Heterodimeric antibody-fc-containing proteins and methods for production thereof

Abstract

To provide heterodimeric antibody-Fc-containing proteins, and to provide methods for producing the same.SOLUTION: Provided is an in vitro method for generating a heterodimeric protein. The method comprises the steps of: a) providing a first homodimeric protein comprising an Fc region of an immunoglobulin, the Fc region comprising a first CH3 region; b) providing a second homodimeric protein comprising an Fc region of an immunoglobulin, the Fc region comprising a second CH3 region, and sequences of the first and second CH3 regions being different and being such that heterodimeric interaction between the first and second CH3 regions is stronger than each of the homodimeric interactions of the first and second CH3 regions; c) incubating the first protein together with the second protein under reducing conditions sufficient to allow cysteines in a hinge region to undergo disulfide-bond isomerization; and d) obtaining the heterodimeric protein.SELECTED DRAWING: None

Description

[Technical field]

[0001] FIELD OF THEINVENTION The present invention relates to novel heterodimeric antibody Fc-containing proteins, such as bispecific antibodies, and novel methods for producing such proteins. [Background technology]

[0002] 2. Background of the Invention Monoclonal antibodies have become successful therapeutic molecules in recent years, especially for the treatment of cancer. Unfortunately, however, monoclonal antibodies often fail to cure disease when used as monotherapy. Bispecific antibodies could potentially overcome some of the limitations of monoclonal antibody therapy, for example, they could be used as mediators for targeting drugs or toxic compounds to target cells, for retargeting effector mechanisms to disease-related sites, or for increasing specificity to tumor cells, for example by binding to a combination of target molecules found exclusively on tumor cells.

[0003] Recently, Chames and Baty (2009) Curr Opin Drug Disc Dev 12: 276 (Non-Patent Document 1) have reviewed different formats and uses of bispecific antibodies. One of the main obstacles in the development of bispecific antibodies has been the difficulty of producing the material in sufficient quality and quantity by conventional techniques such as hybrid hybridomas and chemical conjugation methods (Marvin and Zhu (2005) Acta Pharmacol Sin 26: 649 (Non-Patent Document 2)). Co-expression in a host cell of two antibodies, consisting of different heavy and light chains, results in a mixture of possible antibody products in addition to the desired bispecific antibody.

[0004] Several strategies have been described that favor the formation of heterodimeric, ie bispecific, products by co-expression of different antibody constructs.

[0005] Lindhofer et al. (1995 J Immunol 155:219) describe that fusion of rat and mouse hybridomas producing different antibodies results in enrichment of functional bispecific antibodies due to selective species-restricted heavy / light chain pairing. Another strategy to promote heterodimer formation over homodimers is the "knob-in-hole" strategy, in which a protuberance is introduced at the interface of a first heavy chain polypeptide and at the location of a corresponding cavity at the interface of a second heavy chain polypeptide, such that the protuberance can be positioned within the cavity to promote heterodimer formation and prevent homodimer formation. The "protuberance" is constructed by exchanging a small amino acid side chain from the interface of the first polypeptide with a larger side chain. A complementary "cavity" of the same or similar size as the protuberance is created at the interface of the second polypeptide by exchanging the large amino acid side chain with a smaller one. EP1870459 (Chugai) and WO2009089004 (Amgen) describe other strategies to favor heterodimerization by co-expression of different antibody domains in a host cell. In these methods, one or more residues constituting the CH3-CH3 interface in both CH3 domains are replaced with charged amino acids such that homodimerization is electrostatically unfavorable and heterodimerization is electrostatically favorable. WO2007110205 (Merck) describes yet another strategy, which exploits the difference between the IgA CH3 domain and the IgG CH3 domain to promote heterodimerization.

[0006] Dall'acqua et al. (1998 Biochemistry 37:9266) (Non-Patent Document 4) identified five energetically important amino acid residues (366, 368, 405, 407 and 409) involved in CH3-CH3 contacts at the CH3 homodimer interface.

[0007] WO2008119353 (Genmab) describes an in vitro method for producing bispecific antibodies, in which a bispecific antibody is formed by "Fab arm" or "half molecule" exchange (swapping of heavy and attached light chains) between two monospecific IgG4 or IgG4-like antibodies by incubation under reducing conditions. This Fab arm exchange reaction is the result of a disulfide bond isomerization reaction and dissociation / association of CH3 domains, in which the heavy chain disulfide bonds in the hinge region of the parent (originally monospecific) antibody are reduced and the resulting free cysteines form inter-heavy chain disulfide bonds with cysteine ​​residues of another parent antibody molecule (also with a different specificity), while the CH3 domains of the parent antibody are released and reorganized by dissociation / association. The resulting product is a bispecific antibody with two Fab arms potentially composed of different sequences. It should be noted, however, that this process is random and Fab arm exchange can occur between two molecules with identical sequences, or two bispecific molecules can participate in Fab arm exchange to regenerate an antibody containing the specificity of the original monospecific parent antibody.

[0008] It has now surprisingly been found that by introducing asymmetric mutations in the CH3 regions of two monospecific starting proteins, the Fab arm exchange reaction can be forced to become directed and thus to give rise to highly stable heterodimeric proteins. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] EP1870459 [Patent Document 2] WO2009089004 [Patent Document 3] WO2007110205 [Patent Document 4] WO2008119353 [Non-patent literature]

[0010] [Non-Patent Document 1] Chames and Baty (2009) Curr Opin Drug Disc Dev 12: 276 [Non-Patent Document 2] Marvin and Zhu (2005) Acta Pharmacol Sin 26: 649 [Non-Patent Document 3] Lindhofer et al. (1995 J Immunol 155:219) [Non-Patent Document 4] Dall'acqua et al. (1998 Biochemistry 37:9266) Summary of the Invention

[0011] Thus, in one aspect, the present invention provides an efficient in vitro method for the production of highly stable heterodimeric Fc-containing proteins based on a stable homodimeric Fc-containing starting material. For example, highly stable bispecific antibodies can be formed in high yield and purity based on two stable monospecific antibodies as starting materials.

[0012] Thus, in one aspect, the present invention relates to an in vitro method for producing a heterodimeric protein comprising the steps of: a) providing a first homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a first CH3 region; b) providing a second homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a second CH3 region; wherein the sequences of the first CH3 region and the second CH3 region are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; c) incubating the first protein with the second protein under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization; and d) obtaining said heterodimeric protein.

[0013] This method can be used for in vitro production of heterodimeric proteins, such as bispecific antibodies, for various uses, e.g., therapeutic or diagnostic. The advantage of this in vitro method is that the heavy / light chain pairing remains intact during the reaction, so that undesired combinations of heavy and light chains are not obtained in the product. This is in contrast to some of the co-expression methods described in the prior art (see above), which require finding a common light chain that can form a functional antibody with both heavy chains, to avoid the formation of non-functional heavy / light chain products due to random heavy / light chain pairing in cells. Furthermore, the in vitro process can be performed in the laboratory, which allows for higher control, flexibility and yield of heterodimeric proteins than is possible with co-expression.

[0014] The in vitro methods of the present invention can also be used to generate compound libraries of larger size, for example in screening methods to identify advantageous combinations of specificities. For example, for some combinations of antibody targets, not all bispecific antibodies are functional, i.e., not able to bind both targets simultaneously and mediate the desired functional effect. In such cases, bispecific antibodies with the desired properties, e.g., optimal target binding or cell killing, can be identified by screening the bispecific antibodies. a) providing a first set of homodimeric antibodies having different variable regions, wherein the first set of antibodies comprises a first CH3 region; b) providing a second set of homodimeric antibodies having different variable regions, wherein the second set of antibodies comprises a second CH3 region; wherein the sequences of the first and second CH3 regions are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; c) incubating the combination of said first set and said second set of antibodies under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization, thus generating a set of bispecific antibodies; d) optionally returning the conditions to non-reducing conditions; e) assaying the resulting set of bispecific antibodies for a given desired property; and f) selecting bispecific antibodies with desired properties It can be identified by:

[0015] In a further aspect, the present invention relates to a heterodimeric protein obtained or obtainable by the method of the invention and to a method for producing the heterodimeric protein of the invention by co-expression in a suitable host cell. The present invention also relates to the following: [Item 1] 1. An in vitro method for producing a heterodimeric protein comprising the steps of: a) providing a first homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a first CH3 region; b) providing a second homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a second CH3 region; wherein the sequences of the first CH3 region and the second CH3 region are different and such that a heterodimer interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimer interactions of the first CH3 region and the second CH3 region; c) incubating the first protein with the second protein under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization; and d) obtaining said heterodimeric protein. [Item 2] 2. The in vitro method of item 1, wherein the first homodimeric protein and the second homodimeric protein are selected from the group consisting of: (i) an Fc region, (ii) an antibody, (iii) a fusion protein comprising an Fc region, and (iv) an Fc region conjugated to a prodrug, peptide, drug, or toxin. [Item 3] 2. The in vitro method of item 1, wherein the first homodimeric protein is a full-length antibody. [Item 4] The in vitro method of any one of the preceding paragraphs, wherein the second homodimeric protein is a full-length antibody. [Item 5] The in vitro method of any one of the preceding paragraphs, wherein the first and second homodimeric proteins are both antibodies and bind to different epitopes. [Item 6] The in vitro method of any one of the preceding paragraphs, wherein the Fc region of the first homodimeric protein is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4, and the Fc region of the second homodimeric protein is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4. [Item 7] The in vitro method of any one of the preceding paragraphs, wherein the Fc regions of both the first homodimeric protein and the second homodimeric protein are of the IgG1 isotype. [Item 8] The in vitro method of any one of the preceding paragraphs, wherein one of the Fc regions of the homodimeric protein is of the IgG1 isotype and the other is of the IgG4 isotype. [Item 9] The in vitro method of any one of the preceding paragraphs, wherein the increase in strength of the heterodimeric interaction compared to each of the homodimeric interactions is due to a CH3 modification other than a covalent bond, introduction of a cysteine ​​residue or a charged residue. [Item 10] The in vitro method of any one of the preceding paragraphs, wherein the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is such that Fab arm exchange cannot occur at 0.5 mM GSH under the conditions described in Example 13. [Item 11] The in vitro method of any one of the preceding paragraphs, wherein the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is such that no Fab arm exchange occurs in vivo in mice under the conditions described in Example 14. [Item 12] The in vitro method of any one of the preceding claims, wherein the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is more than two-fold stronger, such as more than three-fold stronger, such as more than five-fold stronger, than the strongest of the two homodimeric interactions, e.g., determined as described in Example 30. [Item 13] The in vitro method of any one of the preceding paragraphs, wherein the sequences of the first CH3 region and the second CH3 region are such that, when assayed as described in Example 30, the dissociation constant of the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is less than 0.05 micromolar. [Item 14] The in vitro method according to any one of the preceding items, wherein the sequences of the first CH3 region and the second CH3 region are such that, when assayed as described in Example 21, the dissociation constant of both homodimeric interactions is greater than 0.01 micromolar, such as greater than 0.05 micromolar, preferably 0.01 to 10 micromolar, such as 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, such as 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, such as 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar. [Item 15] The in vitro method of any one of the preceding paragraphs, wherein the sequences of the first CH3 region and the second CH3 region contain amino acid substitutions at positions that are not identical. [Item 16] 16. The in vitro method of item 15, wherein the amino acid substitute is a natural amino acid or an unnatural amino acid. [Item 17] The in vitro method of any one of the preceding paragraphs, wherein the first homodimeric protein has only one amino acid substitution in the CH3 region and the second homodimeric protein has only one amino acid substitution in the CH3 region compared to the wild-type CH3 region. [Item 18] 17. The in vitro method according to any one of items 1 to 16, wherein the first homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405, 407 and 409, and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405, 407 and 409, wherein the first homodimeric protein and the second homodimeric protein are not substituted at the same positions. [Item 19] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407. [Item 20] 19. The in vitro method according to any one of items 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid other than Phe at position 405. [Item 21] 19. The in vitro method according to any one of items 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an amino acid other than Phe, Arg or Gly at position 405. [Item 22] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises an amino acid other than Phe at position 405 and Lys at position 409. [Item 23] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises an amino acid other than Phe, Arg or Gly at position 405 and Lys at position 409. [Item 24] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises a Leu at position 405 and a Lys at position 409. [Item 25] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises an amino acid other than Phe, Arg or Gly at position 405 and a Lys at position 409. [Item 26] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Leu at position 405 and a Lys at position 409. [Item 27] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises Lys at position 409, Thr at position 370 and Leu at position 405. [Item 28] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises an Arg at position 409, and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370, and a Leu at position 405. [Item 29] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein comprises a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370 and a Leu at position 405. [Item 30] The in vitro method according to any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407. 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407. [Item 32] 19. The in vitro method according to any one of items 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has Gly, Leu, Met, Asn or Trp at position 407. [Item 33] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407 and Lys at position 409. [Item 34] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has Ala, GIy, His, Ile, Leu, Met, Asn, Val or Trp at position 407 and Lys at position 409. [Item 35] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407 and a Lys at position 409. [Item 36] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and the second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407 and a Lys at position 409. [Item 37] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and the second homodimeric protein has an Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407 and a Lys at position 409. [Item 38] 19. The in vitro method of any one of items 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and the second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407 and a Lys at position 409. [Item 39] the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein (i) an amino acid other than Phe, Leu, and Met at position 368, or (ii) Trp at position 370, or (iii) an amino acid other than Asp, Cys, Pro, Glu, or Gln at position 399 19. The in vitro method according to any one of items 1 to 18, comprising: [Item 40] the first homodimeric protein has Arg, Ala, His or Gly at position 409, and the second homodimeric protein has (i) Lys, Gln, Ala, Asp, Glu, Gly, His, Ile, Asn, Arg, Ser, Thr, Val or Trp at position 368, or (ii) Trp at position 370, or (iii) Ala, Gly, Ile, Leu, Met, Asn, Ser, Thr, Trp, Phe, His, Lys, Arg, or Tyr at position 399 19. The in vitro method according to any one of items 1 to 18, comprising: [Item 41] the first homodimeric protein has Arg at position 409 and the second homodimeric protein has (i) Asp, Glu, Gly, Asn, Arg, Ser, Thr, Val, or Trp at position 368, or (ii) Trp at position 370, or (iii) Phe, His, Lys, Arg, or Tyr at position 399 19. The in vitro method according to any one of items 1 to 18, comprising: [Item 42] The in vitro method of any one of the preceding paragraphs, wherein the first and second CH3 regions comprise the sequence set forth in SEQ ID NO:1, except for the specified mutations. [Item 43] The in vitro method of any one of the preceding paragraphs, wherein neither the first homodimeric protein nor the second homodimeric protein contains a Cys-Pro-Ser-Cys sequence in the hinge region. [Item 44] The in vitro method of any one of the preceding paragraphs, wherein both the first homodimeric protein and the second homodimeric protein comprise a Cys-Pro-Pro-Cys sequence in the hinge region. [Item 45] The in vitro method of any one of the preceding paragraphs, wherein the first homodimeric protein and the second homodimeric protein are human antibodies, except for any mutations specified. [Item 46] The in vitro method of any one of the preceding paragraphs, wherein the first homodimeric protein and the second homodimeric protein are heavy chain antibodies. [Item 47] 47. The in vitro method according to any one of items 1 to 46, wherein both the first homodimeric protein and the second homodimeric protein further comprise a light chain. [Item 48] 48. The in vitro method of item 47, wherein the light chains are different. [Item 49] The in vitro method of any one of the preceding paragraphs, wherein the first and / or second homodimeric protein comprises a mutation that removes an acceptor site for Asn-linked glycosylation. [Item 50] 2. The in vitro method according to any one of the preceding paragraphs, wherein the first and second homodimeric proteins provided in steps a) and b) are purified. [Item 51] The in vitro method of any one of the preceding paragraphs, wherein the first and / or second homodimeric protein is conjugated to or comprises an acceptor group for a drug, prodrug or toxin. [Item 52] 52. The in vitro method according to any one of items 4 to 51, wherein the first and / or second epitope is located on a tumor cell. [Item 53] 53. The in vitro method according to any one of items 4 to 52, wherein the first or the second epitope is located on a tumor cell and the other epitope is located on an effector cell. [Item 54] 54. The in vitro method according to any one of items 4 to 53, wherein the epitope is located on a T cell, such as on CD3 expressed on a T cell. [Item 55] 53. The in vitro method according to any one of items 4 to 52, wherein the first or second epitope is located on a tumor cell and the other epitope is located on a radioisotope, a toxin, a drug or a prodrug, which may optionally be coupled or linked to a peptide or hapten. [Item 56] 53. The in vitro method according to any one of items 4 to 52, wherein the first or the second epitope is located on a tumor cell and the other epitope is located on an electron dense vesicle or a minicell. [Item 57] 53. The in vitro method according to any one of items 1 to 52, wherein the homodimeric proteins are both antibodies, and the first antibody and the second antibody bind to different epitopes on the same tumor cell. [Item 58] 53. The in vitro method according to any one of items 1 to 52, wherein the homodimeric proteins are both antibodies, and a first antibody binds to an epitope on a tumor cell, and the other antibody is an unrelated or inactive antibody that does not have any in vivo binding activity relevant to the intended use. [Item 59] The in vitro method according to any one of the preceding paragraphs, wherein the reducing conditions in step c) comprise the addition of a reducing agent, such as a reducing agent selected from the group consisting of 2-mercaptoethylamine, dithiothreitol and tris(2-carboxyethyl)phosphine or a chemical derivative thereof. [Item 60] The in vitro method according to any one of the preceding items, wherein step c) is carried out under reducing conditions with a redox potential of -150 to -600 mV, such as -250 to -400 mV. [Item 61] 20. The in vitro method of any one of the preceding claims, wherein step c) comprises incubation in the presence of at least 25 mM 2-mercaptoethylamine or in the presence of at least 0.5 mM dithiothreitol at a temperature of at least 20°C for at least 90 minutes. [Item 62] 8. The in vitro method according to any one of the preceding items, wherein step d) comprises removal of the reducing agent, for example by desalting. [Item 63] A method for the selection of bispecific antibodies with desired properties comprising the steps of: a) providing a first set of homodimeric antibodies comprising antibodies with different variable regions, wherein the antibodies of the first set comprise an identical first CH3 region; b) providing a second set of homodimeric antibodies comprising antibodies with different variable regions or identical variable regions, wherein the antibodies of the second set comprise an identical second CH3 region; wherein the sequences of the first and second CH3 regions are different and such that a heterodimer interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimer interactions of the first CH3 region and the second CH3 region; c) incubating the combination of the first and second sets of antibodies under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization, thus generating a set of bispecific antibodies; d) optionally returning the conditions to non-reducing conditions; e) assaying the resulting set of bispecific antibodies for a given desired property; and f) selecting bispecific antibodies with desired properties. [Item 64] The method of paragraph 63, wherein the second set of homodimeric antibodies have different variable regions. [Item 65] 64. The method of paragraph 63, wherein the second set of homodimeric antibodies have identical variable regions but different amino acids or structural changes outside the antigen binding region. [Item 66] 1. A method for producing a heterodimeric protein, comprising the steps of: a) providing a first nucleic acid construct encoding a first polypeptide comprising a first Fc region of an immunoglobulin, wherein the first Fc region comprises a first CH3 region; b) providing a second nucleic acid construct encoding a second polypeptide comprising a second Fc region of an immunoglobulin, wherein the second Fc region comprises the first CH3 region; wherein the sequences of the first and second CH3 regions are different and such that a heterodimer interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimer interactions of the first CH3 region and the second CH3 region; and wherein said first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and said second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407; and / or wherein the sequences of the first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, for example 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, for example 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, for example 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar; c) co-expressing the first and second nucleic acid constructs in a host cell; and d) Obtaining said heterodimeric protein from the cell culture. [Item 67] the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe, such as other than Phe, Arg or Gly; or the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr; The method described in item 66. [Item 68] 68. The method of item 66 or 67, wherein the first and second polypeptides are full-length heavy chains of two antibodies that bind to different epitopes. [Item 69] 69. The method according to any one of items 66 to 68, wherein step c) further comprises co-expressing in the host cell one or more nucleic acid constructs encoding the light chain. [Item 70] 70. The method according to any one of items 66 to 69, further comprising the features of any one or more of items 2 to 58. [Item 71] 71. An expression vector comprising the nucleic acid construct specified in any one of items 66 to 70. [Item 72] A host cell comprising a nucleic acid construct as defined in any one of items 66 to 70. [Item 73] A heterodimeric protein obtained or obtainable by the method according to any one of items 1 to 72. [Item 74] A heterodimeric protein comprising a first polypeptide comprising a first Fc region of an immunoglobulin, the first Fc region comprising a first CH3 region, and a second polypeptide comprising a second Fc region of an immunoglobulin, the second Fc region comprising a second CH3 region, wherein the sequences of the first and second CH3 regions are distinct and such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; wherein said first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and said second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407; and / or wherein the sequences of the first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, such as 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, such as 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, such as 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar. The heterodimeric protein. [Item 75] the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe, such as other than Phe, Arg or Gly; or the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr; 74. The heterodimeric protein according to item 73. [Item 76] 76. The heterodimeric protein according to item 74 or 75, wherein the first and second polypeptides are full-length heavy chains of two antibodies that bind to different epitopes. [Item 77] 77. The heterodimeric protein according to item 76, further comprising two full-length light chains. [Item 78] The heterodimeric protein according to any one of items 73 to 77, further comprising the features of any one or more of items 2 to 58. [Item 79] [Item 80] The heterodimeric protein according to any one of Items 74 to 78, for use as a pharmaceutical. 80. The heterodimeric protein according to item 79 for use in treating cancer. [Item 81] 79. A pharmaceutical composition comprising the heterodimeric protein according to any one of items 74 to 78 and a pharma- ceutically acceptable carrier. [Item 82] A method for inhibiting the growth and / or proliferation of tumor cells and / or for killing tumor cells, comprising administering to an individual in need thereof the heterodimeric protein according to any one of items 74 to 78. [Brief description of the drawings]

