BISPECIFIC PROTEINS AND METHODS FOR PREPARING SAME

Abstract

The present invention relates to a bispecific protein and a method preparing the same, wherein mutation is introduced into heavy chains and/or light chains to enhance heterodimerization between a heavy chain (CH3 domain or Fc) and a heavy chain (CH3 domain or Fc) and dimerization between a heavy chain (CH1 domain) and a light chain, both targeting the same material, thereby constructing heterodimeric bispecific proteins of high purity. A bispecific protein according to the present invention can find applications in a variety of fields comprising cancer therapy, singling regulation, diagnosis, etc.

Claims

1. A bispecific protein for targeting two different kinds of targets, the bispecific protein comprising a first CH3 domain or a first Fc region comprising the first CH3 domain and a second CH3 domain or a second Fc region comprising the second CH3 domain, wherein the first CH3 domain and the second CH3 domain are mutated such that at least one selected from among amino acid pairs forming respective amino acid-amino acid bonds between the first CH3 domain and the second CH3 domain is modified by at least one of the following mutations: (1) a mutation in which the two amino acids in at least one amino acid pair between the CH3 domains are swapped with each other (swapping mutation); (2) a mutation in which, of at least one amino acid pair between the CH3 domains, one amino acid is substituted with an amino acid having a positive charge while the other is substituted with an amino acid having a negative charge, at least one of the two amino acid residues in the amino acid pair not being hydrophobic (electrostatic interaction-introduced mutation), wherein the amino acid having a negative charge is aspartic acid or glutamic acid and the amino acid having a positive charge is lysine or arginine; and (3) a mutation in which, of at least one amino acid pair between the CH3 domains, one amino acid is substituted with a large hydrophobic amino acid while the other is substituted with a small hydrophobic amino acid (size mutation), wherein the large hydrophobic amino acid is selected from the group consisting of tryptophan and phenylalanine and the small hydrophobic amino acid is selected from the group consisting of alanine, glycine, and valine, wherein the first CH3 domain and the second CH3 domain are each independently derived from an immunoglobulin selected from the group consisting of IgG1, IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM.

2. (canceled)

3. (canceled)

4. The bispecific protein of claim 1, wherein the electrostatic interaction-introduced mutation is a mutation in which, of the two amino acids constituting: at least one amino acid pair selected from the group consisting of serine at position 364 and leucine at position 368, threonine at position 394 and threonine at position 394, glutamic acid at position 357 and lysine at position 370, glutamic acid at position 357 and tyrosine at position 349, threonine at position 366 and tyrosine at position 407, and threonine at position 394 and valine at position 397 in an IgG1 CH3 domain (EU numbering); or at least ones amino acid pair at a position corresponding to the at least one amino acid pair in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM, one is substituted with an amino acid having a positive charge and the other is substituted with an amino acid having a negative charge.

5. (canceled)

6. The bispecific protein of claim 4, wherein the electrostatic interaction-introduced mutation comprises at least one of the following mutations on the basis of IgG1 (EU numbering), or at least one of corresponding mutations in the CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: substitution of serine at position 364 with an amino acid having a positive charge and leucine at position 368 with an amino acid having a negative charge; substitution of threonine at position 394 with an amino acid having a positive charge and threonine at position 394 with an amino acid having a negative charge; substitution of glutamic acid at position 357 with an amino acid having a positive charge and lysine at position 370 with an amino acid having a negative charge; substitution of glutamic acid at position 357 with an amino acid having a positive charge and tyrosine at position 349 with an amino acid having a negative charge; substitution of threonine at position 366 with an amino acid having a positive charge and tyrosine at position 407 with an amino acid having a negative charge; substitution of threonine at position 394 with an amino acid having a positive charge and valine at position 397 with an amino acid having a negative charge; and substitution of tyrosine at position 349 with an amino acid having a positive charge and glutamic acid at position 357 with an amino acid having a negative charge.

7. (canceled)

8. The bispecific protein of claim 1, wherein the swapping mutation is substitution in which exchange is made between two paired amino acid residues in: at least one amino acid pair selected from the group consisting of a pair of serine at position 364 and lysine at position 370, a pair of phenylalanine at position 405 and lysine at position 409, a pair of glutamine at position 347 and lysine at position 360, a pair of glutamic acid at position 357 and tyrosine at position 349, a pair of serine at position 354 and tyrosine at position 349, a pair of glutamic acid at position 357 and lysine at position 370, a pair of lysine at position 360 and tyrosine at position 349, a pair of serine at position 364 and leucine at position 368, a pair of leucine at position 368 and lysine at position 409, a pair of asparagine at position 390 and serine at position 400, a pair of threonine at position 394 and valine at position 397, a pair of leucine at position 398 and lysine at position 392, a pair of tyrosine at position 407 and threonine at position 366, and a pair of threonine at position 411 and lysine at position 370 on the basis of CH3 domain IgG1 (EU numbering); or at least one amino acid pair at a position corresponding to the at least one amino acid pair in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM.

9. The bispecific protein of claim 8, wherein the swapping mutation comprises at least one of the following mutations on the basis of the CH3 domain of IgG1 (EU numbering) or at least one mutation corresponding to the at least one mutation in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or 1 IgM: substitution of serine at position 364 with lysine and lysine at position 370 with serine; substitution of phenylalanine at position 405 with lysine and lysine at position 409 with phenylalanine; substitution of tyrosine at position 407 with threonine and threonine at position 366 with tyrosine; substitution of glutamic acid at position 357 with lysine and lysine at position 370 with glutamic acid; and substitution of glutamic acid at position 357 with tyrosine and tyrosine at position 349 with serine.

10. The bispecific protein of claim 1, wherein the size mutation comprises substitution in which, of two paired amino acid residues in: at least one amino acid pair selected from the group consisting of a pair of lysine at position 409 and tyrosine at position 407, a pair of lysine at position 409 and phenylalanine at position 405, and a combination thereof on the basis of CH3 domain of IgG1 (EU numbering); or at least one amino acid pair at a position corresponding to the at least one amino acid pair in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM, one is substituted with a large hydrophobic amino acid while the other is substituted with a small hydrophobic amino acid wherein the large hydrophobic amino acid is selected from the group consisting of tryptophan and phenylalanine and the small hydrophobic amino acid is selected from the group consisting of alanine, glycine, and valine.

11. (canceled)

12. The bispecific protein of claim 10, wherein the size mutation comprises: at least one mutation selected from among the following mutations on the basis of the CH3 domain of IgG1 (EU numbering); and at least ones mutation corresponding the at least one mutation in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: substitution of lysine at position 409 with tryptophan and tyrosine at position 407 with alanine; and substitution of lysine at position 409 with tryptophan and phenylalanine at position 405 with alanine.

13. (canceled)

14. The bispecific protein of claim 1, comprising: at least one mutation selected from the group consisting of substitution of one of serine at position 364 and leucine at position 368 with an amino acid having a positive charge and the other with an amino acid having a negative charge, substitution of serine at position 364 with lysine and lysine at position 370 with serine, and substitution of phenylalanine at position 405 with lysine and lysine at position 409 with phenylalanine, on the basis of the CH3 domain of IgG1 (EU numbering); or at least ones mutation corresponding to the at least one mutation in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM.

15. The bispecific protein of claim 14, comprising at least one of the following mutations on the basis of the CH3 of IgG1, or a mutation corresponding to the at least one mutation in CH3 domains of IgG2, IgG3 IgG4 IgA1 IgA2 IgD, IgE, or IgM: (a) substitution serine at position 364 with lysine or arginine, and leucine at position 368 with aspartic acid or glutamic acid; (b) substitution of serine at position 364 with lysine, and lysine at position 370 with serine; and (c) substitution of phenylalanine at position 405 with lysine, and lysine at position 409 with phenylalanine.

16. A bispecific protein for targeting two different kinds of targets, the bispecific protein comprising at least one of the following mutations on the basis of CH3 domain of IgG1 (EU numbering), or a mutation corresponding to the at least one mutation in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: substitution of lysine at position 409 with phenylalanine or tryptophan, and phenylalanine at position 405 with lysine, arginine, glutamine, or asparagine; substitution of leucine at position 368 with aspartic acid or glutamic acid, and serine at position 364 with lysine, arginine, or asparagine; and substitution of lysine at position 370 with serine, and serine at position 364 with lysine, arginine, or asparagine.

17. The bispecific protein of claim 16, comprising at least one of the following mutations on the basis of the CH3 domain of IgG1 (EU numbering), or at least ones mutation corresponding to the at least one mutation in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: substitution of phenylalanine at position 405 with arginine, and lysine at position 409 with tryptophan; substitution of serine at position 364 with lysine, and leucine at position 368 with aspartic acid; substitution of serine at position 364 with lysine, and lysine at position 370 with serine; substitution of phenylalanine at position 405 with lysine, and lysine at position 409 with phenylalanine; substitution of phenylalanine at position 405 with arginine, and lysine at position 409 with phenylalanine; and substitution of phenylalanine at position 405 with lysine, and lysine at position 409 with tryptophan.

18. The bispecific protein of claim 17, comprising the following mutations on the basis of the CH3 domain of IgG1 (EU numbering), or a mutation corresponding to the at least one mutation in CH3 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: substitution of serine at position 364 in the first CH3 domain with lysine and leucine at position 368 in the second CH3 domain with aspartic acid, and substitution of lysine at position 409 in the first CH3 domain with phenylalanine and phenylalanine at position 405 in the second CH3 domain with lysine; substitution of serine at position 364 in the first CH3 domain with lysine and leucine at position 368 in the second CH3 domain with aspartic acid, and substitution of lysine at position 409 in the first CH3 domain with phenylalanine and phenylalanine at position 405 in the second CH3 domain with arginine; substitution of serine at position 364 in the first CH3 domain with lysine and lysine at position 370 in the second CH3 domain with serine, and substitution of lysine at position 409 in the first CH3 domain with phenylalanine and phenylalanine at position 405 in the second CH3 domain with lysine; substitution of serine at position 364 in the first CH3 domain with lysine and lysine at position 370 in the second CH3 domain with serine, and substitution of lysine at position 409 in the first CH3 domain with phenylalanine and phenylalanine at position 405 in the second CH3 domain with arginine; substitution of serine at position 364 in the first CH3 domain with lysine and leucine at position 368 in the second CH3 domain with aspartic acid, and substitution of lysine at position 409 in the first CH3 domain with tryptophan and phenylalanine at position 405 in the second CH3 domain with lysine; substitution of serine at position 364 in the first CH3 domain with lysine and leucine at position 368 in the second CH3 domain with aspartic acid, and substitution of lysine at position 409 in the first CH3 domain with tryptophan and phenylalanine at position 405 in the second CH3 domain with arginine; substitution of serine at position 364 in the first CH3 domain with lysine and lysine at position 370 in the second CH3 domain with serine, and substitution of lysine at position 409 in the first CH3 domain with tryptophan and phenylalanine at position 405 in the second CH3 domain with lysine; or substitution of serine at position 364 in the first CH3 domain with lysine and lysine at position 370 in the second CH3 domain with serine, and substitution of lysine at position 409 in the first CH3 domain with tryptophan and phenylalanine at position 405 in the second CH3 domain with arginine.

