Readily isolable bispecific antibodies with native immunoglobulin format

Abstract

To provide readily isolable bispecific antibodies with a native immunoglobulin format.SOLUTION: A readily isolable bispecific antibody is provided, the format comprising an immunoglobulin heavy chain variable domain differentially modified into a CH3 domain, wherein the differential modifications are non-immunogenic or substantially non-immunogenic with respect to the CH3 modifications, and at least one of these modifications results in a differential affinity of the bispecific antibody for an affinity reagent such as protein A, and the bispecific antibody is isolable from a disrupted cell, from medium, or from a mixture of antibodies, based on its affinity for the protein A.SELECTED DRAWING: None

Description

[Technical field]

[0001] FIELD OF THEINVENTION The present invention relates to antigen binding proteins or antibodies having heavy chain heterodimers, i.e., antigen binding proteins or antibodies having two immunoglobulin heavy chains differing by at least one amino acid, allowing for isolation of antigen binding proteins based on differences in the affinity of the immunoglobulin heavy chains and the affinity of the engineered or mutated immunoglobulin heavy chains for an affinity reagent. The present invention also relates to antigen binding proteins (including bispecific antibodies) having IgG CH2 and CH3 regions with different affinities for Protein A, allowing for rapid isolation due to differences in binding of the IgG regions to Protein A. [Background technology]

[0002] background Antibodies are multifunctional molecules that possess unique binding specificity for target antigens as well as the ability to interact with the immune system through antigen-independent mechanisms. Many biotherapeutics currently used against cancer are monoclonal antibodies against antigens that are typically overexpressed on the targeted cancer cells. Such antibodies, upon binding to tumor cells, can trigger antibody-dependent cellular cytotoxicity (ADCC) or complement-dependent cytotoxicity (CDC). Unfortunately, cancerous cells often develop mechanisms to suppress these normal immune responses.

[0003] In recent years, efforts are underway to develop antibody-like therapeutics with two or more antigen-binding specificities, such as bispecific antibodies. For cancer therapy, the multispecific format may allow the molecule to use one specificity, for example, to target tumor cell antigens, and the other specificity to trigger a response that is not normally available to the immune system. Bispecific antibodies may also find use as surrogate ligands for two-component heterodimeric receptor systems that are normally activated by their natural ligands when they bind to the receptor system and become one with both of its components.

[0004] Numerous formats have been developed in the art to address the therapeutic opportunities offered by molecules with multiple binding specificities. Ideally, such molecules should be well-functioning proteins that are easy to make and purify, have favorable in vivo properties, such as appropriate pharmacokinetics for the intended purpose, minimal immunogenicity, and, if desired, the effector functions of traditional antibodies.

[0005] The most straightforward way to generate bispecific antibodies (expressing two different antibodies in a single cell) is to generate multiple species, because each heavy chain can form both homodimers and heterodimers, but only heterodimers are desired. Also, light and heavy chains can pair inappropriately. Some examples of formats that have attempted to address these issues in various ways are described below.

[0006] One format used for bispecific T cell Engager (BiTE) molecules (see, for example, Wolf, E. et al. (2005) Drug Discovery Today 10:1237-1244) is based on single chain variable fragment (scFv) modules. scFvs usually consist of the variable regions of the light and heavy chains of an antibody fused via a flexible linker that can be folded appropriately, and which can bind to a cognate antigen. In BiTEs, two scFvs with different specificities are tandemly linked in a single chain (see FIG. 1A). This arrangement precludes the generation of molecules with two copies of the same heavy chain variable region. Furthermore, the linker arrangement is designed to ensure the correct pairing of the respective light and heavy chains.

[0007] The BiTE format has several drawbacks. First, scFv molecules are well known to be prone to aggregation. And although the immunogenicity of the scFv linker is generally low, the possibility of the production of antibodies against BiTEs cannot be excluded. Also, the absence of an Fc part in the BiTE format makes its serum half-life very short, with the cumbersome problem of frequent repeated administration or continuous infusion by pump. Finally, the absence of Fc naturally entails the absence of Fc-mediated effector functions, which may be beneficial in some situations.

[0008] The second format (FIG. 1B) is a hybrid of mouse and rat monoclonal antibodies and relies on a modification of traditional protein A affinity chromatography (see, e.g., Lindhofer, H. et al. (1995) J. Immunol. 155:219-225). In this format, mouse IgG2a and rat IgG2b antibodies are produced simultaneously in the same cells (e.g., as a quadroma fusion of the two hybridomas or in engineered CHO cells). Since the light chains of each antibody preferentially associate with the heavy chains of their cognate species, only three different species of antibodies can be assembled: the two parent antibodies, and a heterodimer of the two antibodies containing one heavy / light chain pair of each of the two antibodies associated by the Fc portion. The desired heterodimer can be easily purified from this mixture since its binding characteristics to Protein A are different from those of the parent antibody (i.e. rat IgG2b does not bind to Protein A, whereas mouse IgG2a binds to Protein A). Thus, the mouse-rat heterodimer binds to Protein A but elutes at a higher pH than the mouse IgG2a homodimer, allowing the selective purification of the bispecific heterodimer. Similar to the BiTE format, this hybrid format has two monovalent antigen binding sites.

[0009] A disadvantage of the mouse / rat hybrid is that, because it is non-human, it may have adverse side effects (e.g., "HAMA" or "HARA" reactions) and / or may induce an immune response in the patient that may neutralize the therapeutic agent.

[0010] A third format, called "knobs-into-holes" (Figure 1C), has been discussed in the prior art as potentially useful for the generation of bispecific antibodies (U.S. Pat. No. 7,183,076). In this strategy, the Fc portions of two antibodies are engineered so that one has a protruding "knob" and the other a complementary "hole". When produced in the same cell, their heavy chains are said to preferentially form heterodimers rather than homodimers, by association of the engineered "knobs" with the engineered "holes". By selecting antibodies with different specificities but employing identical light chains, the problem of correct light-heavy chain pairing is addressed.

[0011] A drawback of this format is that the "knob-into-hole" strategy can produce significant amounts of undesired homodimers, necessitating additional purification steps. This difficulty is exacerbated by the fact that many of the properties of the contaminating species are nearly identical to the desired species. Because foreign sequences are introduced by the mutations that generate the "knobs" and "holes," the engineered forms can also be potentially immunogenic.

