18|Novartis Ag|Treatment of cancer using chimeric antigen receptor and protein kinase a blocker|

---

TW201710286A|2015-06-15|2017-03-16|艾伯維有限公司|Binding proteins against VEGF, PDGF, and/or their receptors|

---

CN113929779A|2015-06-24|2022-01-14|豪夫迈·罗氏有限公司|Humanized anti-Tauantibodies and methods of use|

---

ES2878188T3|2015-07-29|2021-11-18|Novartis Ag|Combination therapies comprising antibody molecules against LAG-3|

---

EP3316902A1|2015-07-29|2018-05-09|Novartis AG|Combination therapies comprising antibody molecules to tim-3|

---

EP3865147A1|2015-10-02|2021-08-18|Silver Creek Pharmaceuticals, Inc.|Bi-specific therapeutic proteins for tissue repair|

---

US11161915B2|2015-10-08|2021-11-02|Zymeworks Inc.|Antigen-binding polypeptide constructs comprising kappa and lambda light chains and uses thereof|

---

US10227410B2|2015-12-07|2019-03-12|Xencor, Inc.|Heterodimeric antibodies that bind CD3 and PSMA|

---

MA43567A|2015-12-15|2018-11-14|Amgen Inc|PACAP ANTIBODIES AND THEIR USES|

---

MX2018007423A|2015-12-17|2018-11-09|Novartis Ag|Antibody molecules to pd-1 and uses thereof.|

---

US20200261573A1|2015-12-17|2020-08-20|Novartis Ag|Combination of c-met inhibitor with antibody molecule to pd-1 and uses thereof|

---

WO2017125897A1|2016-01-21|2017-07-27|Novartis Ag|Multispecific molecules targeting cll-1|

---

SG11201807489PA|2016-03-04|2018-09-27|Novartis Ag|Cells expressing multiple chimeric antigen receptormolecules and uses therefore|

---

EP3432924A1|2016-03-23|2019-01-30|Novartis AG|Cell secreted minibodies and uses thereof|

---

MX2018012615A|2016-04-15|2019-05-30|Novartis Ag|Compositions and methods for selective protein expression.|

---

US20200317792A1|2016-05-24|2020-10-08|Karl Edwin Griswold|Antibodies and methods of making same|

---

WO2017210617A2|2016-06-02|2017-12-07|Porter, David, L.|Therapeutic regimens for chimeric antigen receptor - expressing cells|

---

MA45255A|2016-06-14|2019-04-17|Xencor Inc|BISPECIFIC CONTROL POINT INHIBITORS ANTIBODIES|

---

US10793632B2|2016-08-30|2020-10-06|Xencor, Inc.|Bispecific immunomodulatory antibodies that bind costimulatory and checkpoint receptors|

---

CA3028728A1|2016-06-23|2017-12-28|Capital One Services, Llc|Neural network systems and methods for generating distributed representations of electronic transaction information|

---

RU2019102008A3|2016-06-28|2020-11-26|||

---

US11091557B2|2016-07-14|2021-08-17|Genmab A/S|Methods of producing multispecific antibodies against CD40 and CD137|

---

CN110461315A|2016-07-15|2019-11-15|诺华股份有限公司|Cytokines release syndrome is treated and prevented using with the Chimeric antigen receptor of kinase inhibitor combination|

---

EP3491016A1|2016-07-28|2019-06-05|Novartis AG|Combination therapies of chimeric antigen receptors and pd-1 inhibitors|

---

CA3032581A1|2016-08-01|2018-02-08|Novartis Ag|Treatment of cancer using a chimeric antigen receptor in combination with an inhibitor of a pro-m2 macrophage molecule|

---

EP3494991A4|2016-08-05|2020-07-29|Chugai Seiyaku Kabushiki Kaisha|Composition for prophylaxis or treatment of il-8 related diseases|

---

US10525083B2|2016-10-07|2020-01-07|Novartis Ag|Nucleic acid molecules encoding chimeric antigen receptors comprising a CD20 binding domain|

---

AU2017342559A1|2016-10-14|2019-05-02|Xencor, Inc.|Bispecific heterodimeric fusion proteins containing IL-15/IL-15Ralpha Fc-fusion proteins and PD-1 antibody fragments|

---

BR112019008702A2|2016-11-01|2019-07-16|Genmab B.V.|polypeptide, method of decreasing a polypeptide effector function, composition, method of treatment, kit of parts, and use of a polypeptide or composition.|