19. (canceled)

20. (canceled)

21. The bispecific protein of claim 1, wherein the bispecific protein is a bispecific antibody or an antigen-binding fragment thereof comprising a first CH1 domain and a first CL (light chain constant region) domain derived respectively from the heavy chain and light chain of an antibody recognizing a first epitope and a second CH1 domain and a second CL domain derived respectively from the heavy chain and light chain of an antibody recognizing a second epitope, wherein the bispecific antibody or an antigen-binding fragment thereof further comprises at least one of the following mutations on the CH1 domains and the CL domains: a mutation in which, of the two amino acids constituting each pair of one or more first amino acid pairs selected from among amino acid pairs forming respective amino acid-amino acid bonds between the first CH1 domain and the first CL domain, one is substituted with an amino acid having a positive charge and the other is substituted with an amino acid having a negative charge; and a mutation in which, of the two amino acids constituting each pair of one or more second amino acid pairs selected from among amino acid pairs forming respective amino acid-amino acid bonds between the second CH1 domain and the second CL domain, one is substituted with an amino acid having a positive charge and the other is substituted with an amino acid having a negative charge, and wherein the amino acid having a positive charge is lysine or arginine, the amino acid having a negative charge is aspartic acid or glutamic acid.

22. The bispecific protein of claim 21, wherein: the amino acids substituted respectively in the first CH1 domain and the second CH1 domain have opposite charges, the amino acids substituted respectively in the first CL domain and the first CH1 domain have opposite charges, and the amino acids substituted respectively in the second CL domain and the second CH1 domain have opposite charges.

23. The bispecific protein of claim 21, wherein the amino acid to be substituted with an amino acid having a positive or negative charge in the CH1 domain is at least one of the following amino acids on the basis of the CH1 domain of IgG1 (EU numbering), or an amino acid at a position corresponding to the at least one amino acid in CH1 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: leucine at position 145, serine at position 183, lysine at position 147, phenylalanine at position 170, proline at position 171, and valine at position 185, and the amino acid to be substituted with an amino acid having a positive or negative charge in the CL domain is at least one of the following amino acids on the basis of the CL domain of kappa type (EU numbering), or an amino acid at a position corresponding to the at least one amino acid in the CL domain of lambda type: serine at position 131, valine at position 133, leucine at position 135, serine at position 162, and threonine at position 180.

24. The bispecific protein of claim 21, wherein a set of two amino acids forming an amino acid pair between the CH1 domain and the CL domain is at least one of the following amino acid pairs on the basis of the CH1 domain of IgG1 and the CL domain of kappa type (EU numbering), or at least one amino acid pair at a position corresponding to the at least one amino acid pair in CH1 domains and CL domain of lambda type of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: a pair of leucine at position 145 in the CH1 domain and serine at position 131 in the CL domain, a pair of leucine at position 145 in the CH1 domain and valine at position 133 in the CL domain, a pair of serine at position 183 in the CH1 domain and valine at position 133 in the CL domain, a pair of lysine at position 147 in the CH1 domain and threonine at position 180 in the CL domain, a pair of valine at position 185 in the CH1 domain and leucine at position 135 in the CL domain, a pair of phenylalanine at position 170 in the CH1 domain and leucine at position 135 in the CL domain, and a pair of proline at position 171 in the CH1 domain and serine at position 162 in the CL domain.

25. The bispecific protein of claim 16, wherein the bispecific protein is a bispecific antibody or an antigen-binding fragment thereof comprising a first CH1 domain and a first CL (light chain constant region) domain derived respectively from the heavy chain and light chain of an antibody recognizing a first epitope and a second CH1 domain and a second CL domain derived respectively from the heavy chain and light chain of an antibody recognizing a second epitope, wherein the bispecific antibody or an antigen-binding fragment thereof further comprises at least one of the following mutations on the CH1 domains and the CL domains: a mutation in which, of the two amino acids constituting each pair of one or more first amino acid pairs selected from among amino acid pairs forming respective amino acid-amino acid bonds between the first CH1 domain and the first CL domain, one is substituted with an amino acid having a positive charge and the other is substituted with an amino acid having a negative charge; and a mutation in which, of the two amino acids constituting each pair of one or more second amino acid pairs selected from among amino acid pairs forming respective amino acid-amino acid bonds between the second CH1 domain and the second CL domain, one is substituted with an amino acid having a positive charge and the other is substituted with an amino acid having a negative charge, and wherein the amino acid having a positive charge is lysine or arginine, the amino acid having a negative charge is aspartic acid or glutamic acid.

26. The bispecific protein of claim 25, wherein: the amino acids substituted respectively in the first CH1 domain and the second CH1 domain have opposite charges, the amino acids substituted respectively in the first CL domain and the first CH1 domain have opposite charges, and the amino acids substituted respectively in the second CL domain and the second CH1 domain have opposite charges.

27. The bispecific protein of claim 25, wherein the amino acid to be substituted with an amino acid having a positive or negative charge in the CH1 domain is at least one of the following amino acids on the basis of the CH1 domain of IgG1 (EU numbering), or at least one amino acid at a position corresponding to the at least one amino acid in CH1 domains of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: leucine at position 145, serine at position 183, lysine at position 147, phenylalanine at position 170, proline at position 171, and valine at position 185, and the amino acid to be substituted with an amino acid having a positive or negative charge in the CL domain is at least one of the following amino acids on the basis of the CL domain of kappa type (EU numbering), or an amino acid at a position corresponding to the at least one amino acid in the CL domain of lambda type: serine at position 131, valine at position 133, leucine at position 135, serine at position 162, and threonine at position 180.

28. The bispecific protein of claim 25, wherein a set of two amino acids forming an amino acid pair between the CH1 domain and the CL domain is at least one of the following amino acid pairs on the basis of the CH1 domain of IgG1 and the CL domain of kappa type (EU numbering), or at least one amino acid pair at a position corresponding to the at least one amino acid pair in CH1 domains and CL domain of lambda type of IgG2, IgG3, IgG4, IgA1, IgA2, IgE, or IgM: a pair of leucine at position 145 in the CH1 domain and serine at position 131 in the CL domain, a pair of leucine at position 145 in the CH1 domain and valine at position 133 in the CL domain, a pair of serine at position 183 in the CH1 domain and valine at position 133 in the CL domain, a pair of lysine at position 147 in the CH1 domain and threonine at position 180 in the CL domain, a pair of valine at position 185 in the CH1 domain and leucine at position 135 in the CL domain, a pair of phenylalanine at position 170 in the CH1 domain and leucine at position 135 in the CL domain, and a pair of proline at position 171 in the CH1 domain and serine at position 162 in the CL domain.

29-42. (canceled)

43. A method for constructing a bispecific protein for targeting different targets, the method comprising one of the following mutation introducing steps to introduce at least one mutation into at least one selected from amino acid pairs forming amino acid-amino acid bonds between a first CH3 domain and a second CH3 domain: swapping the two amino acids in at least one amino acid pair selected from amino acid pairs forming amino-amino acid bonds between a first CH3 domain and a second CH3 domain with each other; substituting one of the two amino acids in at least one amino acid pair selected from amino acid pairs forming amino-amino acid bonds between the CH3 domains with an amino acid having a positive charge and the other with an amino acid having a negative charge, wherein the amino acid having a negative charge is aspartic acid or glutamic acid and the amino acid having a positive charge is lysine or arginine; and substituting one of the two amino acids in at least one amino acid pair selected from amino acid pairs forming amino-amino acid bonds between the CH3 domains with a large hydrophobic amino acid and the other with a small hydrophobic amino acid, wherein the large hydrophobic amino acid is selected from the group consisting of tryptophan and phenylalanine and the small hydrophobic amino acid is selected from the group consisting of alanine, glycine, and valine.

44. The method of claim 43, further comprising the following CH1 and CL domain mutating steps of: substituting one of the two amino acid residues constituting at least one selected from amino acid pairs forming amino acid-amino acid bonds between a first CH1 domain derived from the heavy chain and a first CL domain of an antibody recognizing a first epitope with an amino acid having a positive charge and the other with an amino acid having a negative charge; and substituting one of the two amino acid residues constituting at least one selected from amino acid pairs forming amino acid-amino acid bonds between a second CH1 domain derived from the heavy chain and a second CL domain of an antibody recognizing a second epitope with an amino acid having a positive charge and the other with an amino acid having a negative charge.

45-48. (canceled)

49. The bispecific protein of claim 21, wherein the antibody recognizing a first epitope is an anti-influenza B antibody comprising the heavy chain variable region of SEQ ID NO: 27 and the light chain variable region of SEQ ID NO: 31; and the antibody recognizing a second epitope is an anti-influenza A antibody comprising the heavy chain variable region of SEQ ID NO: 29 and the light chain variable region of SEQ ID NO: 31.

50. (canceled)

Description

DESCRIPTION OF DRAWINGS

[0214] FIG. 1 is a schematic diagram showing only one perfect heterodimeric bispecific antibody (dotted line circle) among a variety of antibody forms possible for construction of bispecific antibodies (10 in total). A and B represent respective heavy chains different from each other and a and b represent respective light chain different from each other.

[0215] FIG. 2 shows residue positions of antibodies used in electrostatic interaction, swapping, and size methods for inducing heterodimerization in Fc constant regions according to one embodiment, and residue positions related thereto.

[0216] FIG. 3 is an SDS-PAGE profile showing heterodimerization results from electrostatic interaction-induced mutations listed in Table 4.

[0217] FIG. 4 is an SDS-PAGE profile showing heterodimerization results from swapping-associated mutations listed in Table 5.

[0218] FIG. 5 is an SDS-PAGE profile showing heterodimerization results from size-associated mutations listed in Table 6.