[0012] There remains a need for bispecific antibody formats, particularly for therapeutic applications, that minimize some or all of the disadvantages noted above. [Prior art documents] [Patent documents]

[0013] [Patent Document 1] U.S. Patent No. 7,183,076 [Non-patent literature]

[0014] [Non-Patent Document 1] Wolf, E. et al. (2005) Drug Discovery Today 10:1237-1244 [Non-Patent Document 2] Lindhofer, H. et al. (1995) J. Immunol. 155:219-225 Summary of the Invention [Means for solving the problem]

[0015] overview The present invention is based, at least in part, on the use of two immunoglobulin CH3 heavy chain constant domain sequences that differ by at least one amino acid in a bispecific antigen binding protein, which confers a difference in the ability of the CH3 domain sequences to bind an affinity agent, thereby improving the ability to isolate the protein.

[0016] In one aspect, there is provided an antigen binding protein comprising a first and a second polypeptide, the first polypeptide comprising, from N-terminus to C-terminus, a first antigen binding region that selectively binds a first antigen, followed by a constant region comprising a first CH3 region of a human IgG selected from IgG1 (SEQ ID NO:1), IgG2 (SEQ ID NO:3) IgG4 (SEQ ID NO:5) and combinations thereof; and the second polypeptide comprising, from N-terminus to C-terminus, a second antigen binding region that selectively binds a second antigen, followed by a constant region comprising a second CH3 region of a human IgG selected from IgG1, IgG2, IgG4 and combinations thereof, wherein the second CH3 region comprises a modification that reduces or eliminates binding of the second CH3 domain to Protein A.

[0017] In one embodiment, the second CH3 region comprises a 95R modification (in IMGT exon numbering; 435R in EU numbering). In another embodiment, the second CH3 region further comprises a 96F modification (436F in IMGT; EU). In a specific embodiment, the second CH3 region is selected from SEQ ID NO:2, SEQ ID NO:4 and SEQ ID NO:6.

[0018] In one embodiment, the second CH3 region is derived from a modified human IgG1 (SEQ ID NO: 2) and further comprises modifications selected from the group consisting of D16E, L18M, N44S, K52N, V57M and V82I (IMGT; D356E, L358M, N384S, K392N, V397M and V422I in EU).

[0019] In one embodiment, the second CH3 region is derived from a modified human IgG2 (SEQ ID NO: 4) and further comprises modifications selected from the group consisting of N44S, K52N and V82I (IMGT; N384S, K392N and V422I in EU).

[0020] In one embodiment, the second CH3 region is derived from a modified human IgG4 (SEQ ID NO: 6) and further comprises modifications selected from the group consisting of Q15R, N44S, K52N, V57M, R69K, E79Q and V82I (IMGT; Q355R, N384S, K392N, V397M, R409K, E419Q and V422I in EU).

[0021] In one embodiment, the CH3 domain is a chimeric domain comprising the sequences of two or more of human IgG1, human IgG2, human IgG3 and human IgG4.

[0022] In one embodiment, the CH3 domain is derived from human IgG1, human IgG2 or human IgG4, and the antigen binding protein further comprises a CH1 domain and a CH2 domain, wherein the CH1 domain and the CH2 domain are independently selected from the group consisting of a human IgG1 CH1 or CH2 domain, a human IgG2 CH1 or CH2 domain, or a chimeric human / human IgG1 / IgG2 or a chimeric human / human IgG1 / IgG3 or a chimeric human / human IgG2 / IgG3 domain or a chimeric human / human IgG1 / IgG4 or a chimeric IgG3 / IgG4 or a chimeric IgG2 / IgG4 domain. In a specific embodiment, the chimeric IgG1 / IgG2, IgG1 / IgG3, IgG2 / IgG3, IgG1 / IgG4, IgG3 / IgG4 and IgG2 / IgG4 domain are not or substantially not immunogenic in humans.

[0023] In one embodiment, the antigen binding protein further comprises an immunoglobulin light chain, hi one embodiment, the immunoglobulin light chain is selected from a human lambda light chain and a human kappa light chain.

[0024] In one embodiment, each of the first and second antigen-binding regions comprises at least one CDR, in another embodiment, at least two CDRs, and in another embodiment, each comprises three CDRs. In a specific embodiment, the CDRs are derived from an immunoglobulin heavy chain. In another specific embodiment, the heavy chain is a human heavy chain.

[0025] In one embodiment, the first antigen-binding region comprises a first immunoglobulin heavy chain variable domain and the second antigen-binding region comprises a second immunoglobulin heavy chain variable domain.

[0026] In one embodiment, the first and second immunoglobulin heavy chain variable domains are independently selected from mouse, rat, hamster, rabbit, monkey, ape and human domains.

[0027] In one embodiment, the first and second immunoglobulin heavy chain variable domains independently comprise human, mouse, rat, rabbit, monkey, ape and humanized CDRs. In one embodiment, the CDRs are human and somatically mutated.

[0028] In one embodiment, the first and second immunoglobulin heavy chain variable domains comprise human framework regions (FRs). In one embodiment, the human FRs are somatically mutated human FRs.

[0029] In one embodiment, the first and / or second antigen-binding region is obtained by screening a phage library containing an antibody variable region for reactivity to an antigen of interest. In another embodiment, the first and / or second antigen-binding region is obtained by immunizing a non-human animal (e.g., a mouse, a rat, a rabbit, a monkey or an ape) with an antigen of interest and identifying a nucleic acid sequence of an antibody variable region that encodes a variable region specific to the antigen of interest. In a particular embodiment, the non-human animal comprises one or more human immunoglobulin variable region genes. In another particular embodiment, the one or more human immunoglobulin variable region genes are present in the non-human animal extrachromosomally, as a replacement in an endogenous immunoglobulin locus, or as a transgene randomly integrated into the genome of the non-human animal. In one embodiment, the first and / or second antigen-binding region is obtained from a hybridoma or quadroma, and in another embodiment, by screening immune cells of the immunized non-human animal using cell sorting.

[0030] In one embodiment, the antigen binding protein is a bispecific antibody, hi one embodiment, the bispecific antibody is a bispecific, fully human antibody, with affinity for each epitope independently in the micromolar, nanomolar or picomolar range.

[0031] In one embodiment, the antigen binding protein is not or substantially not immunogenic in humans. In a particular embodiment, the antigen binding protein does not have a non-natural human T cell epitope. In one embodiment, the modification of the CH3 region is not or substantially not immunogenic in humans. In a particular embodiment, the modification of the CH3 region does not result in a non-natural human T cell epitope.