[0016] [Figure 1] Generation of bispecific antibodies by interspecies Fab arm exchange. Generation of bispecific antibodies after GSH-induced in vitro Fab arm exchange between the indicated EGFR (2F8) IgG4 and CD20 (7D8) IgG4 antibodies was determined by ELISA. A concentration series from 0 to 1 μg / mL (total antibody) was analyzed in the ELISA. Bispecific binding was higher after Fab arm exchange between a rhesus (Rh) and a human (Hu) IgG4 antibody than between two antibodies of the same species. [Diagram 2]Alignment of the amino acid sequences of the core hinge (i.e., the CPPC sequence in human IgG1, which contains the two cysteine ​​residues that potentially form inter-heavy chain disulfide bonds and the corresponding residues in other human isotypes or other species) and the CH3-CH3 interface of human and rhesus antibody isotypes. [Diagram 3] Generation of bispecific antibodies using mutant human IgG1 involved in Fab arm exchange. Generation of bispecific antibodies after GSH-induced in vitro Fab arm exchange between human CD20 (7D8) IgG4 antibody and the indicated human EGFR (2F8) IgG1 antibodies was determined by ELISA. The graph presented shows the average number of three independent Fab arm exchange experiments, in which a total antibody concentration of 1 μg / mL was used for the ELISA. Bispecific binding was higher after Fab arm exchange between IgG1-2F8-CPSC-ITL and IgG4-7D8 than between the two IgG4 antibodies. Combining IgG4-7D8 with either IgG1-2F8-CPSC or IgG1-2F8-ITL did not yield bispecific antibodies under the conditions used. [Figure 4] Generation of bispecific antibodies by in vivo Fab arm exchange of human IgG4 and mutant IgG1 antibodies. Generation of bispecific antibodies after in vivo Fab arm exchange in immunodeficient mice between human CD20 (7D8) IgG4 antibody and the indicated human EGFR (2F8) IgG1 and IgG4 mutant antibodies was determined by ELISA. The graph presented shows the average number (n=4). Bispecific reactivity is presented as the concentration (percentage) of bispecific antibody relative to the total IgG concentration. Human IgG4 with R409K mutation in the CH3 domain or stabilized hinge (CPPC) cannot participate in Fab arm exchange. IgG1 with both CPSC sequence in the hinge and R409K mutation in the CH3 domain participates in Fab arm exchange. (*) Bispecific binding to mixtures containing either IgG1-2F8, IgG4-2F8-CPPC or IgG4-2F8-R409K was below the detection limit and was therefore arbitrarily set to zero. [Diagram 5]Generation of bispecific antibodies using 2-mercapto-ethylamine·HCl (2-MEA)-induced Fab arm exchange between human IgG1 and IgG4 antibodies. Generation of bispecific antibodies after 2-MEA-induced in vitro Fab arm exchange between the indicated human EGFR (2F8) and CD20 (7D8) antibodies was determined by ELISA. A dilution series of 0-40 mM 2-MEA was tested. The graphs presented show the results of the ELISA, where a total antibody concentration of 20 μg / mL was used. 2-MEA efficiently induced Fab arm exchange, even between antibodies containing stabilized hinges (CPPC). A higher level of bispecific reactivity was obtained with the combination of human IgG1 × human IgG4 with the triple mutation T350I-K370T-F405L in the CH3 domain compared to the two wild-type IgG4 antibodies. [Figure 6A] Generation of bispecific antibodies using 2-MEA-induced Fab arm exchange between human IgG1 and IgG4 antibodies. Generation of bispecific antibodies after 2-MEA-induced in vitro Fab arm exchange between the indicated human EGFR (2F8) and CD20 (7D8) antibodies was determined by mass spectrometry for all samples in a concentration series of 0-40 mM 2-MEA. (A) Representative mass spectrometry profiles for samples of Fab arm exchange reactions between IgG1-2F8-ITL x IgG4-7D8-CPPC at 0 mM, 7 mM, and 40 mM 2-MEA are shown. IgG4-2F8 x IgG4-7D8 yielded approximately 50% bispecific antibody. IgG1-2F8-ITL x IgG4-7D8-CPPC yielded approximately 95% bispecific antibody. [Figure 6B]Generation of bispecific antibodies using 2-MEA-induced Fab arm exchange between human IgG1 and IgG4 antibodies. Generation of bispecific antibodies after 2-MEA-induced in vitro Fab arm exchange between the indicated human EGFR (2F8) and CD20 (7D8) antibodies was determined by mass spectrometry for all samples in a concentration series of 0-40 mM 2-MEA. (B) After quantification of the mass spectrometry data, the percentage of bispecific antibodies was calculated and plotted against the concentration of 2-MEA in the Fab arm exchange reaction. IgG4-2F8×IgG4-7D8 yielded approximately 50% bispecific antibodies. IgG1-2F8-ITL×IgG4-7D8-CPPC yielded approximately 95% bispecific antibodies. [Figure 7] Stability analysis of heterodimeric bispecific antibodies obtained by 2-MEA-induced Fab arm exchange. The stability of bispecific samples generated by 2-MEA-induced Fab arm exchange by combining either IgG1-2F8-ITL x IgG4-7D8-CPPC (A) or IgG4-2F8 x IgG4-7D8 (B) was tested by measuring EGFR / CD20 bispecific binding in ELISA after GSH-induced Fab arm exchange reaction in the presence of the indicated concentrations of irrelevant IgG4. Bispecific binding is presented relative to the bispecific binding of the starting material (control), which was set to 100%. (A) The bispecific binding of the 2-MEA-induced bispecific product derived from IgG1-2F8-ITL x IgG4-7D8-CPPC was preserved, suggesting a stable product that did not participate in Fab arm exchange under GSH conditions. (B) Bispecific EGFR / CD20 binding of the 2-MEA-induced bispecific product derived from IgG4-2F8 × IgG4-7D8 was impaired, suggesting that this product was involved in Fab arm exchange with irrelevant IgG4 under GSH conditions. [Figure 8]Plasma clearance rate of heterodimeric bispecific antibodies generated by 2-MEA-induced Fab arm exchange. Three groups of mice (3 mice per group) were injected with the indicated antibodies: (1) 100 μg of bispecific antibody generated by in vitro 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL x IgG4-7D8-CPPC; (2) 100 μg of bispecific antibody + 1,000 μg of irrelevant IgG4; (3) 50 μg of IgG1-2F8-ITL + 50 μg of IgG4-7D8-CPPC. (A) Total antibody concentration over time as determined by ELISA. The total antibody plasma concentration curve was the same for all antibodies. (B) Bispecific antibody concentration as determined by ELISA. The bispecificity of the injected antibodies was the same with or without the addition of excess irrelevant IgG4. (*) Bispecific binding to the IgG1-2F8-ITL + IgG4-7D8-CPPC mixture was below the detection limit, therefore the corresponding symbol could not be plotted in this graph. The average value of two ELISA experiments is shown. [Figure 9A] Purity of bispecific antibodies generated by Fab arm exchange between human IgG1-2F8 and IgG4-7D8-CPPC. (A) Reducing SDS-PAGE (a) shows heavy and light chain bands for both the bispecific sample and the IgG1 control sample. Non-reducing SDS-PAGE (b). [Figure 9B] Purity of bispecific antibodies generated by Fab arm exchange between human IgG1-2F8 and IgG4-7D8-CPPC. (B) Peak results from HP-SEC analysis show that the bispecific sample was more than 98% homogeneous and, notably, no antibody aggregates could be detected. [Figure 9C-1] Purity of bispecific antibodies generated by Fab arm exchange between human IgG1-2F8 and IgG4-7D8-CPPC. (C) Mass spectrometry shows that the Fab arm exchange yielded approximately 100% bispecific product. [Figure 9C-2] This is a continuation of Figure 9C-1. [Figure 10]Comparison of triple mutant (ITL), double mutant (IT, IL, TL) and single mutant (L) human IgG1-2F8 in the generation of bispecific antibodies by Fab arm exchange with human IgG4-7D8. The generation of bispecific antibodies after 2-MEA-induced in vitro Fab arm exchange between human IgG1-2F8 triple and double mutants and wild type IgG4-7D8 with CPSC hinge (A) or mutant IgG4-7D8-CPPC with stabilized hinge (B), or between single mutant IgG1-2F8-F405L and IgG4-7D8 with wild type CPSC hinge or stabilized CPPC hinge (C) was determined by ELISA. A concentration series of 0-20 μg / mL or 0-10 μg / mL (total antibody) was analyzed by ELISA for experiments involving double mutants and single mutants, respectively. Combination with the double mutants IgG1-2F8-IL and IgG1-2F8-TL results in bispecific EGFR / CD20 binding similar to the triple mutant IgG1-ITL. Combination with IgG1-2F8-IT does not result in a bispecific product. Combination with the single mutant IgG1-2F8-F405L results in bispecific EGFR / CD20 binding. [Figure 11] Generation of bispecific antibodies using 2-MEA-induced Fab arm exchange at different temperatures. Generation of bispecific antibodies by combining the indicated human EGFR (2F8) and CD20 (7D8) antibodies in a 2-MEA-induced in vitro Fab arm exchange reaction at 0° C., 20° C. and 37° C. was followed by ELISA over time. Bispecific binding was most efficient at 37° C. and slower at 20° C. No generation of bispecific binding was measured at 0° C. [Figure 12]Generation of bispecific antibodies by in vitro Fab arm exchange induced by different reducing agents. ELISA was used to measure the generation of bispecific antibodies by combining human IgG1-2F8-ITL and IgG4-7D8-CPPC in a reduction reaction with a concentration series of the indicated reducing agents. Bispecific binding was measured after reaction with DTT (maximum obtained at 2.5 mM DTT) and 2-MEA (maximum obtained at 25 mM 2-MEA), but not with GSH. (*) Data for GSH concentrations above 10 mM were excluded due to the formation of antibody aggregates. [Figure 13] 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL and IgG1-7D8-K409X mutants. Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between IgG1-2F8-ITL and the indicated IgG1-7D8-K409X mutants was determined by ELISA. (A) A concentration series (total antibody) from 0 to 20 μg / mL was analyzed. The positive control is a purified batch of bispecific antibody derived from IgG1-2F8-ITL x IgG4-7D8-CPPC. (B) Exchange is presented as bispecific binding at 20 μg / mL against the positive control (black bars). Dark grey bars represent bispecific binding between IgG4 control (IgG4-7D8 x IgG4-2F8), negative control (IgG1-2F8 x IgG1-7D8-K409R) and between IgG1-2F8-ITL and IgG4-7D8-CPPC. Open grey bars represent the results of Fab arm exchange reactions carried out simultaneously between the indicated IgG1-7D8-K409X mutants and IgG1-2F8-ITL. [Figure 14]Antibody deglycosylation does not affect bispecific antibody generation by 2-MEA-induced Fab arm exchange. Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between the indicated EGFR (2F8) and CD20 (7D8) antibodies was determined by ELISA. Exchange with the 7D8 antibody was compared to its enzymatically deglycosylated variants. A concentration series from 0 to 20 μg / mL (total antibody) was analyzed in the ELISA. Fab arm exchange reactions with deglycosylated antibodies showed identical bispecific binding curves as the glycosylated variants from which they were derived. [Figure 15A] The ability to participate in Fab arm exchange correlates with the strength of the CH3-CH3 interaction. (A) Generation of bispecific antibodies by GSH-induced Fab arm exchange between IgG1-2F8 and IgG1-7D8 constructs (A) with the indicated mutations displayed as bispecific binding in ELISA over time. Bispecificity is displayed relative to IgG4-2F8 x IgG4-7D8 control after 24 hours. [Figure 15B] The ability to participate in Fab arm exchange correlates with the strength of the CH3-CH3 interaction. (B) Generation of bispecific antibodies by GSH-induced Fab arm exchange between IgG4-2F8 and IgG4-7D8 constructs (B) with the indicated mutations displayed as bispecific binding in ELISA over time. Bispecificity is displayed relative to the IgG4-2F8 x IgG4-7D8 control after 24 hours. [Figure 15C] The ability to participate in Fab arm exchange correlates with the strength of the CH3-CH3 interaction. (C) Generation of bispecific antibodies by GSH-induced Fab arm exchange between IgG4-2F8 and IgG4-7D8 constructs (C) with the indicated mutations displayed as bispecific binding in ELISA over time. Bispecificity is displayed relative to the IgG4-2F8 x IgG4-7D8 control after 24 hours. [Figure 15D]The ability to participate in Fab arm exchange correlates with the strength of the CH3-CH3 interaction. (D) Relationship between apparent KD (Table 2) and bispecific antibody generation after 24 hours (Figure 15A / B / C) for IgG1-based molecules (D). [Figure 15E] The ability to participate in Fab arm exchange correlates with the strength of the CH3-CH3 interaction. (E) Relationship between apparent KD (Table 2) after 24 hours (Figure 15A / B / C) and bispecific antibody generation for IgG4-based molecules (E). [Figure 16] Sequence alignment of anti-EGFr antibody 2F8 in IgG1, IgG4 and (partial) IgG3 backbones, depicting the amino acid numbering according to Kabat and the amino acid numbering according to the EU index (both as described in Kabat et al., Sequences of Proteins of Immunological Interest, 5th Ed. Public Health Service, National Institutes of Health, Bethesda, MD. (1991)). [Figure 17] Generation of bispecific antibodies by in vitro Fab arm exchange induced by different reducing agents. ELISA was used to measure the generation of bispecific antibodies by combining human IgG1-2F8-F405L and IgG1-7D8-K409R in reduction reactions with a concentration series of the indicated reducing agents. The measured OD values ​​were normalized to the signal of the bispecific control sample derived from 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL x IgG4-7D8-CPPC, which was set to 100%. Maximum bispecific binding was measured after reaction with DTT in the concentration range of 0.5-50 mM, 2-MEA in the concentration range of 25-50 mM, and tris(2-carboxyethyl)phosphine (TCEP) in the concentration range of 0.5-5.0 mM, but not with GSH. (*) Data at GSH concentrations above 25 mM were excluded due to the formation of antibody aggregates. [Figure 18]Generation of bispecific antibodies using 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L and IgG1-7D8-K409R. (A) Generation of bispecific antibodies after 2-MEA-induced in vitro Fab arm exchange was determined by ELISA. The graph presented shows the results of the ELISA with a total antibody concentration of 20 μg / mL. 2-MEA efficiently induced Fab arm exchange. (B) Generation of bispecific antibodies after 2-MEA-induced in vitro Fab arm exchange was determined by mass spectrometry for all samples in the concentration series of 0-40 mM 2-MEA. After quantification of the mass spectrometry data, the percentage of bispecific antibodies was calculated and plotted against the concentration of 2-MEA in the Fab arm exchange reaction. IgG1-2F8-F405L x IgG1-7D8-K409R yielded approximately 100% bispecific antibodies, confirming the ELISA data. [Figure 19-1] Purity of bispecific antibodies generated by Fab arm exchange between human IgG1-2F8-F405L x IgG1-7D8-K409R. Mass spectrometry shows that the Fab arm exchange yielded approximately 100% bispecific product. [Figure 19-2] This is a continuation of Figure 19-1. [Figure 20] Plasma clearance of bispecific antibodies generated by 2-MEA-induced Fab arm exchange. Two groups of mice (3 mice per group) were injected with the indicated antibodies: (1) 100 μg of bispecific antibody generated by in vitro 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R; (2) 100 μg of bispecific antibody + 1,000 μg of irrelevant IgG4. (A) Total antibody concentration over time as determined by ELISA. The total antibody plasma concentration curve was the same for all antibodies. (B) Bispecific antibody concentration as determined by ELISA. The bispecificity of the injected antibodies was the same with or without the addition of excess irrelevant IgG4. [Figure 21]CDC-mediated cell killing of CD20-expressing cells by bispecific antibodies generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R. A concentration series of the indicated antibodies was used to test their ability to mediate CDC on Daudi cells (A) and Raji cells (B). Both cell lines express CD20 but not EGFR. Introduction of K409R in IgG1-7D8 did not affect its ability to induce CDC. The bispecific antibody derived from 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R was still able to induce CDC. [Figure 22] ADCC-mediated cell killing of EGFR-expressing cells by bispecific antibodies generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R. A concentration series of the indicated antibodies was used to test their ability to mediate ADCC against A431 cells. IgG1-7D8 was unable to bind to CD20-negative A431 cells and therefore did not induce ADCC. ADCC was induced by the EGFR antibody IgG1-2F8 even after introduction of the F405L mutation in the CH3 domain. The ADCC effector function of IgG1-2F8-F405L was retained in the bispecific format obtained by Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R. [Diagram 23]2-MEA-induced Fab arm exchange between IgG1-2F8-F405X mutants and IgG1-7D8-K409R. Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between the indicated IgG1-2F8-F405X mutants and IgG1-7D8-K409R was determined by ELISA. (A) A concentration series from 0 to 20 μg / mL (total antibody) was analyzed in the ELISA. The positive control is a purified batch of bispecific antibody derived from IgG1-2F8-F405L x IgG1-7D8-K409R. (B) The exchange is presented as bispecific binding at an antibody concentration of 20 μg / mL relative to the positive control (black bars). The dark grey bars represent bispecific binding between the IgG4 control (IgG4-7D8 x IgG4-2F8) and the negative control (IgG1-2F8 x IgG1-7D8-K409R). Open grey bars represent the results of Fab arm exchange reactions performed simultaneously between the indicated IgG1-2F8-F405X mutants and IgG1-7D8-K409R or controls. [Figure 24] 2-MEA-induced Fab arm exchange between IgG1-2F8-Y407X mutants and IgG1-7D8-K409R. Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between the indicated IgG1-2F8-Y407X mutants and IgG1-7D8-K409R was determined by ELISA. (A) A concentration series from 0 to 20 μg / mL (total antibody) was analyzed in the ELISA. The positive control is a purified batch of bispecific antibody derived from IgG1-2F8-F405L x IgG1-7D8-K409R. (B) The exchange is presented as bispecific binding at an antibody concentration of 20 μg / mL relative to the positive control (black bars). The dark grey bars represent bispecific binding between the IgG4 control (IgG4-7D8 x IgG4-2F8) and the negative control (IgG1-2F8 x IgG1-7D8-K409R). Open grey bars represent the results of Fab arm exchange reactions performed simultaneously between the indicated IgG1-2F8-Y407X mutants and IgG1-7D8-K409R or controls. [Diagram 25]Analysis of bispecific antibodies generated by 2-MEA-induced Fab arm exchange by SDS-PAGE under non-reducing (Figure 25(A)) and reducing (Figure 25(B)) conditions. [Figure 26] HP-SEC profiles of the homodimeric starting material IgG1-2F8-F405L (Figure 26(B)), the homodimeric starting material IgG1-7D8-K409R (Figure 26(A)), a mixture of both dimers (1:1) (Figure 26(C)), and a bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R (Figure 26(D)). [Figure 27] Mass spectrometry (ESI-MS) of the homodimeric starting material IgG1-2F8-F405L (Figure 27(B)), the homodimeric starting material IgG1-7D8-K409R (Figure 27(A)), a mixture of both dimers (1:1) (Figure 27(C)), and a bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R (Figure 27(D)). [Figure 28A] Capillary isoelectric focusing (cIEF) profile of the homodimeric starting material IgG1-2F8-F405L (Figure 28(A)). [Figure 28B] Capillary isoelectric focusing (cIEF) profile of the homodimeric starting material IgG1-7D8-K409R (Figure 28(B)). [Figure 28C] Capillary isoelectric focusing (cIEF) profile of a mixture of both dimers (1:1) (Figure 28(C)). [Figure 28D] Capillary isoelectric focusing (cIEF) profile of a bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R (Figure 28(D)). [Figure 29]HPLC-CIEX profiles of the homodimeric starting material IgG1-2F8-F405L (Figure 29(A)), the homodimeric starting material IgG1-7D8-K409R (Figure 29(B)), a mixture of both dimers (1:1) (Figure 29(C)), and a bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R (Figure 29(D)). [Diagram 30] Electrospray ionization mass spectrometry of IgG obtained by co-transfection of expression vectors encoding the heavy and light chains of IgG1-7D8-K409R or IgG1-2F8-F405. Heterodimer peaks are indicated with *. Homodimer peaks are indicated with †. [Diagram 31] Exchange reactions of homodimers IgG1-2F8-F405L and IgG1-7D8-K409R as monitored by high pressure liquid chromatography cation exchange (HPLC-CIEX) after injections at different intervals. [Diagram 32] Residual homodimers (indicated by arrows) of the exchange reaction shown in FIG. 32 as detected by CIEX. [Figure 33A] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA. [Figure 33B] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA. [Figure 33C] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA. [Figure 33D] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA. [Figure 33E] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA. [Figure 33F]Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA. [Figure 34A] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA and compared to an arbitrary control set at 100%. [Figure 34B] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA and compared to an arbitrary control set at 100%. [Figure 34C] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA and compared to an arbitrary control set at 100%. [Fig. 34D] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as determined by ELISA and compared to an arbitrary control set at 100%. [Figure 35A] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as analyzed by HPLC-CIEX. [Figure 35B] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as analyzed by HPLC-CIEX. [Figure 35C] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as analyzed by HPLC-CIEX. [Figure 35D] Generation of bispecific antibodies at various IgG concentrations, 2-MEA concentrations, incubation temperatures and times as analyzed by HPLC-CIEX. [Diagram 36]Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between the indicated IgG1-2F8-L368X mutants and IgG1-7D8-K409R was determined by ELISA using a concentration series from 0 to 20 μg / mL (total antibody) (Figure 37(A)). The positive control is a purified batch of bispecific antibody derived from IgG1-2F8-F405L x IgG1-7D8-K409R. Figure 37(B) shows bispecific binding at 20 μg / mL against the positive control (black bars). The dark grey bars represent bispecific binding between the IgG4 control (IgG4-7D8 x IgG4-2F8) and the negative control (IgG1-2F8 x IgG1-7D8-K409R). The open grey bars represent the results of a Fab arm exchange reaction performed simultaneously between the indicated IgG1-2F8-L368X mutants and IgG1-7D8-K409R. [Figure 37] Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between the indicated IgG1-2F8-K370X mutants and IgG1-7D8-K409R was determined by ELISA using a concentration series from 0 to 20 μg / mL (total antibody) (Figure 37(A)). The positive control is a purified batch of bispecific antibody derived from IgG1-2F8-F405L x IgG1-7D8-K409R. Figure 37(B) shows bispecific binding at 20 μg / mL against the positive control (black bars). The dark grey bars represent bispecific binding between the IgG4 control (IgG4-7D8 x IgG4-2F8) and the negative control (IgG1-2F8 x IgG1-7D8-K409R). The open grey bars represent the results of a Fab arm exchange reaction performed simultaneously between the indicated IgG1-2F8-D370X mutants and IgG1-7D8-K409R. [Figure 38]Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between the indicated IgG1-2F8-D399X mutants and IgG1-7D8-K409R was determined by ELISA using a concentration series from 0 to 20 μg / mL (total antibody) (Figure 38(A)). Figure 38(B) shows bispecific binding at an antibody concentration of 20 μg / mL against the positive control (black bars). Dark grey bars represent bispecific binding between the IgG4 control (IgG4-7D8 x IgG4-2F8) and the negative control (IgG1-2F8 x IgG1-7D8-K409R). Open grey bars represent the results of a Fab arm exchange reaction performed simultaneously between the indicated IgG1-2F8-D399X mutants and IgG1-7D8-K409R. [Figure 39] 2-MEA-induced Fab arm exchange between four different IgG1 mutant combinations after 0, 30, 60, 105 and 200 min of incubation at 15° C. as determined by sandwich ELISA. [Diagram 40] 2-MEA-induced Fab arm exchange between different IgG1 mutant combinations after 90 min of antibody incubation at 15° C. as determined by sandwich ELISA. [Diagram 41] Phosphorylation of c-Met by c-Met specific antibodies. A549 cells are incubated with HGF or a panel of different antibodies for 15 min. Proteins are separated by SDS-page gel electrophoresis and transferred to membranes by Western blotting. Phosphorylated c-Met, total c-Met and β-actin are detected by antibodies against phosphorylated c-Met, total c-Met or β-actin. [Diagram 42] Proliferation assay in NCI-H441 cells. NCI-H441 cells were incubated for 7 days with monovalent bispecific IgG1 069 / b12, control antibodies (IgG1-069, UniBody-069, IgG1-b12) and left untreated. Cell mass was determined and plotted as a percentage of the untreated sample (set as 100%). [Diagram 43]CDC-mediated cell killing of CD20-expressing cells by bispecific antibodies generated by 2-MEA-induced Fab arm exchange between IgG1-7D8-F405L or IgG1-2F8-F405L and IgG1-7D8-K409R. A concentration series of the indicated antibodies was used to test their ability to mediate CDC on Daudi cells (A) and Raji cells (B). Both cell lines express CD20 but not EGFR. Bispecific antibodies generated by 2-MEA-induced Fab arm exchange between IgG1-7D8-F405L x IgG1-7D8-K409R were as effective as IgG1-7D8 in inducing CDC-mediated cell killing. Bispecific antibodies derived from 2-MEA-induced Fab arm exchange between IgG2-2F8-F405L × IgG1-7D8-K409R resulted in monovalent CD20-binding bispecific antibodies that had little effect on the induction of CDC-mediated cell killing. [Diagram 44] Killing of A431 cells induced by anti-κ-ETA' pre-incubated HER2xHER2 bispecific antibody. Viability of A431 cells after 3 days of incubation with HER2 antibody pre-incubated with anti-κ-ETA'. Cell viability was quantified using Alamar Blue. Data shown are the fluorescence intensity (FI) of one experiment with A431 cells treated with anti-κ-ETA' conjugated HER2 antibody and HER2xHER2 bispecific antibody. Staurosporine was used as a positive control, while an isotype control antibody was used as a negative control. [Diagram 45] HER2xHER2 bispecific molecules induced downregulation of HER2 receptor. Relative percentage of HER2 expression levels in AU565 cell lysates after 3 days of incubation with 10 μg / mL mAb. The amount of HER2 was quantified using a HER2-specific capture ELISA and depicted as % inhibition compared to untreated cells. Data shown are the mean + standard deviation of two experiments. [Figure 46]Colocalization analysis of HER2xHER2 bispecific antibodies (FITC) with the lysosomal marker LAMP1 (Cy5). FITC pixel intensity overlapping with Cy5 for various monospecific HER2 antibodies and HER2xHER2 bispecific antibodies (Figure 46(B)). FITC pixel intensity of LAMP1 / Cy5 positive pixels in three different images was plotted for each antibody tested. Lower FITC pixel intensity of LAMP1 / Cy5 positive pixels is evident for monospecific antibodies compared to bispecific antibodies. Figure 46(B) represents the average value of FITC pixel intensity per LAMP1 / Cy5 positive pixel calculated from three different images. Taken together, these results suggest that after internalization, higher levels of bispecific antibodies are localized to Lamp1 / Cy5 positive vesicles compared to monospecific antibodies. [Figure 47] Inhibition of proliferation by HER-2 monospecific and bispecific antibodies. AU565 cells were seeded in the presence of 10 μg / mL of HER2 antibody or HER2xHER2 bispecific antibody in serum-free cell culture medium. After 3 days, the amount of live cells was quantified with Alamar Blue and cell viability was presented as a percentage of untreated cells. Isotype control antibody was used as a negative control. Data shown are % AU565 live cells compared to untreated cells, measured at 5-fold ± standard deviation. * indicates that only one data point was depicted. [Figure 48A] Binding of monospecific and bispecific IgG1 antibodies and hinge-deleted IgG1 antibodies to human and mouse FcRn at different pH. Plates with human and mouse FcRn were incubated with different monospecific and bispecific IgG1 antibodies or hinge-deleted IgG1 molecules. Binding to FcRn was analyzed by ELISA at 405 nm. (A) Binding of monospecific and bispecific IgG1 antibodies and hinge-deleted IgG1 (Uni-G1) molecules to human FcRn at pH 7.4 and 6.0. Binding to human FcRn is very low at neutral pH. At pH 6.0, (bispecific) antibodies bind efficiently to human FcRn unless they contain the H435A mutation. Hinge-deleted IgG1 (Uni-G1) molecules bind human FcRn with low efficiency. [Figure 48B] Binding of monospecific and bispecific IgG1 antibodies and hinge-deleted IgG1 antibodies to human and mouse FcRn at different pH. Plates with human and mouse FcRn were incubated with different monospecific and bispecific IgG1 antibodies or hinge-deleted IgG1 molecules. Binding to FcRn was analyzed by ELISA at 405 nm. (B) Binding of monospecific and bispecific IgG1 antibodies and hinge-deleted IgG1 (Uni-G1) molecules to mouse FcRn at pH 7.4 and 6.0. Binding to mouse FcRn is very low at neutral pH. At pH 6.0, (bispecific) antibodies bind very efficiently to mouse FcRn unless they contain the H435A mutation in both Fab arms. Bispecific molecules with the H435A mutation in only one Fab arm can still bind mouse FcRn. Hinge-deleted IgG1 (Uni-G1) molecules bind mouse FcRn with moderate efficiency, and hinge-deleted IgG1 (Uni-G1) bispecific molecules carrying the H435A mutation in only one Fab arm bind slightly less efficiently. [Figure 49-1] T cell-mediated cytotoxicity of AU565 cells by Her2xCD3 bispecific antibody and the N297Q mutant of Her2xCD3 bispecific antibody. [Figure 49-2] T cell-mediated cytotoxicity of AU565 cells by Her2xCD3 bispecific antibody and the N297Q mutant of Her2xCD3 bispecific antibody. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] Detailed Description of the Invention definition The term "immunoglobulin" refers to a class of structurally related glycoproteins consisting of two pairs of polypeptide chains, one pair of low molecular weight light (L) chains and one pair of heavy (H) chains, all four of which are interconnected by disulfide bonds. The structure of immunoglobulins has been well characterized. See, for example, Fundamental Immunology Ch. 7 (Paul, W., ed., 2nd ed. Raven Press, NY (1989)). Briefly, each heavy chain typically consists of a heavy chain variable region (abbreviated herein as VH) and a heavy chain constant region. The heavy chain constant region typically consists of three domains, CH1, CH2, and CH3. The heavy chains are interconnected by disulfide bonds in the so-called "hinge region". Each light chain typically consists of a light chain variable region (abbreviated herein as VL) and a light chain constant region. The light chain constant region typically consists of one domain, CL. Typically, the numbering of amino acid residues in the constant region is done according to the EU index as described in Kabat et al., Sequences of Proteins of Immunological Interest, 5th Ed. Public Health Service, National Institutes of Health, Bethesda, MD. (1991). Figure 16 shows an overview of the EU and Kabat numbering for the different isotype forms of antibody 2F8 (WO 02 / 100348). The VH and VL regions can be further subdivided into regions of hypervariability (or hypervariable regions that may be hypervariable in sequence and / or in the form of structurally defined loops), also called complementarity determining regions (CDRs), interspersed with more conserved regions called framework regions (FRs). Each VH and VL is typically composed of three CDRs and four FRs, arranged from amino-terminus to carboxy-terminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4 (see also Chothia and Lesk J. Mol. Biol. 196, 901 917 (1987)).

[0018] As used herein, the term "Fab arm" refers to one heavy-light chain pair.

[0019] As used herein, the term "Fc region" refers to the region of an antibody that comprises at least the hinge region, the CH2 domain, and the CH3 domain.

[0020] The term "antibody" (Ab) in the context of the present invention refers to an immunoglobulin molecule, a fragment of an immunoglobulin molecule, or a derivative of either, that has the ability to specifically bind to an antigen under typical physiological conditions with a half-life of a significant period of time, such as at least about 30 minutes, at least about 45 minutes, at least about 1 hour, at least about 2 hours, at least about 4 hours, at least about 8 hours, at least about 12 hours (h), about 24 hours or more, about 48 hours or more, about 3, 4, 5, 6, 7 days or more, etc., or any other relevant, functionally defined period of time (such as a time sufficient to induce, promote, enhance, and / or modulate a physiological response associated with antibody binding to the antigen and / or a time sufficient for the antibody to recruit effector activity). The variable regions of the heavy and light chains of an immunoglobulin molecule contain the binding domains that interact with the antigen. The constant region of an antibody (Ab) may mediate the binding of the immunoglobulin to host tissues or factors, including various cells of the immune system (such as effector cells) and components of the complement system, such as C1q, the first component of the classical pathway of complement activation. An antibody may also be a bispecific antibody, a diabody, or a similar molecule. The term "bispecific antibody" refers to an antibody having specificity for at least two different epitopes, typically non-overlapping epitopes. As stated above, the term antibody herein includes fragments of antibodies that retain the ability to specifically bind to an antigen, unless otherwise specified or clearly contradicted by the context. Such fragments can be provided by any known technique, such as enzymatic cleavage, peptide synthesis, and recombinant expression techniques. It has been shown that the antigen-binding function of an antibody can be performed by fragments of a full-length antibody, for example, the F(ab')2 fragment. The term antibody should also be understood to include antibody-like polypeptides, such as polyclonal antibodies, monoclonal antibodies (mAbs), chimeric antibodies, and humanized antibodies, unless otherwise specified. The antibody thus generated may have any isotype.

[0021] As used herein, the term "full length antibody" refers to an antibody that contains all the heavy and light chain constant and variable domains normally found in an antibody of that isotype.

[0022] As used herein, "isotype" refers to the immunoglobulin class (e.g., IgG1, IgG2, IgG3, IgG4, IgD, IgA, IgE, or IgM) that is encoded by heavy chain constant region genes.

[0023] The term "human antibody" as used herein is intended to include antibodies having variable and constant regions derived from human germline immunoglobulin sequences. The human antibodies of the present invention may contain amino acid residues not encoded by human germline immunoglobulin sequences (e.g., mutations introduced by in vitro random or site-specific mutagenesis or by in vivo somatic mutation). However, the term "human antibody" as used herein is not intended to include antibodies in which CDR sequences derived from the germline of another mammalian species, such as a mouse, are grafted onto human framework sequences.

[0024] As used herein, the term "heavy chain antibody" refers to an antibody that consists of only two heavy chains and lacks the two light chains normally found in antibodies. Heavy chain antibodies, such as those naturally occurring in camelids, are capable of binding to antigens despite only having a VH domain.

[0025] The term "epitope" refers to a protein determinant capable of specific binding to an antibody. Epitopes usually consist of surface groups of molecules such as amino acids or sugar side chains and usually have specific three-dimensional structural characteristics and specific charge characteristics. Conformational and non-conformational epitopes are distinguished in that the binding to the former is lost in the presence of denaturing solvents, whereas the binding to the latter is not lost. Epitopes may include amino acid residues that are directly involved in binding (also referred to as the immunodominant components of the epitope) and other amino acid residues that are not directly involved in binding, such as amino acid residues that are effectively blocked by the specific antigen-binding peptide (i.e., the amino acid residues are within the footprint of the specific antigen-binding peptide).

[0026] As used herein, the term "binding" in the context of binding of an antibody to a given antigen typically refers to a binding rate of about 10, as determined, for example, by surface plasmon resonance (SPR) technology in a BIAcore 3000 instrument using the antigen as the ligand and the antibody as the analyte. -6 M or less, e.g., about 10 -7 M or less, e.g., about 10 -8 M or less, e.g., about 10 -9 M or less, about 10 -10 M or less, or about 10 -11 M or lower K D and the K is at least 10 times lower, e.g., at least 100 times lower, e.g., at least 1000 times lower, e.g., at least 10,000 times lower, e.g., at least 100,000 times lower, e.g., at least 100,000 times lower, than the affinity of binding to a non-specific antigen other than the given antigen or a closely related antigen (e.g., BSA, casein). D The amount of the antibody that binds to a given antigen with an affinity corresponding to the K D and therefore the K D If the affinity for an antigen is very low (i.e., the antibody is highly specific), the amount by which the affinity for the antigen is lower than the affinity for a non-specific antigen may be at least 10,000 times.D The term "" (M) refers to the dissociation equilibrium constant of a particular antibody-antigen interaction.

[0027] As used herein, the term "heterodimeric interaction between a first CH3 region and a second CH3 region" refers to the interaction between a first CH3 region and a second CH3 region in a first CH3 / second CH3 heterodimeric protein.

[0028] As used herein, the term "homodimeric interaction between a first CH3 region and a second CH3 region" refers to an interaction between a first CH3 region and another first CH3 region in a first CH3 / first CH3 homodimeric protein, as well as an interaction between a second CH3 region and another second CH3 region in a second CH3 / second CH3 homodimeric protein.

[0029] As used herein, an "isolated antibody" means that the material has been removed from its original environment (e.g., the natural environment if naturally occurring or the host cell if recombinantly expressed). It is also convenient for the antibody to be in a purified form. The term "purified" does not require absolute purity, but rather is intended as a relative definition indicating an increase in antibody concentration relative to the concentration of contaminants in the composition when compared to the starting material.

[0030] The term "host cell" as used herein is intended to refer to a cell into which an expression vector, such as an expression vector encoding an antibody of the invention, has been introduced. Recombinant host cells include transfectomas, such as, for example, CHO cells, HEK293 cells, NS / 0 cells, and lymphocytic cells.

[0031] As used herein, the term "co-expression" of two or more nucleic acid constructs refers to the expression of the two constructs in a single host cell.

[0032] The term "tumor cell protein" refers to a protein that is located on the cell surface of a tumor cell.

[0033] As used herein, the term "effector cell" refers to an immune cell that is involved in the effector phase of an immune response, as opposed to the recognition and activation phases of an immune response. Exemplary immune cells include cells of myeloid or lymphoid origin, such as lymphocytes (such as B cells and T cells, including cytotoxic T cells (CTLs)), killer cells, natural killer cells, macrophages, monocytes, polymorphonuclear cells such as eosinophils, neutrophils, granulocytes, mast cells, and basophils. Some effector cells express specific Fc receptors and perform specific immune functions. In some embodiments, effector cells can induce antibody-dependent cellular cytotoxicity (ADCC), such as natural killer cells that can induce ADCC. In some embodiments, effector cells can phagocytose target antigens or target cells.

[0034] The term "reducing conditions" or "reducing environment" refers to conditions or circumstances in which cysteine ​​residues in the hinge region of a substrate, herein an antibody, are more likely to become reduced than oxidized.

[0035] The term "disulfide bond isomerization" refers to the exchange of disulfide bonds between different cysteines, i.e., the rearrangement of disulfide bonds.

[0036] Further aspects and embodiments of the present invention As mentioned above, in a first aspect, the present invention relates to an in vitro method for producing a heterodimeric protein comprising the steps of: a) providing a first homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a first CH3 region; b) providing a second homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a second CH3 region; wherein the sequences of the first CH3 region and the second CH3 region are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; c) incubating the first protein with the second protein under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization; and d) obtaining said heterodimeric protein.

[0037] The bispecific format can be used in many ways to create a desired combination of bispecific antibodies. In addition to being able to combine antibodies targeting different antigens in a very selective way, it can be used to change a desired property, for example to increase CDC, by combining two different antibodies targeting the same antigen. Furthermore, it can be used to remove the partial agonist activity of an antagonist antibody, or to convert an agonist antibody into an antagonist antibody by creating the bispecific antibody with an unrelated (inactive) antibody.

[0038] In one embodiment, the homodimeric protein is selected from the group consisting of: (i) an Fc region, (ii) an antibody, (iii) a fusion protein comprising an Fc region, such as an Fc region fused to a receptor, cytokine or hormone, and (iv) an Fc region conjugated to a prodrug, peptide, drug or toxin.

[0039] In some embodiments, the first and / or second homodimeric protein comprises, in addition to the Fc region, one or more or all of the other regions of the antibody, namely, the CH1 region, the VH region, the CL region and / or the VL region.Thus, in one embodiment, the first homodimeric protein is a full-length antibody.In another embodiment, the second homodimeric protein is a full-length antibody.

[0040] In an important embodiment, the first and second homodimeric proteins are both antibodies, preferably full-length antibodies, and bind to different epitopes.In such an embodiment, the heterodimeric protein produced is a bispecific antibody.The epitopes can be located on different antigens or on the same antigen.