[0219] FIG. 6a shows comparison of 12 mutations that allow outstanding heterodimerization among the single mutations lusted in Table 7.

[0220] FIG. 6b shows heterodimerization results after S364K of two key mutations was substituted with other amino acids according to one embodiment.

[0221] FIG. 6c shows heterodimerization results after F405K of two key mutations was substituted with other amino acids according to one embodiment.

[0222] FIG. 6d shows heterodimerization results after additional mutations were introduced to the three selected key-lock mutation pairs S364K-L368D, S364K-K370S, and F405K-K409F according to one embodiment, wherein x-axis accounts for molecular weights (kD).

[0223] FIG. 7 shows heterodimerization efficiency of the single mutations S364K-L368D, S364K-K370S, and F405K-K409F in comparison with conventional techniques (KiH, CPC, and AzS controls). When amino acid residues corresponding to each key are changed to other different amino acids, K is best for S364 and K and R (arginine) show almost the same effect for F405. Positions at which single mutations are made on A and B chains are indicated and heterodimerization quantity is numerically expressed.

[0224] FIG. 8 shows heterodimization efficiencies of a total of four double mutation pairs in comparison with conventional techniques (KiH, CPC, AzS controls), wherein the four double mutations are obtained by selecting two double mutation pairs from combinations of the three mutation pairs S364K-L368D, S364K-K370S, and F405K-K409F and mutating F405 into two types K and R for each pair.

[0225] FIGS. 9a to 9c are SDS-PAGE profiles after L368, K370, and K409 corresponding to lock mutations in double mutation pairs are substituted with other amino acids as indicated in Table 6 in order to identify better effects when amino acid residues corresponding to lock mutations are changed to other mutations.

[0226] FIG. 10 shows positions of electrostatic interaction-associated mutations, size-associated mutations, and swap-associated mutations in Fab of antibodies.

[0227] FIG. 11 is a schematic diagram of a competitive pairing (CPP) assay procedure.

[0228] FIG. 12 shows pairing modes established through the process of FIG. 11 between heavy and light chains and visualized on SDS-PAGE after 4D9 and 2B9 having conventional mutation pairs Cl, DuetMab, and V23 between heavy and light chains were cloned and co-expressed.

[0229] FIG. 13 is an SDS-PAGE profile showing pairing modes when an electrostatic interaction-associated mutation between heavy and light chains is established by substituting K(lysine) for target amino acids on A chain (2B9 heavy chain) and D (aspartic acid) for target amino acids on B chain (4D9 heavy chain).

[0230] FIG. 14 shows pairing modes of the mutation at position 30 resulting from substituting heavy chain L145 and light chain V133 with K and D, respectively, which exhibit relatively high pairing accuracy among the list of mutation pairs as the 4D9 and 2B9 antibodies are compared on SDS-PAGE.

[0231] FIGS. 15 and 16 show pairing modes after mutation at position 29 S131D and/or S131K is introduced to the light chain of the antibody in which heavy chain L145 is substituted with E or D and light chain V133 is substituted with R, as analyzed by SDS-PAGE.

[0232] FIG. 17 shows pairing modes of mutation pairs at position 48 (heavy chain S183 and light chain V133 were substituted with K(R) and D(E), respectively), as analyzed by SDS-PAGE.

[0233] FIG. 18 shows pairing modes analyzed by SDS-PAGE after the mutation pair c29c30c48FΦ29f30f48 shown in Table 19 and variations thereof are introduced.

[0234] FIG. 19 shows pairing modes analyzed by SDS-PAGE after the mutation pairs of Table 20 are introduced.

[0235] FIG. 20 shows pairing modes analyzed by SDS-PAGE after the mutation pairs of Table 21 are introduced.

[0236] FIG. 21 shows paring ratios of chains after combinations of the mutation pairs at position 34 and 51 selected from among the heavy and light chain mutation pairs are introduced according to one embodiment.

[0237] FIG. 22 shows paring ratios of chains after c34Φf51 mutation pair selected from among the heavy and light chain mutation pairs is introduced according to one embodiment.

[0238] FIG. 23 shows paring ratios of chains after c40Φf44 mutation pair selected from among the heavy and light chain mutation pairs is introduced according to one embodiment.

[0239] FIG. 24 is a schematic diagram of a bispecific antibody in which the heavy and light chains are mutated according to one embodiment.

[0240] FIG. 25 is a graph showing thermal stability of A chain and B chain for the heavy chain and a chain and b chain for the light chain in antibodies constructed according to one embodiment.

[0241] FIG. 26 is a cleavage map of pcDNA3.

[0242] FIG. 27 is a graph showing hydrophobic interaction chromatography (HIC) results of the bispecific antibodies Trabev and Adabev constructed according to one embodiment, each having c′29c′30c48Φf′29f′30f′48 mutation pair and AWBB mutation pair introduced thereto.

[0243] FIG. 28 is a graph showing size exclusion chromatography (SEC) analysis results of the bispecific antibody Trabev, constructed according to one embodiment, having c′29c′30c48Φf′29f′30f′48 mutation pair and AWBB mutation pair introduced thereto.

[0244] FIG. 29 is a graph showing size exclusion chromatography (SEC) analysis results of the bispecific antibodies Trabev and Adabev constructed according to one embodiment, each having c′29c′30c48Φf′29f′30f′48 mutation pair and AWBB mutation pair introduced thereto.

[0245] FIG. 30 shows dimerization modes of the bispecific antibodies Trabev and Adabev constructed according to one embodiment, each having c′29c′30c48Φf′29f′30f′48 mutation pair and AWBB mutation pair introduced thereto.

[0246] FIG. 31 is a graph showing affinity for antigens (Her2 and VEGF) of the bispecific antibody Trabev, constructed according to one embodiment, having c′29c′30c48Φf′29f′30f′48 mutation pair and AWBB mutation pair introduced thereto.

[0247] FIG. 32 is a graph showing of for antigens (TNF-alpha and VEGF) of the bispecific antibody Adabev, constructed according to one embodiment, having c′29c′30c48Φf′29f′30f′48 mutation pair and AWBB mutation pair introduced thereto.

[0248] FIG. 33a shows sequence alignment between the human IgG1 heavy chain constant region and the human IgA1 heavy chain constant region (human IgG1 heavy chain constant region: SEQ ID NO: 33; human IgA1 heavy chain constant region: SEQ ID NO: 34).

[0249] FIG. 33b shows sequence alignment between the kappa constant region and lambda constant region of human immunoglobulin light chain (kappa constant region of human immunoglobulin light chain: SEQ ID NO: 35; and lambda constant region of human immunoglobulin light chain: SEQ ID NO: 36).

[0250] FIGS. 33c and 33d show sequence alignment of heavy chain constant regions among human, mouse, and rat IgG subtypes (CH1 domain sequences in FIG. 33c and CH3 domain sequences in FIG. 33d) (Human IgG1: SEQ ID NO: 1, Human IgG2: SEQ ID NO: 2, Human IgG3: SEQ ID NO: 3, Human IgG4: SEQ ID NO: 4, Mouse IgG1: SEQ ID NO: 37, Mouse IgG2a.sup.b: SEQ ID NO: 38, Mouse IgG2a.sup.a: SEQ ID NO: 39, Mouse IgG2b: SEQ ID NO: 40, Mouse IgG3: SEQ ID NO: 41, Rat IgG1: SEQ ID NO: 42, Rat IgG2a: SEQ ID NO: 43, Rat IgG2b: SEQ ID NO: 44, Rat IgG2c: SEQ ID NO: 45 in FIG. 33c (CH1 domain); and Human IgG1: SEQ ID NO: 15, Human IgG2: SEQ ID NO: 16, Human IgG3: SEQ ID NO: 17, Human IgG4: SEQ ID NO: 18, Mouse IgG1: SEQ ID NO: 46, Mouse IgG2a.sup.b: SEQ ID NO: 47, Mouse IgG2a.sup.a: SEQ ID NO: 48, Mouse IgG2b: SEQ ID NO: 49, Mouse IgG3: SEQ ID NO: 50, Rat IgG1: SEQ ID NO: 51, Rat IgG2a: SEQ ID NO: 52, Rat IgG2b: SEQ ID NO: 53, and Rat IgG2c: SEQ ID NO: 54 in FIG. 33d (CH3 domain)).

[0251] FIG. 34 shows paring levels between light and heavy chains in antibodies, constructed according to one embodiment, having c34Φf51 mutation pair introduced thereto.

[0252] FIG. 35 is a graph showing a HIC result of the bispecific antibody Trabev, constructed according to one embodiment, having c34Φf51 mutation pair and AWBB mutation pair.

[0253] FIG. 36 shows graphs of HIC results of the bispecific antibody Adabev, constructed according to one embodiment, having c34Φf51 mutation pair and AWBB mutation pair.

[0254] FIG. 37 shows dimerization modes of the bispecific antibodies Trabev and Adabev, constructed according to one embodiment, each having c34Φf51 mutation pair and AWBB mutation pair introduced thereto.

[0255] FIG. 38 shows pairing levels between light and heavy chains in an antibody, constructed according to one embodiment, having c40Φf44 mutation pair introduced thereto.

[0256] FIG. 39 shows graphs of HIC results of the bispecific antibody Trabev, constructed according to one embodiment, having c40Φf44 mutation pair and AWBB mutation pair.

[0257] FIG. 40 shows graphs of HIC results of the bispecific antibody Adabev, constructed according to one embodiment, having c40Φf44 mutation pair and AWBB mutation pair.

[0258] FIG. 41 shows dimerization modes of the bispecific antibodies Trabev and Adabev, constructed according to one embodiment, each having c40Φf44 mutation pair and AWBB mutation pair introduced thereto.

[0259] FIG. 42 is an SDS-PAGE profile showing homodimerization levels upon single transfection of Enb-Fc and Fas-Fc, separately, according to a Comparative Example.

MODE FOR INVENTION

[0260] Hereinafter, the present invention will be described in further detail with reference to examples. It is to be understood, however, that these examples are for illustrative purposes only and are not to be construed to limit the scope of the present invention.

[0261] For the practice of the present invention, all samples were prepared after the following procedure.

[0262] Protein Expression

[0263] 1. A target gene was cloned to an expression vector (pcDNA3 (Invitrogen).

[0264] 2. HEK293E cells (ATCC) were cultured in high-glucose DMEM (Dulbecco's modified Eagle's medium) supplemented with 5% PBS (fetal bovine serum) in a humidified, CO.sub.2 incubator.