[0032] In one embodiment, the antigen-binding protein comprises a heavy chain, wherein the heavy chain is not or substantially not immunogenic in humans. In one embodiment, the heavy chain has an amino acid sequence that does not comprise a non-natural T cell epitope. In one embodiment, the heavy chain comprises an amino acid sequence that cannot be proteolytically degraded to an amino acid sequence of about 9 amino acids that is immunogenic in humans. In a particular embodiment, the human is a human treated with the antigen-binding protein. In one embodiment, the heavy chain comprises an amino acid sequence that cannot be proteolytically degraded to an amino acid sequence of about 13 to about 17 amino acids that is immunogenic in humans. In a particular embodiment, the human is a human treated with the antigen-binding protein.

[0033] In one aspect, a bispecific binding protein is provided that comprises a CH2 modification and / or a CH3 modification as described herein, wherein the bispecific binding protein comprises a first binding moiety that specifically recognizes an antigen on a B cell and a second binding moiety that specifically recognizes an antigen on a T cell.

[0034] In one embodiment, the binding protein is a bispecific antibody. In a particular embodiment, the bispecific antibody comprises a human IgG1 heavy chain and a human IgG1ΔAdp heavy chain. In one embodiment, the first binding moiety is a human heavy chain variable domain that specifically recognizes CD20. In one embodiment, the second binding moiety is a human heavy chain variable domain that specifically recognizes CD3. In one embodiment, the bispecific antibody has a killing activity of about 2.8-3.2×10 in a Raji killing assay. -12 M or approx. 2.8-3.0 x 10 -12 M and exhibits no more than about 1-10% or 1-5% bystander killing in a bystander cell killing assay, where the bystander cells do not contain the CD20 epitope. In a specific embodiment, the bystander cells are 293 cells. In another specific embodiment, bystander cell killing in the above assay is no more than about 10 -8 M~about 10 -14 M is measured over a range of bispecific antibody concentrations.

[0035] In one aspect, a method for making a bispecific antibody is provided, comprising: obtaining a nucleic acid sequence encoding a first immunoglobulin heavy chain comprising a first variable domain that recognizes a first epitope, wherein the first immunoglobulin heavy chain comprises an IgG1, IgG2 or IgG4 isotype constant domain or a chimeric isotype constant domain thereof; obtaining a second nucleic acid sequence encoding a second immunoglobulin heavy chain comprising a second variable domain that recognizes a second epitope, wherein the second immunoglobulin heavy chain comprises an IgG1, IgG2 or IgG4 isotype constant domain or a chimeric isotype constant domain thereof that comprises a modification in its CH3 domain that abolishes or reduces binding to Protein A; obtaining a third nucleic acid sequence encoding an immunoglobulin light chain that pairs with the first and second immunoglobulin heavy chains; introducing the first, second and third nucleic acid sequences into a mammalian cell; expressing an immunoglobulin in the cell, and isolating the immunoglobulin using Protein A.

[0036] In one embodiment, the cells are selected from CHO, COS, 293, HeLa and retinal cells expressing a viral nucleic acid sequence (eg, PERC.6™ cells).

[0037] In one aspect, a bispecific antigen binding protein is provided with a first specificity which binds to an antigen and a second specificity which activates a receptor, wherein the bispecific antigen binding protein comprises a first polypeptide comprising a first IgG1, IgG2 or IgG4 CH3 domain which comprises a Protein A-binding determinant, and a second polypeptide comprising a second IgG1, IgG3 or IgG4 CH3 domain which lacks the Protein A-binding determinant.

[0038] In one embodiment, by virtue of the second specificity that activates the receptor, the receptor has a K that is in the molar, millimolar, micromolar, nanomolar or picomolar range. D Combine with.

[0039] In one embodiment, the second specificity binds to a receptor selected from a G protein-coupled receptor, a tyrosine kinase receptor, an integrin receptor, and a toll-like receptor.

[0040] In one embodiment, the second specificity causes the receptor, or a subunit or protein physically associated therewith, to contact and phosphorylate serine, threonine, or tyrosine; cause cyclicization of nucleotides (e.g., cAMP, cADP, or cGMP); cause the production of phosphatidylinositol or its derivatives (e.g., IP3 or PIP3); cause the production of lipid second messengers (e.g., diacylglycerol, ceramide, lysophosphatidylcholine, phosphatidylinositol ... These receptors can cause the production of proteases (diphosphate ions, eicosanoids, etc.); dephosphorylation (e.g., phosphatase activity); phosphorylation of lipids to form second messengers; hydrolysis of second messengers; proteolysis; redox signaling; translocation of proteins to organelles (e.g., the nucleus); receptor aggregation (with itself) to form homomultimers or heteromultimers (with other receptors); or the opening or closing of transmembrane channels.

[0041] In one aspect, a method for making a bispecific antibody is provided comprising: isolating a bispecific antibody of interest from a quadroma, where the bispecific antibody of interest comprises a first heavy chain that is an IgG1, IgG2, or IgG4 isotype, a second heavy chain that is an IgG1, IgG2, or IgG4 isotype having a constant domain that comprises a modification in its CH3 domain that abolishes or reduces binding to Protein A, where the bispecific antibody of interest is isolated using Protein A.

[0042] In one aspect, a method for making bispecific antibodies is provided comprising isolating from disrupted cells or a mixture of antibodies a bispecific antibody having differentially modified IgG1, IgG2 or IgG4 CH3 domains, which are non-immunogenic or substantially non-immunogenic in humans, which modification results in a bispecific antibody having heterodimeric heavy chains whose monomers differ in affinity for Protein A, and which bispecific antibody is isolated from the disrupted cells or mixture using Protein A.

[0043] In one embodiment, the bispecific antibodies are isolated using a Protein A affinity support, where the bispecific antibodies are eluted at a pH of about 3.9 to about 4.4, about 4.0 to about 4.3, about 4.1 to about 4.2, or at a pH of about 4.2. In one embodiment, the bispecific antibodies are eluted at a pH of about 4, 4.1, 4.2, 4.3, 4.4, or 4.5.

[0044] In one embodiment, the bispecific antibody is isolated using a protein A affinity support and a pH gradient or pH step, wherein the pH gradient or pH step comprises an ionic modifier. In certain embodiments, the ionic modifier is present at a concentration of about 0.5 to about 1.0 molar. In certain embodiments, the ionic modifier is a salt. In one embodiment, the ionic modifier is selected from the group consisting of beryllium, lithium, sodium and potassium salts of acetate; sodium and potassium bicarbonate; lithium, sodium, potassium and cesium carbonate; lithium, sodium, potassium and cesium chloride; sodium and potassium fluoride; sodium, potassium and calcium nitrate; sodium and potassium phosphate; and calcium and magnesium sulfate. In certain embodiments, the ionic modifier is an alkali or alkaline earth metal halide salt. In certain embodiments, the ionic modifier is sodium chloride.