[0041] However, in other embodiments, only one of the homodimeric proteins is a full-length antibody, and the other homodimeric protein is not a full-length antibody, for example, it is an Fc region without a variable region, which is expressed together with another protein or peptide sequence, such as a receptor, cytokine or hormone, or is bound to a prodrug, peptide, drug or toxin.In further embodiments, neither of the homodimeric proteins is a full-length antibody.For example, either homodimeric protein can be an Fc region that is fused to another protein or peptide sequence, such as a receptor, cytokine or hormone, or is bound to a prodrug, peptide, drug or toxin.

[0042] In one embodiment, the Fc region of the first homodimeric protein is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4, and the Fc region of the second homodimeric protein is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4. In a preferred embodiment, the Fc regions of both the first and second homodimeric proteins are of the IgG1 isotype. In another preferred embodiment, one of the Fc regions of the homodimeric proteins is of the IgG1 isotype and the other is of the IgG4 isotype. In the latter embodiment, the resulting heterodimer contains an IgG1 Fc region and an IgG4 Fc region and can therefore have interesting intermediate properties with respect to activation of effector functions. Similar products can be obtained if the first and / or second homodimeric protein contains mutations that remove acceptor sites for Asn-linked glycosylation or are otherwise engineered to change the glycosylation properties.

[0043] In further embodiments, one or both of the homodimeric proteins are glycoengineered to reduce fucose and thus enhance ADCC, for example by adding compounds to the culture medium during antibody production as described in US2009317869 or van Berkel et al. (2010) Biotechnol. Bioeng. 105:350, or by using FUT8 knockout cells as described in Yamane-Ohnuki et al. (2004) Biotechnol. Bioeng 87:614. Alternatively, ADCC can be optimized using the methods described by Umana et al. (1999) Nature Biotech 17:176.

[0044] In further embodiments, one or both of the homodimeric proteins are engineered to enhance complement activation, e.g., as described in Natsume et al. (2009) Cancer Sci. 100:2411.

[0045] In further embodiments, one or both of the homodimeric proteins were engineered to reduce or increase binding to the neonatal Fc receptor (FcRn) to manipulate the serum half-life of the heterodimeric protein.

[0046] In further embodiments, one of the homodimeric starting proteins is engineered to not bind Protein A, thus allowing the heterodimeric protein to be separated from the homodimeric starting protein by passing the product through a Protein A column. This may be particularly useful in embodiments where an excess of one homodimeric protein is used relative to the other homodimeric protein as starting material. In such embodiments, it may be useful to engineer the homodimeric protein in excess to reduce its ability to bind Protein A. After the heterodimerization reaction, the heterodimeric protein can then be separated from the excess non-exchanged homodimeric protein by passing it through a Protein A column.

[0047] In further embodiments, one of the homodimeric proteins is a full-length antibody that recognizes an Fc region or an unrelated epitope, or a full-length antibody that does not undergo somatic hypermutation and does not bind to self-antigens and comprises a germline-derived sequence. In such embodiments, the heterodimeric protein functions as a monovalent antibody. In another embodiment, both homodimeric proteins contain the same heavy chain, but only one of the homodimeric proteins contains a light chain that forms a functional antigen-binding site with the heavy chain, and the other homodimeric protein contains a non-functional light chain that does not bind to any antigen in combination with the heavy chain. In such embodiments, the heterodimeric protein functions as a monovalent antibody. Such a non-functional light chain can be, for example, a germline-derived sequence that does not undergo somatic hypermutation and does not bind to self-antigens.

[0048] The antibodies used as homodimer starting materials of the present invention may be produced by the hybridoma method, for example as first described by Kohler et al., Nature 256, 495 (1975), or by recombinant DNA methods. Monoclonal antibodies may also be isolated from phage antibody libraries, for example using the techniques described in Clackson et al., Nature 352, 624 628 (1991) and Marks et al., J. Mol. Biol. 222, 581 597 (1991). Monoclonal antibodies may be obtained from any suitable source. Thus, for example, monoclonal antibodies may be obtained from hybridomas prepared from murine splenic B cells obtained from mice immunized with the antigen of interest, for example in the form of cells expressing the antigen on their surface, or in the form of nucleic acid encoding the antigen of interest. Monoclonal antibodies may also be obtained from hybridomas derived from antibody-expressing cells of immunized humans or non-human mammals such as rats, dogs, primates, etc.

[0049] The antibody used as the homodimeric starting material of the present invention may be, for example, a chimeric antibody or a humanized antibody. In another embodiment, except for any specific mutation, one or both of the homodimeric starting proteins are human antibodies. Human monoclonal antibodies can be generated using transgenic or transchromosomic mice, such as HuMAb mice, that carry parts of the human immune system rather than the mouse immune system. HuMAb mice contain human immunoglobulin gene minilocus encoding unrearranged human heavy (μ and γ) and κ light immunoglobulin sequences, with targeted mutations that inactivate endogenous μ and κ chain loci (Lonberg, N. et al., Nature 368, 856 859 (1994)). Thus, the mice exhibit reduced expression of mouse IgM or κ, and in response to immunization, the introduced human heavy and light chain transgenes undergo class switching and somatic mutation to generate high affinity human IgG, κ monoclonal antibodies (Lonberg, N. et al. (1994), supra: Lonberg, N. Handbook of Experimental Pharmacology 113, 49-101 (1994); Lonberg, N. and Huszar, D., Intern. Rev. Immunol. Vol. 13 65-93 (1995); and Harding, F. and Lonberg, N. Ann. NY Acad. Sci 764 536-546 (1995)).The preparation of HuMAb mice is described in detail in Taylor, L. et al., Nucleic Acids Research 20, 6287-6295 (1992); Chen, J. et al., International Immunology 5, 647-656 (1993); Tuaillon et al., J. Immunol. 152, 2912-2920 (1994); Taylor, L. et al., International Immunology 6, 579-591 (1994); Fishwild, D. et al., Nature Biotechnology 14, 845-851 (1996). See also US 5,545,806, US 5,569,825, US 5,625,126, US 5,633,425, US 5,789,650, US 5,877,397, US 5,661,016, US 5,814,318, US 5,874,299, US 5,770,429, US 5,545,807, WO 98 / 24884, WO 94 / 25585, WO 93 / 1227, WO 92 / 22645, WO 92 / 03918, and WO 01 / 09187. Splenocytes from these transgenic mice can be used to generate hybridomas secreting human monoclonal antibodies by well known techniques.

[0050] Furthermore, human antibodies of the invention or antibodies of the invention derived from other species can be identified through display-type technologies, including but not limited to phage display, retroviral display, ribosome display, mammalian display, and other techniques, using techniques well known in the art, and the resulting molecules can be subjected to further maturation, such as affinity maturation, as such maturation techniques are well known in the art.

[0051] In further embodiments of the invention, the antibody or portion thereof, such as one or more CDRs, is of a species of the Camelidae family, or of a species of cartilaginous fish, such as the nurse shark, or is a heavy chain or domain antibody, see WO2010001251.

[0052] In one embodiment of the method of the invention, the first and second homodimeric proteins provided in steps a) and b) are purified.

[0053] In one embodiment, the first and / or second homodimeric protein is conjugated to or contains an acceptor group for a drug, prodrug, or toxin. Such an acceptor group may be, for example, an unnatural amino acid.

[0054] As described above, the sequences of the first and second CH3 regions of the homodimeric starting protein are different such that the heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region.

[0055] In one embodiment, the increase in strength of the heterodimeric interaction compared to each of the homodimeric interactions is due to CH3 modifications other than covalent bonds, introduction of cysteine ​​residues or charged residues.

[0056] In some embodiments, the products of the invention are highly stable and do not undergo Fab arm exchange under mild reducing conditions in vitro or, importantly, in vivo upon administration to humans. Thus, in one embodiment, the heterodimeric interaction between the first and second proteins in the resulting heterodimeric protein is such that Fab arm exchange cannot occur at 0.5 mM GSH under the conditions described in Example 13.

[0057] In another embodiment, the heterodimeric interaction between the first and second proteins in the resulting heterodimeric protein is such that no Fab arm exchange occurs in vivo in mice under the conditions described in Example 14.

[0058] In another embodiment, the heterodimeric interaction between the first and second proteins in the resulting heterodimeric protein is more than two-fold stronger, such as more than three-fold stronger, such as more than five-fold stronger, than the strongest of the two homodimeric interactions, e.g., as determined as described in Example 30.

[0059] In a further embodiment, the sequences of the first CH3 region and the second CH3 region are such that the dissociation constant of the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is less than 0.05 micromolar when assayed as described in Example 30.

[0060] In further embodiments, the sequences of the first CH3 region and the second CH3 region are such that the dissociation constants of both homodimeric interactions are greater than 0.01 micromolar, such as greater than 0.05 micromolar, preferably 0.01-10 micromolar, such as 0.05-10 micromolar, more preferably 0.01-5 micromolar, such as 0.05-5 micromolar, even more preferably 0.01-1 micromolar, such as 0.05-1 micromolar, 0.01-0.5 micromolar or 0.01-0.1 micromolar, when assayed as described in Example 21. Embodiments in which the homodimeric starting protein is relatively stable can have the advantage that it is easier to produce large amounts of the starting protein and to avoid, for example, aggregation or misfolding.

[0061] In some embodiments, stable heterodimeric proteins can be obtained in high yields using the methods of the invention based on two homodimeric starting proteins containing only a few, fairly conservative, asymmetric mutations in the CH3 region.

[0062] Thus, in one embodiment, the sequences of the first CH3 region and the second CH3 region contain amino acid substitutions at positions that are not identical.

[0063] The amino acid substitutes may be natural or unnatural amino acids. Examples of unnatural amino acids are disclosed, for example, in Xie J and Schultz PG, Current Opinion in Chemical Biology (2005), 9:548-554, and Wang Q. et al., Chemistry & Biology (2009), 16:323-336.

[0064] In one embodiment, the amino acid is a naturally occurring amino acid.

[0065] In one embodiment, the first homodimeric protein has only one amino acid substitution in the CH3 region and the second homodimeric protein has only one amino acid substitution in the CH3 region compared to the wild-type CH3 region.

[0066] In one embodiment, a first homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405, 407 and 409, and a second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405, 407 and 409, wherein said first homodimeric protein and said second homodimeric protein are not substituted at the same positions.

[0067] In one embodiment, the first homodimeric protein has an amino acid substitution at position 366 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 368, 370, 399, 405, 407 and 409. In one embodiment, the amino acid at position 366 is selected from Arg, Lys, Asn, Gln, Tyr, Glu and Gly.

[0068] In one embodiment, the first homodimeric protein has an amino acid substitution at position 368 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 370, 399, 405, 407 and 409.

[0069] In one embodiment, the first homodimeric protein has an amino acid substitution at position 370 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 399, 405, 407 and 409.

[0070] In one embodiment, the first homodimeric protein has an amino acid substitution at position 399 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 405, 407 and 409.

[0071] In one embodiment, the first homodimeric protein has an amino acid substitution at position 405 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 407 and 409.

[0072] In one embodiment, the first homodimeric protein has an amino acid substitution at position 407 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 409.

[0073] In one embodiment, the first homodimeric protein has an amino acid substitution at position 409 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407.

[0074] Thus, in one embodiment, the sequences of the first CH3 region and the second CH3 region contain asymmetric mutations, i.e., mutations at different positions in the two CH3 regions, for example a mutation at position 405 in one of the CH3 regions and a mutation at position 409 in the other CH3 region.

[0075] In one embodiment, the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407.

[0076] In one such embodiment, the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid other than Phe at position 405. In a further embodiment thereof, the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid other than Phe, Arg or Gly at position 405.

[0077] In another embodiment, the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises an amino acid other than Phe at position 405 and Lys at position 409. In a further embodiment of this, the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises an amino acid other than Phe, Arg or Gly at position 405 and Lys at position 409.

[0078] In another embodiment, the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises a Leu at position 405 and a Lys at position 409. In a further embodiment of this, the first homodimeric protein comprises a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises an amino acid other than Phe, Arg or Gly at position 405 and a Lys at position 409.

[0079] In another embodiment, the first homodimeric protein comprises a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Leu at position 405 and a Lys at position 409.

[0080] In a further embodiment, the first homodimeric protein comprises an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370 and a Leu at position 405.

[0081] In a further embodiment, the first homodimeric protein comprises an Arg at position 409 and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370 and a Leu at position 405.

[0082] In a further embodiment, the first homodimeric protein comprises a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370 and a Leu at position 405.

[0083] In another embodiment, the first homodimeric protein comprises an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein comprises a Lys at position 409 and: a) an Ile at position 350 and a Leu at position 405, or b) a Thr at position 370 and a Leu at position 405.

[0084] In another embodiment, the first homodimeric protein comprises an Arg at position 409 and the second homodimeric protein comprises a Lys at position 409 and: a) an Ile at position 350 and a Leu at position 405, or b) a Thr at position 370 and a Leu at position 405.

[0085] In another embodiment, the first homodimeric protein comprises a Thr at position 350, a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Lys at position 409 and: a) an Ile at position 350 and a Leu at position 405, or b) a Thr at position 370 and a Leu at position 405.

[0086] In another embodiment, the first homodimeric protein comprises a Thr at position 350, a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises an Ile at position 350, a Thr at position 370, a Leu at position 405 and a Lys at position 409.

[0087] In another embodiment, the first homodimeric protein has an amino acid at position 409 other than Lys, Leu or Met and the second homodimeric protein has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr.

[0088] In another embodiment, the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407.

[0089] In another embodiment, the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407.

[0090] In another embodiment, the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407 and a Lys at position 409.

[0091] In another embodiment, the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407 and a Lys at position 409.

[0092] In another embodiment, the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407 and a Lys at position 409.

[0093] In another embodiment, a first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and a second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407 and a Lys at position 409.

[0094] In another embodiment, a first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and a second homodimeric protein has an Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407 and a Lys at position 409.

[0095] In another embodiment, a first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and a second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407 and a Lys at position 409.

[0096] In one embodiment, the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has (i) an amino acid other than Phe, Leu, and Met at position 368, or (ii) Trp at position 370, or (iii) an amino acid other than Asp, Cys, Pro, Glu, or Gln at position 399 has.

[0097] In one embodiment, the first homodimeric protein has Arg, Ala, His or Gly at position 409 and the second homodimeric protein has (i) Lys, Gln, Ala, Asp, Glu, Gly, His, Ile, Asn, Arg, Ser, Thr, Val or Trp at position 368, or (ii) Trp at position 370, or (iii) Ala, Gly, Ile, Leu, Met, Asn, Ser, Thr, Trp, Phe, His, Lys, Arg, or Tyr at position 399.

[0098] In one embodiment, the first homodimeric protein has Arg at position 409 and the second homodimeric protein has (i) Asp, Glu, Gly, Asn, Arg, Ser, Thr, Val or Trp at position 368, or (ii) Trp at position 370, or (iii) Phe, His, Lys, Arg, or Tyr at position 399 has.

[0099] In addition to the amino acid substitutions specified above, the first and second homodimeric proteins may contain further amino acid substitutions, deletions or insertions compared to the wild-type Fc sequence.

[0100] In a further embodiment, the first and second CH3 regions comprise the sequence set forth in SEQ ID NO:1 (IgG1m(a)), except for the specified mutations. TIFF2025063202000001.tif20169

[0101] In a further embodiment, the first and second CH3 regions comprise the sequence set forth in SEQ ID NO:2 (IgG1m(f)), except for the specified mutations. TIFF2025063202000002.tif20170

[0102] In a further embodiment, the first and second CH3 regions comprise the sequence set forth in SEQ ID NO:3 (IgG1m(ax)), except for the specified mutations. TIFF2025063202000003.tif20170

[0103] In further embodiments, the homodimeric proteins provided may be rat and mouse antibodies, which exhibit favorable pairing, as described by Lindhofer et al. (1995) J Immunol 155:21 (see above), or so-called knobs-in-holes variant antibodies, as described in US Pat. No. 5,731,168 (see above). In some cases, however, the latter homodimeric starting proteins may be more difficult to produce due to extremely weak homodimeric CH3-CH3 interactions. Thus, variants described herein with mutations at positions 350, 370, 405 and 409 may be preferred.

[0104] The sequence of the hinge region of the homodimeric starting proteins may vary, however the resulting heterodimeric protein may be more stable under some circumstances if the hinge region is not IgG4-like, and preferably is IgG1-like.

[0105] Thus, in one embodiment, neither the first nor the second homodimeric protein comprises a Cys-Pro-Ser-Cys sequence in the (core) hinge region.

[0106] In a further embodiment, both the first homodimeric protein and the second homodimeric protein comprise a Cys-Pro-Pro-Cys sequence in the (core) hinge region.

[0107] In many embodiments where the first and second homodimeric proteins are antibodies, the antibodies further comprise light chains.As explained above, the light chains can be different, i.e., have different sequences, and each can form a functional antigen-binding domain with only one of the heavy chains.However, in other embodiments, the first and second homodimeric proteins are heavy chain antibodies, which do not require light chains for antigen binding, see, for example, Hamers-Casterman (1993) Nature 363:446.

[0108] As mentioned above, step c) of the method of the present invention comprises incubating the first protein with the second protein under reducing conditions sufficient to allow the cysteine ​​in the hinge region to undergo disulfide bond isomerization. Examples of suitable conditions are given herein. The minimum requirements for the cysteine ​​in the hinge region to undergo disulfide bond isomerization may vary depending on the homodimeric starting protein, in particular depending on the exact sequence in the hinge region. It is important that each homodimeric interaction of the first and second CH3 regions is weak enough to allow the cysteine ​​in the hinge region to undergo disulfide bond isomerization under given conditions.

[0109] In one embodiment, the reducing conditions in step c) include the addition of a reducing agent, such as a reducing agent selected from the group consisting of 2-mercaptoethylamine (2-MEA), dithiothreitol (DTT), dithioerythritol (DTE), glutathione, tris(2-carboxyethyl)phosphine (TCEP), L-cysteine ​​and β-mercaptoethanol, preferably a reducing agent selected from the group consisting of 2-mercaptoethylamine, dithiothreitol and tris(2-carboxyethyl)phosphine.

[0110] In one embodiment, the reducing conditions that allow for controlled Fab arm exchange are described in terms of the required redox potential. The tripeptide glutathione (GSH) is the major low molecular weight thiol in cells and controls the thiol-disulfide redox state that is essential for normal redox signaling in vivo. The dynamics of the redox balance of the cell is achieved by maintaining the thiol-to-disulfide state of reduced GSH and its oxidized form GSSG. The value of the reduction potential can be measured as in Rost and Rapoport, Nature 201: 185 (1964) and Aslund et al., J. Biol. Chem. 272: 30780-30786 (1997). The redox potential E, which takes into account the stoichiometry of two GSHs oxidized per one GSSG, is h is a quantitative measure of the redox state. E h is the Nernst equation: E h = E O + (RT / nF)ln ([GSSG (ox)] / [GSH (red)] 2 ), where Eo is the standard potential of the redox couple at a defined pH, R is the gas constant, T is absolute temperature, F is the Faraday constant, and n is the number of electrons transferred. In vivo estimates for the Eh of the GSH / GSSG couple range from -260 to -200 mV (Aw, T., News Physiol. Sci. 18:201-204 (2003)). Terminally differentiated cells thus maintain an Eh of approximately -200 mV, while actively proliferating cells maintain a further reduced Eh of approximately -260 mV.

[0111] The standard redox potential of DTT is -330 mV (Cleland et al. Biochemistry 3: 480-482 (1964)). TCEP has been shown to reduce DTT in solution and therefore has a more negative redox potential than DTT. However, precise values ​​have not been reported. The reducing conditions that allow for controlled Fab arm exchange conditions can therefore be described in terms of a required redox potential Eh, which is optimally below the value achieved under normal plasma conditions in vivo and above the redox potential that reduces antibody disulfide bonds other than those located in the hinge region and involved in the formation of the inter-heavy chain disulfide bonds.

[0112] Thus, in a further embodiment, step c) is carried out under reducing conditions with a redox potential below -50 mV, such as below -150 mV, preferably ranging from -150 to -600 mV, such as from -200 to -500 mV, more preferably from -250 to -450 mV, such as from -250 to -400 mV, even more preferably from -260 to -300 mV.

[0113] In a further embodiment, step c) comprises incubation for at least 90 minutes in the presence of at least 25 mM 2-mercaptoethylamine or in the presence of at least 0.5 mM dithiothreitol at a temperature of at least 20° C. The incubation can be carried out at a pH of 5 to 8, for example at pH 7.0 or at pH 7.4.

[0114] In a further embodiment, step d) comprises restoring the conditions to non-reducing or less reducing conditions, for example by removal of the reducing agent, for example by desalting.

[0115] In some embodiments, the methods of the invention result in an antibody product in which more than 80%, such as more than 90%, such as more than 95%, such as more than 99% of the antibody molecules are the desired bispecific antibody.

[0116] Post-production techniques are more flexible and easier to control than prior art methods based on co-expression.

[0117] The post-production nature of the technique of generating bispecific antibodies by Fab exchange under reducing conditions (such as by adding 2-MEA) disclosed herein makes this a highly suitable strategy for (high throughput) screening of multiple combinations of specificities for bispecific antibody discovery. Furthermore, the in vitro steps can be performed in the laboratory, allowing for higher control, flexibility and yield of heterodimeric proteins than is possible by co-expression. An additional advantage of this strategy is that screening can be performed in the final therapeutic format, eliminating the need for genetic engineering during lead selection.

[0118] As explained above, in a further aspect, the method of the present invention can be used for "matrix" screening, i.e. to generate a large number of different combinations of binding specificities based on two sets of antibodies, one set having an identical first CH3 region and another set having an identical second CH3 region, the sequences of the first and second CH3 regions being different such that the heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region.

[0119] Thus, in one embodiment, the present invention relates to a method for the selection of heterodimeric proteins with desired properties, said method comprising the steps of: a) providing a first set of homodimeric proteins comprising an Fc region, wherein the homodimeric proteins have an identical first CH3 region; b) providing a second set of homodimeric proteins comprising an Fc region, wherein the homodimeric proteins have an identical second CH3 region; wherein the sequences of the first and second CH3 regions are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; c) incubating the combination of said first set and said second set of homodimeric proteins under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization, thus generating a set of bispecific antibodies; d) optionally returning the conditions to non-reducing conditions; e) assaying the resulting set of heterodimeric proteins for a given desired property; and f) selecting heterodimeric proteins with desired properties.

[0120] In one embodiment, the present invention relates to a method for the selection of bispecific antibodies with desired properties, said method comprising the steps of: a) providing a first set of homodimeric antibodies comprising antibodies with different variable regions, wherein the antibodies of the first set comprise an identical first CH3 region; b) providing a second set of homodimeric antibodies comprising antibodies with different variable regions or identical variable regions, wherein the antibodies of the second set comprise an identical second CH3 region; wherein the sequences of the first and second CH3 regions are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; c) incubating the combination of said first set and said second set of antibodies under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization, thus generating a set of bispecific antibodies; d) optionally returning the conditions to non-reducing conditions; e) assaying the resulting set of bispecific antibodies for a given desired property; and f) selecting bispecific antibodies with desired properties.

[0121] In one embodiment, the second set of homodimeric antibodies have different variable regions.

[0122] In one embodiment, the second set of homodimeric antibodies have identical variable regions but different amino acids or structural changes outside the antigen binding region.

[0123] The two sets can be configured in many different ways as required. Thus, the two sets can target the same epitope or different epitopes on the same antigen. The two sets can also target different antigens, and each set can include antibodies that bind to the same epitope or different epitopes on the antigen in question. Furthermore, one or both sets can each include antibodies that target different antigens.

[0124] In another embodiment, the desired property is cell killing, cell lysis, inhibition of cell proliferation, or binding to cells expressing both antigen targets.

[0125] The screening strategy involves two panels of antibody vectors encompassing a range of specificities, where one panel is cloned into a backbone that can participate in Fab-arm exchange under reducing conditions (such as by addition of 2-MEA) with the backbone of the other panel, e.g., the first panel is cloned into an IgG1-F405L backbone and the second panel is cloned into an IgG1-K409R backbone (see also Examples 19, 28, 29, 30, 35, 36, 37, 38, and 39 for other possible backbone combinations).

[0126] Each member of the two panels of antibody vectors is then expressed individually at small scale. For example, all antibody vectors are transiently transfected into HEK293 cells and expressed in 2.3 mL cultures in 24-well plates. Alternatively, other suitable (small scale) production systems known in the art can be used.

[0127] The expressed antibodies of the two antibody panels are then mixed pairwise in equimolar ratios. For example, all individual antibodies are purified by small-scale Protein A chromatography and the antibody concentration is measured by absorbance at 280 nm wavelength. Alternatively, other suitable (small-scale) purification methods for determining protein concentration known in the art can be used. In another embodiment, if the expression medium does not affect downstream applications, the purification step can be omitted. The antibody concentrations are then normalized so that the appropriate volume contains equimolar amounts of both antibodies. For example, a panel of 8 antibodies in the F405L backbone is mixed separately with 8 antibodies in the K409R backbone so that a 100 μl 64-mix contains 80 μg / mL antibody A (F405L) and 80 μg / mL antibody B (K409R). Alternatively, the step of normalizing the antibody amounts can be omitted if the strategy includes bispecific antibody-specific downstream purification steps.

[0128] The mixture of antibodies is added with an appropriate amount of reducing agent and incubated at a permissive temperature for an appropriate time. For example, 100 μl containing 80 μg / mL of antibody A (F405L) and 80 μg / mL of antibody B (K409R) is added with 25 μl of 125 mM 2-MEA (final concentration of 25 mM 2-MEA) and incubated overnight at 25° C.

[0129] The reducing agent is then immediately removed from the mixture (now containing the bispecific antibody) to facilitate oxidation of disulfide bonds and to avoid interference of the reducing agent with the screening assay. For example, 2-MEA is removed by buffer exchange of the 64 mixtures using Zeba Spin 96-well desalting plates (Pierce Biotechnology, #89807). Alternatively, other methods suitable for removing reducing agents known in the art can be used.

[0130] The bispecific antibodies are then characterized biochemically or functionally to identify lead candidates. For example, the 64 bispecific antibodies are assayed for inhibition of growth of or binding to a suitable cell line. The identified lead candidates are then produced on a larger scale and characterized in more detail.

[0131] Production by co-expression The heterodimeric protein of the invention can also be obtained by co-expression in a single cell of constructs encoding the first and second polypeptides.

[0132] Thus, in a further aspect, the present invention relates to a method for producing a heterodimeric protein comprising the steps of: a) providing a first nucleic acid construct encoding a first polypeptide comprising a first Fc region of an immunoglobulin, wherein the first Fc region comprises a first CH3 region; b) providing a second nucleic acid construct encoding a second polypeptide comprising a second Fc region of an immunoglobulin, wherein the second Fc region comprises a second CH3 region; wherein the sequences of the first and second CH3 regions are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; and wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407; and / or wherein the sequences of the first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, for example 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, for example 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, for example 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar; c) co-expressing the first and second nucleic acid constructs in a host cell; and d) Obtaining said heterodimeric protein from the cell culture.

[0133] Suitable expression vectors, including promoters, enhancers, etc., and host cells suitable for producing antibodies are well known in the art. Examples of host cells include yeast cells, bacterial cells and mammalian cells, such as CHO or HEK cells.

[0134] In one embodiment of this method, the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe; and / or The sequences of the first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, for example 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, for example 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, for example 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar.

[0135] In another aspect of this method: the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe, such as an amino acid at position 405 other than Phe, Arg or Gly; or The first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr.

[0136] In some embodiments, the first and second polypeptides are full-length heavy chains of two antibodies that bind to different epitopes (i.e., the first and second nucleic acid constructs encode full-length heavy chains of two antibodies that bind to different epitopes), and thus the heterodimeric protein is a bispecific antibody. The bispecific antibody can be a heavy chain antibody, or the host cell may further express one or more nucleic acid constructs encoding a light chain. If only one light chain construct is co-expressed with the heavy chain construct, a functional bispecific antibody will only be formed if the light chain sequence is such that it can form a functional antigen-binding domain with each of the heavy chains. If two or more different light chain constructs are co-expressed with the heavy chain, multiple products will be formed.

[0137] In a further embodiment, the co-expression method according to the invention comprises any of the additional features described under the in vitro methods above.

[0138] In a further aspect, the present invention relates to an expression vector comprising the first and second nucleic acid constructs as specified herein above. In a further aspect, the present invention relates to a host cell comprising the first and second nucleic acid constructs as specified herein above.

[0139] Heterodimeric Proteins In a further aspect, the present invention relates to a heterodimeric protein obtained or obtainable by the method of the present invention.

[0140] Furthermore, the methods of the invention allow for the formation of asymmetric molecules, molecules with different characteristics for each of the Fab arms or for each of the CH3 domains, or molecules with different modifications throughout the molecule, such as molecules with non-natural amino acid substitutions for binding. Such asymmetric molecules can be created in any suitable combination. This is further illustrated below by some non-limiting examples.

[0141] Bispecific antibodies can be used to pretarget target cells of interest, including but not limited to tumor cells. Pretargeting of target cells could be used for imaging studies or for immunotherapeutic purposes.

[0142] In one embodiment of the method of the invention, the first Fab arm of the bispecific molecule binds to a tumor cell, such as a tumor cell surface protein or tumor cell surface glycan, such as one of the tumor cell surface proteins described herein, and the second Fab arm recognizes a radioactive effector molecule, including but not limited to a radiolabel coupled or linked (via a chelator) to a peptide or hapten. One example of such a radiolabeled peptide is indium-labeled diethylenetriaminepentaacetic acid (anti-DTPA(In) van Schaijk et al. Clin. Cancer Res. 2005; 11 : 7230s-7126s). Another example is the use of hapten-labeled colloidal particles, such as nanoparticles of polymeric micelles, liposomes, carrying a radionuclide, such as technetium-99 (Jestin et al. QJ Nucl Med Mol Imaging 2007; 51:51-60).

[0143] In another embodiment, another cytostatic molecule coupled to a hapten, such as a toxin, is used.

[0144] In a further embodiment of the method of the invention, the first Fab arm of the bispecific molecule is glycosylated at position N297 (EU numbering) and the second Fab arm of the bispecific molecule is not glycosylated (e.g. aglycosylated by mutating N297 to a Q or A or E mutation (Bolt S et al., Eur J Immunol 1993, 23: 403-411)). Asymmetric glycosylation in the Fc region affects the interaction with Fcγ-receptors and affects the interaction with other effector function molecules such as C1q, as well as the antibody-dependent cellular cytotoxicity of the antibody (Ha et al., Glycobiology 2011, April 5).

[0145] In another embodiment of the methods of the invention, the first Fab arm of the bispecific molecule interacts with the neonatal Fc receptor FcRn (Roopenian DC, et al. Nat. Rev. Immunol. 2007, 7: 715-725) and the second Fab arm is impaired in binding to FcRn by mutation of the FcRn interaction site on the molecule, for example by creating an H435A mutation (Shields, RL, et al, J Biol Chem, 2001; Firan, M., et al, Int Immunol, 2001).