---

MA46753A|2016-11-14|2019-09-18|Amgen Inc|BISPECIFIC OR BIPARATOPIC ANTIGEN BINDING PROTEINS AND USES THEREOF|

---

US20180230218A1|2017-01-04|2018-08-16|Immunogen, Inc.|Met antibodies and immunoconjugates and uses thereof|

---

WO2018140725A1|2017-01-26|2018-08-02|Novartis Ag|Cd28 compositions and methods for chimeric antigen receptor therapy|

---

EP3580233A1|2017-02-10|2019-12-18|Genmab B.V.|Polypeptide variants and uses thereof|

---

EP3589647A1|2017-02-28|2020-01-08|Novartis AG|Shp inhibitor compositions and uses for chimeric antigen receptor therapy|

---

MX2019010583A|2017-03-09|2019-11-25|Genmab As|Antibodies against pd-l1.|

---

SG11201908772TA|2017-03-31|2019-10-30|Genmab Holding B V|Bispecific anti-cd37 antibodies, monoclonal anti-cd37 antibodies and methods of use thereof|

---

US20200055948A1|2017-04-28|2020-02-20|Novartis Ag|Cells expressing a bcma-targeting chimeric antigen receptor, and combination therapy with a gamma secretase inhibitor|

---

US20200179511A1|2017-04-28|2020-06-11|Novartis Ag|Bcma-targeting agent, and combination therapy with a gamma secretase inhibitor|

---

TW201919692A|2017-06-07|2019-06-01|荷蘭商珍美寶公司|Pharmaceutical compositions comprising antibodies and methods of use thereof|

---

SG11201912473PA|2017-06-22|2020-01-30|Novartis Ag|Antibody molecules to cd73 and uses thereof|

---

JP2020525483A|2017-06-27|2020-08-27|ノバルティス アーゲー|Dosing regimens for anti-TIM-3 antibodies and uses thereof|

---

CN111315872A|2017-06-30|2020-06-19|纪念斯隆-凯特林癌症中心|Compositions and methods for adoptive cell therapy of cancer|

---

WO2019006465A1|2017-06-30|2019-01-03|Memorial Sloan Kettering Cancer Center|Compositions and methods for adoptive cell therapy for cancer|

---

EP3645122A1|2017-06-30|2020-05-06|Xencor, Inc.|Targeted heterodimeric fc fusion proteins containing il-15/il-15ra and antigen binding domains|

---

RU2020107432A|2017-07-20|2021-08-20|Новартис Аг|DOSING MODES OF ANTI-LAG-3 ANTI-BODIES AND THEIR APPLICATION|

---

JP2019030231A|2017-08-04|2019-02-28|シスメックス株式会社|Antibody and production method thereof, as well as method for improving thermal stability of antibody|

---

SG11202000198QA|2017-08-04|2020-02-27|Genmab As|Binding agents binding to pd-l1 and cd137 and use thereof|

---

AU2018335231A1|2017-09-22|2020-03-19|F. Hoffmann-La Roche Ag|Multivalent mono- or bispecific recombinant antibodies for analytic purpose|

---

CA3078974A1|2017-10-12|2019-04-18|Immunowake Inc.|Vegfr-antibody light chain fusion protein|

---

US20210179709A1|2017-10-31|2021-06-17|Novartis Ag|Anti-car compositions and methods|

---

US10981992B2|2017-11-08|2021-04-20|Xencor, Inc.|Bispecific immunomodulatory antibodies that bind costimulatory and checkpoint receptors|

---

KR20200089286A|2017-11-16|2020-07-24|노파르티스 아게|Combination therapy|

---

EP3737699A1|2018-01-12|2020-11-18|Amgen Inc.|Pac1 antibodies and uses thereof|

---

WO2019145455A1|2018-01-24|2019-08-01|Genmab B.V.|Polypeptide variants and uses thereof|

---

US20210038659A1|2018-01-31|2021-02-11|Novartis Ag|Combination therapy using a chimeric antigen receptor|

---

WO2019165411A1|2018-02-26|2019-08-29|Just Biotherapeutics, Inc.|Determining impact on properties of proteins based on amino acid sequence modifications|

---

MA52152A|2018-03-12|2021-01-20|Genmab As|ANTIBODY|

---

WO2019195623A2|2018-04-04|2019-10-10|Xencor, Inc.|Heterodimeric antibodies that bind fibroblast activation protein|

---

US20210147547A1|2018-04-13|2021-05-20|Novartis Ag|Dosage Regimens For Anti-Pd-L1 Antibodies And Uses Thereof|

---

US20210047405A1|2018-04-27|2021-02-18|Novartis Ag|Car t cell therapies with enhanced efficacy|

---

AU2019264217A1|2018-05-03|2020-12-17|Genmab B.V.|Antibody variant combinations and uses thereof|