[0265] 3. A prepared plasmid DNA was introduced by transient transfection into HEK393E cells grown to full confluency. Before transfection, the cells were washed with PBS (phosphate-buffered saline) and, followed by exchanging the culture medium with serum-free high-glucose DMEM.

[0266] 4. After incubation for one week, a conditioned medium was used to harvest the proteins which were then filtered. Fc-fusion proteins and antibodies were isolated by protein A chromatography.

[0267] 5. Isolated proteins were quantitatively determined as analyzed at 280 nm.

Example 1: Selection of Mutation for Heterodimerization of Two Fe Regions

[0268] Mutation positions for heterodimerization of two Fc regions (CH3 domain; SEQ ID NO: 15) (on the basis of IgG1) are depicted in FIG. 2. Section was made of 39 positions for electrostatic interaction-associated mutation (electrostatic interaction-introduced mutation) (represented by “Charge J”), 14 positions for swapping-associated mutation (swapping mutation) (represented by “Swap 0”), and 40 positions for size-associated mutation (size mutation) (represented by “Size B”). Finally selected positions in CH3 domains and mutated amino acids thereat are summarized in Table 3. Each of the mutation pairs was applied two Fc regions (respectively represented by A chain and B chain) derived from different antibodies, followed by cloning for co-expression. For electrostatic interaction-associated mutation, mutations were carried out to substitute the corresponding amino acids on A chain with K (lysine) representative of amino acids having a positive charge and on B chain with D (aspartic acid) representative of amino acids having a negative charge. Swapping-associated mutation was carried out to exchange the corresponding amino acids on A chain and B chain with each other. For size-associated mutation, the corresponding amino acids were substituted with W (tryptophan) on A chain and with the small size amino acid A (alanine) on B chain

TABLE-US-00003 TABLE 3 Amino acid Charge (J) Swap (O) Size (B) pair No. Chain A Chain B Chain A Chain B Chain A Chain B 1 Q347K K360D Q347K K360Q Q347W K360A 2 Y349K S354D Y349W S354A 3 Y349K E357 Y349W E357A 4 Y349K K360D Y349W K360D 5 L351K L351D L351W L351A 6 P352K P352D P352W P352A 7 S354K Y349D S354Y Y349S S354W Y349A 8 D356K K439D D356W K439A 9 E357K Y349D E357Y Y349E E357W Y349A 10 E357K K370D E357K K370E E357W K370A 11 K360K Q347D K360W Q347A 12 K360K Y349D K360Y Y349K K360W Y349A 13 S364K L368D S364L L368S S364W L368A 14 S364K K370D S364K K370S S364W K370A 15 T366K T366D T366W T366A 16 T366K Y407D T366W Y407A 17 L368K S364D L368W S364A 18 L368K K409D L368K K409L L368W K409A 19 K370 E357D K370W E357A 20 K370 S364D K370W S364A 21 K370 T411D K370W T411A 22 N390K S400D N390S S400N N390W S400A 23 K392 L398D K392W L398A 24 T394K T394D T394W T394A 25 T394K V397D T394V V397T T394W V397A 26 P395K P395D P395W P395A 27 P395K V397D P395W V397A 28 V397K T394D V397W T394A 29 V397K P395D V397W P395A 30 L398K K392D L398K K392L L398W K392A 31 S400K N390D S400W N390A 32 F405K K409D F405K K409F F405W K409A 33 Y407K T366D Y407T T366Y Y407W T366A 34 Y407K Y407D Y407W Y407A 35 Y407K K409D Y407W K409A 36 K409 L368D K409W L368A 37 K409 F405D K409W F405A 38 K409 Y407D K409W Y407A 39 T411K K370D T411K K370T T411W K370A 40 K439W D356A

[0269] In order to easily identify homodimers or heterodimers of Fc-fusion proteins through SDS-PAGE, fusion was made of the ectodomain (coding sequence of region 1-771 of SEQ ID NO: 24) of TNF-alpha receptor (TNFRSF1B: NP_001057.1 (coding gene: CDS of NM_001066.2; SEQ ID NO: 24)) to Fc (coding gene: SEQ ID NO: 26) on one chain (chain A: Enbrel) and the ectodomain (coding sequence of region 1-519 of SEQ ID NO: 25) of Fas receptor (NP_000034.1 (coding gene: CDS of NM_000043.5; SEQ ID NO: 25)) to Fc (coding gene: SEQ ID NO: 26) on the other chain (chain B: Fas), using pcDNA3 vector (see FIG. 26; Invitrogen) as a backbone. The monomer of chain A has a size of 53 kD while the monomer of chain B has a size of about 44 kD. Because chains A and B, each having an Fc-ectodomain fusion, were different in size from each other, two homodimers (AA and BB) and one heterodimer (AB) could be easily discriminated on SDS-PAGE.

[0270] Percentages (%) of dimerization between chains A and between chains B (homodimerization) are expressed as S.sub.AA and S.sub.BB, respectively whereas S accounts for percentages (%) of dimerization between chains A and B (heterodimerization).

[0271] When modes of dimerization were observed by SDS-PAGE, ratios (%) of homodimerization (AA and BB) and heterodimerization (AB) between chains A and B having the mutations of Table 3 introduced thereinto were compared with those between chains A and B that had not been mutated (represented by WT). Results are given Table 4 and FIG. 3 for the electrostatic interaction-induced mutation, Table 5 and FIG. 4 for the size mutation, and Table 6 and FIG. 5 for the swapping-associated mutation.

TABLE-US-00004 TABLE 4 A WT Y349K Y349K S354K E356K E357K E357K S364K T366K T394K T394K T411K B WT E357D K360D Y349D K439D Y349D K370D L368D Y407D T394D V397D K370D S.sub.AA 28 11 5 27 10 15 12 0 0 11 10 15 S.sub.AB 46 70 60 63 64 75 78 100 75 79 73 67 S.sub.BB 26 19 35 10 26 10 10 0 25 10 17 18

TABLE-US-00005 TABLE 5 A E357Y E357K S364L S364K F405K Y407T T411K B Y349E K370E L368S K370S K409F T366Y K370T S.sub.AA 8 4 8 0 0 10 13 S.sub.AB 71 80 61 90 70 83 60 S.sub.BB 21 16 31 10 30 7 27

TABLE-US-00006 TABLE 6 A WT K409W K409W B WT F405A Y407A S.sub.AA 22  3  2 S.sub.AB 56 60 85 S.sub.BB 23 37 13

[0272] In Tables 4 to 6, mutations accounting for a heterodimerization rate (%) of 70% or higher are expressed in bold. As is understood from data of Tables 4 to 6, the tested mutation pairs exhibited a heterodimerization rate of 60% or higher.

[0273] Through the results, selection was made of 12 mutation pairs that gave higher rates to heterodimerization than homodimerization, with the heterodimerization rate being 70% or higher (electrostatic interaction-introduced mutation: Y349K-E357D, E357K-Y349D, E357K-K370D, S364K-L368D, T366K-Y407D, T394K-T394D, T394K-V397D; and swapping mutation: E357Y-Y349E, E357K-K370E, S364K-K370S, F405K-K409F, Y407T-T366Y) (expressed in bold in Tables 4 and 5).

[0274] Each of the 12 amino acid residue mutations contained in the selected 12 amino acid pairs was introduced into the Fas-Fc fusion protein to express mutant Fas-Fc fusion proteins that contained single mutations at the amino acid positions, respectively. In the same condition, homodimer and monomer survival rates (S.sub.sm; 1=100%) were compared, and the results are depicted in Table 7 and FIG. 6a.

TABLE-US-00007 TABLE 7 SM S.sub.SM E357Y 0.03 Y349D 0.03 K370D 0.06 L368D 0.06 S364K 0.77 K370S 0.05 S400N 0.05 T394D 0.11 F405K 0.67 K409F 0.05 T366A 0.09 T366Y 0.46

[0275] In Table 7 and FIG. 6a, two mutations S364K and F405K, which were low in homodimerization rate but high in monomer survival rate (50% or higher), were selected.

[0276] It was postulated that the two selected mutations acted as “key” mutations while the CH3 domain sites interacting therewith on the other chain were “lock” mutations. As combinations therebetween, three key-lock pairs S364K-L368D, S364K-K370S, and F405K-K409F were selected. These mutation pairs were introduced into chain A (key mutation) and chain B (lock mutation) to make three different single mutation pairs (see TABLE 8).

TABLE-US-00008 TABLE 8 Key Single Lock Single Mutation type Sample Mutation Mutation Introduction of J13 S364K L368D electrostatic interaction Swapping O14 S364K K370S Swapping O32 F405K K409F

[0277] The amino acids at respective key mutation positions were substituted with different amino acid residues and tested for the single mutation-induced heterodimerization effect, as described above, so as to identify mutation types effective for heterodimerization at the positions. For S364, substitution with K (lysine) was observed to bring about the highest heterodimerization effect. An outstanding effect was detected upon substitution with N (asparagine) and R (arginine) (see FIG. 6b). F405 exhibited excellent similar heterodimerization effects upon substitution with K and R and outstanding heterodimerization effects upon substitution with N and Q (glutamine) (see FIG. 6c).

[0278] The additional mutations S364N, S364R, F405R, F405N, and F405Q, which were identified in FIGS. 6b and 6c, were applied to the three selected key-lock mutation pairs S364K-L368D, S364K-K370S, and F405K-K409F, to introduce the lock mutations to chain A (TNFR2-Fc) and the key mutations to chain B (Fas-Fc), followed by analyzing heterodimerization effects on SDS-PAGE.

[0279] The results thus obtained are depicted in Table 9 and FIG. 6d:

TABLE-US-00009 TABLE 9 A(TNFR2) B(Fas) Key Single Lock Single Sample Mutation Mutation S.sub.AA S.sub.AB S.sub.BB J13(K:D) S364K L368D  8 92  0 J13(N:D) S364N L368D  8 82  9 J13(R:D) S364R L368D 20 79  1 O14(K:S) S364K K370S 25 65  9 O14(N:S) S364N K370S 13 84  3 O14(R:S) S364R K370S 23 76  1 O32(K:F) F405K K409F  0 75 25 O32(N:F) F405N K409F 21 62 17 O32(Q:F) F405Q K409F  5 68 27 O32(R:F) F405R K409F  5 89  6 (S.sub.AA: AA homodimerization rate (%); S.sub.AB: AB heterodimerization rate (%); S.sub.BB: BB homodimerization rate (%))

Example 2: Test for Heterodimerization of Fe Region by Single Mutation

[0280] The three key-lock mutation pairs S364K-L368D, S364K-K370S, and F405K-K409F, which were selected in Example 1, and the mutation pair F405R-K409F, which resulted from substituting F405 with R instead of K, were tested for heterodimerization on SDS-PAGE. The heterodimerization effects were compared with those obtained with the conventional heterodimer Fc mutation pairs KiH, CPC, and AzS used as controls.