[0146] In another embodiment of the method of the invention, the first Fab arm of the bispecific molecule interacts with staphylococcal protein A (Protein A, Deisenhofer et al, Biochemistry 20, 2361-2370 (1981) and streptococcal protein G (Protein G, Derrick et al., Nature 359, 752-754 (1992), which are frequently used for antibody purification, and the second Fab arm of the bispecific molecule is impaired in interacting with Protein A or G. As a result, removal of the amount of impaired Protein A or G binding homodimers remaining after exchange into heterodimers can be easily obtained by purification of the bispecific molecule with Protein A or G.

[0147] In another embodiment, binding to either Fcγ-receptors or FcRn is improved or reduced in one of the two Fab arms of the bispecific molecule.

[0148] In another embodiment, binding to C1q is improved or reduced in one of the two Fab arms of the bispecific molecule.

[0149] In another embodiment, the protein is engineered to enhance complement activation in one or both of the two Fab arms of the molecule.

[0150] In another embodiment, each of the Fab arms present in the bispecific molecule is derived from a different IgG subclass.

[0151] In another embodiment, each of the Fab arms present in the bispecific molecule carries a different allotypic mutation (Jefferis & Lefranc, 2009, MABs 1 :332-8).

[0152] In another embodiment, another class of asymmetric immunotherapeutic molecules is created by replacing one of the Fab arms of the bispecific molecule with an immune activating, stimulating or inhibitory cytokine. Non-limiting examples of such cytokines are IL-2, IFN-α, IFN-β, IFN-γ, TNF-α, G-CSF, GM-CSF, IL-10, IL-4, IL-6, IL-13. Alternatively, (growth) factors or hormone stimulators or inhibitors are included in the molecule.

[0153] In another embodiment, one Fab of the Fab arm is replaced by a cytolytic peptide, i.e., a peptide capable of lysing tumor cells, bacteria, fungi, etc., including but not limited to antimicrobial peptides such as magainin, melittin, cecropin, KLAKKLAK and variants thereof (Schweizer et al. Eur. J. Pharmacology 2009; 625: 190-194; Javadpour, J. Med. Chem., 1996, 39: 3107-3113; Marks et al, Cancer Res 2005; 65:2373-2377; Rege et al, Cancer Res. 2007; 67:6368-6375) or a cationic cytolytic peptide (CLYP technology, US2009 / 0269341).

[0154] In another embodiment, one or both of the Fabs of the Fab arms are replaced by receptors for cytokines and / or growth factors to create so-called decoy receptors, of which Enbrel® (etanercept) targeting TNF-α and VEGF-trap targeting VEGF are well-known examples. Combining these two decoy receptors into one molecule has shown superior activity compared to single decoy receptors (Jung, J.Biol. Chem. 2011; 286: 14410-14418).

[0155] In another embodiment, another class of asymmetric immunotherapeutic molecules is created by fusing immune active, stimulatory or inhibitory cytokines to the N-terminus or C-terminus of one or both of the Fab arms present in the bispecific molecule. This can positively affect the antitumor activity of the bispecific molecule. However, examples of such molecules, including but not limited to the following list, are IL-2 (Fournier et al., 2011, Int. J. Oncology, doi: 10.3892 / ijo.2011.976), IFN-α, IFN-β or IFN-γ (Huan et al., 2007; J. Immunol. 179: 6881-6888, Rossie et al., 2009; Blood 114: 3864-3871), TNF-α. Alternatively, N- or C-terminal fusions of cytokines such as G-CSF, GM-CSF, IL-10, IL-4, IL-6 or IL-13 may positively influence bispecific antibody molecule effector functions. Alternatively, (growth) factors or hormone stimulators or inhibitors are included in the molecule at the N- or C-terminus.

[0156] In another embodiment, N- or C-terminal fusion of cytolytic peptides (Schweizer et al. Eur. J. Pharmacology 2009; 625: 190-194; Javadpour, J. Med. Chem., 1996, 39: 3107-3113; Marks et al, Cancer Res 2005; 65: 2373-2377; Rege et al, Cancer Res. 2007; 67: 6368-6375) or cationic cytolytic peptides (CLYP technology, US2009 / 0269341), e.g., antimicrobial peptides such as magainin, melittin, cecropin, KLAKKLAK and variants thereof, on one or both of the Fab arms may enhance the activity of the molecule.

[0157] In another embodiment, another class of asymmetric immunotherapeutic molecules is monovalent antibodies, i.e., molecules that interact with a target of choice with one Fab arm. In such molecules, one of the Fab arms present in the bispecific molecule is directed against a target molecule of choice, and the other Fab arm of the molecule has no Fab or a non-binding / non-functional Fab, as described for MetMab (Genentech; WO 96 / 38557). Alternatively, monomeric Fc fusion proteins, such as those described for Factor VIII and Factor IX (Peters et al., Blood 2010; 115: 2057-2064), can be created.

[0158] Alternatively, combinations of any of the above asymmetric molecules can be generated by the methods of the present invention.

[0159] In a further aspect, the present invention relates to a heterodimeric protein comprising a first polypeptide comprising a first Fc region of an immunoglobulin, the first Fc region comprising a first CH3 region, and a second polypeptide comprising a second Fc region of an immunoglobulin, the second Fc region comprising a second CH3 region, wherein the sequences of the first and second CH3 regions are different such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; and wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407; and / or Here, the sequences of the first and second CH3 regions are such that the dissociation constant of the homodimer interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, for example 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, for example 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, for example 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar.

[0160] In one embodiment, the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe; and / or The sequences of the first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, for example 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, for example 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, for example 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar.

[0161] In a further embodiment of the heterodimeric protein, the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe, such as other than Phe, Arg or Gly; or The first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr.

[0162] In a further embodiment, the heterodimeric protein according to the invention comprises any of the additional features described above for the production method.

[0163] Thus, in a further embodiment of the heterodimeric protein of the present invention, the first polypeptide is a full-length heavy chain of an antibody, preferably a human antibody.

[0164] In another embodiment of the heterodimeric protein of the present invention, the second polypeptide is a full-length heavy chain of an antibody, preferably a human antibody.

[0165] In a further embodiment of the heterodimeric protein of the invention, the first and second polypeptides are both full-length heavy chains of two antibodies, preferably both human antibodies binding to different epitopes, and the resulting heterodimeric protein is thus a bispecific antibody. This bispecific antibody can be a heavy chain antibody or an antibody comprising, in addition to a heavy chain, two full-length light chains, which can be the same or different.

[0166] In a further embodiment of the heterodimeric protein of the invention, the Fc region of the first polypeptide is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4 (except for certain mutations) and the Fc region of the second polypeptide is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4 (except for certain mutations).

[0167] In a further embodiment of the heterodimeric protein of the present invention, the Fc region of both the first and second polypeptides is of the IgG1 isotype.

[0168] In a further embodiment of the heterodimeric protein of the present invention, one of the Fc regions of the polypeptide is of the IgG1 isotype and the other is of the IgG4 isotype.

[0169] In a further embodiment of the heterodimeric protein of the present invention, the increase in strength of the heterodimeric interaction compared to each of the homodimeric interactions is due to CH3 modifications other than covalent bonds, introduction of cysteine ​​residues or charged residues.

[0170] In a further embodiment of the heterodimeric protein of the present invention, the heterodimeric interaction between the first and second polypeptides in the heterodimeric protein is such that Fab arm exchange cannot occur at 0.5 mM GSH under the conditions described in Example 13.

[0171] In a further embodiment of the heterodimeric protein of the invention, the heterodimeric interaction between the first and second polypeptides in the resulting heterodimeric protein is such that no Fab arm exchange occurs in vivo in mice under the conditions described in Example 14.

[0172] In a further embodiment of the heterodimeric protein of the invention, the first CH3 region comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second CH3 region comprises an amino acid other than Phe at position 405 and a Lys at position 409.

[0173] In a further embodiment of the heterodimeric protein of the invention, the first CH3 region comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second CH3 region comprises a Leu at position 405 and a Lys at position 409.

[0174] In a further embodiment of the heterodimeric protein of the present invention, the first CH3 region comprises a Phe at position 405 and an Arg at position 409, and the second CH3 region comprises a Leu at position 405 and a Lys at position 409.

[0175] In a further embodiment of the heterodimeric protein of the invention, the first CH3 region comprises an amino acid other than Lys, Leu or Met at position 409 and the second CH3 region comprises a Lys at position 409 and: a) an Ile at position 350 and a Leu at position 405, or b) a Thr at position 370 and a Leu at position 405.

[0176] In a further embodiment of the heterodimeric protein of the present invention, the first CH3 region comprises an Arg at position 409 and the second CH3 region comprises a Lys at position 409 and: a) an Ile at position 350 and a Leu at position 405, or b) a Thr at position 370 and a Leu at position 405.

[0177] In a further embodiment of the heterodimeric protein of the invention, the first CH3 region comprises a Thr at position 350, a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second CH3 region comprises a Lys at position 409 and: a) an Ile at position 350 and a Leu at position 405, or b) a Thr at position 370 and a Leu at position 405.

[0178] In a further embodiment of the heterodimeric protein of the present invention, the first CH3 region comprises a Thr at position 350, a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second CH3 region comprises an Ile at position 350, a Thr at position 370, a Leu at position 405 and a Lys at position 409.

[0179] In a further embodiment of the heterodimeric protein of the present invention, neither the first nor the second polypeptide comprises a Cys-Pro-Ser-Cys sequence in the hinge region.

[0180] In a further embodiment of the heterodimeric protein of the present invention, both the first and second polypeptides comprise a Cys-Pro-Pro-Cys sequence in the hinge region.

[0181] In a further embodiment of the heterodimeric protein of the invention, the first and / or second polypeptide comprises a mutation that removes an acceptor site for Asn-linked glycosylation.

[0182] target antigen As explained above, in an important embodiment of the present invention, the heterodimeric protein is a bispecific antibody comprising two variable regions with different binding specificities, i.e. binding to different epitopes.

[0183] In principle, any combination of specificity is possible. As mentioned above, bispecific antibodies can potentially be used to overcome some of the limitations of monospecific antibodies. One possible limitation of monospecific antibodies is the lack of specificity for the desired target cells due to the expression of the target antigen on other cell types where antibody binding is not desired. For example, the target antigen that is overexpressed on tumor cells may also be expressed in healthy tissues, which may cause undesirable side effects from treatment with antibodies made against that antigen. Bispecific antibodies with additional specificity for proteins that are exclusively expressed on target cell types could potentially improve specific binding to tumor cells.

[0184] Thus, in one embodiment of the invention, the first and second epitopes are located on the same cell, e.g., a tumor cell. Suitable targets on tumor cells include, but are not limited to, erbB1 (EGFR), erbB2 (HER2), erbB3, erbB4, MUC-1, CD19, CD20, CD4, CD38, CD138, CXCR5, c-Met, HERV-envelope protein, periostin, Bigh3, SPARC, BCR, CD79, CD37, EGFrvIII, L1-CAM, AXL, tissue factor (TF), CD74, EpCAM, and MRP3. Possible combinations of tumor cell targets include, but are not limited to, erbB1 + erbB2, erbB2 + erbB3, erbB1 + erbB3, CD19 + CD20, CD38 + CD34, CD4 + CXCR5, CD38 + RANKL, CD38 + CXCR4, CD20 + CXCR4, CD20 + CCR7, CD20 + CXCR5, CD20 + RANKL, erbB2 + AXL, erbB1 + cMet, erbB2 + c-Met, erbB2 + EpCAM, c-Met + AXL, c-Met + TF, CD38 + CD20, CD38 + CD138.

[0185] In a further embodiment, the first and second epitopes can be located on the same target antigen, where the location of the two epitopes on the target antigen is such that the binding of the antibody to one epitope does not interfere with the binding of the antibody to the other epitope. In this further embodiment, the first and second homodimeric proteins are antibodies that bind to two different epitopes located on the same target antigen, but have different mechanisms of action for killing target cells, such as tumor cells. For example, in one embodiment, the target antigen is erbB2 (HER2), and the bispecific antibody combines the antigen-binding sites of pertuzumab and trastuzumab. In another embodiment, the target antigen is erbB1 (EGFr), and the bispecific antibody combines the antigen-binding sites of zalutumumab and nimotuzumab.

[0186] Bispecific antibodies can also be used as mediators to retarget effector mechanisms to disease-related tissues, such as tumors. Thus, in a further embodiment, the first or second epitope is located on a tumor cell, such as a tumor cell protein or tumor cell glycan, and the other epitope is located on an effector cell.

[0187] In one embodiment, the effector cell is a T cell.

[0188] Possible targets on effector cells include: FcgammaRI (CD64), expressed on monocytes and macrophages and activated neutrophils; FcgammaRIII (CD16), expressed on natural killer and macrophages; CD3, expressed on circulating T cells; CD89, expressed on PMNs (polymorphonuclear neutrophils), eosinophils, monocytes and macrophages; CD32a, expressed on macrophages, neutrophils, eosinophils; FcεRI, expressed on basophils and mast cells. In one embodiment, the epitope is located on CD3 expressed on T cells.

[0189] In another embodiment, a first antibody has binding specificity for a pathogenic microorganism and a second antibody has binding specificity for an effector cell protein, such as CD3, CD4, CD8, CD40, CD25, CD28, CD16, CD89, CD32, CD64, FcεRI, or CD1.

[0190] Furthermore, bispecific antibodies can be used to more specifically target chemotherapeutic agents to the cells on which they should act. Thus, in one embodiment, one of the homodimeric proteins is an antibody that recognizes a small molecule or peptide or can form a covalent bond with such a molecule, for example, according to the principles described in Rader et al, (2003) PNAS 100: 5396. In a further embodiment of the method of the present invention, the first antibody has binding specificity for (i.e., binds to an epitope of) a tumor cell or tumor cell surface protein, such as erbB1, erbB2, erbB3, erbB4, EGFR3vIII, CEA, MUC-1, CD19, CD20, CD4, CD38, EPCAM, c-Met, AXL, L1-CAM, tissue factor, CD74 or CXCR5, and the second antibody has binding specificity for a chemotherapeutic agent, such as a toxin (including a radiolabeled peptide), a drug or a prodrug.

[0191] Bispecific antibodies can also be used to target vesicles, such as electron-dense vesicles or minicells, containing toxins, drugs or prodrugs to tumors. See, e.g., MacDiarmid et al. (2009) Nature Biotech 27:643. Minicells are achromosomal cells that do not contain chromosomal DNA and are the product of abnormal cell division. Thus, in another embodiment, the first or second epitope is located on a tumor cell, such as a tumor cell protein or tumor cell carbohydrate, and the other epitope is located on the electron-dense vesicle or minicell.

[0192] Furthermore, the serum half-life of the antibody can be modified by including a binding specificity for a serum protein in the bispecific antibody. For example, the serum half-life can be extended by including a binding specificity for serum albumin in the bispecific antibody. Thus, in a further embodiment of the method of the present invention, the first antibody has binding specificity for a tumor cell or tumor cell protein, such as erbB1 (EGFR), erbB2 (HER2), erbB3, erbB4, MUC-1, CD19, CD20, CD4, CD38, CD138, CXCR5, c-Met, HERV-envelope protein, periostin, Bigh3, SPARC, BCR, CD79, CD37, EGFrvIII, L1-CAM, AXL, tissue factor (TF), CD74, EpCAM or MRP3, CEA, and the second antibody has binding specificity for a blood protein, such as serum albumin. The second binding specificity can also be used to target the antibody to a particular tissue, such as the central nervous system or brain (across the blood-brain barrier). Thus, in a further embodiment of the method of the invention: No. 1 the antibodies have binding specificity for brain specific targets, such as amyloid-β (e.g. for the treatment of Alzheimer's disease), Her-2 (e.g. for the treatment of breast cancer metastases in the brain), EGFr (e.g. for the treatment of primary brain tumors), Nogo A (e.g. for the treatment of brain injury), TRAIL (e.g. for the treatment of HIV), α-synuclein (e.g. for the treatment of Parkinson's), Htt (e.g. for the treatment of Huntington's), prion (e.g. for the treatment of Mad Cow Disease), West Nile Virus proteins, SecondThe antibodies have binding specificity for blood-brain barrier proteins such as transferrin receptor (TfR), insulin receptor, melanotransferrin receptor (MTfR), lactoferrin receptor (LfR), apolipoprotein E receptor 2 (ApoER2), LDL-receptor-related proteins 1 and 2 (LRP1 and LRP2), receptor for advanced glycation end products (RAGE), diphtheria toxin-receptor = heparin-binding epidermal growth factor-like growth factor (DTR = HB-EGF), gp190 (Abbott et al, Neurobiology of Disease 37 (2010) 13-25).

[0193] The binding specificity for blood-brain barrier proteins can also be used to target other non-antibody molecules to specific tissues, such as the central nervous system or the brain (across the blood-brain barrier). Thus, in a further embodiment, one of the homodimeric proteins is a full-length antibody with binding specificity for a blood-brain barrier protein (such as TfR, insulin receptor, MTfR, LfR, ApoER2, LRP1, LRP2, RAGE, DTR (= HB-EGF) or gp190), and the other homodimeric protein is a cytokine, soluble receptor or other protein, such as VIP (vasoactive intestinal peptide), BDNF (brain-derived neurotrophic factor), FGF (fibroblast growth factor), multiple FGFs, EGF (epidermal growth factor), PNA (peptide nucleic acid), NGF (peptide-derived neurotrophic factor), or the like. The Fc region is linked at the N-terminus or C-terminus to another protein, such as nerve growth factor, neurotrophin (NT)-3, NT-4 / 5, glial-derived neurotrophic factor, ciliary neurotrophic factor, neurturin, neuregulin, interleukins, transforming growth factor (TGF)-α, TGF-β, erythropoietin, hepatocyte growth factor, platelet-derived growth factor, artemin, persephin, netrin, cardiotrophin-1, stem cell factor, midkine, pleiotrophin, bone morphogenetic protein, saposin, semaphorin, leukocyte inhibitory factor, α-L-iduronidase, iduronate-2-sulfatase, N-acetyl-galactosamine-6-sulfatase, arylsulfatase B, acid α-glucosidase, or sphingomyelinase (Pardridge, Biopharmaceutical drug targeting to the brain, Journal of Drug Targeting 2010, 1-11; Pardridge, Re-engineering Biopharmaceuticals for delivery to brain with molecular Trojan horses. Bioconjugate Chemistry 2008, 19: 1327-1338).

[0194] Furthermore, the second binding specificity can be used to target the blood clotting factor to a specific desired site of action. For example, a bispecific antibody having a first binding specificity for a tumor cell and a second binding specificity for a blood clotting factor could target blood clotting to the tumor and thus stop tumor growth. Thus, in a further embodiment of the method of the present invention, the first antibody has binding specificity for a tumor cell or a tumor cell protein, such as erbB1, erbB2, erbB3, erbB4, MUC-1, CD19, CD20, CD4 or CXCR5, and the second antibody has binding specificity for a protein involved in blood clotting, such as tissue factor.

[0195] Further particularly interesting binding specificity combinations include CD3 + HER2, CD3 + CD20, IL-12 + IL18, IL-1a + IL-1b, VEGF + EGFR, EpCAM + CD3, GD2 + CD3, GD3 + CD3, HER2 + CD64, EGFR + CD64, CD30 + CD16, NG2 + CD28, HER2 + HER3, CD20 + CD28, HER2 + CD16, Bcl2 + CD3, CD19 + CD3, CEA + CD3, EGFR + CD3, IgE + CD3, EphA2 + CD3, CD33 + CD3, MCSP + CD3, PSMA + CD3, TF + CD3, CD19 + CD16, CD19 + CD16a, CD30 + CD16a, CEA + HSG, CD20 + HSG, MUC1 + HSG, CD20 + CD22, HLA-DR + These include CD79, PDGFR + VEGF, IL17a + IL23, CD32b + CD25, CD20 + CD38, HER2 + AXL, CD89 + HLA class II, CD38+CD138, TF + cMet, Her2 + EpCAM, HER2 + HER2, EGFR + EGFR, EGFR + c-Met, c-Met + non-binding arm and G protein-coupled receptor combinations.

[0196] In a further embodiment, bispecific antibodies according to the invention can be used to remove pathogens, pathogenic autoantibodies or harmful compounds, such as venoms and toxins, from the circulating blood by targeting to red blood cells essentially as described in Taylor et al. J. Immunol. 158: 842-850 (1997) and Taylor and Ferguson, J. Hematother. 4: 357-362, 1995. The first epitope is located on a red blood cell protein, including but not limited to red blood cell complement receptor 1, and the second epitope is located on the compound or organism targeted for removal.

[0197] In a further embodiment, the second Fab arm comprises a fusion protein that targets an autoantigen or binding site to which an autoantigen such as dsDNA is attached. Targeting of pathogens, autoantibodies or harmful compounds by the bispecific antibodies of the invention and their subsequent clearance via red blood cells may thus have therapeutic utility in the treatment of a variety of diseases and syndromes.

[0198] join In a further embodiment of the present invention, the first and / or second homodimeric protein is linked to a compound selected from the group consisting of toxins (including radioisotopes), prodrugs or drugs. Such compounds can further enhance the killing of target cells, for example in cancer therapy. The resulting heterodimeric protein is thus an immunoconjugate. Alternatively, a compound may be coupled to the resulting heterodimeric protein, i.e. after Fab arm exchange has taken place.

[0199] Compounds suitable for forming the immunoconjugates of the present invention include taxol, cytochalasin B, gramicidin D, ethidium bromide, emetine, mitomycin, etoposide, tenoposide, vincristine, vinblastine, colchicine, doxorubicin, daunorubicin, dihydroxyanthracin dione, mitoxantrone, mithramycin, actinomycin D, 1-dehydro-testosterone, glucocorticoids, procaine, tetracaine, lidocaine, propranolol, tetracaine, tetrahydrocannabinol, tetrahydrofuran ... nolol, and puromycin, antimetabolites (e.g., methotrexate, 6-mercaptopurine, 6-thioguanine, cytarabine, fludarabine, 5-fluorouracil, decarbazine, hydroxyurea, asparaginase, gemcitabine, cladribine), alkylating agents (e.g., mechlorethamine, thioepa, chlorambucil, melphalan, carmustine (BSNU), lomustine (CCNU), cyclophosphamide, busulfan, dibromomannitol, Streptozotocin, dacarbazine (DTIC), procarbazine, mitomycin C, cisplatin and other platinum derivatives, e.g., carboplatin), antibiotics (e.g., dactinomycin (formerly actinomycin), bleomycin, daunorubicin (formerly daunomycin), doxorubicin, idarubicin, mithramycin, mitomycin, mitoxantrone, plicamycin, anthramycin (AMC)), diphtheria toxin and related molecules (e.g., diphtheria toxin ... phtheria A chain and its active fragments and hybrid molecules), ricin toxin (e.g., ricin A or deglycosylated ricin A chain toxin), cholera toxin, Shiga-like toxins (SLT-I, SLT-II, SLT-IIV), LT toxin, C3 toxin, Shiga toxin, pertussis toxin, tetanus toxin, soybean Bowman-Birk protease inhibitor, Pseudomonas aeruginosa exotoxin, allorin, saporin, modeccin, geranin, abrin A chain, modeccin A chain, α-sarcin, Aleurites fordii proteins, dianthin proteins, Phytolacca americana proteins (PAPI, PAPII, and PAP-S), momordica charantia inhibitor, curcin, crotin, sapaonaria officinalisofficinalis inhibitors, xeronine, mitogenin, restrictocin, phenomycin, and enomycin toxin. Other suitable binding molecules include ribonuclease (RNase), DNase I, Staphylococcal enterotoxin-A, pokeweed antiviral protein, diphtherin toxin, Pseudomonas aeruginosa endotoxin, maytansinoids, auristatins (MMAE, MMAF), calicheamicin and duocarmycin analogs (Ducry and Stump, Bioconjugate Chem. 2010, 21: 5-13), dolastatin-10, dolastatin-15, irinotecan or its active metabolite SN38, pyrrolobenzodiazepines (PBD's).

[0200] In a further embodiment of the invention, the first and / or second homodimeric protein are linked to an alpha emitter, including but not limited to thorium-227, radium-223, bismuth-212 and actinium-225.

[0201] In a further embodiment of the invention, the first and / or second homodimeric protein are linked to a beta-emitting radionuclide, including but not limited to, iodium-313, yttrium-90, fluorine-18, rhenium-186, gallium-68, technetium-99, indium-111, and lutetium-177.

[0202] In another embodiment, the conjugated compound comprises a nucleic acid or a nucleic acid-associated molecule. In one such aspect of the invention, the conjugated nucleic acid is a cytotoxic ribonuclease, an antisense nucleic acid, an inhibitory RNA molecule (e.g., an siRNA molecule) or an immunostimulatory nucleic acid (e.g., an immunostimulatory CpG motif-containing DNA molecule).

[0203] Hunter et al., Nature 144 , 945 (1962), David et al., Biochemistry 13, 1014 (1974), Pain et al., J. Immunol. Meth. 40, 219 (1981) and Nygren, J. Histochem. and Cytochem. 30 Any method known in the art for conjugation can be utilized, including the method described by Ducry and Stump (2010) Bioconjugate Chem., 407 (1982). Conjugates can be made by chemically linking the other moiety to the N- or C-terminus of the protein (see, for example, Antibody Engineering Handbook, edited by Osamu Kanemitsu and published by Chijin Shokan (1994)). Such conjugated antibody derivatives may be made by conjugation at internal residues or sugars, as appropriate. Drugs may be coupled directly or indirectly to the proteins of the invention. An example of indirect coupling of a second drug is coupling through a spacer moiety. Drug-conjugate conjugation techniques have recently been described in Ducry and Stump (2010) Bioconjugate Chem. 21 :5 is summarised by:

[0204] Compositions and Uses In a further main aspect, the present invention relates to a pharmaceutical composition comprising the heterodimeric protein according to the invention as described herein and a pharma- ceutically acceptable carrier.

[0205] Pharmaceutical compositions can be formulated according to conventional techniques, such as those disclosed in Remington: The Science and Practice of Pharmacy, 19th Edition, Gennaro, Ed., Mack Publishing Co., Easton, PA, 1995. Pharmaceutical compositions of the invention can include, for example, diluents, bulking agents, salts, buffers, surfactants (e.g., non-ionic surfactants such as Tween-20 or Tween-80), stabilizers (e.g., sugars or protein-free amino acids), preservatives, tissue fixatives, solubilizing agents, and / or other materials suitable for inclusion in a pharmaceutical composition.

[0206] Pharmaceutically acceptable carriers include any and all suitable solvents, dispersion media, coatings, antibacterial and antifungal agents, isotonic agents, antioxidants, and absorption delaying agents that are physiologically compatible with the compounds of the present invention. Examples of suitable aqueous and non-aqueous carriers that can be used in the pharmaceutical compositions of the present invention include water, saline, phosphate buffered saline, ethanol, dextrose, polyols (e.g., glycerol, propylene glycol, polyethylene glycol). Pharmaceutically acceptable carriers include sterile aqueous solutions or dispersions and sterile powders for the extemporaneous preparation of sterile injectable solutions or dispersions. Proper fluidity may be maintained by the use of coating materials, such as lecithin, by maintaining the required particle size in the case of dispersions, and by the use of surfactants.

[0207] The pharmaceutical compositions of the present invention may also contain pharma- ceutically acceptable antioxidants, for example: (1) water-soluble antioxidants such as ascorbic acid, cysteine ​​hydrochloride, sodium bisulfate, sodium metabisulfite, sodium bisulfite, and the like; (2) oil-soluble antioxidants such as ascorbyl palmitate, butylated hydroxyanisole, butylated hydroxytoluene, lecithin, propyl gallate, α-tocopherol, and the like; and (3) metal chelators such as citric acid, ethylenediaminetetraacetic acid (EDTA), sorbitol, tartaric acid, phosphoric acid, and the like.

[0208] The pharmaceutical compositions of the present invention may also contain isotonic agents, such as sugars, polyalcohols such as mannitol, sorbitol, glycerol, or sodium chloride in the composition.

[0209] The pharmaceutical composition of the present invention can also contain one or more adjuvants suitable for the selected route of administration, such as preservatives, wetting agents, emulsifiers, dispersing agents, preservatives or buffers, which can enhance the shelf life or effectiveness of the pharmaceutical composition.The compound of the present invention can be prepared with a carrier that prevents the compound from being released quickly, such as a controlled release formulation, including implants, transdermal patches and microencapsulated delivery systems.Such carriers can include biodegradable, biocompatible polymers, such as gelatin, glyceryl monostearate, glyceryl distearate, ethylene vinyl acetate alone or with wax, polyanhydrides, polyglycolic acid, collagen, polyorthoesters and polylactic acid, or other materials well known in the art.Methods for preparing such formulations are generally known to those skilled in the art.

[0210] Sterile injectable solutions can be prepared by incorporating the active compound in a required amount in an appropriate solvent with one or a combination of ingredients, e.g., as enumerated above, as required, followed by microfilter sterilization.

[0211] The actual dosage level of the active ingredient in the pharmaceutical composition can be varied to obtain an amount of the active ingredient effective to achieve the therapeutic response desired for a particular patient, composition, and mode of administration without being toxic to the patient.The dosage level selected will depend on various pharmacokinetic factors, including the activity of the particular composition of the present invention used, the route of administration, the time of administration, the excretion rate of the particular compound used, the duration of treatment, other drugs, compounds and / or materials used in combination with the particular composition used, the age, sex, weight, condition, general health and past medical history of the patient being treated, and similar factors well known in the medical field.

[0212] Pharmaceutical compositions can be administered by any suitable route and mode.In one embodiment, pharmaceutical compositions of the present invention are administered parenterally.As used herein, "administered parenterally" refers to the mode of administration other than enteral and topical administration, usually by injection, including epidermal, intravenous, intramuscular, intraarterial, intrathecal, intracapsular, intraorbital, intracardiac, intradermal, intraperitoneal, intratendinous, transtracheal, subcutaneous, subcuticular, intraarticular, subcapsular, subarachnoid, intraspinal, intracranial, intrathoracic, epidural and intrasternal injection and infusion.

[0213] In one embodiment, the pharmaceutical composition is administered by intravenous or subcutaneous injection or infusion.

[0214] In a main aspect, the present invention relates to a heterodimeric protein according to the present invention, such as a bispecific antibody according to the present invention, for use as a medicament. The heterodimeric protein according to the present invention can be used for several purposes. In particular, as explained above, the heterodimeric protein according to the present invention can be used for the treatment of various forms of cancer, including metastatic cancer and refractory cancer.

[0215] Thus, in one aspect, the present invention relates to a method for inhibiting the growth and / or proliferation of tumor cells and / or for killing tumor cells comprising the administration of a heterodimeric protein according to the invention as described herein to an individual in need thereof.

[0216] In another embodiment, the heterodimeric proteins of the present invention are used for the treatment of immune and autoimmune diseases, inflammatory diseases, infectious diseases, cardiovascular diseases, CNS and musculoskeletal diseases.

[0217] Dosage regimens in the above methods of treatment and use are adjusted to provide the optimum desired response (e.g., a therapeutic response). For example, a single bolus may be administered, several divided doses may be administered over time, or the dose may be proportionally reduced or increased as indicated by the exigencies of the therapeutic situation.