---

TW202003580A|2018-05-21|2020-01-16|美商坎伯斯治療有限責任公司|Compositions and methods for enhancing the killing of target cells by NK cells|

---

WO2019226658A1|2018-05-21|2019-11-28|Compass Therapeutics Llc|Multispecific antigen-binding compositions and methods of use|

---

WO2019227003A1|2018-05-25|2019-11-28|Novartis Ag|Combination therapy with chimeric antigen receptortherapies|

---

WO2019232244A2|2018-05-31|2019-12-05|Novartis Ag|Antibody molecules to cd73 and uses thereof|

---

SG11202011830SA|2018-06-13|2020-12-30|Novartis Ag|Bcma chimeric antigen receptors and uses thereof|

---

JP2021528996A|2018-06-19|2021-10-28|アターガ，エルエルシー|Antibody molecule for complement fifth component and its use|

---

JP2021528439A|2018-06-22|2021-10-21|ゲンマブ ホールディング ビー．ブイ．|Anti-CD37 and anti-CD20 antibodies, compositions, and how to use them|

---

AR116109A1|2018-07-10|2021-03-31|Novartis Ag|DERIVATIVES OF 3-PIPERIDINE-2,6-DIONA AND USES OF THE SAME|

---

WO2020014306A1|2018-07-10|2020-01-16|Immunogen, Inc.|Met antibodies and immunoconjugates and uses thereof|

---

GB201811368D0|2018-07-11|2018-08-29|Ucb Biopharma Sprl|Antibody|

---

EP3820890A1|2018-07-13|2021-05-19|Genmab A/S|Trogocytosis-mediated therapy using cd38 antibodies|

---

JP2021524276A|2018-07-13|2021-09-13|ゲンマブ エー／エス|CD38 antibody variant and its use|

---

WO2020021465A1|2018-07-25|2020-01-30|Advanced Accelerator ApplicationsS.R.L.|Method of treatment of neuroendocrine tumors|

---

JP2022512608A|2018-10-04|2022-02-07|ゲンマブ ホールディング ビー．ブイ．|A pharmaceutical composition comprising a bispecific anti-CD37 antibody|

---

EP3877409A1|2018-11-06|2021-09-15|Genmab A/S|Antibody formulation|

---

TW202031288A|2018-11-13|2020-09-01|美商坎伯斯治療有限責任公司|Multispecific binding constructs against checkpoint molecules and uses thereof|

---

KR20210106483A|2018-12-20|2021-08-30|노파르티스 아게|Extended low-dose regimen for MDM2 inhibitors|

---

KR20210106437A|2018-12-20|2021-08-30|노파르티스 아게|Dosage regimens and pharmaceutical combinations comprising 3-piperidine-2,6-dione derivatives|

---

CN113329792A|2019-02-15|2021-08-31|诺华股份有限公司|Substituted 3-piperidine-2, 6-dione derivatives and uses thereof|

---

US10871640B2|2019-02-15|2020-12-22|Perkinelmer Cellular Technologies Germany Gmbh|Methods and systems for automated imaging of three-dimensional objects|

---

AU2020222345A1|2019-02-15|2021-07-08|Novartis Ag|3-isoindolin-2-yl)piperidine-2,6-dione derivatives and uses thereof|

---

WO2020172553A1|2019-02-22|2020-08-27|Novartis Ag|Combination therapies of egfrviii chimeric antigen receptors and pd-1 inhibitors|

---

EP3946593A1|2019-03-29|2022-02-09|Atarga, LLC|Anti fgf23 antibody|

---

CN110164506B|2019-04-19|2021-02-26|浙江工业大学|Multi-domain protein structure assembly method based on inter-domain residue contact|

---

CA3138861A1|2019-05-02|2020-12-10|Board Of Regents, The University Of Texas System|System and method for increasing synthesized protein stability|

---

WO2020225456A1|2019-05-09|2020-11-12|Genmab B.V.|Dosage regimens for a combination of anti-dr5 antibodies for use in treating cancer|

---

AU2020283890A1|2019-05-30|2021-12-16|Amgen Inc.|Engineering the hinge region to drive antibody dimerization|

---

AU2020304671A1|2019-06-28|2022-01-20|Amgen Inc.|Anti-CGRP receptor/anti-PAC1 receptor bispecific antigen binding proteins|

---

TW202128166A|2019-10-21|2021-08-01|瑞士商諾華公司|Combination therapies|

---

TW202128191A|2019-10-21|2021-08-01|瑞士商諾華公司|Tim-3 inhibitors and uses thereof|

---

WO2021089850A1|2019-11-06|2021-05-14|Genmab B.V.|Antibody variant combinations and uses thereof|