[0281] SDS-PAGE data in this Example and all the following Examples were obtained by quantitating the band intensities with the aid of GelQuant.NET Software.

[0282] The results are depicted in Table 10 and FIG. 7:

TABLE-US-00010 TABLE 10 Chain A Sample (eTNFR2) Chain B (eFas) S.sub.AB S.sub.A S.sub.B S.sub.M T.sub.m J13 S364K L368D 0.93 0.86 0.10 0.48 66.8 O14 S364K K370S 0.87 0.86 0.07 0.47 67.6 O32 F405K K409F 0.72 0.91 0.08 0.49 65.2 O32′ F405R K409F 0.72 0.91 0.08 0.49 65.2 KiH T366S/L368A/ T366W 0.90 0.68 0.64 0.66 67.4 Y407V CPC K392D/K409D E356K/D399K 0.74 0.85 0.30 0.58 66.4 AzS T350V/T366L/ T350V/L351Y/ 0.84 0.84 0.64 0.74 69.2 K392L/T394W F405A/Y407V (S.sub.AB: AB heterodimerization rate (%); S.sub.A; AA homodimerization rate (%); S.sub.B: BB homodimerization rate (%); S.sub.M: monomer survival rate (%))

[0283] Tm was measured as follows:

[0284] Reagent: Invitrogen 4461146 “Protein Thermal Shift” Dye Kit

[0285] Instrument: Chromo4-PTC 200 (MJ Research)

[0286] Reaction mixture: 20 in total

TABLE-US-00011 Protein 10 μl DW  3 μl Protein Thermal Shift ™ Buffer  5 μl 1/100 diluted Protein Thermal Shift ™ Dye  2 μl

[0287] Protocol:

[0288] 1. Incubate at 50.0□ for 30 sec;

[0289] 2. Melting Curve from 50.0□ to 90.0□, read every 0.2□, hold for 2 sec;

[0290] 3. Incubate at 90.0□ for 2 min

[0291] 4. Incubate at 10.0□ forever

[0292] 5. End

[0293] As shown in Table 10 and FIG. 7, when chain A and chain B were each separately expressed, chain A retaining the key did not form a homodimer, but is expressed as a monomer. A sample resulting from co-expression was observed to contain no homodimers, but mostly heterodimers (see FIG. 7).

Example 3: Heterodimerization of Fe Region by Double Mutation

[0294] Less plausibility of homodimerization might result from existence of the key mutations on both of chains A and B than on either of the chains Two double mutation pairs were made from combinations of the three mutation pairs selected in Example 1. For each double mutation pair, F405 was mutated into two types K and R. Thus, a total of four double mutation pairs were obtained. These found double mutations pairs were analyzed for heterodimerization on SDS-PAGE and the dimerization effects were compared with those of the controls KiH, CPC, and AzS.

[0295] The results are given in Table 11 and FIG. 8:

TABLE-US-00012 TABLE 11 Chain A DMP (eTNFR2) Chain B (eFas) S.sub.AB S.sub.A S.sub.B S.sub.M T.sub.m J13/O32 S364K/K409F L368D/F405K 0.96 1.00 0.74 0.87 58.2 J13/O32′ S364K/K409F L368D/F405R 0.95 1.00 0.73 0.87 61.5 O14PO32 S364K/K409F K370S/F405K 0.95 1.00 0.72 0.86 64.1 O14/O32′ S364K/K409F K370S/F405R 0.93 1.00 0.71 0.86 64.4 KiH T366S/L368A/Y407V T366W 0.91 0.64 0.61 0.63 67.4 CPC K392D/K409D E356K/D399K 0.73 0.81 0.30 0.58 66.4 AzS T350V/T366L/ T350V/L351Y/ 0.84 0.82 0.62 0.72 69.2 K392L/T394W F405A/Y407V

[0296] As shown in Table 11 and FIG. 8, the key mutations present in both chains allowed the homodimerization of the chains for none of the four mutation pairs upon separate expression. All of the proteins in a sample resulting from co-expression were observed to be heterodimers (See FIG. 8). In addition, higher thermal stability was measured in a double-mutation pair having F405R introduced thereto than F405K introduced thereto (Table 11).

[0297] In order to examine whether a better effect was obtained when amino acids corresponding to lock mutation were substituted with other residues, L368, K370, and K409, which were the lock in the double mutation pairs, were substituted with other amino acid residues. Mutation combinations obtained by variously mutating L368, K370, and K409 are summarized in Table 12. The mutation combinations were tested for heterodimerization on SDS-PAGE (NR: 8% SDS-PAGE gel; Sample: 24 ul Loading), and the results are depicted in FIGS. 9a to 9c.

TABLE-US-00013 TABLE 12 Lock variants Chain Chain Chain Chain Chain X A Chain B Y A B Z A B UA S364K/ L368A/ XA S364K/ K370A/ ZA S364K/ K370S/ K409F F405R K409F F405R K409A F405R UC S364K/ L368C/ XC S364K/ K370C/ ZC S364K/ K370S/ K409F F405R K409F F405R K409C F405R UD S364K/ L368D/ XD S364K/ K370D/ ZD S364K/ K370S/ K409F F405R K409F F405R K409D F405R UE S364K/ L368E/ XE S364K/ K370E/ ZE S364K/ K370S/ K409F F405R K409F F405R K409E F405R UF S364K/ L368F/ XF S364K/ K370F/ ZF S364K/ K370S/ K409F F405R K409F F405R K409F F405R UG S364K/ L368G/ XG S364K/ K370G/ ZG S364K/ K370S/ K409F F405R K409F F405R K409G F405R UH S364K/ L368H/ XH S364K/ K370H/ ZH S364K/ K370S/ K409F F405R K409F F405R K409H F405R UI S364K/ L368I/ XI S364K/ K370I/ ZI S364K/ K370S/ K409F F405R K409F F405R K409I F405R UK S364K/ L368K/ XK S364K/ K370/ ZK S364K/ K370S/ K409F F405R K409F F405R K409K F405R UL S364K/ L368/ XL S364K/ K370L/ ZL S364K/ K370S/ K409F F405R K409F F405R K409L F405R UM S364K/ L368M/ XM S364K/ K370M/ ZM S364K/ K370S/ K409F F405R K409F F405R K409M F405R UN S364K/ L368N/ XN S364K/ K370N/ ZN S364K/ K370S/ K409F F405R K409F F405R K409N F405R UQ S364K/ L368Q/ XQ S364K/ K370Q/ ZQ S364K/ K370S/ K409F F405R K409F F405R K409Q F405R UR S364K/ L368R/ XR S364K/ K370R/ ZR S364K/ K370S/ K409F F405R K409F F405R K409R F405R US S364K/ L368S/ XS S364K/ K370S/ ZS S364K/ K370S/ K409F F405R K409F F405R K409S F405R UT S364K/ L368T/ XT S364K/ K370T/ ZT S364K/ K370S/ K409F F405R K409F F405R K409T F405R UV S364K/ L368V/ XV S364K/ K370V/ ZV S364K/ K370S/ K409F F405R K409F F405R K409V F405R UW S364K/ L368W/ XW S364K/ K370W/ ZW S364K/ K370S/ K409F F405R K409F F405R K409W F405R UY S364K/ L368Y/ XY S364K/ K370Y/ ZY S364K/ K370S/ K409F F405R K409F F405R K409Y F405R

[0298] The combinations were measured for thermal stability (see Example 2) and the results are given in Table 13.

TABLE-US-00014 TABLE 13 Chain A Chain B Tm UC S364K/K409F L368C/F405R 63.8 UD S364K/K409F L368D/F405R 60.8 UL S364K/K409F L368/F405R 63.0 UW S364K/K409F L368W/F405R 61.2 UY S364K/K409F L368Y/F405R 61.2 ZF S364K/K409F K370S/F405R 65.6 ZH S364K/K409H K370S/F405R 64.4 ZI S364K/K409I K370S/F405R 62.4 ZN S364K/K409N K370S/F405R 61.0 ZR S364K/K409R K370S/F405R 62.8 ZT S364K/K409T K370S/F405R 65.4 ZV S364K/K409V K370S/F405R 66.0 ZW S364K/K409W K370S/F405R 67.0 ZY S364K/K409Y K370S/F405R 63.8

[0299] Analysis showed higher thermal stability of 1(409W (tryptophan) than K409F.

[0300] Based on the data obtained above, mutations were made of S364K and 1(409W into chain A and K370S and F405R into chain B to produce a double mutation pair, termed AWBB mutation pair, for use in the following Fc heterodimerization test of CH3 domain.

Example 4: Selection of Mutation for Heavy Chain and Light Chain in Antibody Fab

[0301] To select mutations in heavy and light chains of antibodies, electrostatic interaction-associated mutations, size-associated mutations, and swapping-associated mutations were carried out. Positions of interaction at heavy and light chains are depicted in FIG. 10.

[0302] In order to easily identify mutations in heavy and light chains, antibodies having the same light chain were cloned. 4D9 antibody (anti-Influenza A antibody) and 2B9 antibody (anti-Influenza B antibody), which are both anti-influenza antibodies, have the same light chain (common light chain CLC). Because they have the same light chain, SDS-PAGE analysis allows interaction between heavy and light chains to be easily understood. Although identical in the amino acid sequence of light chain, the two antibodies 4D9 and 2B9 are different from each other with respect to the sequence and size of the heavy chain (the heavy chain of 2B9 has more amino acid residues by six than that of 4D9 and a size (50130.62 Daltons) greater than that of 4D9 (49499.98 Daltons): they can be clearly discriminated on SDS-PAGE). Thus, size analysis on SDS-PAGE makes it possible to understand which of the two chains interacts with the light chain.

[0303] Amino acid sequences and coding nucleic acid sequences thereof in the heavy chain variable regions and light chain variable regions of the two antibodies 4D9 and 2B9 are listed in Table 14, below.