[0218] Efficient doses and dosage regimens for the heterodimeric protein depend on the disease or condition being treated and can be determined by one of skill in the art. Exemplary, non-limiting ranges for a therapeutically effective amount of a bispecific antibody of the invention are about 0.1-100 mg / kg, such as about 0.1-50 mg / kg, such as about 0.1-20 mg / kg, such as about 0.1-10 mg / kg, such as about 0.5 mg / kg, such as about 0.3 mg / kg, about 1 mg / kg, about 3 mg / kg, about 5 mg / kg, or about 8 mg / kg.

[0219] A physician or veterinarian having ordinary skill in the art can easily determine and prescribe the effective amount of pharmaceutical composition required.For example, a physician or veterinarian can start the dosage of the heterodimeric protein utilized in the pharmaceutical composition at a level lower than that required to achieve the desired therapeutic effect, and gradually increase the dosage until the desired effect is achieved.In general, the appropriate daily dosage of the composition of the present invention will be the amount of the compound that is the minimum dosage effective to produce a therapeutic effect.Administration can be parenteral, for example, intravenous, intramuscular or subcutaneous.

[0220] The heterodimeric proteins of the present invention can also be administered prophylactically to reduce the risk of developing a disease, such as cancer, to delay the onset of occurrence of events in the disease progression, and / or to reduce the risk of recurrence when a disease, such as cancer, is in remission.

[0221] The heterodimeric protein, such as the bispecific antibody of the present invention, can also be administered in combination therapy, i.e., combined with other therapeutic agents suitable for the disease or condition to be treated. Thus, in one embodiment, the medicament containing the heterodimeric protein is intended for combination with one or more additional therapeutic agents, such as cytotoxic agents, chemotherapeutic agents or angiogenesis inhibitors. Such combined administration can be simultaneous, separate or sequential. In a further embodiment, the present invention provides a method for treating or preventing a disease, such as cancer, comprising administering to a subject in need thereof a therapeutically effective amount of the heterodimeric protein, such as the bispecific antibody of the present invention, in combination with radiation therapy and / or surgery.

[0222] Heterodimeric proteins, such as bispecific antibodies, of the present invention can also be used for diagnostic purposes. EXAMPLES

[0223] Example 1: Expression vectors for the expression of human IgG1-2F8 and IgG1-7D8 The VH and VL coding regions of HuMab 2F8 (WO 02 / 100348) and HuMab 7D8 (WO 04 / 035607) were cloned into the expression vector pConG1f (containing the genomic sequence of the human IgG1f allotype constant region (Lonza Biologics)) for the production of human IgG1 heavy chains and pConKappa (containing the human kappa light chain constant region, Lonza Biologics) for the production of kappa light chains. For IgG4 antibodies, the VH regions were inserted into the pTomG4 vector (containing the genomic sequence of the human IgG4 constant region in the pEE12.4 vector (Lonza Biologics)). Alternatively, in follow-up studies of the constructs, vectors containing fully codon-optimized coding regions for the heavy chain (IgG1 or IgG4) in the pEE12.4 vector or the human kappa light chain of HuMab 2F8 or HuMab 7D8 in the pEE6.4 vector (Lonza Biologics) were used.

[0224] Example 2 Expression vectors for the expression of hinge-deleted IgG1-2F8 and human IgG1 and IgG4 CH2-CH3 fragments containing specific mutations To introduce mutations into the hinge and CH3 regions of the antibody heavy chain, the Quickchange site-directed mutagenesis kit (Stratagene, La Jolla, Calif.) was used according to the manufacturer's recommendations. Alternatively, constructs were fully synthesized or VH regions were cloned into vectors that already contained substitutions encoding specific amino acids.

[0225] Constructs encoding the CH2 and CH3 fragments were constructed by PCR or total synthesis and codon optimized. These constructs had an N-terminal signal peptide and a 6 amino acid His tag and contained amino acids 341-447 of the human IgG1 / 4 constant region. The constructs were cloned into pEE12.4.

[0226] To construct a hinge-deleted IgG1 (Uni-G1) molecule, a synthetic DNA construct was generated encoding the Uni-G1 form of the human IgG1 isotype with EGFR specificity. In this construct, the native hinge region (defined by the hinge exon) was deleted. An additional mutation from Ser to Cys at position 158 was made in the IgG1 construct to take advantage of the Cys bond between the HC and LC chains in this subtype. The protein sequence is shown below. The construct was inserted into the pEE6.4 vector and named pHG1-2F8. TIFF2025063202000004.tif49170

[0227] Example 3: Expression vectors for the expression of rhesus IgG4-2F8 and IgG4-7D8 Vectors containing the coding regions for Chinese rhesus IgG4 heavy and kappa light chains and the VH and VL regions of Humab 2F8 and 7D8 were synthesized, fully codon-optimized, and inserted into pEE12.4 (heavy chain) and pEE6.4 (light chain). The sequences of the heavy chain constant regions used (based on the sequences described by Scinicariello et al., Immunology 111: 66-74, 2004) were as follows (aligned against the human sequence): TIFF2025063202000005.tif128162

[0228] The rhesus monkey light chain constant region (CL) sequence used was as follows: TIFF2025063202000006.tif14170

[0229] Example 4: Antibody production by transient expression in HEK-293F cells Antibodies were produced under serum-free conditions by co-transfecting the relevant heavy and light chain expression vectors in HEK-293F cells (Invitrogen) using 293fectin (Invitrogen) according to the manufacturer's instructions.

[0230] Example 5: Purification of IgG1 and IgG4 antibodies IgG1 and IgG4 antibodies were purified by protein A affinity chromatography. Cell culture supernatants were filtered over a 0.20 μM dead-end filter followed by loading onto a 5 mL protein A column (rProtein A FF, GE Healthcare, Uppsala, Sweden) and elution of IgG with 0.1 M citrate-NaOH, pH 3. The eluate was immediately neutralized with 2 M Tris-HCl, pH 9 and dialyzed overnight against 12.6 mM sodium phosphate, 140 mM NaCl, pH 7.4 (B. Braun, Oss, The Netherlands). After dialysis, samples were sterile filtered over a 0.20 μM dead-end filter. The concentration of purified IgG was determined by turbidimetry and absorbance at 280 nm. The purified protein was analyzed by SDS-PAGE, IEF, mass spectrometry and glycoanalysis.

[0231] Example 6: Purification of CH2-CH3 fragment His-tagged CH2-CH3 proteins were incubated with immobilized metal ions (Ni 2+ The purified proteins were purified by affinity chromatography (Macherey-Nagel GmbH, Düren, Germany), desalted using a PD-10 column (GE Healthcare) equilibrated with PBS, and filter-sterilized onto a 0.2 μM dead-end filter. The concentration of the purified proteins was determined by absorbance at 280 nm. The quality of the purified proteins was analyzed by SDS-PAGE.

[0232] Example 7: Generation of bispecific antibodies by GSH-induced Fab arm exchange between human IgG4 and rhesus IgG4 antibodies As mentioned above, WO 2008119353 (Genmab) describes an in vitro method for producing bispecific antibodies, in which bispecific antibodies are formed by "Fab arm" or "half molecule" exchange (swapping of heavy and attached light chains) between two monospecific IgG4 or IgG4-like antibodies by incubation under reducing conditions. This Fab arm exchange reaction is the result of a disulfide bond isomerization reaction in which the inter-heavy chain disulfide bond in the hinge region of a monospecific antibody is reduced and the resulting free cysteine ​​forms a new inter-heavy chain disulfide bond with a cysteine ​​residue of another antibody molecule with a different specificity. The resulting product is a bispecific antibody with two Fab arms with different sequences.

[0233] To test for Fab arm exchange between human IgG4 and rhesus IgG4 antibodies, all possible combinations of the two antibodies were generated using human IgG4-2F8 (anti-EGFR), human IgG4-7D8 (anti-CD20), rhesus IgG4-2F8 and rhesus IgG4-7D8. For in vitro Fab arm exchange, the antibody mixture containing each antibody at a final concentration of 4 μg / mL in 0.5 mL of PBS containing 0.5 mM reduced glutathione (GSH) was incubated at 37° C. for 24 hours. To stop the reduction reaction, 0.5 mL of PBS / 0.05% Tween 20 (PBST) was added to the reaction mixture.

[0234] The presence of bispecific antibodies was tested by determining bispecific binding using a sandwich enzyme-linked immunosorbent assay (ELISA) method. ELISA plates (Greiner bio-one, Frickenhausen, Germany) were coated overnight at 4°C with recombinant extracellular domain of EGFR at 2 μg / mL (100 μL / well) in PBS. Plates were washed once with PBST. A dilution series of antibody samples (0–1 μg / mL in 3-fold dilutions) in PBST / 0.2% BSA (PBSTB) was transferred to the coated ELISA plates (100 μL / well) and incubated for 60 min at room temperature (RT) on a plate shaker (300 rpm). Samples were discarded and plates were washed once with PBS / 0.05% Tween 20 (PBST). Plates were then incubated with mouse anti-idiotypic monoclonal antibody 2F2 SAB1.1 (raised against 7D8; Genmab) (100 μL / well) at 2 μg / mL in PBTB on a plate shaker (300 rpm). Plates were washed once with PBS / 0.05% Tween 20 (PBST). Plates were then incubated with HRP-conjugated goat anti-mouse IgG (15G; Jackson ImmunoResearch Laboratories, Westgrove, PA, USA; 1:5.000) (100 μL / well) in PBSTB on a plate shaker (300 rpm) for 60 min at RT. Plates were washed once with PBS / 0.05% Tween 20 (PBST). ABTS (50 mg / mL; Roche Diagnostics GmbH, Mannheim, Germany) was added (100 μL / well) and incubated for 30 min at RT in the dark. The reaction was stopped with 2% oxalic acid (100 μL / well; Riedel de Haen Seelze, Germany). After 10 min at RT, the absorbance at 405 nm was measured in an ELISA plate reader.

[0235] Figure 1 shows that the combination of human and rhesus IgG4 resulted in higher bispecific binding (higher OD 405 nm) compared to each of the combinations of IgG4 molecules of the same species. These data indicate that Fab arm exchange takes place between human IgG4 and rhesus IgG4. Furthermore, the higher bispecific binding suggests that human IgG4 half molecules show preferential dimerization (heterodimerization) with rhesus IgG4 half molecules, resulting in an equilibrium of the Fab arm exchange reaction shifting towards bispecific heterodimers instead of a stochastic exchange of 50% heterodimers and 50% homodimers.

[0236] Example 8: Sequence analysis of human and rhesus IgG4 It has been described that the ability of antibodies to participate in Fab arm exchange requires a third constant domain (CH3) in addition to a so-called permissive (e.g. CPSC-containing) hinge region, which only requires a reducing environment to be activated (Van der Neut Kolfschoten, 2007, Science). In the case of human antibodies, Fab arm exchange was found to be a unique feature of IgG4, characterized by an arginine (R) at position 409 in the CH3 domain and a permissive hinge (226-CPSC-229) (see WO 2008145142 (Genmab)). In contrast, human IgG1, which does not participate in Fab arm exchange, has a lysine (K) at position 409 and a stable (i.e. non-permissive) hinge (226-CPPC-229) (EU numbering, see also Figure 16).

[0237] In an attempt to elucidate the higher Fab arm exchange between human IgG4 and rhesus IgG4 compared to Fab arm exchange between IgG4 molecules of the same species, the core hinge and CH3-CH3 interface amino acids of human and rhesus antibodies were analyzed (see, e.g., Dall'Acqua, et al (1998) Biochemistry 37:9266 for an overview of human CH3-CH3 interface residues). Figure 2 shows that the core hinge sequence in Chinese rhesus IgG4 is 226-CPAC-229, and that the CH3 domain contains a lysine (K) at position 409. Furthermore, sequence alignment showed that rhesus IgG4 is characterized by three additional amino acid substitutions at the CH3-CH3 interface compared to human IgG4: isoleucine (I) at position 350 in rhesus versus threonine (T) in human; threonine (T) at position 370 in rhesus versus lysine (K) in human; and leucine (L) at position 405 in rhesus versus phenylalanine (F) in human.

[0238] Example 9: Generation of bispecific antibodies using GSH-induced Fab arm exchange between human IgG4 and human IgG1 containing rhesus IgG4 CH3 sequence It has been described for human antibodies that replacing the IgG1 core hinge sequence (CPPC) with the human IgG4 sequence (CPSC) by the P228S substitution to allow Fab arm exchange in IgG1 molecules had no effect, but mutating CH3 to an IgG4-like sequence was required for Fab arm exchange activity (Van der Neut Kolfschoten, 2007, Science).

[0239] Based on the Fab arm exchange between human IgG4 and rhesus IgG4 described in Example 7, we analyzed whether the IgG4 CH3 sequence of Chinese rhesus macaques could involve human IgG1 in Fab arm exchange. Therefore, in addition to the P228S mutation resulting in the hinge sequence CPSC, the triple mutation T350I-K370T-F405L (hereafter referred to as ITL) was introduced into human IgG1-2F8. The human IgG1-2F8 mutant was combined with human IgG4-7D8 for GSH-induced Fab arm exchange in vitro. The antibody mixtures containing each antibody at a final concentration of 4 μg / mL in 0.5 mL of PBS with 0.5 mM GSH were incubated at 37°C for 0-3-6-24 hours. To stop the reduction reaction, 0.5 mL of PBS / 0.05% Tween 20 (PBST) was added to the reaction mixture. Measurement of bispecific binding in ELISA was performed as described in Example 7.

[0240] Figure 3 confirms that introduction of the CPSC hinge alone does not allow human IgG1-2F8 to participate in GSH-induced Fab arm exchange when combined with human IgG4-7D8. Also, introduction of the rhesus IgG4-specific CH3 interface amino acids (ITLs) into human IgG1-2F8, while retaining the wild-type IgG1 hinge, did not result in participation in Fab arm exchange when combined with human IgG4-7D8 under these conditions. In contrast, a variant human IgG1-2F8 backbone sequence carrying both the CPSC sequence in the hinge and the rhesus IgG4-specific CH3 interface amino acids (ITLs) showed increased bispecific binding after GSH-induced Fab arm exchange with human IgG4-7D8 compared to the two human IgG4 antibodies. These data indicate that the CPSC-containing hinge, in combination with a CH3 domain containing I, T and L at positions 350, 370 and 405, respectively, is sufficient to engage human IgG1 in GSH-induced Fab arm exchange, and that the equilibrium of the exchange reaction shifts towards exchanged bispecific products when combined with human IgG4.

[0241] Example 10: Generation of bispecific antibodies by in vivo Fab arm exchange between human IgG4 and IgG1 or IgG4 mutants To further identify the characteristics required for participation in Fab arm exchange, human IgG4 and IgG1 variants were analyzed in vivo. Four female SCID mice per group (Charles River, Maastricht, The Netherlands) were injected iv with an antibody mixture containing 600 μg of antibody (500 μg 7D8 + 100 μg 2F8) in a total volume of 300 μL. Blood samples were taken from the saphenous vein 3, 24, 48 and 72 hours after injection. Blood was collected in heparin-containing vials and centrifuged at 10,000 g for 5 minutes to separate plasma from cells. Bispecific antibody generation was followed by evaluating CD20 and EGFR bispecific reactions by ELISA using plasma samples serially diluted in PBSTB as described in Example 7. Bispecific antibodies were quantified in plasma samples by nonlinear regression curve fitting (GraphPad Software, San Diego, CA) using the in vitro exchanged antibody mixture as a reference standard.

[0242] Figure 4 shows that human IgG4-2F8, in which either the hinge or CH3 sequence has been converted to its corresponding human IgG1 sequence (CPPC or R409K, respectively), no longer participates in Fab arm exchange in vivo. Conversely, human IgG1, in which both the hinge region and the CH3 interface sequence have been converted to its corresponding human IgG4 sequence (CPSC and K409R), can participate in Fab arm exchange in vivo. These data indicate that a hinge containing CPSC (S at position 228) in combination with a CH3 domain containing an arginine (R) at position 409 is sufficient to enable Fab arm exchange by human IgG1 in vivo.

[0243] Example 11: 2-MEA-induced Fab arm exchange: creating bispecific antibodies by bypassing / disrupting a stabilizing hinge 2-Mercaptoethylamine·HCl (2-MEA) is a mild reducing agent that has been described to selectively cleave disulfide bonds in the hinge region of antibodies while preserving the disulfide bond between the heavy and light chains. Therefore, a concentration series of 2-MEA was tested for its ability to induce the generation of bispecific antibodies by Fab arm exchange between two antibodies containing CPSC or CPPC hinge regions. An antibody mixture containing each antibody at a final concentration of 0.5 mg / mL was incubated for 90 min at 37°C with a concentration series of 2-MEA (0, 0.5, 1.0, 2.0, 5.0, 7.0, 10.0, 15.0, 25.0, and 40.0 mM) in a total volume of 100 μL of TE. To stop the reduction reaction, the reducing agent 2-MEA was removed by desalting the samples using spin columns (Microcon centrifugal filters, 30k, Millipore) according to the manufacturer's recommendations. Bispecific binding was measured by ELISA as described in Example 7.

[0244] 2-MEA-induced Fab arm exchange was tested for the combination IgG4-2F8×IgG4-7D8, which contained the CPSC hinge region and was found to be involved in GSH-induced Fab arm exchange, and for the combination IgG1-2F8-ITL×IgG4-7D8-CPPC, which was not involved in GSH-induced Fab arm exchange due to the stabilized hinge region (described in Example 9, FIG. 3). Surprisingly, 2-MEA was found to induce separation of the light chain from the heavy chain as judged by non-reducing SDS-PAGE (data not shown). Nevertheless, functional bispecific antibodies were generated as shown in FIG. 5. The maximum level of bispecific binding after Fab arm exchange between wild-type human IgG4-2F8 and IgG4-7D8 was achieved at a concentration of 2.0 mM 2-MEA, comparable to the level achieved with 0.5 mM GSH as described in Example 9 (FIG. 3). However, 2-MEA was able to induce Fab arm exchange between human antibodies IgG1-2F8-ITL and IgG4-7D8-CPPC (with stabilized hinge regions) in a dose-dependent manner. Little or no bispecific antibodies were formed at low 2-MEA concentrations, likely due to the presence of CPPC sequences in the hinge regions of both antibodies, whereas at higher 2-MEA concentrations bispecific antibody generation was very efficient. Maximal bispecific binding was achieved at 25 mM 2-MEA, exceeding the maximum binding after Fab arm exchange between two wild-type IgG4 antibodies. These maximum binding levels were comparable to those described in Example 9 (Figure 3) for GSH treatment of the corresponding antibody with a CPSC hinge (IgG1-2F8-CPSC-ITL). As both IgG1-2F8-ITL and IgG4-7D8-CPPC contain CPPC hinges, these data suggest that 2-MEA was able to circumvent the requirement of a CPSC hinge for Fab arm exchange in vitro.

[0245] Example 12: Mass spectrometry to follow the generation of bispecific antibodies by 2-MEA-induced Fab arm exchange Generation of bispecific antibodies by 2-MEA-induced Fab arm exchange is described in Example 11, where bispecific binding was demonstrated by ELISA (Figure 5). To confirm that bispecific antibodies were formed, samples were analyzed by electrospray ionization mass spectrometry (ESI-MS) to determine molecular weight. Samples were first deglycosylated by incubating 200 μg of antibody with 0.005 U of N-glycanase (catalog number GKE-5006D; Prozyme) in 180 μL of PBS overnight at 37°C. Samples were desalted at 60°C on an Aquity UPLC™ (Waters, Milford, USA) equipped with a BEH300 C18, 1.7 μm, 2.1 × 50 mm column and eluted with a gradient of a mixture of MQ water (eluent A) and LC-MS grade acetonitrile (eluent B) (Biosolve, Valkenswaard, The Netherlands) containing 0.05% formic acid (Fluka Riedel-de Haen, Buchs, Germany). Time-of-flight electrospray ionization mass spectra were recorded online on a micrOTOF™ mass spectrometer (Bruker, Bremen, Germany) operated in positive ion mode. Prior to analysis, the 500-4000 m / z scale was calibrated with ES tuning mix (Agilent Technologies, Santa Clara, USA). Mass spectra were deconvoluted using Maximal Entropy with DataAnalysis™ software v. 3.4 (Bruker, Bremen, Germany). Based on the molecular weight of the antibodies used for Fab arm exchange in this experiment, bispecific antibodies could be distinguished from the original antibodies (IgG1-2F8-ITL×IgG4-7D8-CPPC is also described in Example 15, FIG. 9C). For the bispecific antibody peaks, the area under the curve was determined and divided by the total area under the curve to calculate the percentage of bispecific antibody in each sample.Figure 6A shows three representative mass spectrometry profiles of Fab arm exchange reactions between IgG1-2F8-ITL and IgG4-7D8-CPPC at 0 mM 2-MEA (two peaks corresponding to parental antibodies), 7 mM 2-MEA (three peaks corresponding to parental and bispecific antibodies), and 40 mM 2-MEA (one peak corresponding to bispecific antibodies). The homogeneous peak of the bispecific product suggests that no mispairing of the light chains occurred, which would have resulted in the subdivided peaks. Quantified data is presented in Figure 6B and shows that Fab arm exchange between IgG1-2F8-ITL and IgG4-7D8-CPPC produced nearly 100% bispecific antibodies. In contrast, Fab arm exchange between wild-type IgG4 antibodies produced less than 50% bispecific products. These data confirm the results of the bispecific binding ELISA described in Example 11 (Figure 5).

[0246] Example 13: Stability of bispecific antibodies generated by 2-MEA-induced Fab arm exchange The stability of bispecific antibodies generated by 2-MEA-induced in vitro Fab arm exchange was tested. Therefore, 2 μg of bispecific sample generated from IgG1-2F8-ITL and IgG4-7D8-CPPC in 7.0 mM 2-MEA (as described in Example 11, Figure 5) was used in a GSH-induced Fab arm exchange reaction in the presence of a concentration series (0, 2, 20, 100 μg) of irrelevant IgG4 (IgG4-MG against acetylcholine receptor), corresponding to 0-fold, 1-fold, 10-fold, and 50-fold excess of IgG4-MG compared to 2 μg of bispecific test sample. Fab arm exchange in this reaction would cause loss of bispecific EGFR / CD20 binding. The conditions of the GSH reduction reaction were the same as described in Example 7 (24 hours at 37° C. in 0.5 mL of PBS / 0.5 mM GSH). To stop the reduction reaction, 0.5 mL of PBSTB was added to the reaction mixture. Bispecific binding was measured by ELISA as described in Example 7. Bispecific binding after the GSH reduction reaction is presented relative to the bispecific binding measured in the starting material (control), which was set to 100%.

[0247] FIG. 7A shows that for bispecific samples derived from IgG1-2F8-ITL×IgG4-7D8-CPPC, EGFR / CD20 bispecific binding is not significantly altered after GSH-induced Fab arm exchange in the presence of irrelevant IgG4, suggesting that the bispecific product is stable, i.e., does not participate in GSH-induced Fab arm exchange. In contrast, FIG. 7B shows that samples derived from IgG4-2F8×IgG4-7D8 show reduced EGFR / CD20 bispecific binding after GSH-induced Fab arm exchange in the presence of irrelevant IgG4, suggesting that the product is not stable. These data indicate that heterodimers consisting of a human IgG1 heavy chain containing the triple mutation T350I-K370T-F405L in the CH3 domain and a human IgG4 heavy chain containing the S228P substitution resulting in a stabilized hinge (CPPC) are stable.

[0248] Example 14: In vivo analysis of pharmacokinetics and stability of bispecific antibodies generated by 2-MEA-induced Fab arm exchange Bispecific antibodies generated by in vitro 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL × IgG4-7D8-CPPC were injected into SCID mice to analyze their stability (in vivo Fab arm exchange) and pharmacokinetic properties (plasma clearance rate) in comparison with the parent antibodies IgG1-2F8-ITL and IgG4-7D8-CPPC. Three groups of mice (3 mice per group) were intravenously injected with 200 μL of purified antibody into the tail vein: (1) 100 μg of bispecific antibody; (2) 100 μg of bispecific antibody + 1,000 μg of irrelevant IgG4 (natalizumab, anti-α4-integrin); (3) 50 μg of IgG1-2F8-ITL + 50 μg of IgG4-7D8-CPPC. Blood samples (50-100 μL) were collected by buccal puncture at predetermined time intervals after antibody administration (10 min, 3 h, 1 day, 2 days, 7 days, 14 days, and 21 days). Blood was collected into heparin-containing vials and centrifuged at 14,000 g for 10 min. Plasma was stored at -20 °C prior to further analysis.

[0249] Total IgG concentration in plasma samples was assayed by ELISA. The assay conditions for the following steps were the same as for the ELISA described in Example 7. The specific compounds used for total IgG measurement were: coating with 2 μg / mL mouse anti-human IgG (clone MH16-1; CLB; catalog number M1268); serum sample dilution (1:500 and 1:2,500 for group 1 and group 3) and (1:2,500 and 1:10,000 for group 2); conjugate: HRP-conjugated goat anti-human IgG (clone 11H; Jackson; catalog number 109-035-098; 1:10,000). The presence of bispecific antibodies in plasma samples was assayed and quantified by CD20 and EGFR bispecific reactivity in ELISA as described in Example 10.

[0250] Figure 8A shows the total antibody plasma concentration. The shape of the plasma clearance curves was identical in all groups, suggesting that the plasma clearance of the bispecific antibody was the same as for the parental antibodies IgG1-2F8-ITL and IgG4-7D8-CPPC over the analyzed time interval. Figure 8B shows the plasma concentration of the bispecific antibody over time. Addition of a 10-fold excess of irrelevant IgG4 to the bispecific antibody did not affect the bispecific antibody concentration, suggesting that no Fab arm exchange occurred in vivo. After injection of the parental antibodies (IgG1-2F8-ITL + IgG4-7D8-CPPC), the bispecific antibody was not detectable in the plasma, confirming that these antibodies do not participate in Fab arm exchange in vivo. These data suggest that the bispecific antibody products generated by in vitro 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL x IgG4-7D8-CPPC were stable in vivo (absence of Fab arm exchange) and exhibited comparable pharmacokinetic properties (plasma clearance rate) as the parental monovalent antibodies.

[0251] Example 15: Purity of bispecific antibodies generated by 2-MEA-induced Fab arm exchange between two antibodies A batch of bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-ITL x IgG4-7D8-CPPC was purified on a PD-10 desalting column (catalog no. 17-0851-01; GE Healthcare). The purity of the bispecific products was then analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), high performance size exclusion chromatography (HP-SEC) and mass spectrometry. The functionality of the generated bispecific antibodies was confirmed by bispecific binding in ELISA (data not shown).

[0252] Samples were run at neutral pH and subjected to SDS-PAGE under reducing and non-reducing conditions on 4-12% NuPAGE Bis-Tris gels (Invitrogen, Breda, The Netherlands) using a modified Laemli method (Laemli 1970 Nature 227(5259): 680-5). SDS-PAGE gels were stained with Coomassie and digitally processed images were acquired using GeneGenius (Synoptics, Cambridge, UK). Figure 9A shows that only a few half molecules (H1L1) were detectable on the non-reducing gel, indicating that the antibody sample after Fab arm exchange consisted of intact IgG (Figure 9A-b).

[0253] The HP-SEC fraction was analyzed using a TSK HP-SEC column (G3000SW XI ; Toso Biosciences, via Omnilabo, Breda, The Netherlands) and a Waters Alliance 2695 Separator (Waters, Etten-Leur, The Netherlands) connected to a Waters 2487 Dual λ Absorbance Detector (Waters). Samples were run at 1 mL / min. Results were processed using Empower software version 2002 and expressed peak by peak as a percentage of total peak height. Figure 9B shows that virtually no aggregates were formed and more than 98% of the sample consisted of intact IgG.

[0254] Mass spectrometry was performed as described in Example 12. Figure 9C shows the mass spectrometry profiles of the starting materials IgG1-2F8-ITL and IgG4-7D8-CPPC and the bispecific product generated by Fab arm exchange between IgG1-2F8-ITL x IgG4-7D8-CPPC. The product in the Fab arm exchange sample was 145,901 kDa, which is a perfect match with the bispecific product derived from IgG1-2F8-ITL (146,259.5 / 2=73,130) + IgG4-7D8-CPPC (145,542.0 / 2=72,771). Furthermore, the bispecific antibody product showed a homogenous peak, suggesting that no mispairing of the light chains occurred, which would have resulted in a subdivided peak. These data indicate that the Fab arm exchange generated 100% bispecific antibodies. The minor peaks detected in addition to the major peak (K0) in the IgG4-7D8-CPPC and bispecific samples can be attributed to the presence of one (K1) or two (K2) C-terminal lysines.

[0255] These data indicate that approximately 100% functional bispecific antibody samples were generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL x IgG4-7D8-CPPC.

[0256] Example 16: Elucidation of the requirement of T350I, K370T and F405L substitutions for participation in Fab arm exchange of human IgG1 To further identify determinants in the IgG1 CH3 domain required for IgG1 to participate in Fab arm exchange, IgG1 containing the triple mutation T350I-K370T-F405L (ITL) was compared with the double mutants T350I-K370T (IT), T350I-F405L (IL) and K370T-F405L (TL). Also, the single mutant F405L (L) was tested. 2-MEA was used as a reducing agent to induce Fab arm exchange in vitro (50 μg of each antibody in 100 μL of PBS / 25 mM 2-MEA for 90 min at 37° C.). For single mutant F405L antibodies, unpurified antibodies from the supernatant of transient transfection were used after buffer exchange into PBS using an Amicon Ultra centrifugation device (30k, Millipore, Cat. No. UFC803096). To stop the reduction reaction, the reducing agent 2-MEA was removed by desalting the samples using spin columns as described in Example 11. Bispecific antibody production was determined by bispecific binding measured by ELISA as described in Example 7.

[0257] Triple (ITL), double (IT, IL and TL) and single (L) mutations were introduced into IgG1-2F8. These mutants were combined with IgG4-7D8 containing a CPSC hinge (wild type) or IgG4-7D8 containing a stabilized hinge (IgG4-7D8-CPPC) for Fab arm exchange with 25 mM 2-MEA for 90 min at 37°C. Figures 10A-B show that IgG1-2F8-IL and IgG1-2F8-TL mutants showed Fab arm exchange to the same level as the triple mutant ITL, regardless of the IgG4-7D8 (CPSC or CPPC hinge) combined. In contrast, no bispecific binding was observed in combination with the IgG1-2F8-IT mutant. Figure 10C shows that similarly, the IgG1-2F8-F405L mutant showed Fab arm exchange, regardless of the IgG4-7D8 (CPSC or CPPC hinge) combined. These data suggest that the F405L mutation is sufficient to engage human IgG1 in Fab arm exchange under the conditions described above.