---

WO2021092355A1|2019-11-08|2021-05-14|Amgen Inc.|Engineering charge pair mutations for pairing of hetero-igg molecules|

---

WO2021102049A1|2019-11-19|2021-05-27|Amgen Inc.|Novel multispecific antibody format|

---

WO2021108661A2|2019-11-26|2021-06-03|Novartis Ag|Chimeric antigen receptors and uses thereof|

---

TW202135859A|2019-12-20|2021-10-01|瑞士商諾華公司|Combination therapies|

---

UY38995A|2019-12-20|2021-06-30|Amgen Inc|MESOTHELIN TARGETING CD40 AGONIST MULTI-SPECIFIC ANTIBODY CONSTRUCTS FOR THE TREATMENT OF SOLID TUMORS|

---

GB201919061D0|2019-12-20|2020-02-05|Ucb Biopharma Sprl|Multi-specific antibody|

---

GB201919062D0|2019-12-20|2020-02-05|Ucb Biopharma Sprl|Antibody|

---

WO2021144457A1|2020-01-16|2021-07-22|Genmab A/S|Formulations of cd38 antibodies and uses thereof|

---

US20210222244A1|2020-01-17|2021-07-22|Becton, Dickinson And Company|Methods and compositions for single cell secretomics|

---

WO2021144657A1|2020-01-17|2021-07-22|Novartis Ag|Combination comprising a tim-3 inhibitor and a hypomethylating agent for use in treating myelodysplastic syndrome or chronic myelomonocytic leukemia|

---

GB202001447D0|2020-02-03|2020-03-18|Ucb Biopharma Sprl|Antibodies|

---

WO2021155916A1|2020-02-04|2021-08-12|BioNTech SE|Treatment involving antigen vaccination and binding agents binding to pd-l1 and cd137|

---

WO2021173995A2|2020-02-27|2021-09-02|Novartis Ag|Methods of making chimeric antigen receptor-expressing cells|

---

US20210292418A1|2020-03-18|2021-09-23|Genmab A/S|Antibodies|

---

WO2021216372A1|2020-04-23|2021-10-28|Eli Lilly And Company|Subcutaneous absorption and bioavailability of antibodies|

---

WO2021226351A1|2020-05-08|2021-11-11|Rutgers, The State University Of New Jersey|Methods and systems for stabilizing proteins using intelligent automation|

---

WO2021260528A1|2020-06-23|2021-12-30|Novartis Ag|Dosing regimen comprising 3-piperidine-2,6-dione derivatives|

---

WO2022013787A1|2020-07-16|2022-01-20|Novartis Ag|Anti-betacellulin antibodies, fragments thereof, and multi-specific binding molecules|

---

WO2022018294A1|2020-07-23|2022-01-27|Genmab B.V.|A combination of anti-dr5 antibodies and an immunomodulatory imide drug for use in treating multiple myeloma|

---

WO2022026592A2|2020-07-28|2022-02-03|Celltas Bio, Inc.|Antibody molecules to coronavirus and uses thereof|

---

WO2022029573A1|2020-08-03|2022-02-10|Novartis Ag|Heteroaryl substituted 3-piperidine-2,6-dione derivatives and uses thereof|

---

WO2022029011A1|2020-08-06|2022-02-10|BioNTech SE|Binding agents for coronavirus s protein|

---

WO2022040466A1|2020-08-20|2022-02-24|Amgen Inc.|Antigen binding proteins with non-canonical disulfide in fab region|

法律状态:
2009-02-19| DA3| Amendments made section 104|Free format text: THE NATURE OF THE AMENDMENT IS: ADD THE CO-INVENTOR NAME TO READ LUGOVSKOY, ALEXEY ALEXANDROVICH |

---

2012-08-09| FGA| Letters patent sealed or granted (standard patent)|

---

2021-10-21| MK14| Patent ceased section 143(a) (annual fees not paid) or expired|

优先权:

---

申请号 | 申请日 | 专利标题

---

US78362206P| true| 2006-03-17|2006-03-17||

---

US60/783,622||2006-03-17||

---

US81268806P| true| 2006-06-09|2006-06-09||

---

US60/812,688||2006-06-09||

---

US87380206P| true| 2006-12-08|2006-12-08||

---

US87399606P| true| 2006-12-08|2006-12-08||

---

US60/873,996||2006-12-08||

---

US60/873,802||2006-12-08||

---

PCT/US2007/006883|WO2007109254A2|2006-03-17|2007-03-19|Stabilized polypeptide compositions|

---