TABLE-US-00015 TABLE 14 Amino Acid Sequence Nucleic Acid Sequence 2B9 Heay EVQLVESGGGLVQPGKSLRLSC GAAGTGCAGCTGGTGGAGTCTGGGGGAG Chain AATGFTFDDYAMHWVRQAPG GCTTGGTACAGCCTGGCAAGTCCCTGAG Variable KGLEWVSSLNWKGNSVDYAD ACTCTCCTGTGCAGCCACTGGATTCACAT Region SVRGRFTMSRDNAKKLVYLQM TTGACGATTACGCCATGCACTGGGTCCGC NGLRGDDTAVYFCAKDNKAD CAAGCTCCAGGGAAGGGCCTGGAGTGGG ASMDYYYHHGMDVWGQGTT TCTCAAGTCTTAATTGGAAGGGAAATAGT VTVSS (SEQ ID NO: 27) GTAGACTACGCGGACTCTGTGAGGGGCC GATTCACCATGTCCAGAGACAACGCCAA GAAACTAGTGTATCTGCAAATGAACGGT CTGAGAGGTGACGACACGGCCGTCTATTT TTGTGCAAAAGATAATAAAGCGGATGCA TCTATGGACTACTACTACCACCACGGTAT GGACGTCTGGGGCCAAGGGACCACGGTC ACCGTCTCCTCG (SEQ ID NO: 28) 4D9 Heay QVTLRESGPGLVKPSETLSLTCT CAGGTCACCTTGAGGGAGTCGGGCCCAG Chain ISGASINTDYWSWIRQPPGKGLE GACTGGTGAAGCCTTCGGAGACCCTGTCC Variable WIGYIYYRGRTNYNPSLRSRVTI CTCACCTGCACTATCTCCGGTGCCTCCAT Region SVDTSKNQFSL CAATACTGACTACTGGAGCTGGATCCGG QMTSMTAADTAVYYCARDVT CAGCCCCCAGGGAAGGGACTGGAGTGGA GISRENAFDIWGQGTLVTVSS TTGGCTATATCTATTACAGAGGGCGCACC (SEQ ID NO: 29) AACTACAACCCCTCCCTCAGGAGCCGAG TCACCATATCAGTAGACACGTCCAAGAA TCAATTCTCCCTG CAGATGACGTCTATGACCGCTGCTGACAC GGCCGTATATTACTGTGCGAGAGATGTG ACTGGCATCAGTCGAGAAAATGCTTTTGA TATCTGGGGCCAAGGCACCCTGGTCACC GTCTCCTCG (SEQ ID NO: 30) Light Chain AIRMTQSPSSLSASVGDRVTTTC GCCATCCGGATGACCCAGTCTCCATCCTC (CLC) RASQSISGYLNWYQQKPGKAP CCTGTCTGCATCTGTAGGAGACAGAGTCA Varuable KLLIYAASSLQSGVPSRFSGSGS CCATCACTTGCCGGGCAAGTCAGAGCATT Region GTDFILTISSLQPEDFATYYCQQ AGCGGCTATTTAAATTGGTATCAGCAGA SYSIPTTFGQGTRLEIK (SEQ ID AACCAGGGAAAGCCCCTAAGCTCTTGAT NO: 31) CTATGCTGCATCCAGTTTGCAGAGTGGGG TCCCATCAAGGTTCAGTGGCAGTGGATCT GGGACAGATTTCACTCTCACCATCAGCAG TCTGCAACCTGAAGATTTTGCAACTTACT ACTGTCAACAGAGCTACAGTATCCCCACC ACCTTCGGCCAAGGGACACGACTGGAGA TTAAA (SEQ ID NO: 32)

[0304] The two antibodies 4D9 and 2B9 employ the constant region of IgG1 as a heavy chain constant region and the kappa constant region as a light chain constant region.

[0305] Examination was made to see whether a light chain having a mutation introduced thereinto pairs with only a heavy chain having a mutation interacting with the mutation of the light chain. In this regard, a light chain, a heavy chain pairing with the light chain, and a heavy chain forming a mispair with the light chain were co-expressed to afford antibodies. SDS-PAGE analysis of the antibodies in a reducing condition can identify the extent to which the pairings are accurately formed. For convenience, 4D9 heavy chain is referred to as A chain and a light chain pairing therewith as a chain, and 2B9 heavy chain is referred to as B chain and a light chain pairing therewith as b chain (see the following illustrations).

[0306] First, 4D9 and 2B9 having conventional mutation pairs between heavy and light chains, already known to be effective, were cloned and co-expressed. Pairing modes established through competition between the resulting mutant chains, that is, two heavy chains and one light chain were examined on SDS-PAGE. (Competitive Pairing (CPP) Assay). The process of performing a competitive pairing (CPP) assay is schematically illustrated in FIG. 11, and the result thus obtained is depicted in FIG. 12. As shown in FIG. 12, all the already known mutation pairs undergo normal pairing and abnormal mispairing, concurrently.

[0307] For electrostatic interaction-associated mutations between heavy and light chains, antibodies in which various mutation pairs were introduced with the substitution of corresponding amino acids in B chain (2B9 heavy chain) with K (lysine) and in A chain (4D9 heavy chain) with D (aspartic acid) were subjected to CPP assay on SDS-PAGE (see FIG. 11).

[0308] As a result, seven candidate mutation pairs that were identified to undergo relatively accurate pairing were screened in the electrostatic interaction-associated mutation group. CPP assay results on SDS-PAGE of the antibodies into which the seven screened mutation pairs were introduced are given in Table 15 and depicted in FIG. 13.

TABLE-US-00016 TABLE 15 SOP No (Symmetric Orthogonal Pairs) A (4D9) B (2B9) a b a.sup.CPP b.sup.CPP S.sup.CPP c cΦf L145 L145 S131 S131 1 c29Φf29 L145D L145K S131K S131D 58 50 54 2 c30Φf30 L145D L145K V133K V133D 50 79 65 3 c34Φf34 K147D K147K T180K T180D 49 87 68 4 c40Φf40 F170D F170K L135K L135D 59 61 60 5 c44Φf44 P171D P171K S162K S162D 59 51 55 6 c48Φf48 S183D S183K V133K V133D 62 80 72 7 c51Φf51 V185D V185K L135K L135D 67 73 70 (CPP Score (S.sup.CPP) = ½(a.sup.CPP + b.sup.CPP) a.sup.CPP = 100(C.sub.aA)/(C.sub.aA + C.sub.aB) = 100(C.sub.A)/(C.sub.A + C.sub.B) b.sup.CPP = 100(C.sub.bB)/(C.sub.bB + C.sub.bA) = 100(C.sub.B)/(C.sub.A + C.sub.B))

[0309] Comparison on SDS-PAGE found some mutation pairs that underwent relatively accurate pairing in the list of mutation pairs using the 4D9 (A chain) and 2B9 (B chain) antibodies (expressed in bold in Table 15). For further study, mutation at position 30 (expressed as c30Φf30) was modified as in Table 16.

TABLE-US-00017 TABLE 16 Mutation c30Φf30 Heavy Chain Light Chain Paring Accuracy Aa (4D9) L145D V133R 75% Bb (2B9) L145R V133D 60%

[0310] The mutation at position 30 resulted from substitution heavy chain L145 and light chain V133 with K and D, respectively. As is understood from the data of Table 16, the effect (pairing accuracy: Aa pairing or Bb pairing ratio) was observed to be good (Aa pairing accuracy 75%, Bb pairing accuracy: 60%) for the variations in which heavy chain L145 and light chain V133 were substituted with R and D, respectively (see FIG. 14).

[0311] Addition of mutation S131D on the light chain (termed mutation at position 29) to the mutation at position 30 or variations thereof (heavy chain L145 and light chain V133 were substituted with K or R, and D or E, respectively) improved the accuracy of pairing (see Table 17 and FIG. 15 (c29c.sup.R30Φf.sup.R30 result; Aa pairing accuracy: 80% and Bb pairing accuracy: 70%) and FIG. 16 (c.sup.R29c.sup.RE30Φf.sup.R29f.sup.RE30 result; Aa pairing accuracy: 90% and Bb pairing accuracy: 80%)).

TABLE-US-00018 TABLE 17 Heavy Chain Light Chain Paring Accuracy c29c.sup.R30Φf.sup.R30 Aa (4D9) L145D V133R 80% Bb (2B9) L145R S131D/V133D 70% c.sup.R29c.sup.RE30Φf.sup.R29f.sup.RE30 Aa (4D9) L145E S131K/V133R 90% Bb (2B9) L145R S131D/V133E 80%

[0312] As can be seen in FIGS. 14 and 15, outstanding pairing accuracy was detected in the cases where heavy chain L145 and light chain V133 were substituted with R and D, respectively and where substitution of light chains S131 and V133 were respectively made by K and R, or D and E as well as in mutation at position 30 wherein heavy chain L145 and light chain V133 were substituted with K and D, respectively.

[0313] In addition, a mutation pair in which heavy chain S183 and light chain V133 were respectively substituted with K (or R) and D (or E) (termed mutation at position 48) was also observed to be effective (Table 18 and FIG. 17 (Aa pairing accuracy (low bands): 45%, Bb pairing accuracy (upper bands): 95%).

TABLE-US-00019 TABLE 18 LC V133K (a) V133D (b) 2B9 Heavy Chain (B) S183K 4D9 Heavy Chain (A) S183D mg/L 2.6 2.4

[0314] Antibodies containing a combination of the aforementioned mutation at position 29, mutation at position 30, and mutation at position 48 (c29c30c48FΦ29f30f48) or variations thereof were constructed (see Table 19) and then tested for pairing accuracy. The results are depicted in FIG. 18 (lower bands: Aa pairing; and upper bands: Bb paring):

[0315] In addition, antibodies containing a variant mutation pair of at least one of mutation at position 29, mutation at position 30, and mutation at position 48 (see Table 20) were examined for pairing accuracy and the results are depicted in FIG. 19 (lower bands: Aa pairing; and upper bands: Bb paring):

TABLE-US-00020 TABLE 20 No. 1 2 3 4 5 C c30Φf30 c.sup.κ30Φf.sup.κ30 c29c.sup.R30Φf.sup.R30 c.sup.R29c.sup.RE30Φ c.sup.R29c.sup.RE30f48Φ f.sup.R29f.sup.RE30 f.sup.R29f.sup.RE30f48 LC V133K V133D V133R V133D V133R S131D/ S131K/ S131D/ S131K/ S131D/ (a) (b) V133D V133R V133E V133R V133E 4D9 L145D L145D L145D L145E L145E/S183D (A) 2B9 L145K L145R L145R L145R L145R/S183K (B) Aa 79% 67% 80% 90% 100%  Bb 50% 62% 70% 80% 90% (Aa: Aa pairing accuracy; Bb: Bb pairing accuracy)

[0316] Further, combinations of mutation at position 29, mutation at position 30, and mutation at position 48 were subjected to swapping between D and E to search for combinations that allowed for relative accurate pairing (Table 21 and FIG. 20).