[0045] In one aspect, a binding protein comprising an Fc, wherein the Fc comprises a first CH3 domain modified as described herein to form a heterodimeric Fc, and a second CH3 that is unmodified, wherein the differential modification results in the binding protein eluting from a Protein A affinity material at 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.2, 1.3 or 1.4 pH units higher than a corresponding binding protein without the differential modification.

[0046] In one embodiment, the differentially modified binding protein elutes at a pH of about 4.2 while the unmodified binding protein elutes at a pH of about 3. In one embodiment, the differentially modified binding protein elutes at a pH of about 4.5 while the unmodified binding protein elutes at a pH of about 3.5. In one embodiment, the differentially modified binding protein elutes at a pH of about 4 while the unmodified binding protein elutes at a pH of about 2.8-3.5, 2.8-3.2, or 2.8-3. In one embodiment, the differentially modified binding protein elutes at a pH of about 4.2 while the unmodified binding protein elutes at a pH of about 2.8. In one embodiment, the differentially modified binding protein elutes at a pH of about 4.4 while the unmodified binding protein elutes at a pH of about 3.6. In these embodiments, "unmodified" refers to having no modifications at 435 (EU numbering) or at 435 and 436 (EU numbering) in both CH3 domains.

[0047] Any of the embodiments and aspects described herein can be used in conjunction with one another unless otherwise indicated or clear from the context. Other embodiments will become apparent to those of skill in the art upon review of the ensuing description. In certain embodiments, for example, the following are provided: (Item 1) a first polypeptide comprising, from N-terminus to C-terminus, a first epitope-binding region that selectively binds a first epitope, an immunoglobulin constant region comprising a first CH3 region of a human IgG selected from IgG1, IgG2, and IgG4; and a second polypeptide comprising, from the N-terminus to the C-terminus, a second epitope-binding region that selectively binds a second epitope, an immunoglobulin constant region comprising a second CH3 region of a human IgG selected from IgG1, IgG2, and IgG4; wherein said second CH3 region comprises a modification that reduces or eliminates binding of said second CH3 domain to Protein A. (Item 2) 2. The bispecific protein of claim 1, wherein the first polypeptide and the second polypeptide are human IgG heavy chains. (Item 3) 2. The bispecific protein of item 1, further comprising an immunoglobulin light chain. (Item 4) 4. The bispecific protein of item 3, wherein the immunoglobulin light chain is a human immunoglobulin light chain. (Item 5) 2. The bispecific protein of item 1, wherein the first polypeptide and the second polypeptide are each a human IgG1 heavy chain. (Item 6) 2. The bispecific protein according to item 1, wherein the modification is selected from the group consisting of: (a) 95R and (b) 95R and 96F in the IMGT exon numbering system, or (a') 435R and (b') 435R and 436F in the EU numbering system. (Item 7) 7. The bispecific protein according to item 6, further comprising one to five modifications selected from the group consisting of 16E, 18M, 44S, 52N, 57M and 82I in the IMGT exon numbering system, or 356E, 358M, 384S, 392N, 397M and 422I in the EU numbering system. (Item 8) 7. The bispecific protein according to item 6, wherein the CH3 domain of the bispecific antibody is not immunogenic or substantially not immunogenic in humans. (Item 9) 8. The bispecific protein according to item 7, wherein the CH3 domain of the bispecific antibody is not immunogenic or substantially not immunogenic in humans. (Item 10) 1. A method for making a bispecific antibody, the method comprising: a. obtaining a nucleic acid sequence encoding a first immunoglobulin heavy chain comprising a first variable domain recognizing a first epitope, wherein the first immunoglobulin heavy chain comprises an IgG1, IgG2 or IgG4 isotype constant domain; b. obtaining a second nucleic acid sequence encoding a second immunoglobulin heavy chain comprising a second variable domain recognizing a second epitope, wherein the second immunoglobulin heavy chain comprises an IgG1, IgG2 or IgG4 isotype constant domain that comprises a modification in its CH3 domain that abolishes or reduces binding to Protein A; c. obtaining a third nucleic acid sequence encoding said first immunoglobulin heavy chain and an immunoglobulin light chain that pairs with said second immunoglobulin heavy chain; d. introducing said first nucleic acid sequence, said second nucleic acid sequence and said third nucleic acid sequence into a mammalian cell; e. allowing the cells to express the bispecific antibody; and f. Isolating the bispecific antibody based on its ability to bind Protein A. The method includes: (Item 11) 11. The method of claim 10, wherein the modification is selected from the group consisting of: (a) 95R and (b) 95R and 96F in the IMGT exon numbering system, or (a') 435R and (b') 435R and 436F in the EU numbering system. (Item 12) 12. The method of claim 11, further comprising one to five modifications selected from the group consisting of 16E, 18M, 44S, 52N, 57M and 82I in the IMGT exon numbering system, or 356E, 358M, 384S, 392N, 397M and 422I in the EU numbering system. (Item 13) 12. The method of claim 11, wherein the CH3 domain of the bispecific antibody is not immunogenic or substantially not immunogenic in humans. (Item 14) 13. The method of claim 12, wherein the CH3 domain of the bispecific antibody is not immunogenic or substantially not immunogenic in humans. (Item 15) 11. The method of claim 10, wherein the bispecific antibody is isolated on a solid support comprising Protein A. (Item 16) 16. The method of claim 15, wherein the solid support comprises a Protein A affinity column and the bispecific antibody is isolated using a pH gradient. (Item 17) 17. The method of claim 16, wherein the pH gradient is a step gradient comprising one or more pH steps between pH 3 and pH 5. (Item 18) 1. A method for isolating a bispecific antibody, comprising: 13. A method comprising the step of isolating from the disrupted cells or mixture of antibodies bispecific antibodies having differentially modified IgG1, IgG2 or IgG4 CH3 domains, wherein said differentially modified CH3 domains are non-immunogenic or substantially non-immunogenic in humans, said modifications resulting in bispecific antibodies having heterodimeric heavy chain constant regions whose monomers differ in their affinity for Protein A, and said bispecific antibodies being isolated from the disrupted cells or mixture based on their affinity for Protein A. (Item 19) The method according to item 18, wherein one monomer of the heavy chain constant region of the heterodimer is human IgG1, and the other monomer of the heavy chain constant region of the heterodimer is a modified human IgG1 comprising a modification selected from the group consisting of (a) H95R and (b) H95R and Y96F in the IMGT exon numbering system, or (a') H435R and (b') H435R and Y436F in the EU numbering system. 20. The method of claim 19, wherein the modified human IgG1 further comprises a modification selected from the group consisting of D16E, L18M, N44S, K52N, V57M and V82I in the IMGT exon numbering system, or D356E, L358M, N384S, K392N, V397M and V422I in the EU numbering system. [Brief description of the drawings]