[0258] Example 17: Generation of bispecific antibodies by 2-MEA-induced Fab arm exchange at different temperatures The ability of 2-MEA to induce the generation of bispecific antibodies by Fab arm exchange between two different antibodies was tested at different temperatures. Fab arm exchange reactions were initiated by incubating 160 μg human IgG1-2F8-ITL (0.5 mg / mL final concentration for each antibody) with 160 μg IgG4-7D8-CPPC in 320 μl PBS / 25 mM 2-MEA at either 0° C., 20° C. (RT) or 37° C. From these reactions, 20 μL samples were taken at different time points (0, 2.5, 5, 10, 15, 30, 45, 60, 75, 90, 120, 150, 180 and 240 min). 20 μl PBS was added to each sample before removing the reducing agent 2-MEA by desalting the samples using a Zeba 96-well spin desalting plate (7k, Cat. No. 89808 Thermo Fisher Scientific) according to the manufacturer's recommendations. Total antibody concentrations were determined by measuring absorbance at a wavelength of 280 nm using a Nanodrop ND-1000 spectrophotometer (Isogen Life Science, Maarssen, The Netherlands). A dilution series of antibody samples (total antibody concentrations of 0-20 μg / mL at 25-fold dilutions) was used in an ELISA to measure bispecific binding as described in Example 7.

[0259] Figure 11 shows that bispecific antibody generation by 2-MEA-induced Fab arm exchange between human IgG1-2F8-ITL and IgG4-7D8-CPPC was found to be most efficient at 37°C, with maximum bispecific binding reached after 45 minutes. At room temperature, bispecific antibody generation was lower, with maximum bispecific binding reached after 240 minutes. At 0°C, no bispecific binding generation was observed during the time course analyzed.

[0260] Example 18: Analysis of the ability of various reducing agents to induce in vitro Fab arm exchange to generate bispecific antibodies It was shown above that 0.5 mM GSH can induce in vitro Fab arm exchange between human IgG4 and IgG1-CPSC-ITL, but not between human IgG4 and IgG1-ITL containing a stable hinge (Figure 3). Furthermore, it was found that 2-MEA can induce Fab arm exchange between antibodies with stabilized hinge regions, such as IgG1-ITL x IgG4-CPPC (Figure 5). To test whether other concentrations of GSH or 2-MEA or other reducing agents can induce in vitro Fab arm exchange between two different antibodies, a concentration series of 2-MEA, GSH and DTT (dithiothreitol) was tested. Therefore, a combination of 10 μg human IgG1-2F8-ITL and 10 μg IgG4-7D8-CPPC (final concentration of 0.5 mg / mL for each antibody) in 20 μl PBS was incubated with a concentration series of different reducing agents (0.0, 0.04, 0.1, 0.2, 0.5, 1.0, 2.5, 5.0, 12.5, 25.0 and 50.0 mM) at 37° C. After 90 min, 20 μl PBS was added to each sample and the reducing agent was removed by desalting the samples using a spin desalting plate as described in Example 17. Total antibody concentration was determined as described in Example 17. A dilution series of the antibody samples (total antibody concentration 0-20 μg / mL in 3-fold dilutions) was used in an ELISA to measure bispecific binding as described in Example 7.

[0261] Figure 12 confirms that 2-MEA induces maximum bispecific binding at a concentration of 25 mM 2-MEA. DTT proved to be very effective in generating bispecific antibodies, reaching maximum bispecific binding at 2.5 mM DTT. GSH concentrations ranging from 0 to 5 mM were unable to induce the generation of bispecific antibodies by Fab arm exchange between IgG1-ITL and IgG4-CPPC antibodies, both containing stabilized hinge regions. Higher GSH concentrations (12.5 to 50 mM) caused the formation of antibody aggregates as judged by non-reducing SDS-PAGE (data not shown). Therefore, these samples were excluded from the analysis. These data indicate that the generation of bispecific antibodies by Fab arm exchange between two different antibodies can be induced by various reducing agents.

[0262] Example 19: Determinants at IgG1 position 409 for participation in 2-MEA-induced Fab arm exchange in combination with IgG1-ITL 2-MEA can induce Fab arm exchange between human IgG1-ITL and IgG4-CPPC as described in Example 11 (Figure 5). The CH3 interface residues of human IgG1 and IgG4 differ only at position 409: lysine (K) in IgG1 and arginine (R) in IgG4 (described in Example 8, Figure 2). Therefore, it was tested whether the substitution of lysine with arginine or any other amino acid at position 409 (K409X) could allow IgG1 to participate in 2-MEA-induced Fab arm exchange with IgG1-ITL. A combination of 10 μg human IgG1-2F8-ITL and 10 μg IgG1-7D8-K409X (0.5 mg / mL final concentration for each antibody) in 20 μl PBS / 25 mM 2-MEA was incubated at 37° C. for 90 min. Unpurified antibodies from transient transfection supernatants were used after buffer exchange into PBS using Amicon Ultra centrifugation devices (30k, Millipore, Cat. No. UFC803096). After the Fab arm exchange reaction, 20 μl of PBS was added to each sample and the reducing agent was removed by desalting the samples using spin desalting plates as described in Example 17. A dilution series of antibody samples (total antibody concentrations of 0-20 μg / mL in 3-fold dilutions) was used in an ELISA to measure bispecific binding as described in Example 7.

[0263] Figure 13A shows the results of bispecific binding by 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL x IgG1-7D8-K409X. In Figure 13B, the exchange is presented as bispecific binding for purified batches of bispecific antibodies derived from 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL and IgG4-7D8-CPPC, set at 100%. These data were also scored as (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange or (++) high Fab arm exchange, as presented in Table 1. No Fab arm exchange (-) was considered when position 409 in IgG1-7D8 was K (= wild type IgG1), L or M. Fab arm exchange was found to be moderate (+) when position 409 in IgG1-7D8 was F, I, N or Y, and high (++) when position 409 in IgG1-7D8 was A, D, E, G, H, Q, R, S, T, V or W.

[0264] Table 1. 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL and IgG1-7D8-K409X mutants Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between IgG1-2F8-ITL and IgG1-7D8-K409X mutants was determined by sandwich ELISA: (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange, (++) high Fab arm exchange. TIFF2025063202000007.tif110140

[0265] Example 20: Antibody deglycosylation does not affect the generation of bispecific antibodies by 2-MEA-induced Fab arm exchange IgG4-7D8 and IgG4-7D8-CPPC samples were deglycosylated by incubating 200 μg of antibody with 0.005 U of N-glycanase (catalog no. GKE-5006D; Prozyme) in 180 μL of PBS overnight at 37°C. These samples were used directly in Fab arm exchange reactions. Fab arm exchange was performed by incubating 50 μg of each antibody (0.5 mg / mL final concentration for each antibody) in 100 μl of PBS / 25 mM 2-MEA for 90 min at 37°C. The reducing agent 2-MEA was removed by desalting the samples using spin columns as described in Example 11. A dilution series of antibody samples (total antibody concentrations of 0-20 μg / mL in 3-fold dilutions) was used in a sandwich ELISA to measure bispecific binding as described in Example 7.

[0266] Mass spectrometry showed that the deglycosylation reaction resulted in 100% deglycosylated antibody product (data not shown). Figure 14 shows that Fab arm exchange with deglycosylated antibodies was not different from Fab arm exchange with the corresponding glycosylated antibodies (IgG4-2F8 x IgG4-7D8-deglycosylated vs IgG4-2F8 x IgG4-7D8 and IgG1-2F8-ITL x IgG4-7D8-CPPC-deglycosylated vs IgG1-2F8-ITL x IgG4-7D8-CPPC). These data suggest that deglycosylation did not affect the generation of bispecific antibodies by 2-MEA-induced Fab arm exchange.

[0267] Example 21: Quantification of non-covalent CH3-CH3 interactions The strength of the interaction at the CH3 interface should be such that both heavy chains in the parent antibody are able to dissociate in a Fab arm exchange reaction and to subsequently associate in a heterodimerization reaction. Therefore, a correlation between the ability to participate in Fab arm exchange and the strength of the non-covalent CH3-CH3 interaction (dissociation constant, K D) were analyzed. GSH-induced Fab arm exchange was performed as described in Example 9 (0.5 mM GSH at 37° C.) for the following human antibody combinations: TIFF2025063202000008.tif149140 Bispecific antibody production was measured by determining bispecific binding in a sandwich ELISA as described in Example 7. Figures 15A / B / C show the results of bispecific binding after the Fab arm exchange reaction.

[0268] To measure the effect of the above CH3 mutations on the strength of the CH3-CH3 interaction, fragments consisting of only the CH2-CH3 domains were generated. The lack of the hinge region in these fragments prevented the covalent inter-heavy chain disulfide bond. The fragments were analyzed by non-denaturing mass spectrometry. Samples were buffer exchanged into 100 mM ammonium acetate pH 7 using a 10 kDa MWCO spin filter column. Aliquots (approximately 1 μL) of serially diluted samples (20 μM-25 nM; monomer equivalents) were loaded onto gold-plated borosilicate capillaries for analysis on an LCT mass spectrometer (Waters). The monomer signal M s is the monomer peak area as a percentage of the total peak area in the spectrum (M S / (M S +D S ), D in the formula s = dimer signal). The concentration of monomer at equilibrium [M] eq is M S .[M] 0 In the formula, [M] 0 is the total protein concentration in units of monomer. The dimer concentration at equilibrium [D] eq ([M] 0 -[M] eq ) / 2. Then, K D [D] eq vs [M] eq 2 The K of the non-covalent CH3-CH3 interaction was extracted from the slope of the plot. D are presented in Table 2.

[0269] The correlation between the ability to participate in Fab arm exchange and the strength of non-covalent CH3-CH3 interactions was analyzed. Figure 15D / E shows the measured K D Figure 15 shows the percentage of bispecific binding after Fab arm exchange plotted against the apparent K for CH3-CH3 interactions that allow efficient Fab arm exchange under the conditions tested (Figure 15D for IgG1; Figure 15E for IgG4). D A certain range of values ​​is suggested.

[0270] (Table 2) K of non-covalent CH3-CH3 interaction D TIFF2025063202000009.tif114140 * Compared to the corresponding CH2-CH3 fragments of wild-type IgG1 or IgG4

[0271] Example 22: Analysis of the ability of various reducing agents to induce the generation of bispecific antibodies by in vitro Fab arm exchange between IgG1-2F8-F405L and IgG1-7D8-K409R 2-MEA and DTT were found to induce in vitro Fab arm exchange between human IgG1-ITL and IgG4-CPPC (FIG. 12). We tested whether these reducing agents could similarly induce in vitro Fab arm exchange between human IgG1-2F8-F405L and IgG1-7D8-K409R. A concentration series of 2-MEA, DTT, GSH and TCEP (tris(2-carboxyethyl)phosphine) was tested. Fab arm exchange was performed as described in Example 18. The concentration series of the various reducing agents tested was as follows: 0.0, 0.04, 0.1, 0.2, 0.5, 1.0, 5.0, 25.0, 50.0 mM of 2-MEA, GSH, DTT or TCEP.

[0272] From FIG. 17, it was confirmed that 2-MEA induced maximum Fab arm exchange at a concentration of 25 mM 2-MEA, which persisted at higher concentrations of 50.0 mM 2-MEA. DTT was found to be very effective in generating bispecific antibodies, reaching maximum Fab arm exchange at 0.5 mM DTT, which persisted over higher DTT concentrations (1.0-50.0 mM). Similarly, TCEP was found to be very effective in generating bispecific antibodies, reaching maximum Fab arm exchange at 0.5 mM. At concentrations above 25.0 mM, Fab arm exchange by TCEP was prevented. GSH concentrations in the range of 0.0-5.0 mM were unable to induce generation of bispecific antibodies through Fab arm exchange. Higher GSH concentrations (25.0-50.0 mM) caused the formation of antibody aggregates (data not shown). Therefore, these samples were excluded from the analysis. These data indicate that the generation of bispecific antibodies by Fab arm exchange between two different antibodies can be induced by a variety of reducing agents.

[0273] Example 23: Generation of bispecific antibodies by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L and IgG1-7D8-K409R To confirm the formation of bispecific antibodies by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L and IgG1-7D8-K409R, the molecular weights of samples from Fab arm exchange reactions with a concentration series of 2-MEA were determined by ESI-MS. The concentration series tested was as follows: 0.0, 0.5, 1.0, 2.0, 5.0, 7.0, 10.0, 15.0, 25.0 and 40.0 mM 2-MEA. Fab arm exchange (in PBS) and sandwich ELISA were performed as described in Example 11. ESI-MS was performed as described in Example 12.

[0274] Figure 18A shows that 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L and IgG1-7D8-K409R efficiently led to the generation of bispecific antibodies in a dose-dependent manner, with the level of bispecific binding maximal at a concentration of 15.0 mM 2-MEA. Quantified ESI-MS data is presented in Figure 18B and shows that Fab arm exchange between IgG1-2F8-F405L and IgG1-7D8-K409R yielded nearly 100% bispecific antibodies, confirming the bispecific binding ELISA results.

[0275] Example 24: Purity of bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L x IgG1-7D8-K409R Batches of bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L x IgG1-7D8-K409R were purified using a PD-10 desalting column (catalog no. 17-0851-01; GE Healthcare). The purity of the bispecific products was then analyzed by mass spectrometry as described in Example 12.

[0276] Figure 19 shows the mass spectrometry profiles of the bispecific products generated by Fab arm exchange between starting materials IgG1-2F8-F405L and IgG1-7D8-K409R and IgG1-2F8-F405L x IgG1-7D8-K409R. The product in the Fab arm exchange sample was 146,160.7 kDa, which is consistent with the bispecific product derived from IgG1-2F8-F405L (146,606.8 / 2 = 73,303.3) x IgG1-7D8-K409R (146,312.2 / 2 = 73,156.1) = 146,459.4 kDa. Furthermore, the bispecific antibody product showed a uniform peak, suggesting that no mispairing of the light chains occurred, which would have resulted in a subdivided peak. These data indicate that the Fab arm exchange produced approximately 100% bispecific antibodies.

[0277] Example 25: In vivo analysis of stability and pharmacokinetics of a bispecific antibody generated from IgG1-2F8-F405L x IgG1-7D8-K409R by 2-MEA-induced Fab arm exchange Bispecific antibodies generated by in vitro 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R were injected into SCID mice to analyze their stability (in vivo Fab arm exchange) and pharmacokinetic properties as described in Example 14. Two groups of mice (3 mice per group) were analyzed: (1) 100 μg bispecific antibody; (2) 100 μg bispecific antibody + 1,000 μg irrelevant IgG4 (IgG4-637, described in WO2007068255). In this example, total IgG concentrations were assayed in plasma samples by ELISA as described in Example 14, except that HRP-conjugated goat anti-human IgG (Jackson, Cat. No. 109-035-098, 1:10,000) was used as the detection conjugate. The presence of bispecific antibodies in plasma samples was assayed and quantified by CD20 and EGFR bispecific reactivity in a sandwich ELISA as described in Example 14.

[0278] Figure 20A shows the total antibody plasma concentration over time. The shape of the plasma clearance curves was identical in both groups. Figure 20B shows the plasma concentration of the bispecific antibody over time. Addition of a 10-fold excess of irrelevant IgG4 to the bispecific antibody did not affect the bispecific antibody concentration, suggesting that no Fab arm exchange occurred in vivo. These data suggest that the bispecific antibody product generated by in vitro 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R was stable in vivo (no Fab arm exchange).

[0279] Example 26: CDC-mediated cell killing by bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L x IgG1-7D8-K409R The CD20 antibody IgG1-7D8 can efficiently kill CD20-expressing cells by complement-dependent cytotoxicity (CDC). In contrast, the EGFR antibody IgG1-2F8 does not mediate CDC against target cells expressing EGFR. We tested whether bispecific antibodies generated by 2-MEA-induced Fab arm exchange between the mutants IgG1-7D8-K409R and IgG1-2F8-F405L x IgG1-7D8-K409R were still able to induce CDC against CD20-expressing cells. 10 5 Daudi or Raji cells were preincubated with a concentration series of antibodies in 80 μL RPMI medium supplemented with 0.1% BSA for 15 min on a shaker at room temperature. 20 μL normal human serum (NHS) was added as a source of complement (20% NHS final concentration) and incubated at 37° C. for 45 min. CDC reaction was stopped by adding 30 μL ice-cold RPMI medium supplemented with 0.1% BSA. 10 μL 10 μg / mL propidium iodide (PI) (1 μg / mL final concentration) was added and dead and live cells were differentiated by FACS analysis.

[0280] Figure 21 shows that CDC-mediated cell killing of CD20-expressing Daudi cells (Figure 21A) and CD20-expressing Raji cells (Figure 21B) by IgG1-7D8 was not affected by the introduction of the K409R mutation. Neither Daudi nor Raji cells express EGFR, causing monovalent binding of the bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R. Nevertheless, the bispecific antibody still induced CDC-mediated cell killing of CD20-expressing cells. These data suggest that the CDC capacity of the parent antibody was maintained in the bispecific format.

[0281] Example 27: ADCC-mediated cell killing by bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L x IgG1-7D8-K409R The EGFR antibody IgG1-2F8 can kill EGFR-expressing cells, such as A431, by antibody-dependent cellular cytotoxicity (ADCC). A431 cells do not express CD20, and therefore the CD20 antibody IgG1-7D8 does not induce ADCC against these cells. We tested whether bispecific antibodies generated by 2-MEA-induced Fab arm exchange between the mutants IgG1-2F8-F405L and IgG1-2F8-F405L x IgG1-7D8-K409R were still able to induce ADCC against A431 cells. For effector cell isolation, peripheral blood mononuclear cells (PBMCs) were isolated from whole blood of healthy donors using Leucosep® tubes (Greiner Bio-one, Cat. No. 227290) according to the manufacturer's recommendations. 5 x 10 6 cells were isolated in 1 mL of RPMI medium supplemented with 0.1% BSA. 6 100 μCi per A431 cell 51 Target cells were labeled by adding Cr and incubating for 60 min in a shaking water bath at 37°C. Labeled cells were washed and resuspended in RPMI supplemented with 0.1% BSA. 5 × 10 cells were collected in RPMI supplemented with 0.1% BSA. 4 Labeled target cells were pre-incubated in 100 μL with a concentration series of antibodies (3-fold dilutions with a range of 0–10 μg / mL final concentrations in the ADCC assay) at room temperature for 15 min. The ADCC assay was performed with effector cells (5 × 10 cells). 6 After 4 hours at 37°C, the results were analyzed from triplicate experiments. 51 Cr release was measured in a scintillation counter as counts per minute (cpm). Percent cytotoxicity was calculated using the following formula: Percent specific lysis = (experimental cpm - basal cpm) / (maximum cpm - basal cpm) x 100. 50 μL of 5% Triton X-100 was added to 50 μL of target cells (5 x 10 cells). 4 pcs) to get the maximum 51 Cr release was determined and basal release was measured in the absence of sensitizing antibody and effector cells.

[0282] Figure 22 shows that the CD20-specific antibody IgG1-7D8 did not induce ADCC against CD20-negative A431 cells. Both IgG1-2F8 and mutant IgG1-2F8-F405L were able to induce ADCC against A431 cells, suggesting that the introduction of the F405L mutation in IgG1-2F8 did not affect its ADCC effector function. In addition, the bispecific antibody derived from IgG1-2F8-F405L x IgG1-7D8-K409R induced ADCC against A431 cells in a dose-dependent manner, suggesting that the ADCC effector function was maintained in the bispecific format.

[0283] Example 28: Determinants at IgG1 position 405 for participation in 2-MEA-induced Fab arm exchange in combination with IgG1-K409R In Example 16, it was described that the F405L mutation is sufficient to allow human IgG1 to participate in Fab arm exchange when combined with IgG4-7D8. To further investigate the determinants of IgG1 position 405 for participation in 2-MEA-induced Fab arm exchange in combination with human IgG1-K409R, all possible IgG1-2F8-F405X mutants (except C and P) were combined with IgG1-7D8-K409R. The procedure was performed with purified antibodies as described in Example 19.

[0284] Figure 23 shows the results of bispecific binding by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405X x IgG1-7D8-K409R. These data were also scored as (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange or (++) high Fab arm exchange, as presented in Table 3. No Fab arm exchange (-) was observed when position 405 in IgG1-2F8 was F (= wild type IgG1). Fab arm exchange was found to be low (+ / -) when position 405 in IgG1-2F8 was G or R. Fab arm exchange was found to be high (++) when position 405 in IgG1-2F8 was A, D, E, H, I, K, L, M, N, Q, S, T, V, W or Y. These data suggest that specific mutations at IgG1 position 405, when combined with IgG1-K409R, engage IgG1 in 2-MEA-induced Fab arm exchange.

[0285] Table 3: 2-MEA-induced Fab arm exchange between IgG1-2F8-F405X mutant and IgG1-7D8-K409R Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between IgG1-2F8-F405X mutants and IgG1-7D8-K409R was determined by sandwich ELISA: (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange, (++) high Fab arm exchange. TIFF2025063202000010.tif110140

[0286] Example 29: Determinants at IgG1 position 407 for participation in 2-MEA-induced Fab arm exchange in combination with IgG1-K409R In Example 28, it was described that certain single mutations at position F405 were sufficient to allow human IgG1 to participate in Fab arm exchange when combined with IgG1-K409R. To investigate whether other determinants involved in the location of the Fc:Fc interface in the CH3 domain could similarly mediate the Fab arm exchange mechanism, mutagenesis was performed at IgG1 position 407 and this mutant was examined for its participation in 2-MEA-induced Fab arm exchange in combination with human IgG1-K409R. All possible IgG1-2F8-Y407X mutants (except C and P) were combined with IgG1-7D8-K409R. The procedure was performed with purified antibodies as described in Example 19.

[0287] Figure 24 shows the results of bispecific binding by 2-MEA-induced Fab arm exchange between IgG1-2F8-Y407X x IgG1-7D8-K409R. These data were also scored as (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange or (++) high Fab arm exchange, as presented in Table 4. No Fab arm exchange (-) was observed when position 407 in IgG1-2F8 was Y (= wild type IgG1), E, ​​K, Q or R. Fab arm exchange was found to be low (+ / -) when position 407 in IgG1-2F8 was D, F, I, S or T, moderate (+) when position 407 in IgG1-2F8 was A, H, N or V and high (++) when position 407 in IgG1-2F8 was G, L, M or W. These data suggest that a specific single mutation at IgG1 position 407 in combination with IgG1-K409R engages IgG1 in 2-MEA-induced Fab arm exchange.

[0288] Table 4: 2-MEA-induced Fab arm exchange between IgG1-2F8-Y407X mutant and IgG1-7D8-K409R Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between IgG1-2F8-Y407X mutant and IgG1-7D8-K409R was determined by sandwich ELISA: (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange, (++) high Fab arm exchange. TIFF2025063202000011.tif110140

[0289] Example 30: Quantification of non-covalent CH3-CH3 interactions in IgG1 heterodimers It was described in Example 21 that there is a certain range of strength of interaction of CH3-CH3 homodimers that allows efficient Fab arm exchange. The strength of interaction at the CH3 interface must be such that both heavy chains in the parent antibody (homodimer) can dissociate in the Fab arm exchange reaction and then associate in the heterodimerization reaction. To generate a stable heterodimer, the strength of the heterodimer interaction must be higher than that of the homodimer interaction and must be such that it favors heterodimerization over homodimerization. To confirm this, the strength of CH3-CH3 interaction in heterodimers was measured and compared with that in homodimers. The K of CH2-CH3 fragments derived from IgG1-K409R, IgG1-F405L and IgG1-ITL homodimers was 0.01 to 0.01. D was measured as described in Example 21. K D For the determination of K, CH2-CH3 domain fragments (G1-F405L and G1-ITL) were mixed with the IgG1Δhinge fragment of IgG1-7D8-K409R, which contains all antibody domains except the hinge. The absence of the hinge region in both fragments prevents covalent inter-heavy chain disulfide bonds. The fragments were mixed and analyzed after 24 hours by non-denaturing mass spectrometry as described in Example 21. K of non-covalent CH3-CH3 interactions in the indicated CH2-CH3 fragments or mixtures of CH2-CH3 fragments with IgG1Δhinge DThe values ​​are presented in Table 5. From these data, it can be seen that under the conditions tested, the strength of the heterodimeric interactions is higher (lower K D ) is suggested.

[0290] (Table 5) TIFF2025063202000012.tif33140

[0291] Example 31: Biochemical analysis of bispecific antibodies generated by 2-MEA-induced Fab arm exchange Batches of bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L × IgG1-7D8-K409R were purified on a PD-10 desalting column (catalog no. 17-0851-01; GE Healthcare). The purity of the bispecific products was then analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), high performance size exclusion chromatography (HP-SEC), mass spectrometry, HPLC cation exchange chromatography (HPLC-CIEX), and capillary isoelectric focusing (cIEF).

[0292] SDS-PAGE was performed under non-reducing (Figure 25A) and reducing (Figure 25B) conditions as described in Example 15. Figure 25A shows that only a small amount of half molecule (H1L1) was detectable on the non-reducing gel, indicating that the antibody sample after 2-MEA-induced Fab arm exchange consisted of intact IgG.

[0293] HP-SEC was performed as described in Example 15. Figures 26(B) and 26(A) show the HP-SEC profiles of the starting materials IgG1-2F8-F405L and IgG1-7D8-K409R, respectively. A mixture of both antibodies (1:1) and the bispecific product generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R are shown in Figures 26C and 26D, respectively. Furthermore, Figure 26D shows that virtually no aggregates are formed and more than 99% of the sample consists of intact IgG.

[0294] Mass spectrometry (ESI-MS) was performed as described in Example 12. Figures 27(B) and 27(A) show the mass spectrometry profiles of the starting materials IgG1-2F8-F405L and IgG1-7D8-K409R, respectively. The mixture of both antibodies (1:1) and the bispecific product generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R are shown in Figures 27C and 27D, respectively. The product in the 2-MEA-induced Fab arm exchange sample was 146,159.7 kDa, which is perfectly compatible with the bispecific product derived from IgG1-2F8-F405L (146,289.0 / 2=73,145) x IgG1-7D8-K409R (146,028.0 / 2=73,014). Furthermore, the bispecific antibody product showed a homogenous peak, suggesting that mispairing of the light chains did not occur, which would have produced subdivided peaks. These data indicate that 2-MEA-induced Fab arm exchange produced bispecific IgG. * The small peak indicated by a ) resulted from incomplete deglycosylation prior to analysis. These data indicate that the bispecific antibody sample was generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R.

[0295] Capillary isoelectric focusing (cIEF) was performed using an iCE280 analyzer (Convergent Biosciences). Figures 28A and 28B show the cIEF profiles of the starting materials IgG1-2F8-F405L and IgG1-7D8-K409R, respectively. A mixture of both antibodies (1:1) and the bispecific product generated by Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R are shown in Figures 28C and 28D, respectively. All samples were desalted before use. The final concentration in the assay mixture was 0.3 mg / mL IgG (0.35% methylcellulose; 2% carrier ampholyte 3-10; 6% carrier ampholyte 8-10.5; 0.5% pI marker 7.65 and 0.5% pI marker 10.10). Isoelectric focusing was performed at 3000 V for 7 min, and the whole capillary absorption image was captured by a charge-coupled device camera. After peak profile calibration, the data were analyzed by EZChrom software. pI markers were ( * ) These data show that bispecific antibody samples were generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R.

[0296] Another technique to investigate the charge isoforms of monoclonal antibodies is high pressure liquid chromatography cation exchange (HPLC-CIEX). Figures 29A and 29B show the HPLC-CIEX profiles of the starting materials IgG1-2F8-F405L and IgG1-7D8-K409R, respectively. A mixture of both antibodies (1:1) and the bispecific product generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R are shown in Figures 29C and 29D, respectively. Samples were diluted to 1 mg / mL in mobile phase A (10 mM NaPO4, pH 7.0) for injection into the HPLC. Differentially charged IgG molecules were separated using a ProPac® WCX-10, 4 mm x 250 mm analytical column at a flow rate of 1 mL / min. Mobile phase A was diluted to mobile phase B (10 mM NaPO4 Elution was performed with a gradient of 1000 mM NaCl (pH 7.0, 0.25 M NaCl) and detection was performed at 280 nm. These data show that bispecific antibody samples were generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R. They also show that cation exchange is a powerful tool to separate remaining homodimers from heterodimers. Another application of cation exchange chromatography is therefore the final polishing of bispecific heterodimers, i.e., purifying away any homodimers remaining after exchange.

[0297] Example 32: Recombinant expression of a heterodimer by simultaneous co-expression of both homodimers To illustrate that heterodimerization also occurs when two homodimers are recombinantly co-expressed, four expression vectors encoding the heavy and light chains of IgG1-7D8-K409R and IgG1-2F8-F405 were co-transfected into HEK-293F cells in a ratio of 1:1:1:1 (see Example 1). Antibodies were transiently produced under serum-free conditions as described in Example 4. IgG was then purified by Protein A chromatography as described in Example 5. Purified IgG was deglycosylated and then analyzed by electrospray ionization mass spectrometry as described in Example 12.

[0298] The theoretical masses of the heavy and light chains of IgG1-7D8-K409R and IgG1-2F8-F405 are shown in Table 6.

[0299] Table 6: Theoretical masses of the heavy and light chains of IgG1-7D8-K409R and IgG1-2F8-F405 TIFF2025063202000013.tif23140

[0300] Based on these masses the following IgG molecules could theoretically be detected (Table 7): The measured masses (Figure 30) are shown in the last column.

[0301] Table 7. Theoretical detection of heavy and light chains of IgG1-7D8-K409R and IgG1-2F8-F40 TIFF2025063202000014.tif64126

[0302] The two most abundant peaks at 146345 Da and 146159 Da corresponded to heterodimers incorporating one (from IgG1-7D8-K409R) or both light chains, respectively. Homodimers of both heavy chains of IgG1-7D8-K409R or IgG1-2F8-F405 were detected but only in small amounts. These data indicate that heterodimerization also occurs when the two homodimers are coexpressed.

[0303] Example 33: Monitoring the kinetics of 2-MEA-induced Fab arm exchange and quantification of remaining homodimers after exchange using HPLC-CIEX Generation of bispecific antibodies by 2-MEA-induced Fab arm exchange is described in Example 11, where the exchange reaction was monitored by high pressure liquid chromatography cation exchange (HPLC-CIEX; as described in Example 31) at various time points during the exchange reaction.

[0304] Homodimers IgG1-2F8-F405L and IgG1-7D8-K409R were mixed in a 1:1 molar ratio at a concentration of 1 mg / mL each. After addition of 25 mM 2-MEA, the samples were placed in a HPLC autosampler pre-warmed at 25° C. Figures 31A-31H show eight consecutive injections obtained by HPLC-CIEX at different time intervals ranging from t = 0 min to t = 450 min, respectively, after addition of 2-MEA. The data show that bispecific IgG was formed fairly quickly and that the majority of homodimers were exchanged after 135 min. The heterogeneous heterodimer peak appearing after 45 min resolved into a more homogeneous peak after approximately 180 min, suggesting that exchange was taking place in a different phase. Furthermore, Figure 32A shows that approximately 3% of remaining homodimers were detected by the CIEX method (indicated by the arrow). As shown, this method is suitable for quantifying the remaining homodimer content once the exchange reaction is near completion (elution of homodimer is shown in FIG. 32B).