TABLE-US-00021 TABLE 21 Code A B a b Aa Bb 0 L145K/S183K L145D/S183D S131D/V133D S131K/V133K  65% 100% ab L145K/S183K L145E/S183E S131D/V133D S131K/V133K  95%  95% gh L145K/S183K L145D/S183D S131E/V133E S131K/V133K 100% 100% abgh L145K/S183K L145E/S183E S131E/V133E S131K/V133K 100% 100% (Aa: Aa pairing accuracy; Bb: Bb pairing accuracy)

[0317] As can be seen in Table 21 and FIG. 20, all the tested mutation pairs ab, gh, and abgh exhibited a pairing accuracy of 65% or higher, or 95% or higher. Of them, abgh was selected and termed c′29c′30c48Φf′29f′30f′48 mutation pair.

[0318] Chains into which combinations of the mutation pairs at position 34 and 51 selected from among the mutation pairs identified in Table 15 were introduced were tested for pairing and the results are given in Table 22 and FIG. 21.

TABLE-US-00022 TABLE 22 LC L135K/T180K (a) L135D/T180D (b) 4D9 Heavy Chain (A) K147D/V185D 2B9 Heavy Chain (B)  K147/V185K Aa (%) 95% Bb (%) 62%

[0319] In addition, mutation at position 34 and mutation at position 51 of Table 15 were subjected to change from K to R and from D to E, followed by a pairing test to detect combinations having improved pairing accuracy. As a result, a combination in which L135 and

[0320] T180 on the light chain were each substituted with E was improved in pairing accuracy and termed c34Φf51 mutation pair (see Table 23 and FIG. 22).

TABLE-US-00023 TABLE 23 LC L135K/T180K (a) L135E/T180E (b) 4D9 Heavy Chain (A) K147D/V185D 2B9 Heavy Chain (B)  K147/V185K Aa (%) 95% Bb (%) 86%

[0321] Of various mutation pairs found in Table 15, a combination of mutation at position 40 and mutation at position 44 was also obverted to improve in pairing accuracy.

[0322] In addition, all the mutation pairs were subjected to swapping between K and R and between D and E to search for a combination that allowed for the most accurate pairing. The combination was a mutation pair in which light chain L135 was substituted with R and E, and was termed c40Φf44 mutation pair (see Table 24 and FIG. 23):

TABLE-US-00024 TABLE 24 LC L135R/S162K (a) L135E/S162D (b) 4D9 Heavy Chain (A) F170D/P171D 2B9 Heavy Chain (B) F170K/P171K Aa (%) 95% Bb (%) 84%

Example 5: Bispecific Antibody Formation by Coupling Between Heavy Chains and Between Heavy Chain and Light Chain

[0323] 5.1. Bispecific Antibody Having c′29c′30c48Φf′29f′30f′48 Mutation Pair Introduced Thereto

[0324] 5.1.1. Test of Coupling Between Heavy and Light Chain Using 4D9/2B9 Antibody

[0325] Antibodies were constructed by coupling heavy and light chains containing the c′29c′30c48Φf′29f′30f′48 mutation pair (see Table 25):

TABLE-US-00025 TABLE 25 Heavy Chain Light Chain Constant Region Constant Region Antibody (HC:CH1) (LC) 4D9 (Aa) L145E/S183E S131K/V133K 2B9 (Bb) L145K/S183K S131E/V133E

[0326] In order to examine the accuracy of pairing among A chain, B chain, a chain, and b chain in the bispecific antibodies, individual heavy chains and light chains were co-expressed in all possible combinations of normal pairs and abnormal mispairs thereamong. Expression levels measured by SDS-PAGE are given in Table 26, below. Thermal stability (Tm) of the combinations was measured and the results are given in Table 27 and FIG. 25:

TABLE-US-00026 TABLE 26 HC A B LC a b a b Expression 89.8 mg/l 18.8 mg/l 24.5 mg/l 68.5 mg/l Amount

TABLE-US-00027 TABLE 27 Pairs Tm 4D9 WT 69.9 ± 2.00 2B9 WT 69.1 ± 1.50 Aa 70.3 ± 1.70 Ab  58.5 ± 11.96 Bb 68.7 ± 0.92 Ba 58.6 ± 0.53

[0327] Thermal stability (Tm) was measured with reference to the method explained in Example 2.

[0328] As shown in Tables 26 and 27 and FIG. 25, the bispecific antibodies having c′29c′30c48Φf′29f′30f′48 mutation pair introduced thereto were found to be higher in expression level and thermal stability for normal pairing (Aa/Bb) than for abnormal pairing (Ab/Ba).

[0329] 5.1.2. Trastuzumab/Bevacizumab Bispecific Antibody or Adalimumab/Bevacizumab Bispecific Antibody

[0330] Trastuzumab (Herceptin®; Roche), Bevacizumab (Avastin™; Roche), and Adalimumab (Humira®; AbbVie) were purchased and subjected to amino acid sequencing (the Korea Basic Science Institute, Korea). cDNAs corresponding to the amino acid sequences were synthesized and used to construct bispecific antibodies to which the c′29c′30c48Φf′29f′30f′48 mutation pair and the AWBB mutation pair of CH3 domain selected in Example 3 were introduced in the combinations shown in Table 28, below (pcDNA3 vector (see FIG. 26) used).

TABLE-US-00028 TABLE 28 Heavy Chain Light Chain Constant Region (HC) Constant Antibody CH1 CH3 Region (LC) Trabev Trastuzumab (Aa) L145E/S183E S364K/K409W S131K/V133K Bevacizumab (Bb) L145K/S183K K370S/F405R S131E/V133E Adabev Adalimumab (Aa) L145E/S183E S364K/K409W S131K/V133K Bevacizumab (Bb) L145K/S183K K370S/F405R S131E/V133E

[0331] The bispecific antibodies Trabev and Adabev were subjected to hydrophobic interaction chromatography (HIC) in the following condition and the results are depicted in FIG. 27 (y-axis: Value (mAU); and x-axis: time (min)):

[0332] Instrument: HPLC-U3000

[0333] Column MAbPac Hic-20

[0334] Flow rate: 0.2 mL/min

[0335] Detection: UV, 280 nm

[0336] Mobile Phase: 0.10 M Ammonium acetate, pH 7.0

[0337] Passage of the obtained protein through the HIC column formed peaks at different time points depending on the hydrophobicity of the protein, whereby the construction of accurate bispecific antibodies could be accounted for. As can be seen in FIG. 27, peaks for the heterodimerized bispecific antibodies (Trabev and Adabev) are distinctively observed between peaks for two different homodimer antibodies (Trastuzumab and Bevacizumab, or Adalimumab and Bevacizumab), indicating the fine formation of bispecific antibodies, each composed of halves from two different antibodies.

[0338] In addition, the bispecific antibodies Trabev and Adabev thus obtained were subjected to size exclusion chromatography (SEC) analysis, and the results are given in Table 29 and FIGS. 28 and 29 (x-axis: time (min)):

[0339] Instrument: HPLC-U3000

[0340] Column size-exclusion chromatography T SKgel G3000SWXL Tosoh Bioscience

[0341] Flow rate: 1.0 mL/min

[0342] Detection: UV, 280 nm

[0343] Mobile Phase: 25 mM Tris-HCl (pH 8.5), 150 mM NaCl

TABLE-US-00029 TABLE 29 WT (Monospecific Antibody) BsAb (Bispecific Antibody) 1A. Trastuzumab 8.007 min. 1″ Trabev 7.847 min. 1B. Bevacizumab 7.743 min.  2″ Adabev 7.833 min. 2A. Adalimumab 7.987 min.

[0344] In the SEC analysis, peaks are detected according to protein size and can elucidate protein aggregation. As shown in Table 29 and FIGS. 28 and 29, peaks for the bispecific antibodies are present between peaks of the two corresponding antibodies on the time axis, indicating the fine formation of the bispecific antibodies.

[0345] Further, heterodimization modes of the bispecific antibodies Trabev and Adabev on SDS-PAGE are depicted in FIG. 30. As can be seen in FIG. 30, the bispecific antibodies Trabev and Adabev, which are heterodimers, were detected as single bands at intermediate sizes between Trastuzumab and Bevacizumab and between Adalimumab and Bevacizumab, respectively. These results imply that the bispecific antibodies are not homodimers, but are constructed only as a result of normal pairing.

[0346] Through ELISA, tests were conducted to examine whether the BsAbs bind effectively to respective antigens (Trastuzumab: Her2), Bevacizumab (VEGF), and Adalimumab (TNF-alpha).

[0347] Affinity for antigen was measured as follows:

[0348] Reagent

[0349] Detection antibody: goat anti-human kappa-HRP (southern biotech, 2060-05)

[0350] TMB single solution (LIFE TECHNOLOGY, 002023)

[0351] Instrument

[0352] Emax precision microplate reader (Molecular devices)

[0353] Protocol

[0354] Coating buffer: Carbonate buffer pH 9.6

[0355] Blocking buffer: protein-free(TBS) blocking buffer (Thermo scientific)

[0356] Wash buffer: 0.05% (w/v) Tween20 in TBS, pH7.4 (TBST)

[0357] Diluent: 0.05% (w/v) Tween20 in TBS, pH7.4

[0358] Stop buffer: 1N Hydrochloric acid solution (HCl)

[0359] Protocol

[0360] Coating: dilute antigen in coating buffer, load 100 ul of dilution to each well, and incubate at 4□ overnight (Her2, VEGF: 50 ng/well, TNF-alpha: 100 ng/well);

[0361] Washing 3 times with washing buffer;

[0362] Blocking: load 300 ul of blocking buffer, incubate at room temperature (RT) for 1 hour;

[0363] Washing 3 times with washing buffer;

[0364] Binding: load antibodies at an aliquot of 100 ng/well, and incubate at RT for 1 hr;

[0365] Washing 3 times with washing buffer;

[0366] Detection Antibody: dilute goat anti-human kappa-HRP in TBST at a ratio of 1:4000 and incubate at RT for 1 hr;

[0367] Washing 3 times with washing buffer

[0368] Detection: load 100 ul of TMB solution per well, incubate at RT for 3 min in the dark;

[0369] Stop solution: load 100 ul of 1N HCl per well;

[0370] Reading: read at optical density 450 nm

[0371] The results thus obtained are given in Table 30 and FIG. 31 (for Trabev) and in Table 31 and FIG. 32 (for Adabev):

TABLE-US-00030 TABLE 30 Antigen Her2 VEGF Antibody Trastuzumab Bevacizumab Trabev Trastuzumab Bevacizumab Trabev O.D 2.50 0.05 2.54 0.65 2.89 2.04 2.96 0.05 2.60 0.70 2.88 2.25 2.58 0.04 2.49 0.60 3.12 2.36 Mean 2.68 0.05 2.54 0.65 2.96 2.22 blank 0.05 0.05

TABLE-US-00031 TABLE 31 Antigen TNF-alpha VEGF Antibody Adalimumab Bevacizumab Adabev Adalimumab Bevacizumab Adabev O.D 1.24 0.23 0.93 0.07 2.24 2.37 1.26 0.22 1.16 0.06 2.27 2.28 1.19 0.27 1.09 0.07 2.60 2.47 Mean 1.23 0.24 1.06 0.07 2.37 2.37 blank 0.06 0.06

[0372] As shown in Table 29 and FIG. 31, the bispecific antibody Trabev was found to bind to both Her2 and VEGF, which are antigens of Trastuzumab and Bevacizumab, respectively. Also, data in Table 30 and FIG. 32 proved that the bispecific antibody Adabev binds to TNF-alpha and VEGF, which are antigens of Adalimumab and Bevacizumab, respectively. These results confirmed that the two bispecific antibodies normally exerting desired functions were successfully constructed.