[0048] [Figure 1] Figure 1 illustrates three bispecific antibody formats: (A) Bispecific T cell Engager (BiTE); (B) Mouse-rat hybrid; and (C) Knob-into-hole with conventional light chains. [Diagram 2] FIG. 2 illustrates the FcΔAdp modification: (A) Alignment of the Fc regions of human IgG1 (SEQ ID NO: 1) and IgG3 (SEQ ID NO: 7) (FcΔAdp modification shown in box); (B) Schematic representation of the FcΔAdp bispecific antibody. [Figure 2B] FIG. 2 illustrates the FcΔAdp modification: (A) Alignment of the Fc regions of human IgG1 (SEQ ID NO: 1) and IgG3 (SEQ ID NO: 7) (FcΔAdp modification shown in box); (B) Schematic representation of the FcΔAdp bispecific antibody. [Diagram 3] FIG. 3 illustrates the alignment of the human CH3 domains of IgG1, IgG2 and IgG4 (using IMGT exon numbering and EU numbering) with and without the ΔAdp dipeptide modification, and IgG3. [Figure 4] FIG. 4 shows a Protein A column trace for isolating bispecific antibodies showing the elution profile using a step gradient. [Diagram 5] Figure 5 shows the IL4-Ra and IL-6Ra BIACORE™ binding profiles of column fractions eluted from the chromatographic separation shown in Figure 4. Antibodies in the fractions were captured with immobilized anti-Fc antibody, and soluble IL-4Ra or IL-6Ra was then assayed for binding to the captured antibody. [Figure 6] FIG. 6 shows the pharmacokinetic profiles of the FcΔAdp bispecific antibody (IL-6RΔ / IL-4R), the FcΔAdp homodimer (IL-6RΔ / IL-6RΔ), an IgG1 antibody with a wild-type CH3 sequence (IL-4R / IL-4R) and a control monospecific antibody. [Figure 7] FIG. 7 illustrates the efficacy of the CD20×CD3ΔAdp bispecific antibody in the Raji cell killing assay. [Figure 8] FIG. 8 illustrates a bystander cell (293) killing assay using the CD20×CD3ΔAdp bispecific antibody. [Figure 9] Figure 9 shows the results for an expression experiment using various mFc heterodimers. Panel A: Western blot of pH separation of heterodimeric mIgG2a / mIgG2aPTTTK from homodimeric mIgG2a and homodimeric IgG2aPTTTK; Panel B: Western blot of pH separation of heterodimeric mIgG2a / mIgG2aTTTK from homodimeric mIgG2a and homodimeric IgG2aTTTK; IP = input; FT = flow-through; W2 = 2nd wash (1x PBS pH 7.2); E1 = 1st eluate (20mM Na Citrate, 1M NaCl pH 5.5); E2 = 2nd eluate (20mM Na Citrate, 1M NaCl; 57% pH 5.5 + 43% pH 2.6); E3 = 3rd eluate (20mM Na Citrate, 1M NaCl pH 2.6). [Figure 10]Figure 10 illustrates the preferential heterodimeric form of mutant IgG2a relative to that of a mixture of isotypes (e.g., mIgG2a and mIgG1) when a 4:1 ratio of IFNAR1:IFNAR2 constructs is used. Lane 1: IFNAR1-IgG2a:IFNAR2-IgG1; Lane 2: IFNAR1-IgG2a:IFNAR2-IgG2aTTT; Lane 3: IFNAR1-IgG2a:IFNAR2-IgG2aTTTK; Lane 4: IFNAR1-IgG2a:IFNAR2-IgG2aPTTTK; Lane 5: IFNAR1-IgG2a:IFNAR2-IgG2aRF. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0049] Detailed Description The present invention is not limited to the specific methods and experimental conditions described, since such methods and conditions may vary. Since the scope of the present invention is defined by the claims, it is also understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs. Although any method and material similar or equivalent to the method and material described herein can be used in the practice or testing of this invention, specific methods and materials are described herein. All publications mentioned are incorporated herein by reference.

[0051] The term "antibody" as used herein includes immunoglobulin molecules that contain four polypeptide chains, two heavy (H) chains and two light (L) chains, interconnected by disulfide bonds. Each heavy chain contains a heavy chain variable region (abbreviated herein as HCVR or VH) and a heavy chain constant region. The heavy chain constant region contains three domains CH1, CH2 and CH3. Each light chain contains a light chain variable region (abbreviated herein as LCVR or VL) and a light chain constant region. The light chain constant region contains one domain CL. The VH and VL regions can be further subdivided into regions of hypervariability called complementarity determining regions (CDRs), interspersed with more conserved regions called framework regions (FRs). Each VH and VL is 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 (heavy chain CDRs may be abbreviated as HCDR1, HCDR2 and HCDR3; light chain CDRs may be abbreviated as LCDR1, LCDR2 and LCDR3). The term "high affinity" antibody refers to an antibody that has at least 10 affinity for its target as measured by surface plasmon resonance, e.g., BIACORE™ or solution-affinity ELISA. -9 M, at least 10 -10 M; at least 10 -11 M; or at least 10 -12 It refers to an antibody having the binding affinity of M.

[0052] The phrase "antigen binding protein" refers to an antigen binding protein that has at least one CDR and is capable of selectively recognizing an antigen, i.e., has a K of at least in the micromolar range. D Therapeutic antigen-binding proteins (e.g., therapeutic antibodies) include proteins that can bind to antigens with K in the nanomolar or picomolar range. D"Antigen binding proteins" also include proteins comprising a first and a second CH3 domain as described herein, and a first protein or ligand recognition domain and a second protein or ligand recognition domain, where the first protein or ligand recognition domain and the second protein or ligand recognition domain each independently recognize the same protein or ligand, or together recognize the same protein or ligand, or each independently recognize a different protein or ligand. One example of such a protein is an immunoadhesin comprising a fusion protein (hetero- or homo-) dimer, where the polypeptide of the dimer is a fusion polypeptide comprising a receptor component or a ligand component, and the ligand component comprises an amino acid sequence that binds to a receptor.