[0305] Example 34: Generation of bispecific antibodies by 2-MEA-induced Fab arm exchange at high antibody concentrations with varying 2-MEA concentrations, temperatures and incubation times 2-MEA-induced Fab arm exchange was performed at high IgG concentrations, and the effects of 2-MEA concentration, incubation temperature and time on the amount of exchange were investigated.

[0306] The exchange process was performed using the combination IgG1-7D8-K409R x IgG1-2F8-F405L. Both materials were purified by affinity chromatography using Protein A. After the materials were concentrated to >20 mg / mL, a successive anion exchange step was performed (in flow-through mode) using a HiPrep Q FF 16 / 10 (GE Health Care, #28-9365-43). The final purified material was buffer exchanged into PBS.

[0307] Bispecific exchange was investigated at final IgG concentrations of 20 mg / mL (at a final concentration of 10 mg / mL of each homodimer) and 10 mg / mL (at a final concentration of 5 mg / mL of each homodimer) in PBS. Separate mixtures were prepared for both IgG concentrations containing 2-MEA at final concentrations of 10, 25, 50 and 100 mM. The mixtures were divided into aliquots of 100 μl in Eppendorf tubes and stored at 15, 25 and 37° C. Separate tubes were used for different incubation times at each temperature: 90 min, 5 h and 24 h.

[0308] Mixtures were also prepared without 2-MEA for both IgG concentrations and stored at 4°C as untreated controls. After appropriate incubation times, 90 min and 5 h samples were collected for desalting to remove 2-MEA (90 min samples were first placed on ice to stop the exchange reaction). Samples were desalted using Zeba 96-well desalting plates (7k, Cat. No. 89808, Thermo Fisher Scientific). 24 h samples were desalted separately after 24 h incubation.

[0309] Serial dilutions of antibody samples (total antibody concentrations 10-0.123 μg / mL in 3-fold dilutions for 90 min and 5 h samples; 10-0.041 μg / mL in 3-fold dilutions for 24 h samples) were used in sandwich ELISA to measure bispecific binding as described in Example 7. For each plate, a control of a purified batch of bispecific antibody derived from 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL and IgG4-7D8-CPPC (as described in Example 15) was included. Figures 34(A)-(F) show the results of bispecific binding measured in individual ELISA plates. The highest OD405 value (determined for a concentration of 10 μg / mL in ELISA) was used to calculate bispecific binding relative to the control, which was arbitrarily set to 100%. This resulted in the percentage of controlled Fab arm exchange (% cFAE) relative to the control as shown in Figures 34(A)-(D) for each 2-MEA concentration.

[0310] The data show that the maximum level of bispecific binding (89-109% relative to the control) was reached at a concentration of 100 mM 2-MEA for both IgG concentrations at all temperature-time conditions. At 50 mM 2-MEA, maximum binding (88-107%) was reached at 25°C and 37°C, and also at 15°C after 24 hours of incubation. For the lower concentrations of 25 mM and 10 mM 2-MEA, the exchange was more efficient at higher temperatures and increased with prolonged incubation times, leading to maximum exchange at 37°C with 24 hours of incubation with 25 mM 2-MEA. None of the conditions tested with 10 mM 2-MEA produced 100% bispecific product. The exchange process was slightly faster at an IgG concentration of 10 mg / mL compared to 20 mg / mL total IgG.

[0311] To confirm that bispecific antibodies were formed and to examine the bispecific products in more detail, samples were analyzed by cation exchange (HPLC-CIEX) analysis. HPLC-CIEX analysis was performed as described in Example 31 on samples at 20 mg / mL IgG concentration after 5 and 24 hours of incubation and at all 2-MEA concentrations.

[0312] The CIEX chromatograms in Figure 35(A)-(D) show that the highest yields of bispecific product were obtained with 50 and 100 mM 2-MEA, confirming the bispecific ELISA results. However, small amounts of residual homodimers were still detected with 50 and 100 mM 2-MEA (2-3.5% homodimers, respectively, in samples incubated at 25°C and 37°C). Exchange at higher concentrations, longer (24 h) incubation times and increasing 2-MEA concentrations leads to the appearance of an additional peak at 22-24 min in the CIEX profile.

[0313] Minimal amounts of additional peaks were obtained when the exchange was complete within 5 hours. SDS-PAGE and HP-SEC analyses were performed to identify the nature of these peaks. HP-SEC showed that the amount of aggregates was less than 1% for all conditions, suggesting that the additional peaks do not represent aggregates. However, non-reducing SDS-PAGE suggested that the additional peaks may represent heterodimers missing one or two light chains. Small amounts of half molecules were detected as well.

[0314] Experiments show that the exchange reaction takes place at high homodimer concentrations, making this process attractive on a commercial scale, and that the yield of bispecific antibodies depends on the 2-MEA concentration, temperature and incubation time.

[0315] Example 35: Determinants at IgG1 position 368 for participation in 2-MEA-induced Fab arm exchange in combination with IgG1-K409R In examples 28 and 29, we show that certain single mutations at positions F405 and Y407 are sufficient to allow human IgG1 to participate in Fab arm exchange when combined with IgG1-K409R. As illustrated in this example, further determinants involved in the location of the Fc:Fc interface in the CH3 domain may also mediate the Fab arm exchange mechanism. To this effect, mutagenesis at IgG1 position 368 was performed and this mutant was examined for its participation in 2-MEA-induced Fab arm exchange in combination with human IgG1-K409R. All possible IgG1-2F8-L368X mutants (except C and P) were combined with IgG1-7D8-K409R. The procedure was performed with purified antibodies as described in example 19.

[0316] Figure 36 shows the results of bispecific binding by 2-MEA-induced Fab arm exchange between IgG1-2F8-L368X x IgG1-7D8-K409R. These data were also scored as (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange or (++) high Fab arm exchange, as presented in Table 8. No Fab arm exchange (-) was observed when position 368 in IgG1-2F8 was L (= wild type IgG1), F or M. Fab arm exchange was found to be low (+ / -) when position 368 in IgG1-2F8 was Y. Fab arm exchange was found to be moderate (+) when position 368 in IgG1-2F8 was K and high (++) when position 368 in IgG1-2F8 was A, D, E, G, H, I, N, Q, R, S, T, V or W. These data suggest that specific mutations at IgG1 position 368, when combined with IgG1-K409R, engage IgG1 in 2-MEA-induced Fab arm exchange.

[0317] Table 8: 2-MEA-induced Fab arm exchange between IgG1-2F8-L368X mutant and IgG1-7D8-K409R Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between IgG1-2F8-L368X mutant and IgG1-7D8-K409R was determined by sandwich ELISA: (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange, (++) high Fab arm exchange. TIFF2025063202000015.tif110140

[0318] Example 36: Determinants at IgG1 position 370 for participation in 2-MEA-induced Fab arm exchange in combination with IgG1-K409R In examples 28, 29 and 35, we show that certain single mutations at positions F405, Y407 or L368 are sufficient to allow human IgG1 to participate in Fab arm exchange when combined with IgG1-K409R. As illustrated in this example, further determinants involved in the location of the Fc:Fc interface in the CH3 domain may also mediate the Fab arm exchange mechanism. To this effect, mutagenesis at IgG1 position 370 was performed and this mutant was examined for its participation in 2-MEA-induced Fab arm exchange in combination with human IgG1-K409R. All possible IgG1-2F8-K370X mutants (except C and P) were combined with IgG1-7D8-K409R. The procedure was performed with purified antibodies as described in example 19.

[0319] Figure 37 shows the results of bispecific binding by 2-MEA-induced Fab arm exchange between IgG1-2F8-K370X x IgG1-7D8-K409R. These data were also scored as (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange or (++) high Fab arm exchange, as presented in Table 9. No Fab arm exchange (-) was observed when position 370 in IgG1-2F8 was K (= wild type IgG1), A, D, E, F, G, H, I, L, M, N, Q, R, S, T, V or Y. Only replacement of K370 with W resulted in moderate Fab arm exchange (+). These data suggest that only one mutation at IgG1 position 370 (K370W) can engage IgG1 in 2-MEA-induced Fab arm exchange when combined with IgG1-K409R.

[0320] Table 9. 2-MEA-induced Fab arm exchange between IgG1-2F8-K370X mutant and IgG1-7D8-K409R Bispecific antibody generation after 2-MEA-induced in vitro Fab arm exchange between IgG1-2F8-K370X mutant and IgG1-7D8-K409R was determined by sandwich ELISA: (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange, (++) high Fab arm exchange. TIFF2025063202000016.tif110140

[0321] Example 37: Determinants at IgG1 position 399 for participation in 2-MEA-induced Fab arm exchange in combination with IgG1-K409R In examples 28, 29, 35 and 36, we show that certain single mutations at positions F405, Y407, L368 or K370 are sufficient to allow human IgG1 to participate in Fab arm exchange when combined with IgG1-K409R. As illustrated in this example, further determinants involved in the location of the Fc:Fc interface in the CH3 domain may also mediate the Fab arm exchange mechanism. To this effect, mutagenesis at IgG1 position 399 was performed and this mutant was examined for its participation in 2-MEA-induced Fab arm exchange in combination with human IgG1-K409R. All possible IgG1-2F8-D399X mutants (except C and P) were combined with IgG1-7D8-K409R. The procedure was performed with purified antibodies as described in example 19.

[0322] Figure 38 shows the results of bispecific binding by 2-MEA-induced Fab arm exchange between IgG1-2F8-D399X x IgG1-7D8-K409R. These data were also scored as (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange or (++) high Fab arm exchange, as presented in Table 10. No Fab arm exchange (-) was observed when position 399 in IgG1-2F8 was D (= wild type IgG1), E and Q. Fab arm exchange was found to be low (+ / -) when position 399 in IgG1-2F8 was V and moderate (+) when position 399 in IgG1-2F8 was G, I, L, M, N, S, T or W. Fab arm exchange was found to be high (++) when position 399 in IgG1-2F8 was A, F, H, K, R or Y. These data suggest that specific mutations at IgG1 position 399, when combined with IgG1-K409R, engage IgG1 in 2-MEA-induced Fab arm exchange.

[0323] Table 10: 2-MEA induced Fab arm exchange between IgG1-2F8-D399X mutants and IgG1-7D8-K409R. Generation of bispecific antibodies following 2-MEA induced in vitro Fab arm exchange between IgG1-2F8-D399X mutants and IgG1-7D8-K409R was determined by sandwich ELISA. (-) no Fab arm exchange, (+ / -) low Fab arm exchange, (+) moderate Fab arm exchange, (++) high Fab arm exchange. TIFF2025063202000017.tif110140

[0324] Example 38: Determination of the range of conditions under which 2-MEA-induced Fab arm exchange is performed suboptimally to identify highly efficient IgG1 mutants The process of 2-MEA-induced Fab arm exchange is efficiently carried out at 37° C. with 25 mM 2-MEA. Under these conditions, the majority of permissive IgG1 mutants (IgG1s with certain single mutations at positions 368, 370, 399, 405 and 407 and / or 409 as described in Examples 19, 28, 29 and 35-37) show high levels of 2-MEA-induced Fab arm exchange (80%-100%). To identify experimental conditions that may allow discrimination between IgG1 mutants with the highest efficiency, 2-MEA-induced Fab arm exchange for four different mutant combinations (IgG1-2F8-F405S x IgG1-7D8-K409A, IgG1-2F8-D399R x IgG1-7D8-K409G, IgG1-2F8-L368R x IgG1-7D8-K409H and IgG1-2F8-F405L x IgG1-7D8-K409R) was investigated over time at 15° C. and 20° C., respectively. Apart from the variation of temperature, duration and antibody dilution (20, 2, 0.2 and 0.02 μg / mL), the procedure was performed as described in Example 19.

[0325] At 20°C, 2-MEA-induced Fab arm exchange of the four mutant combinations was performed at various percentages compared to maximum exchange (positive control). After 105 min of incubation, IgG1-2F8-L368R × IgG1-7D8-K409H reached maximum exchange levels, while IgG1-2F8-F405S × IgG1-7D8-K409A, IgG1-2F8-D399R × IgG1-7D8-K409G and IgG1-2F8-F405L × IgG1-7D8-K409R reached a maximum of 90%, 85% and 85%, respectively, after 200 min.

[0326] Incubation of the different IgG1 mutant combinations at 15° C. showed the most striking differences in the exchange rates (shown in FIG. 39). After 60 and 105 min of incubation, 2-MEA-induced Fab arm exchange, the differences between the four mutant combinations were the most extreme. After 200 min of incubation, Fab arm exchange showed 100% efficiency (IgG1-2F8-L368R×IgG1-7D8-K409H), 85% efficiency (IgG1-2F8-F405L×IgG1-7D8-K409R and IgG1-2F8-D399R×IgG1-7D8-K409G) or 65% efficiency (IgG1-2F8-F405S×IgG1-7D8-K409A) compared to the positive control.

[0327] Example 39: Analysis of 2-MEA-induced Fab arm exchange efficiency of mutants under suboptimal conditions The process of 2-MEA-induced Fab arm exchange is efficiently carried out at 37° C. when 25 mM 2-MEA is used. Under these conditions, the majority of permissive IgG1 mutants (IgG1 with certain single mutations at positions 368, 370, 399, 405 and 407 and / or 409 as described in Examples 19, 28, 29 and 35-37) show high levels of 2-MEA-induced Fab arm exchange (80-100%). In Example 38, it was described that the differences in 2-MEA-induced Fab arm exchange efficiency are most pronounced under so-called suboptimal conditions, i.e. after incubation at 15° C. for 60-105 min. A total of 24 IgG1-2F8 mutants (see Table 11) at L368, D399, F405 and Y407 showing more than 90% 2-MEA-induced Fab arm exchange with IgG1-7D8-K409R (Examples 28, 29 and 35-37) were selected and subjected (based on the results reported in Example 19) to Fab arm exchange analysis with IgG1-7D8-K409A, G, H or R. To classify these mutant combinations by their efficiency in generating bispecific antibodies, 2-MEA-induced Fab arm exchange was performed for 90 min at 15° C. (suboptimal conditions). Two IgG1-2F8 mutants Y407Q and D399Q (Examples 29 and 37), which showed weak 2-MEA-induced Fab arm exchange after incubation with IgG1-7D-K409R, were introduced and used as further negative controls to investigate whether incubation with another amino acid (G, H, or W) at position K409 would lead to different results. Apart from varying the temperature and antibody dilution (20, 2, 0.2 and 0.02 ug / mL), the procedure was performed as described in Example 19.

[0328] Incubation of all different IgG1 mutant combinations for 90 min at 15° C. showed a range of different 2-MEA-induced Fab arm exchange efficiencies (as evident from Table 11). The bispecific binding results at an antibody concentration of 20 μg / mL are shown in Table 11. As specified in the legend to Table 11 below, the results were classified into four categories; zero (-), low (+ / -), moderate (+) and high (++) bispecific binding efficiency. These results reveal that under suboptimal conditions, several combinations of amino acid mutations in IgG1 molecules favor 2-MEA-induced Fab arm exchange.

[0329] Table 11. Bispecific binding between permissive IgG1 variants (20 μg / mL) at 15° C. for 90 min (% relative to positive control) TIFF2025063202000018.tif244161TIFF2025063202000019.tif36162

[0330] From the mutant IgG1-2F8 molecules tested (Table 11), six were selected for a secondary analysis to confirm the results obtained previously (Table 11). Some mutants were selected for their high 2-MEA-induced Fab arm exchange efficiency (IgG1-2F8-L368R) as well as intermediate 2-MEA-induced Fab arm exchange efficiency (IgG1-2F8-L368W, IgG1-2F8-F405I, IgG1-2F8-F405L and IgG1-2F8-Y407W). IgG1-2F8-Y407Q was also analyzed a second time, since it showed an unexpected positive 2-MEA-induced Fab arm exchange reaction with IgG1-7D8-K409H. Overall, these results presented in Figure 40 confirm the primary analysis (Table 11) and show that the 2-MEA-induced Fab arm exchange reaction of mutant IgG1-2F8 molecules with IgG1-7D8-K409H showed the highest efficiency. Furthermore, the 2-MEA-induced Fab arm exchange reaction between mutant IgG1-2F8 molecules with IgG1-7D8-K409R, which is reported as negative in Examples 28, 29 and 35-37, is nevertheless interesting as it potentially promotes the 2-MEA-induced Fab arm exchange of IgG1.

[0331] Example 40: Use of bispecific formats to remove unwanted agonist activity of antagonist c-Met antibodies and convert them into monovalent, bispecific formats Some bivalent antibodies developed for monoclonal antibody therapy exhibit undesirable agonist activity upon binding to their target. This is also the case for most IgG1-based antibodies targeting the receptor tyrosine kinase c-Met. These agonist antibodies induce receptor dimerization with subsequent activation of several downstream signaling pathways. As a consequence, (tumor) cell proliferation and differentiation are induced. By using a monovalent antibody format, the induction of receptor dimerization can be prevented. The combination of an anti-c-Met antibody Fab arm with the Fab arm of an unrelated antibody results in a functionally monovalent and therefore fully antagonistic bispecific antibody. Here, we combined some (IgG1-069) or all (IgG1-058) agonist antibodies with IgG1-b12 (originally described in Burton DR, et al, "Efficient neutralization of primary isolates of HIV-1 by a recombinant human monoclonal antibody", Science. 1994 Nov 11; 266(5187):1024-1027) in a bispecific antibody. IgG1-b12 was raised against a viral protein (HIV-gp120) and was therefore considered an irrelevant non-binding antibody. The anti-c-Met antibody used in this example is a fully human monoclonal antibody generated in transgenic mice. IgG1-058 and IgG1-069 bind to different epitopes on c-Met.

[0332] The two anti-c-Met antibodies used are IgG1,κ antibodies with modified Fc regions, as further disclosed. They have the following heavy and light chain variable sequences: TIFF2025063202000020.tif89142

[0333] Receptor phosphorylation Monovalent bispecific c-Met antibodies were generated by Fab arm exchange reaction with IgG1-058-F405L or IgG1-069-F405L and IgG1-b12-K409R as described in Example 23 using 25 mM 2-MEA. The effect of bispecific antibodies on c-Met phosphorylation was evaluated. Upon dimerization of two adjacent c-Met receptors by the natural ligand HGF or by an agonistic bivalent antibody, three tyrosine residues (positions 1230, 1234 and 1235) in the intracellular domain of c-Met are cross-phosphorylated. This leads to phosphorylation of several other amino acids in the intracellular domain of c-Met and activation of several signaling cascades. Dimerization and activation of c-Met can be monitored by using antibodies specific for the receptor phosphorylated at these positions, which serve as a readout for the potential receptor activating effect of anti-c-Met antibodies.

[0334] A549 cells, CCL-185 obtained from ATCC, were grown in serum-containing DMEM medium until they reached 70% confluency. The cells were trypsinized, washed, and cultured in serum-containing medium for 1 h. *Cells were plated at 10e6 cells / well in 6-well culture plates. After overnight incubation, cells were treated with HGF (R&D systems; Cat. No. 294-HG) (50 ng / mL) or a panel of antibodies (30 μg / mL) and incubated for 15 min at 37°C. Cells were washed twice with ice-cold PBS and lysed in lysis buffer (Cell Signaling; Cat. No. 9803) supplemented with a protease inhibitor cocktail (Roche; Cat. No. 11836170001). Cell lysate samples were stored at -80°C. Receptor activation was determined by detection of c-Met phosphorylation on Western blot using a phospho-c-Met specific antibody. Proteins present in the cell lysates were separated on a 4-12% SDS-PAGE gel and transferred to a nitrocellulose membrane, which was then stained with an antibody specific for phosphorylated c-Met (Y1234 / 1235) (Cell Signaling, Catalog No: 3129). As a control for gel loading, total β-actin and c-Met levels were determined using anti-c-Met antibody (Cell Signaling, Catalog No. 3127) and anti-β-actin antibody (Cell Signaling, Catalog No. 4967). Western blot results are shown in Figure 41.

[0335] Cells treated with tissue culture medium control and the monovalent form of antibody 5D5 (Genentech; WO 96 / 38557) UniBody® (Genmab, WO2007059782 and WO2010063785) did not show any c-Met receptor phosphorylation. The monovalent UniBody form used herein is an IgG4 with a deleted hinge region and mutated CH3 region at positions 405 and 407. In contrast, Western blot analysis of cells treated with positive control HGF or agonist antibody IgG1-058 showed a clear band at the expected height of phosphorylated c-Met. Partial agonist antibody IgG1-069 showed lower but detectable receptor phosphorylation, suggesting that some cross-linking of the receptor occurs. However, neither the bispecific IgG1 058 / b12 nor the bispecific 069 / b12 antibodies induced any c-Met phosphorylation, indicating a complete lack of agonistic activity associated with the parent antibodies (Figure 41).

[0336] Effect of c-Met antibody on the proliferation of NCI-H441 in vitro The potential proliferative agonist activity of c-Met antibodies was tested in vitro in the lung adenocarcinoma cell line NCI-H441 (ATCC, HTB-174™). This cell line expresses high levels of c-Met but does not produce its ligand HGF. NCI-H441 cells were seeded in serum-free RPMI (Lonza) in 96-well tissue culture plates (Greiner bio-one, Frickenhausen, Germany) (5,000 cells / well). Anti-c-Met antibodies were diluted to 66.7 nM in serum-free RPMI and added to the cells. Incubated at 37°C / 5% CO 2After 7 days of incubation at 4°C, the amount of viable cells was quantified with Alamar Blue (BioSource International, San Francisco, US) according to the manufacturer's instructions. Fluorescence was monitored using an EnVision 2101 Multilabel reader (PerkinElmer, Turku, Finland) with standard Alamar Blue settings.

[0337] In contrast to IgG1-069, incubation of NCI-H441 cells with bispecific IgG1-069 / b12 did not induce proliferation, as shown in Figure 42. Also, the UniBody-069 control did not induce proliferation, which was comparable to no treatment or IgG1-b12 treatment.

[0338] Example 41: CDC-mediated cell killing by bispecific antibodies generated by 2-MEA-induced Fab arm exchange between human IgG1-2F8-F405L or IgG1-7D8-F405L and IgG1-7D8-K409R The CD20 antibody IgG1-7D8 can efficiently kill CD20-expressing cells by complement-dependent cytotoxicity (CDC). In contrast, the EGFR antibody IgG1-2F8 does not mediate CDC against target cells expressing EGFR. Both IgG1-7D8-K409R and the bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-F405L x IgG1-7D8-K409R can induce CDC against CD20-expressing cells (as described in Example 26). We tested whether the bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG1-7D8-F405L and IgG1-7D8-K409R can similarly induce CDC against CD20-expressing cells. 10 5Daudi or Raji cells were preincubated with a concentration series of antibodies in 100 μL RPMI medium supplemented with 0.1% BSA for 15 min on a shaker at room temperature. 25 μL normal human serum (NHS) was added as a source of complement (20% NHS final concentration) and incubated at 37°C for 45 min. After incubation, the plates were placed on ice to stop the CDC reaction. 10 μL of 10 μg / mL propidium iodide (PI) (0.6 μg / mL final concentration) was added and dead and live cells were differentiated by FACS analysis.

[0339] Figure 43 shows that the bispecific product generated by 2-MEA-induced Fab arm exchange between IgG1-7D8 and IgG1-7D8-F405L and IgG1-7D8-K409R has the same potency to induce CDC-mediated cell killing of CD20-expressing Daudi (Figure 43A) and CD20-expressing Raji (Figure 43B). Neither Daudi nor Raji cells express EGFR, causing monovalent binding of the bispecific antibody generated by 2-MEA-induced Fab arm exchange between IgG2-2F8-F405L x IgG1-7D8-K409R. This bispecific product also induced CDC-mediated cell killing, although with slightly lower efficiency. These data suggest that the CDC capacity of the parent antibody was maintained in the bispecific format. The bivalent bispecific product (IgG1-7D8-F405L x IgG1-7D8-K409R) induced CDC-mediated cell killing slightly more efficiently than the monovalent bispecific product (IgG2-2F8-F405L x IgG1-7D8-K409R). The 11B8 antibody targeting CD20 is unable to induce CDC-mediated cell killing and serves as a negative control.

[0340] Example 42: HER2xHER2 bispecific antibodies tested in an in vitro kappa-targeted ETA' killing assay The examples show that HER2xHER2 bispecific antibodies can deliver cytotoxic agents to tumor cells after internalization in a universal in vitro cell-based killing assay using kappa-targeted Pseudomonas aeruginosa exotoxin A (anti-kappa-ETA'). The assay utilizes a high affinity anti-kappa domain antibody conjugated to a truncated form of Pseudomonas aeruginosa exotoxin A. Similar fusion proteins of antibody binding proteins (IgG binding motifs from streptococcal protein A or protein G) and diphtheria toxin or Pseudomonas aeruginosa exotoxin A have been described (Mazor Y. et al., J. Immunol. Methods 2007; 321: 41-59); Kuo SR. et al., 2009 Bioconjugate Chem. 2009; 20:1975-1982). In contrast to anti-kappa-ETA', these molecules bound the Fc portion of the complete antibody. Upon internalization and endocytic sorting, anti-κ-ETA' domain antibodies undergo proteolysis and reduction of disulfide bonds, separating the catalytic domain from the binding domain. The catalytic domain is then transported from the Golgi to the endoplasmic reticulum by the KDEL retention motif and subsequently translocated to the cytosol where it inhibits protein synthesis and induces apoptosis (Kreitman RJ . et. al., BioDrugs 2009; 23: 1-13).

[0341] The anti-HER2 antibodies used in this Example and in Examples 43-45 below are fully human monoclonal antibodies generated in transgenic mice. They bind to different epitopes on HER2.

[0342] They are all IgG1,κ antibodies whose Fc regions have been modified, as further disclosed. They have the following heavy and light chain variable sequences: TIFF2025063202000021.tif167165

[0343] The HER2xHER2 bispecific antibody was preincubated with anti-κ-ETA' prior to incubation with A431 cells, which express approximately 15,000 HER2 antibodies per cell (as determined by Qifi analysis) and are not sensitive to treatment with "naked" HER2 antibodies.

[0344] First, the optimal concentration of anti-κ-ETA' was determined for each cell line, i.e. the maximum tolerated dose that did not result in the induction of nonspecific cell death. A431 cells (2500 cells / well) were seeded in normal cell culture medium in 96-well tissue culture plates (Greiner bio-one) and allowed to adhere for at least 4 hours. The cells were incubated with a dilution series of anti-κ-ETA' at 100, 10, 1, 0.1, 0.01, 0.001 and 0 μg / mL in normal cell medium. After 3 days, the amount of viable cells was quantified with Alamar Blue (BioSource International, San Francisco, US) according to the manufacturer's instructions. Fluorescence was monitored using an EnVision 2101 Multilabel reader (PerkinElmer, Turku, Finland) with standard Alamar Blue settings. The highest concentration of anti-κ-ETA' that did not kill cells with anti-κ-ETA' alone (1 μg / mL in the case of A431 cells) was used in the following experiments.

[0345] Next, the effect of HER2xHER2 bispecific and HER2 monospecific antibodies preincubated with anti-κ-ETA' was tested for their ability to induce cell death. A431 cells were seeded as described above. A dilution series of HER2 specific antibodies (monospecific and bispecific) was made and added to the cells after preincubation with the indicated concentrations of anti-κ-ETA' for 30 min. After 3 days of incubation at 37°C, the amount of viable cells was quantified as described above. The Alamar Blue signal of cells treated with anti-κ-ETA' preincubated with the antibody was plotted compared to cells treated without antibody treatment. EC was calculated using GraphPad Prism 5 software. 50Values ​​and maximum cell death were calculated. Staurosporine (23.4 μg / mL) was used as a positive control for cell killing. An isotype control antibody (IgG1 / κ; IgG1-3G8-QITL) was used as a negative control.

[0346] Figure 44 shows that all anti-κ-ETA' pre-incubated HER2 bispecific antibodies were able to kill A431 cells in a dose-dependent manner. These results demonstrate that most HER2 bispecific antibodies tested were more effective than the monospecific antibodies included in the combination in this anti-κ-ETA' assay. Furthermore, the potency of bispecific antibodies 005X169, 025X169 and 153X169 demonstrated that the potency of a monospecific antibody lacking activity in this in vitro kappa-targeted ETA' killing assay, namely HER2-specific antibody 169, can be increased through bispecific combination with another HER2-specific antibody.

[0347] Example 43: Down-modulation of the HER2 receptor by incubation with bispecific antibodies targeting different HER2 epitopes HER2xHER2 bispecific antibodies can bind two different epitopes on two spatially distinct HER2 receptors. This allows other HER2xHER2 bispecific antibodies to bind to the remaining epitopes on these receptors. This can lead to multivalent receptor cross-linking (compared to the dimerization induced by monovalent antibodies), which can result in enhanced down-regulation of the receptor. To examine whether HER2xHER2 bispecific antibodies induce enhanced down-regulation of HER2, AU565 cells were incubated with antibodies and bispecific antibodies for 3 days. Total HER2 levels and antibody-bound HER2 levels were determined.

[0348] AU565 cells were seeded in 24-well tissue culture plates (100.000 cells / well) in normal cell medium and cultured for 3 days at 37°C in the presence of 10 μg / mL of HER2 antibody or HER2×HER2 bispecific antibody. After washing with PBS, cells were lysed by incubating them with 25 μL of Surefire Lysis buffer (Perkin Elmer, Turku, Finland) for 30 min at room temperature. Total protein levels were quantified using bicinchoninic acid (BCA) protein assay reagent (Pierce) according to the manufacturer's protocol. HER2 protein levels in the lysates were analyzed using a HER2-specific sandwich ELISA. HER2 was captured using a rabbit anti-human HER2 intracellular domain antibody (Cell Signaling) and bound HER2 was detected using a biotinylated goat anti-human HER2 polyclonal antibody (R&D systems, Minneapolis, USA) followed by streptavidin-poly-HRP. The reaction was visualized with 2,2'-azino-bis-3-ethylbenzothiazoline-6-sulfonic acid (1 ABTS tablet diluted in 50 mL of ABTS buffer [Roche Diagnostics, Almere, The Netherlands]) and stopped with oxalic acid (Sigma-Aldrich, Zwijndrecht, The Netherlands). Fluorescence at 405 nm was measured in a microtiter plate reader (Biotek Instruments, Winooski, USA) and the amount of HER2 was expressed as a percentage of untreated cells.

[0349] The results are shown in Figure 45. The figure demonstrates that all HER2xHER2 bispecific antibodies tested induced 40% or more of HER2 downregulation. Interestingly, all HER2xHER2 bispecific antibodies demonstrated increased HER2 downregulation compared to both of their monospecific counterparts.