[0373] 5.2. Bispecific Antibody Having c34Φf51 Mutation Pair Introduced Thereto

[0374] With reference to Example 5.1.1, antibodies having c34Φf51 mutation pair (see Table 23) introduced to 4D9 (Aa) and 2B9 (Bb) thereof were constructed. For each heavy chain, a and b light chains were co-expressed, and inter-light/heavy chain pairing ratios were measured on SDS-PAGE (conducted for heavy chains A and B, each). Tm was measured as in Example 2, and the results are given in Table 32, below and FIG. 34.

TABLE-US-00032 TABLE 32 Test 1 Test 2 HC A(K147D/V185D) B (K147/V185K) LC a b b a (L135K/T180K) (L135E/T180E) (L135E/T180E) (L135K/T180K) q 90% 10% 89% 11% Tm 65.8 N/A 65.8 N/A (q: light chain/heavy chain paring ratio (%); N/A: not available)

[0375] As shown in Table 32 and FIG. 34, normal heavy chain/light chain pairs (Aa and Bb) were formed at a ratio of as high as 90% and 89%, respectively, and exhibited high thermal stability.

[0376] In addition, bispecific antibodies Trabev (Aa: Trastuzumab; Bb: Bevacizumab) and Adabev (Aa: Adalimumab; Bb: Bevacizumab) to each of which the c34Φf51 mutation pair and the CH3 domain AWBB mutation pair (Aa: S364K/K409W; Bb: K370S/F405R) selected in Example 3 were introduced were constructed using Trastuzumab (Herceptin®; Roche), Bevacizumab (Avastin™; Roche), and Adalimumab (Humira®; AbbVie) in a manner similar to the construction procedure for bispecific antibodies of Example 5.1.2.

[0377] The constructed bispecific antibodies Trabev and Adabev, each having the c34Φf51 mutation pair and AWBB mutation pair introduced thereinto were analyzed using hydrophobic interaction chromatography (HIC) with reference to the method of Example 5.1.2.

[0378] The resulting analysis data are given in Table 33 and FIG. 35 (for Trabev) and in Table 34 and FIG. 36 (for Adabev).

TABLE-US-00033 TABLE 33 # Set 1 Antibody Time (min.) Aa Trastuzumab 23.41 Bb Bevacizumab 38.05 1AB Trabev 29.13

TABLE-US-00034 TABLE 34 # Set 2 Antibody Time (min.) Aa Adalimumab 22.07 Bb Bevacizumab 38.05 2AB Adabev 30.21

[0379] As can be seen in Table 33 and FIG. 35 and in Table 34 and FIG. 36, the peak of each of the heterodimeric bispecific antibodies (Trabev and Adabev) is distinctively observed between the peaks of two corresponding monospecific antibodies (Trastuzumab and Bevacizumab, or Adalimumab and Bevacizumab). The result implies the fine formation of bispecific antibodies having the c34Φf51 mutation pair and AWBB mutation pair introduced thereto, each composed of halves from two different antibodies.

[0380] Further, heterodimization modes of the bispecific antibodies having the c34Φf51 mutation pair and AWBB mutation pair introduced thereto are depicted in FIG. 37 (SDS-PAGE gel 6%, Non-Reducing condition). As can be seen in FIG. 37, concurrent introduction of the c34Φf51 mutation pair and AWBB mutation pair into already known antibodies was found to construct pure heterodimers as analyzed by SDS-PAGE.

[0381] 5.3. Bispecific Antibody Having c40Φf44 Mutation Pair Introduced Thereto

[0382] With reference to Example 5.1.1, antibodies having c40Φf44 mutation pair (see Table 24) introduced to 4D9 (Aa) and 2B9 (Bb) thereof were constructed. For each heavy chain, a and b light chains were co-expressed, and inter-light/heavy chain pairing ratios were measured on SDS-PAGE (conducted for heavy chains A and B, each). Tm was measured as in Example 2, and the results are given in Table 35, below and FIG. 38.

TABLE-US-00035 TABLE 35 HC A (F170D/P171D) B (F170K/P171K) LC a b b a (L135R/S162K) (L135E/S162D) (L135E/S162D) (L135R/S162K) q 99% 1% 99% 1% Tm 65.0 N/A 63.4 N/A (q: light chain/heavy chain pairing ratio (%); N/A: not available)

[0383] As shown in Table 35 and FIG. 38, normal heavy chain/light chain pairs (Aa and Bb) were both formed at a ratio of as high as 99%, and exhibited high thermal stability.

[0384] In addition, bispecific antibodies Trabev (Aa: Trastuzumab; Bb: Bevacizumab) and Adabev (Aa: Adalimumab; Bb: Bevacizumab) to each of which the c40Φf44 mutation pair and the AWBB mutation pair (A: S364K/K409W; B: K370S/F405R) were introduced were constructed using Trastuzumab (Herceptin®; Roche), Bevacizumab (Avastin™; Roche), and Adalimumab (Humira®; AbbVie) in a manner similar to the construction procedure for bispecific antibodies of Example 5.1.2.

[0385] The constructed bispecific antibodies Trabev and Adabev, each having the c40Φf44 mutation pair and AWBB mutation pair introduced thereinto, were analyzed using hydrophobic interaction chromatography (1-11C) with reference to the method of Example 5.1.2.

[0386] The resulting analysis data are given in Table 36 and FIG. 39 (for Trabev) and in Table 37 and FIG. 40 (for Adabev).

TABLE-US-00036 TABLE 36 # Set 1 Antibody Time (min.) Aa Trastuzumab 23.41 Bb Bevacizumab 38.05 1AB Trabev 26.11

TABLE-US-00037 TABLE 37 # Set 2 Antibody Time (min.) Aa Adalimumab 22.07 Bb Bevacizumab 38.05 2AB Adabev 31.97

[0387] As can be seen in Table 36 and FIG. 39 and in Table 37 and FIG. 40, the peak of each of the heterodimeric bispecific antibodies (Trabev and Adabev) is distinctively observed between the peaks of two corresponding monospecific antibodies (Trastuzumab and Bevacizumab, or Adalimumab and Bevacizumab). The result implies the fine formation of bispecific antibodies having the c40Φf44 mutation pair and AWBB mutation pair introduced thereto, each composed of halves from two different antibodies. In addition, only one peak distinctively appearing between the two monospecific antibodies indicates that only one kind of normal pairing was made without mispairs between two heavy chains and between heavy and light chains

[0388] Further, heterodimization modes of the bispecific antibodies having the c40Φf44 mutation pair and AWBB mutation pair introduced thereto are depicted in FIG. 41 (SDS-PAGE gel 6%, Non-Reducing condition). As can be seen in FIG. 41, concurrent introduction of the c40Φf44 mutation pair and AWBB mutation pair into already known antibodies was found to construct pure heterodimers as analyzed by SDS-PAGE.

Comparative Example 1

[0389] Heterodimerization rates were examined for cases where the mutations of CH3 domain proposed herein, which were different from the mutations suggested in Examples 1 to 3 although being identical amino acid pairs, were introduced.

[0390] For comparison, TNFRSF1B-Fc fusion protein (Enbrel; Enb) and Fas-Fc fusion protein (Fas) to which mutations of Table 38 were introduced were constructed with reference to Example 1 (expressed as BEAT-A and BEAT-B).

TABLE-US-00038 TABLE 38 Enb Fas AW/BB AW BB (S364K/K409W) (K370S/F405R) BEAT-A S364K/K409W K370T/F405A BEAT-B S364T/K409R K370T/F405S

[0391] TNFRSF1B-Fc fusion protein (A chain) and Fas-Fc fusion protein having CH3 domain mutations selected in Example 3, which are representative of CH3 domain mutations proposed in the description (e.g., A chain having S364K and 1(409W introduced thereto and B chain having K370S and F405R introduced thereto) (expressed as AW/BB in Table 38) were prepared.

[0392] When Enb-Fc and Fas-Fc which had the mutation pairs listed in Table 38 were subjected to single transfection, homodimization modes were observed on SDS-PAGE. The results are given in FIG. 42.

[0393] As can be seen in FIG. 42, AW/BB exhibited high percentages of monomers whereas higher percentages of homodimers than monomers were detected in BEAT-A and BEAT-B. On the basis of the result, BEAT-A and BEAT-B are more likely to form homodimers than AW/BB when there is a difference in expression level between A chain and B chain. If a large quantity of homodimers is produced when heavy and light chains are combined so as to construct bispecific antibodies, an accurate heterodimer is difficult to separate from the homodimers because there are almost no differences in physical properties between the heterodimer and the homodimers. Accordingly, minimalizing homodimerization is very important for constructing a bispecific antibody of high purity. A combination of mutations that permits heterodimerization as little as possible is advantageous for the construction of bispecific antibodies.

[0394] Taken together, the data show that BEAT-A and BEAT-B have low heterodimerization potentials due to high homodimizerization rates compared to AW/BB and make it difficult to isolate accurate heterodimers.