[0053] The phrase "bispecific antibody" includes antibodies that can selectively bind to two or more epitopes. Bispecific antibodies generally comprise two different heavy chains, each of which specifically binds to a different epitope (an epitope on two different molecules (e.g., antigens) or an epitope on the same molecule (e.g., the same antigen). When a bispecific antibody can selectively bind to two different epitopes (a first epitope and a second epitope), the affinity of the first heavy chain for the first epitope is generally at least one to two or three or four orders of magnitude lower than the affinity of the first heavy chain for the second epitope, or vice versa. The epitopes recognized by bispecific antibodies can be on the same target or on different targets (e.g., on the same or different proteins). Bispecific antibodies can be made, for example, by combining heavy chains that recognize different epitopes of the same antigen. For example, nucleic acid sequences encoding heavy chain variable sequences that recognize different epitopes of the same antigen can be fused to nucleic acid sequences encoding various heavy chain constant regions, and such sequences can be expressed in cells expressing immunoglobulin light chains. A typical bispecific antibody has two heavy chains (each having three heavy chain CDRs followed (N-terminus to C-terminus) by a CH1 domain, a hinge, a CH2 domain, and a CH3 domain) and an immunoglobulin light chain that does not provide antigen binding specificity but can associate with each heavy chain, or an immunoglobulin light chain that can associate with each heavy chain and bind to one or more epitopes bound by the heavy chain antigen binding region, or an immunoglobulin light chain that can associate with each heavy chain and allow binding of one or both heavy chains to one or both epitopes.

[0054] The term "cell" includes any cell suitable for expressing a recombinant nucleic acid sequence. Cells include prokaryotic and eukaryotic cells (unicellular or multicellular), bacterial cells (e.g., strains of E. coli, Bacillus spp., Streptomyces spp., etc.), mycobacterial cells, fungal cells, yeast cells (e.g., S. cerevisiae, S. pombe, P. pastoris, P. methanolica, etc.), plant cells, insect cells (e.g., SF-9, SF-21, baculovirus-infected insect cells, Trichoplusia ni, etc.), non-human animal cells, human cells or cell fusions (e.g., hybridomas or quadromas). In some embodiments, the cells are human, monkey, ape, hamster, rat or mouse cells. In some embodiments, the cell is a eukaryotic cell and is selected from the following cells: CHO (e.g., CHO K1, DXB-11CHO, Veggie-CHO), COS (e.g., COS-7), retinal cells, Vero, CV1, kidney (e.g., HEK293, 293EBNA, MSR293, MDCK, HaK, BHK), HeLa, HepG2, WI38, MRC5, Colo205, HB8065, HL-60 (e.g., BHK21), Jurkat, Daudi, A431 (epidermis), CV-1, U937, 3T3, L cells, C127 cells, SP2 / 0, NS-0, MMT060562, Sertoli cells, BRL3A cells, HT1080 cells, myeloma cells, tumor cells, and cell lines derived from the above-mentioned cells. In some embodiments, the cells comprise one or more viral genes (eg, retinal cells that express viral genes (eg, PER.C6™ cells)).

[0055] The phrase "complementarity determining region" or term "CDR" includes amino acid sequences that are normally (i.e., in wild-type animals) found between two framework regions in the variable region of a light or heavy chain of an immunoglobulin molecule (e.g., an antibody or T cell receptor) that are encoded by a nucleic acid sequence of an organism's immunoglobulin genes. A CDR can be encoded, for example, by a germline sequence or a rearranged or unrearranged sequence, for example, by a naive or mature B cell or a naive or mature T cell. In some situations (e.g., in the case of a CDR3), a CDR can be encoded by two or more sequences (e.g., germline sequences) that are not contiguous (e.g., not contiguous in an unrearranged nucleic acid sequence) but are contiguous in a B cell's nucleic acid sequence, for example, as a result of splicing or joining of those sequences (e.g., VDJ recombination to form a heavy chain CDR3).

[0056] The phrase "heavy chain" or "immunoglobulin heavy chain" includes immunoglobulin heavy chain constant region sequences from any organism, including a heavy chain variable domain unless otherwise specified. The heavy chain variable domain includes three heavy chain CDRs and four FR regions, unless otherwise specified. Fragments of heavy chains include CDRs, CDRs and FRs, and combinations thereof. A typical heavy chain has, after the variable domain (from N-terminus to C-terminus), a CH1 domain, a hinge, a CH2 domain, and a CH3 domain. A functional fragment of a heavy chain includes a fragment capable of specifically recognizing an antigen (e.g., a K in the micromolar, nanomolar, or picomolar range). D The term "antigen recognition fragment" includes fragments that are capable of being expressed and secreted from a cell and that contain at least one CDR.

[0057] The phrase "Fc-containing protein" includes antibodies, bispecific antibodies, immunoadhesins, and other binding proteins that contain at least a functional portion of an immunoglobulin CH2 and CH3 region. "Functional portion" refers to the CH2 and CH3 regions that can bind to an Fc receptor (e.g., FcγR; or FcRn, i.e., the neonatal Fc receptor) and / or participate in the activation of complement. If the CH2 and CH3 regions contain deletions, substitutions, and / or insertions or other modifications that render them incapable of binding to any Fc receptor and activating complement, the CH2 and CH3 regions are not functional.

[0058] The Fc-containing protein may contain modifications in the immunoglobulin domain, including those that affect one or more effector functions of the binding protein (e.g., modifications that affect FcγR binding, FcRn binding and thus half-life, and / or CDC activity). Such modifications include the following modifications and combinations thereof according to the EU numbering of immunoglobulin constant regions: 238, 239, 248, 249, 250, 252, 254, 255, 256, 258, 265, 267, 268, 269, 270, 272, 276, 278, 280, 283, 285, 286, 289, 290, 292, 293, 294, 295, 296, 297, 298, 301, 303, 305, 307, 308, 309, 311, 312, 315, These include, but are not limited to, 318, 320, 322, 324, 326, 327, 328, 329, 330, 331, 332, 333, 334, 335, 337, 338, 339, 340, 342, 344, 356, 358, 359, 360, 361, 362, 373, 375, 376, 378, 380, 382, ​​383, 384, 386, 388, 389, 398, 414, 416, 419, 428, 430, 433, 434, 435, 437, 438 and 439.

[0059] For example, but not limited to, the binding protein is an Fc-containing protein that exhibits an extended serum half-life (relative to the same Fc-containing protein without the recited modifications) and has a modification at position 250 (e.g., E or Q); 250 and 428 (e.g., L or F); 252 (e.g., L / Y / F / W or T), 254 (e.g., S or T) and 256 (e.g., S / R / Q / E / D or T); or has a modification at position 428 and / or 433 (e.g., L / R / SI / P / Q or K) and / or 434 (e.g., H / F or Y); or has a modification at position 250 and / or 428; or has a modification at position 307 or 308 (e.g., 308F, V308F) and 434. In another example, the modifications may include 428L (e.g., M428L) and 434S (e.g., N434S) modifications; 428L, 259I (e.g., V259I) and 308F (e.g., V308F) modifications; 433K (e.g., H433K) and 434 (e.g., 434Y) modifications; 252, 254 and 256 (e.g., 252Y, 254T and 256E) modifications; 250Q and 428L modifications (e.g., T250Q and M428L); 307 and / or 308 modifications (e.g., 308F or 308P).