[0350] Example 44: Colocalization of HER2xHER2 bispecific antibodies with the lysosomal marker LAMP1 analyzed by confocal microscopy HER2 downregulation assays as described in Example 43 suggested that HER2xHER2 bispecific antibodies could increase lysosomal degradation of HER2. To confirm these findings, confocal microscopy techniques were applied. AU565 cells were grown on glass coverslips (1.5 micron thickness, Thermo Fisher Scientific, Braunschweig, Germany) in standard tissue culture medium for 3 days at 37°C. Cells were preincubated with leupeptin (Sigma) for 1 hour to block lysosomal activity, after which 10 ug / mL of HER2 monospecific or HER2xHER2 bispecific antibodies were added. Cells were incubated for another 3 or 18 hours at 37°C. They were then washed with PBS and incubated with 4% formaldehyde (Klinipath) for 30 minutes at room temperature. Slides were washed with blocking buffer (PBS supplemented with 0.1% saponin [Roche] and 2% BSA [Roche]) and incubated with 20 mM NH 4Formaldehyde was quenched by incubation with blocking buffer containing Cl for 20 min. Slides were washed again with blocking buffer and incubated with mouse anti-human CD107a (LAMP1) (BD Pharmingen) for 45 min at room temperature to stain lysosomes. After washing with blocking buffer, slides were incubated with a cocktail of secondary antibodies; goat anti-mouse IgG-Cy5 (Jackson) and goat anti-human IgG-FITC (Jackson) for 30 min at room temperature. Slides were washed again with blocking buffer and mounted overnight on microscope slides with 20 μL of mounting medium (6 grams of glycerol [Sigma] and 2.4 grams of Mowiol 4-88 [Omnilabo] were dissolved in 6 mL of distilled water to which 12 mL of 0.2 M Tris [Sigma] pH 8.5 was added, followed by incubation at 50-60 °C for 10 min. The mounting medium was aliquoted and stored at -20 °C). Slides were imaged with a Leica SPE-II confocal microscope (Leica Microsystems) equipped with a 63×1.32-0.6 oil immersion objective and LAS-AF software. To allow quantification of overlapping pixel intensities, pixel saturation should be avoided. Therefore, the FITC laser intensity was reduced to 10%, smart gain was set to 830 V, and smart offset was set to −9.48%. Using these settings, bispecific antibodies were clearly visualized without pixel saturation, whereas monospecific antibodies were sometimes difficult to detect. To compare lysosomal colocalization between monospecific and bispecific antibodies, these settings were kept the same for all confocal slides analyzed.

[0351] 12-bit images were analyzed for colocalization using MetaMorph® software (Meta Series 6.1 version, Molecular Devices Inc, Sunnyvale California, USA). FITC and Cy5 images were imported as stacks and background subtracted. Identical threshold settings were used for all FITC and all Cy5 images (set manually). Colocalization was depicted as pixel intensity of FITC in the overlap region (ROI), where the ROI is composed of all Cy5-positive areas. To compare different slides stained with several HER2 antibodies or HER2xHER2 bispecific antibodies, images were normalized using pixel intensity of Cy5. Goat anti-mouse IgG-Cy5 was used to stain the lysosomal marker LAMP1 (CD107a). The pixel intensity of LAMP1 should not differ between the different HER2 antibodies or HER2xHER2 bispecific antibodies tested (one cell had a Cy5 pixel intensity of approximately 200.000).

[0352] Normalization value for FITC and Cy5 colocalization = [(TPI-FITC×%FITC-Cy5 colocalization) / 100]×[200.000 / TPI-Cy5]

[0353] In this formula, TPI represents total pixel intensity.

[0354] The above formula represents the percentage of live cells as measured by FITC pixel intensity overlapping with Cy5 for various HER2 monospecific antibodies and HER2xHER2 bispecific antibodies. For each depicted antibody or bispecific molecule, three different images were analyzed from one slide containing approximately 1, 3 or more than 5 cells. Significant differences were observed between different images within each slide. However, it was clear that all HER2xHER2 bispecific antibodies showed increased colocalization with the lysosomal marker LAMP1 when compared to their monospecific counterparts. These results suggest that HER2xHER2 bispecific antibodies, once internalized, are efficiently localized towards the lysosomal compartment, making them suitable for bispecific antibody drug conjugate approaches.

[0355] Example 45: Inhibition of proliferation of AU565 cells by incubation with HER2 monospecific antibody or HER2xHER2 bispecific antibody HER2 bispecific antibodies were tested for their ability to inhibit the proliferation of AU565 cells in vitro. Due to the high HER2 expression level on AU565 cells (approximately 1.000.000 copies per cell as determined by the Qifi kit), HER2 is constitutively active in these cells and thus does not depend on ligand-induced heterodimerization. AU565 cells were seeded at 9.000 cells per well in 96-well tissue culture plates (Greiner bio-one, Frickenhausen, Germany) in the presence of 10 μg / mL of HER2 antibody or HER2×HER2 bispecific antibody in serum-free cell culture medium. As a control, cells were seeded in serum-free medium without antibody or bispecific antibody. After 3 days, the amount of viable cells was quantified using Alamar Blue (BioSource International, San Francisco, US) according to the manufacturer's instructions. Fluorescence was monitored using an EnVision 2101 Multilabel reader (PerkinElmer, Turku, Finland) with standard Alamar Blue settings. Alamar Blue signal of antibody-treated cells was plotted as a percentage of untreated cells.

[0356] Figure 47 depicts the Alamar Blue fluorescence intensity of AU565 cells after incubation with HER2 antibodies and HER2xHER2 bispecific antibodies. Herceptin® (Trastuzumab) was included as a positive control to demonstrate inhibition of proliferation as described by Juntilla TT. et al., Cancer Cell 2009; 15: 429-440. All HER2xHER2 bispecific antibodies were able to inhibit proliferation of AU565 cells. Bispecific antibodies: IgG1-005-ITL x IgG1-169-K409R and IgG1-025-ITL x IgG1-005-K409R were effective in this assay compared to their monospecific antibody counterparts.

[0357] Example 46: In vitro and in vivo analysis of FcRn binding by hinge-deleted IgG1 bispecific and bispecific IgG1 antibodies containing one or two FcRn sites in the Fc region This example illustrates the generation of asymmetric bispecific molecules, i.e. molecules with different features in each Fab arm, according to the present invention.

[0358] The neonatal Fc receptor (FcRn) is responsible for the long plasma half-life of IgG by protecting it from degradation. After antibody internalization, FcRn binds to the antibody Fc region in endosomes where the interaction is stable in a slightly acidic environment (pH 6.0). Upon recycling to the plasma membrane where the environment is neutral (pH 7.4), internalization is lost and the antibody is once again released into the circulation. The antibody Fc region contains two FcRn binding sites, one for each heavy chain at the CH2-CH3 interface. The H435A mutation in the antibody Fc region abolishes binding to FcRn (Shields, RL, et al, J Biol Chem, 2001, Firan, M., et al, Int Immunol, 2001), and the hinge region appears to affect FcRn binding (Kim, JK, et al., Mol Immunol., 1995). Furthermore, a role for bivalent over monovalent antibody binding to FcRn has been suggested in efficient recycling (Kim, JK, et al., Scand J Immunol., 1994).

[0359] In this example, the influence of FcRn binding valency is assessed with an asymmetric bispecific IgG1 molecule, which contains a single FcRn binding site, and the additional contribution of the hinge region is assessed with an asymmetric bispecific hinge-deleted IgG1 (Uni-G1) molecule.

[0360] FcRn binding of bispecific IgG1 molecules containing zero, one or two FcRn binding sites or hinge-deleted IgG1 (Uni-G1) molecules was measured by human and mouse FcRn ELISA. The antibodies IgG1-2F8-ITL, IgG1-7D8-K409R and IgG1-7D8-K409R-H435A monospecific molecules were generated as described in Examples 2, 3, 4 and 5. The hinge-deleted IgG1 molecules Uni-G1-2F8-ITL, Uni-G1-7D8-K409R and Uni-G1-7D8-K409R-H435A monospecific molecules were generated as described in Example 11. Bispecific IgG1 molecules were generated by 2-MEA-induced Fab arm exchange between IgG1-2F8-ITL and IgG1-7D8-K409R or IgG1-7D8-K409R-H435A molecules. Bispecific hinge-deleted IgG1 molecules were generated by incubation of Uni-G1-2F8-ITL with Uni-G1-7D8-K409R or Uni-G1-7D8-K409R-H435A. Monospecific and bispecific IgG1 and hinge-deleted IgG1 molecules in a 3-fold dilution series were added to biotinylated human or mouse FcRn captured on streptavidin-coated elisa plates, followed by incubation at pH 6.0 and 7.4 for 1 hour. Bound antibodies and hinge-deleted IgG1 molecules were incubated with horseradish peroxidase-labeled goat anti-human (Fab') as a conjugate. 2 and visualized using ABTS as substrate. Results were measured as optical density at a wavelength of 405 nm using an EL808-Elisa reader.

[0361] Figure 48 shows the binding results of monospecific or bispecific IgG1 molecules and hinge-deleted IgG1 molecules to human FcRn (A) and mouse FcRn (B) at pH 6.0 and pH 7.4. As expected, all antibodies tested, neither (bispecific) nor hinge-deleted IgG1 molecules, bind efficiently to FcRn (both human and mouse) at pH 7.4. At weakly acidic conditions (pH 6.0), monospecific IgG1-2F8-ITL and bispecific IgG1 made from IgG1-2F8-ITL and IgG1-7D8-K409R show bivalent binding efficiency to FcRn, reproducing the positive control for FcRn binding (IgG1-2F8), albeit three times higher for mouse FcRn compared to human. This suggests that the ITL mutations and K409R do not interfere with binding to FcRn.

[0362] When comparing the binding of IgG1 molecules to human and mouse FcRn at pH 6.0, a clear effect of 2 vs. 1 vs. 0 FcRn interaction sites can be seen (Fig. XXA and B, pH 6, left panel). IgG1-2F8-ITL, IgG1-7D8-K409R and IgG1-2F8-ITL / IgG1-7D8-K409R (2 FcRn binding sites) bind equally well to the control (IgG1-2F8). The molecule IgG1-7D8-K409R-H435A with 0 FcRn binding sites does not show any binding. The molecule IgG1-2F8-ITL / IgG1-7D8-K409R-H435A with 1 FcRn binding site shows intermediate binding when compared to the molecule with 2 FcRn binding sites.

[0363] Figure 48(A), pH 6.0, right panel shows the binding of hinge-deleted IgG1 molecules (Uni-G1) to human FcRn. All hinge-deleted molecules are impaired in their interaction with human FcRn when compared to the control IgG1 molecule (IgG1-2F8), suggesting that the hinge does indeed affect the interaction with FcRn as assessed in the FcRn binding ELISA. No clear effect of 2 vs. 1 vs. 0 FcRn interaction sites can be seen when binding to human FcRn at pH 6.0 is constructed from these hinge-deleted molecules.

[0364] However, since the binding of human IgG to mouse FcRn is stronger, a clear effect of 2 vs. 1 vs. 0 FcRn interaction sites can be seen when comparing the binding of these hinge-deleted IgG molecules to mouse FcRn at pH 6.0 (Figure 48(B), pH 6.0, right panel). The binding of Uni-G1-7D8-K409R-H435A / Uni-G1-2F8-ITL (1 FcRn binding site) is moderate when compared to the binding of Uni-G1-2F8-ITL, Uni-G1-7D8-409R and Uni-G1-2F8-ITL / Uni-G1-7D8-K409R (2 FcRn binding sites), and Uni-G1-2F8-ITL-H435A (0 FcRn binding sites, no binding).

[0365] Example 47: Her2xCD3 bispecific antibodies tested in an in vitro cytotoxicity assay CD3 is a co-receptor in the T cell receptor complex expressed on mature T cells. The combination of CD3-specific antibody Fab arm and tumor antigen-specific antibody Fab arm in bispecific antibody will lead to specific targeting of T cells to tumor cells, leading to T cell-mediated tumor cell lysis. Similarly, CD3-positive T cells could be targeted to other derailed cells in the body, to infected cells, or directly to pathogens.

[0366] A Her2xCD3 bispecific antibody was generated. The heavy and light chain variable region sequences of the Her2-specific Fab arm were as shown for antibodies 153 and 169 in Example 42. The following heavy and light chain variable region sequences for the CD3-specific Fab arm were used: YTH12.5 (Sequence described by Routledge et al., Eur J Immunol. 1991, 21(11):2717-25) TIFF2025063202000022.tif30165huCLB-T3 / 4 (Sequence described by Parren et al., Res Immunol. 1991, 142(9):749-63. Minor amino acid substitutions were introduced to make the sequence resemble the closest human germline.) TIFF2025063202000023.tif30166

[0367] All antibodies were expressed as IgG1,κ with their Fc region modified as described below: IgG1-Her2-153-K409R and IgG1-Her2-153-N297Q-K409R, IgG1-Her2-169-K409R, IgG1-hu-CLB-T3 / 4-F405L and IgG1-hu-CLB-T3 / 4-N297Q-F405L, IgG1-YTH12.5-F405L and IgG1-YTH12.5-N297Q-F405L.

[0368] Bispecific antibodies derived from these Her2- and CD3-specific antibodies were generated as described in Example 11 and tested in an in vitro cytotoxicity assay using AU565 cells.

[0369] AU565 cells were grown to near confluence. Cells were washed twice with PBS and trypsinized for 5 min at 37°C. Trypsin was inactivated by adding 12 mL of medium and cells were spun down at 800 rpm for 5 min. Cells were resuspended in 10 mL of medium and a single cell suspension was made by passing the cells through a cell strainer. 5 × 10 cells 5100 μL of the cells / mL suspension was added to each well of a 96-well culture plate and the cells were incubated at 37° C., 5% CO 2 for at least 3 hours to allow them to adhere to the plate.

[0370] Peripheral blood mononuclear cells (PBMCs) were isolated from the blood of healthy volunteers using Leucosep 30 mL tubes according to the manufacturer's protocol (Greiner Bio-one). T cells were isolated from the PBMC preparations by negative selection using the Untouched Human T-cells Dynabead kit (Dynal). Isolated cells were 7 × 10 cells. 6 The cells were resuspended in medium to a final concentration of 100 cells / mL.

[0371] The medium was removed from the adherent AU565 cells and 50 μl / well of 2× concentrated antibody-dilution and 50 μl / well of 7×10 T cells were added. 6 The plates were incubated at 37°C, 5% CO 2 The plates were incubated at 37°C, 5% CO for 3 days. The supernatant was removed and the plates were washed twice with PBS. 150 μL of medium and 15 μL of Alamar Blue were added to each well. The plates were incubated at 37°C, 5% CO 2 The mixture was incubated at 4°C for 4 hours and the absorbance was measured (Envision, Perkin Elmer).

[0372] Figure 49 shows that control antibodies (Her2 monospecific IgG1-Herceptin, CD3 monospecific IgG1-YTH12.5 and monospecific IgG1-huCLB-T3 / 4, irrelevant antigen monospecific IgG1-b12, and CD3xb12 bispecific antibody) did not induce T cell-mediated cytotoxicity, whereas bispecific (dyadic) Her2xCD3 antibodies huCLB / Her2-153, huCLB / Her2-169, YTH12.5 / Her2-153 and YTH12.5 / Her2-169 induced dose-dependent T cell-mediated cytotoxicity of AU565 cells. Bispecific antibodies that included Her2-169 were more potent than those that included Her2-153.

[0373] Mutants of IgG1-hu-CLB-T3 / 4, IgG1-YTH12.5 and Her2-153 were created that contained the N297Q mutation to remove the glycosylation site; glycosylation at this site is important for IgG-Fcγ receptor interaction (Bolt S et al., Eur J Immunol 1993, 23:403-411). Figure 49 shows that the N297Q mutation, and therefore the lack of Fc glycosylation, of the Her2xCD3 bispecific antibodies YTH12.5 / Her2-153 and huCLB / Her2-153 did not affect their potential to induce dose-dependent T cell-mediated cytotoxicity of AU565 cells.

Claims

1. 1. An in vitro method for producing a heterodimeric protein comprising the steps of: a) providing a first homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a first CH3 region; b) providing a second homodimeric protein comprising an Fc region of an immunoglobulin, wherein the Fc region comprises a second CH3 region; wherein the sequences of the first CH3 region and the second CH3 region are different and such that a heterodimer interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimer interactions of the first CH3 region and the second CH3 region; c) incubating the first protein with the second protein under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization; and d) obtaining said heterodimeric protein.

2. 2. The in vitro method of claim 1, wherein the first homodimeric protein and the second homodimeric protein are selected from the group consisting of: (i) an Fc region, (ii) an antibody, (iii) a fusion protein comprising an Fc region, and (iv) an Fc region conjugated to a prodrug, peptide, drug, or toxin.

3. The in vitro method of claim 1 , wherein the first homodimeric protein is a full-length antibody.

4. 2. The in vitro method of any one of the preceding claims, wherein the second homodimeric protein is a full-length antibody.

5. 2. The in vitro method of any one of the preceding claims, wherein the first and second homodimeric proteins are both antibodies and bind to different epitopes.

6. 2. The in vitro method of any one of the preceding claims, wherein the Fc region of the first homodimeric protein is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4, and the Fc region of the second homodimeric protein is of an isotype selected from the group consisting of IgG1, IgG2, IgG3 and IgG4.

7. 2. The in vitro method of any one of the preceding claims, wherein the Fc regions of both the first homodimeric protein and the second homodimeric protein are of the IgG1 isotype.

8. 10. The in vitro method according to any one of the preceding claims, wherein one of the Fc regions of the homodimeric protein is of the IgG1 isotype and the other is of the IgG4 isotype.

9. The in vitro method of any one of the preceding claims, wherein the increase in strength of the heterodimeric interaction compared to each of the homodimeric interactions is due to a CH3 modification other than a covalent bond, introduction of a cysteine ​​residue or a charged residue.

10. 2. The in vitro method of any one of the preceding claims, wherein the heterodimeric interaction between the first and second proteins in the resulting heterodimeric protein is such that Fab arm exchange cannot occur at 0.5 mM GSH under the conditions described in Example 13.

11. 20. The in vitro method of any one of the preceding claims, wherein the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is such that no Fab arm exchange occurs in vivo in mice under the conditions described in Example 14.

12. 2. The in vitro method of any one of the preceding claims, wherein the heterodimeric interaction between the first and second proteins in the resulting heterodimeric protein is more than two-fold stronger, such as more than three-fold stronger, such as more than five-fold stronger, than the strongest of the two homodimeric interactions, e.g. determined as described in Example 30.

13. The in vitro method of any one of the preceding claims, wherein the sequences of the first CH3 region and the second CH3 region are such that, when assayed as described in Example 30, the dissociation constant of the heterodimeric interaction between the first protein and the second protein in the resulting heterodimeric protein is less than 0.05 micromolar.

14. 20. The in vitro method according to any one of the preceding claims, wherein the sequences of the first CH3 region and the second CH3 region are such that the dissociation constant of both homodimeric interactions, when assayed as described in Example 21, is greater than 0.01 micromolar, such as greater than 0.05 micromolar, preferably 0.01 to 10 micromolar, such as 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, such as 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, such as 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar.

15. The in vitro method of any one of the preceding claims, wherein the sequences of the first CH3 region and the second CH3 region contain amino acid substitutions at non-identical positions.

16. 16. The in vitro method of claim 15, wherein the amino acid substitute is a natural amino acid or an unnatural amino acid.

17. The in vitro method of any one of the preceding claims, wherein the first homodimeric protein has only one amino acid substitution in the CH3 region and the second homodimeric protein has only one amino acid substitution in the CH3 region compared to the wild-type CH3 region.

18. 17. The in vitro method of any one of claims 1 to 16, wherein a first homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405, 407 and 409, and a second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405, 407 and 409, wherein said first homodimeric protein and said second homodimeric protein are not substituted at the same positions.

19. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407.

20. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid other than Phe at position 405.

21. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has an amino acid other than Phe, Arg or Gly at position 405.

22. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises an amino acid other than Phe at position 405 and Lys at position 409.

23. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises an amino acid other than Phe, Arg or Gly at position 405 and Lys at position 409.

24. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises a Leu at position 405 and a Lys at position 409.

25. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises an amino acid other than Phe, Arg or Gly at position 405 and a Lys at position 409.

26. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Leu at position 405 and a Lys at position 409.

27. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein comprises Lys at position 409, Thr at position 370 and Leu at position 405.

28. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises an Arg at position 409 and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370 and a Leu at position 405.

29. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein comprises a Lys at position 370, a Phe at position 405 and an Arg at position 409, and the second homodimeric protein comprises a Lys at position 409, a Thr at position 370 and a Leu at position 405.

30. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid at position 409 other than Lys, Leu or Met, and the second homodimeric protein has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr.

31. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407.

32. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409 and the second homodimeric protein has Gly, Leu, Met, Asn or Trp at position 407.

33. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407 and Lys at position 409.

34. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has Ala, GIy, His, Ile, Leu, Met, Asn, Val or Trp at position 407 and Lys at position 409.

35. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407 and a Lys at position 409.

36. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and the second homodimeric protein has an amino acid other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr at position 407 and a Lys at position 409.

37. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and the second homodimeric protein has an Ala, Gly, His, Ile, Leu, Met, Asn, Val or Trp at position 407 and a Lys at position 409.

38. 19. The in vitro method of any one of claims 1 to 18, wherein the first homodimeric protein has a Tyr at position 407 and an Arg at position 409, and the second homodimeric protein has a Gly, Leu, Met, Asn or Trp at position 407 and a Lys at position 409.

39. the first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and the second homodimeric protein (i) an amino acid other than Phe, Leu, and Met at position 368, or (ii) Trp at position 370, or (iii) an amino acid other than Asp, Cys, Pro, Glu, or Gln at position 399 19. The in vitro method according to any one of claims 1 to 18, comprising:

40. the first homodimeric protein has Arg, Ala, His or Gly at position 409, and the second homodimeric protein has (i) Lys, Gln, Ala, Asp, Glu, Gly, His, Ile, Asn, Arg, Ser, Thr, Val or Trp at position 368, or (ii) Trp at position 370, or (iii) Ala, Gly, Ile, Leu, Met, Asn, Ser, Thr, Trp, Phe, His, Lys, Arg, or Tyr at position 399 19. The in vitro method according to any one of claims 1 to 18, comprising:

41. the first homodimeric protein has Arg at position 409 and the second homodimeric protein has (i) Asp, Glu, Gly, Asn, Arg, Ser, Thr, Val, or Trp at position 368, or (ii) Trp at position 370, or (iii) Phe, His, Lys, Arg, or Tyr at position 399 19. The in vitro method according to any one of claims 1 to 18, comprising:

42. The in vitro method of any one of the preceding claims, wherein the first and second CH3 regions comprise the sequence set forth in SEQ ID NO:1, except for the specified mutations.

43. The in vitro method of any one of the preceding claims, wherein neither the first homodimeric protein nor the second homodimeric protein contains a Cys-Pro-Ser-Cys sequence in the hinge region.

44. 2. The in vitro method of claim 1, wherein both the first homodimeric protein and the second homodimeric protein comprise a Cys-Pro-Pro-Cys sequence in the hinge region.

45. 2. The in vitro method of any one of the preceding claims, wherein the first homodimeric protein and the second homodimeric protein are human antibodies, except for any mutations specified.

46. 2. The in vitro method of any one of the preceding claims, wherein the first homodimeric protein and the second homodimeric protein are heavy chain antibodies.

47. 47. The in vitro method of any one of claims 1 to 46, wherein both the first homodimeric protein and the second homodimeric protein further comprise a light chain.

48. 48. The in vitro method of claim 47, wherein the light chains are different.

49. 2. The in vitro method of any one of the preceding claims, wherein the first and / or second homodimeric protein comprises a mutation that removes an acceptor site for Asn-linked glycosylation.

50. 2. The in vitro method according to any one of the preceding claims, wherein the first and second homodimeric proteins provided in steps a) and b) are purified.

51. 20. The in vitro method of any one of the preceding claims, wherein the first and / or second homodimeric protein is conjugated to or comprises an acceptor group for a drug, prodrug or toxin.

52. 52. The in vitro method according to any one of claims 4 to 51, wherein the first and / or second epitope is located on a tumor cell.

53. The in vitro method according to any one of claims 4 to 52, wherein the first or the second epitope is located on a tumor cell and the other epitope is located on an effector cell.

54. 54. The in vitro method according to any one of claims 4 to 53, wherein the epitope is located on a T cell, such as on CD3 expressed on a T cell.

55. 53. The in vitro method according to any one of claims 4 to 52, wherein the first or the second epitope is located on a tumor cell and the other epitope is located on a radioisotope, a toxin, a drug or a prodrug, which may optionally be coupled or linked to a peptide or hapten.

56. The in vitro method of any one of claims 4 to 52, wherein the first or the second epitope is located on a tumor cell and the other epitope is located on an electron dense vesicle or a minicell.

57. 53. The in vitro method according to any one of claims 1 to 52, wherein the homodimeric proteins are both antibodies, and the first antibody and the second antibody bind to different epitopes on the same tumor cell.

58. 53. The in vitro method of any one of claims 1 to 52, wherein the homodimeric proteins are both antibodies, and wherein a first antibody binds to an epitope on a tumor cell and the other antibody is an unrelated or inactive antibody that does not have any in vivo binding activity relevant to the intended use.

59. 2. The in vitro method according to any one of the preceding claims, wherein the reducing conditions in step c) comprise the addition of a reducing agent, such as a reducing agent selected from the group consisting of 2-mercaptoethylamine, dithiothreitol and tris(2-carboxyethyl)phosphine or a chemical derivative thereof.

60. 2. The in vitro method according to any one of the preceding claims, wherein step c) is carried out under reducing conditions with a redox potential of -150 to -600 mV, such as -250 to -400 mV.

61. 10. The in vitro method according to any one of the preceding claims, wherein step c) comprises incubation in the presence of at least 25 mM 2-mercaptoethylamine or in the presence of at least 0.5 mM dithiothreitol at a temperature of at least 20°C for at least 90 minutes.

62. 10. The in vitro method according to any one of the preceding claims, wherein step d) comprises removal of the reducing agent, for example by desalting.

63. A method for the selection of bispecific antibodies with desired properties comprising the steps of: a) providing a first set of homodimeric antibodies comprising antibodies with different variable regions, wherein the antibodies of the first set comprise an identical first CH3 region; b) providing a second set of homodimeric antibodies comprising antibodies with different variable regions or identical variable regions, wherein the antibodies of the second set comprise an identical second CH3 region; wherein the sequences of the first and second CH3 regions are different and such that a heterodimer interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimer interactions of the first CH3 region and the second CH3 region; c) incubating the combination of the first and second sets of antibodies under reducing conditions sufficient to allow cysteines in the hinge region to undergo disulfide bond isomerization, thus generating a set of bispecific antibodies; d) optionally returning the conditions to non-reducing conditions; e) assaying the resulting set of bispecific antibodies for a given desired property; and f) selecting bispecific antibodies with desired properties.

64. The method of claim 63, wherein the second set of homodimeric antibodies have different variable regions.

65. The method of claim 63, wherein the second set of homodimeric antibodies have identical variable regions but different amino acids or structural changes outside the antigen binding region.

66. 1. A method for producing a heterodimeric protein, comprising the steps of: a) providing a first nucleic acid construct encoding a first polypeptide comprising a first Fc region of an immunoglobulin, wherein the first Fc region comprises a first CH3 region; b) providing a second nucleic acid construct encoding a second polypeptide comprising a second Fc region of an immunoglobulin, wherein the second Fc region comprises the first CH3 region; wherein the sequences of the first and second CH3 regions are different and such that a heterodimer interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimer interactions of the first CH3 region and the second CH3 region; and wherein said first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and said second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407; and / or wherein the sequences of said first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is between 0.01 and 10 micromolar, such as between 0.05 and 10 micromolar, more preferably between 0.01 and 5 micromolar, such as between 0.05 and 5 micromolar, even more preferably between 0.01 and 1 micromolar, such as between 0.05 and 1 micromolar, 0.01 and 0.5 micromolar or 0.01 and 0.1 micromolar; c) co-expressing the first and second nucleic acid constructs in a host cell; and d) Obtaining said heterodimeric protein from the cell culture.

67. the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe, such as other than Phe, Arg or Gly; or the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr; 67. The method of claim 66.

68. 68. The method of claim 66 or 67, wherein the first and second polypeptides are full-length heavy chains of two antibodies that bind to different epitopes.

69. 69. The method of any one of claims 66 to 68, wherein step c) further comprises co-expressing in the host cell one or more nucleic acid constructs encoding the light chain.

70. 70. The method according to any of claims 66 to 69, further comprising the features of any one or more of claims 2 to 58.

71. 71. An expression vector comprising a nucleic acid construct as defined in any one of claims 66 to 70.

72. A host cell comprising a nucleic acid construct as defined in any one of claims 66 to 70.

73. A heterodimeric protein obtained or obtainable by the method according to any one of claims 1 to 72.

74. A heterodimeric protein comprising a first polypeptide comprising a first Fc region of an immunoglobulin, the first Fc region comprising a first CH3 region, and a second polypeptide comprising a second Fc region of an immunoglobulin, the second Fc region comprising a second CH3 region, wherein the sequences of the first and second CH3 regions are distinct and such that a heterodimeric interaction between the first CH3 region and the second CH3 region is stronger than each of the homodimeric interactions of the first CH3 region and the second CH3 region; wherein said first homodimeric protein has an amino acid other than Lys, Leu or Met at position 409, and said second homodimeric protein has an amino acid substitution at a position selected from the group consisting of 366, 368, 370, 399, 405 and 407; and / or wherein the sequences of said first and second CH3 regions are such that the dissociation constant of the homodimeric interaction of each of the CH3 regions, when assayed as described in Example 21, is 0.01 to 10 micromolar, such as 0.05 to 10 micromolar, more preferably 0.01 to 5 micromolar, such as 0.05 to 5 micromolar, even more preferably 0.01 to 1 micromolar, such as 0.05 to 1 micromolar, 0.01 to 0.5 micromolar or 0.01 to 0.1 micromolar; The heterodimeric protein.

75. the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 405 other than Phe, such as other than Phe, Arg or Gly; or the first CH3 region has an amino acid at position 409 other than Lys, Leu or Met, and the second CH3 region has an amino acid at position 407 other than Tyr, Asp, Glu, Phe, Lys, Gln, Arg, Ser or Thr; The heterodimeric protein of claim 73.

76. 76. The heterodimeric protein of claim 74 or 75, wherein the first and second polypeptides are full-length heavy chains of two antibodies that bind to different epitopes.

77. 77. The heterodimeric protein of claim 76, further comprising two full-length light chains.

78. 78. A heterodimeric protein according to any one of claims 73 to 77, further comprising the features of any one or more of claims 2 to 58.

79. The heterodimeric protein according to any one of claims 74 to 78 for use as a medicament.

80. 80. The heterodimeric protein of claim 79 for use in the treatment of cancer.

81. 79. A pharmaceutical composition comprising the heterodimeric protein of any one of claims 74 to 78 and a pharma- ceutically acceptable carrier.

82. A method for inhibiting the growth and / or proliferation of tumor cells and / or for killing tumor cells, comprising administration of the heterodimeric protein of any one of claims 74 to 78 to an individual in need thereof.

Images