[0060] The phrase "ionomodulator" includes moieties that reduce or disrupt the effects of non-specific (i.e., non-affinity) ionic interactions between proteins. "Ionomodulators" include, for example, ionic combinations of salts, acetates, bicarbonates, carbonates, halogens (e.g., chlorides or fluorides), nitrates, phosphates, or sulfates with group I and group II metals. A non-limiting, exemplary list of "ionomodulators" includes beryllium, lithium, sodium, and potassium salts of acetate; sodium and potassium bicarbonates; lithium, sodium, potassium, and cesium carbonates; lithium, sodium, potassium, and cesium chlorides; sodium and potassium fluorides; sodium, potassium, and calcium nitrates; sodium and potassium phosphates; and calcium and magnesium sulfates. An "ionic modifier" comprises a moiety that affects ionic interactions when added to a pH gradient or pH step, or upon equilibration of a Protein A support with the "ionic modifier", such that application of a pH step or pH gradient increases the pH unit interval between the elution of homodimeric and heterodimeric IgG (e.g., between the elution of wild-type human IgG and the elution of the same IgG except that it has one or more modifications of its CH3 domain as described herein). The appropriate concentration of "ionic modifier" can be determined by using the same column, pH step or pH gradient with increasing concentrations of "ionic modifier" until the maximum pH interval is reached for a given pH step or pH gradient.

[0061] The phrase "light chain" includes immunoglobulin light chain constant region sequences from any organism, including human kappa light chains and human lambda light chains, unless otherwise specified. A light chain variable (VL) domain typically includes three light chain CDRs and four framework (FR) regions, unless otherwise specified. Generally, a full-length light chain includes a VL domain including, from the amino terminus to the carboxyl terminus, FR1-CDR1-FR2-CDR2-FR3-CDR3-FR4, and a light chain constant domain. Light chains that can be used with the present invention include, for example, light chains that do not selectively bind to the first or second antigens selectively bound by the antigen binding protein. Suitable light chains include light chains that can be identified by screening for the most commonly used light chains in existing antibody libraries (wet libraries or in silico), where the light chain does not substantially interfere with the affinity and / or selectivity of the antigen binding domain of the antigen binding protein. Suitable light chains include those capable of binding to one or both epitopes bound by the antigen-binding region of the antigen-binding protein.

[0062] The phrase "micromolar range" is intended to mean 1 to 999 micromolar; the phrase "nanomolar range" is intended to mean 1 to 999 nanomolar; the phrase "picomolar range" is intended to mean 1 to 999 picomolar.

[0063] The phrase "somically mutated" includes reference to a nucleic acid sequence from a B cell that has undergone class switching, where the nucleic acid sequence of an immunoglobulin variable region (e.g., a heavy chain variable domain or a region including a heavy chain CDR sequence or a heavy chain FR sequence) in the class switched B cell is not identical to the nucleic acid sequence in the B cell prior to class switching (e.g., the CDR nucleic acid sequence or framework nucleic acid sequence differs between the non-class switched and class switched B cell). "Somatically mutated" includes reference to a nucleic acid sequence from an affinity matured B cell that is not identical to the corresponding sequence in a B cell that has not undergone affinity maturation (i.e., a sequence in the genome of a germline cell). The phrase "somically mutated" also includes reference to a nucleic acid sequence from a B cell after exposure of the B cell to an antigen of interest, where the nucleic acid sequence differs from the corresponding nucleic acid sequence prior to exposure of the B cell to the antigen of interest. The phrase "somically mutated" refers to a sequence from an antibody that has been produced in an animal, e.g., a mouse, that has a nucleic acid sequence of a human immunoglobulin variable region in response to an antigen challenge, resulting from a selection process that is essentially effective in such an animal.

[0064] The phrase "variable domain" includes an amino acid sequence of an immunoglobulin light or heavy chain (modified as desired) that comprises the following amino acid regions, in order from N-terminus to C-terminus (unless otherwise indicated): FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. A "variable domain" includes an amino acid sequence that can fold into a canonical domain (VH or VL) having a double beta-sheet structure, in which the beta-sheets are connected by disulfide bonds between residues of the first and second beta-sheets.

[0065] Bispecific antibodies with engineered IgG CH3 regions The present inventors have developed a novel format that combines the conventional light chain strategy with the implementation of an alternative Protein A purification scheme that can be used with components of human antibodies.

[0066] It has been known for some time that human IgG3 does not bind to protein A and therefore may be used together with any of the other three human IgG subclasses in a purification strategy similar to that used for mouse-rat hybrids (Lindhofer, H. et al. (1995) J. Immunol. 155:219-225). However, despite the high sequence homology of all four human IgG subclasses, it is not known how readily the Fc portions of IgG1, IgG2 and IgG4 form heterodimers with IgG3; in certain circumstances (e.g., isolation from quadromas), the preferential formation of homodimers alone may have a detrimental effect on the overall yield of the desired heterodimer. Additional modifications may be required to compensate for differences between the hinge region of IgG3 and that of other subclasses. Due to potential effects on effector functions, it may be preferable in some circumstances not to require the presence of the complete IgG3 Fc.

[0067] Therefore, we have devised a "minimal" format that exploits a fortuitously simple determinant of protein A binding. It has been reported that the inability of IgG3 to bind protein A is determined by a single amino acid residue, Arg435 (EU numbering; Arg95 in IMGT) (in other IgG subclasses the corresponding position is occupied by a histidine residue) (Jendeberg, L. et al. (1997) J. Immunological Meth. 201:25-34)). Therefore, instead of IgG3, it is possible to use an IgG1 sequence in which His435 is mutated to Arg. Thus, a single point mutation in IgG1 should be sufficient to result in a different binding affinity that can be applied to new purification schemes. This modification is called IgG1ΔA (and similarly IgG2ΔA and IgG4ΔA, or more commonly FcΔA), which denotes the inability to bind protein A.

[0068] However, the specified point mutation could potentially be immunogenic since it introduces a new peptide sequence. This point mutation could theoretically be placed on MHC class II molecules and presented to T cells, thereby inducing an immune response. To circumvent this unforeseen difficulty, the dipeptide mutation H435R / Y436F (EU numbering; H95R / Y96F in IMGT) could be used. The resulting sequence in the vicinity of this change would be identical to that of IgG3 (see FIG. 2A) and would therefore be expected to be immunologically "hidden" since there would be no unnatural short peptides available for presentation to T cells. It has been reported that this double mutant still does not bind to protein A (Jendeberg, L. et al. (1997) J. Immunological Meth. 201:25-34). Finally, the dipeptide mutation does not involve any of the residues that form the interface of the Fc dimer, so heterodimer formation is unlikely to be interfered with. This dipeptide mutation is designated "IgG1ΔAdp" (and similarly IgG2ΔAdp, IgG4ΔAdp and FcΔAdp). The location of the dipeptide modification in IgG1, IgG2 and IgG4 is shown in the sequences designated IgG1ΔAdp, IgG2ΔAdp and IgG4ΔAdp in FIG. 3, which are shown together with the wild-type human IgG CH3 domain sequence, and hIgG3, with the IMGT exon numbering and EU numbering indicated.

[0069] The FcΔAdp modification is unlikely to interfere with heterodimer formation, since it does not include any of the residues believed to form the Fc dimer interface. As such, the FcΔAdp is so minimal that it may be similarly incorporated into other engineered Fc forms. IgG2ΔAdp and IgG4ΔAdp may be beneficial in situations where the effector functions (or lack thereof) associated with each of the latter are desired.

[0070] In summary, the bispecific antibody format described above comprises two antibodies of different specificity using the same light chain, where the Fc region of one of them is modified to the FcΔAdp format (see FIG. 2B). The configuration is that of a natural human antibody and should therefore share its favorable properties, including low tendency to aggregate, in vivo stability, minimal immunogenicity, biodistribution properties similar to those of antibodies, good pharmacokinetics, and, optionally, effector functions. A relatively quick and simple to perform method for isolating such bispecific antibodies is provided.

[0071] Bispecific binding proteins with modified mouse IgG CH regions The present inventors have devised a method for easily isolating binding proteins comprising immunoglobulin heavy chains (or functional CH2- and CH3-containing fragments thereof) that are heterodimers with respect to one or more amino acids in the CH3 domain. By carefully selecting modifications of the mouse IgG CH domain and applying specific separation techniques, it becomes possible to easily isolate binding proteins that contain two differentially modified mouse CH regions from homodimers and heterodimers that do not contain those modifications.

[0072] Mouse IgG1, which contains proline at 247, threonines at 252, 254 and 256, and lysine at 258, binds only weakly to Protein A. However, mouse IgG2a and IgG2b contain different residues at those positions (except positions 256 and 258 of IgG2b), and mouse IgG2a and 2b bind well to Protein A. By differentially modifying the CH regions of two mouse IgGs in a manner to generate antibodies whose heavy chains are heterodimers, such antibodies can be endowed with differential Protein A binding properties. In this way, a differential Protein A isolation scheme is devised that can be ready to separate the modified heterodimer from any mouse IgG homodimer, whether it is an IgG1 homodimer (which binds Protein A only very weakly, if at all), or a mouse IgG2a homodimer, a mouse IgG2b homodimer, or an IgG2a / IgG2b heterodimer. For example, a bispecific antibody having two different heavy chain variable domains but of the same isotype, e.g., IgG2a, can be expressed in a suitable expression system using the heavy chain sequences, where only one of the IgG2a CH regions is modified to reduce or eliminate Protein A binding determinants. In this way, only one of the IgG2a CH regions exhibits substantial affinity for Protein A, and any antibody formed from a dimer of unmodified and modified IgG2a can be easily isolated from the modified heterodimer.

[0073] In various embodiments, the antibody comprises a single CH region of the Fc dimer that contains an altered CH region, whereas the other CH of the Fc dimer lacks them, wherein the mouse IgG CH region is altered to contain specific amino acids at specific positions (EU numbering) selected from the group consisting of: 252T, 254T and 256T; 252T, 254T, 256T and 258K; 247P, 252T, 254T, 256T and 258K; 435R and 436F; 252T, 254T, 256T, 435R and 436F; 252T, 254T, 256T, 258K, 435R and 436F; 24tP, 252T, 254T, 256T, 258K, 435R and 436F; and 435R. In certain embodiments, modifications are made to specific groups selected from the group consisting of M252T, S254T, S256T; M252T, S254T, S256T, I258K; I247P, M252T, S254T, S256T, I258K; H435R, H436F; M252T, S254T, S256T, H435R, H436F; M252T, S254T, S256T, I258K, H435R, H436F; I247P, M252T, S254T, S256T, I258K, H435R, H436F; and H435R.

[0074] Heterodimeric mouse IgG-based binding proteins can be used for a variety of applications, for example, they enable methods to isolate bispecific antibodies with mouse constant domains, where the modifications do not interfere or do not substantially interfere with the binding of the antibody to one or more mouse Fc receptors, and where the antibody can, for example, participate in ADCC or CDC and can bind to two or more epitopes on the same or different targets.

[0075] In one aspect, a method for isolating a binding protein comprising a first mouse IgG CH region and a second mouse IgG CH region, where the first IgG CH region is modified to reduce or eliminate the Protein A binding affinity of the first mouse IgG CH region but not the Protein A binding affinity of the second mouse IgG CH region (the second IgG CH region is not modified), and where the binding protein comprises a first binding moiety that binds a first epitope and a second binding moiety that binds a second epitope.

[0076] In one embodiment, the modification does not alter or does not substantially alter the binding affinity of the binding protein to an Fc receptor, hi one embodiment, the binding protein contains a modification that increases or decreases the affinity of the binding protein to an Fc receptor.

[0077] In one embodiment, the modification does not alter or does not substantially alter the serum half-life of the binding protein in a mouse containing a native mouse FcγR receptor and / or a native mouse FcRn compared to the corresponding binding protein without that modification.

[0078] In one embodiment, the modifications do not alter or do not substantially alter the serum half-life of the binding protein in mice that contains replacement of the native high and low affinity mouse FcγR receptors and / or FcRn receptors compared to the corresponding binding protein without the modifications.

[0079] In one embodiment, the first and second epitopes are different and are present on different cells or different proteins. In one embodiment, the first and second epitopes are different and are present on the same cell or the same protein.

[0080] In one embodiment, the Fc receptor is selected from a high affinity Fc receptor, a low affinity Fc receptor, and FcRn. In a specific embodiment, the Fc receptor is selected from one or more of mouse FcRn, mouse FcγR, mouse FcγRIIB, mouse FcγRIII, mouse FcγRIV, and combinations thereof. In a specific embodiment, the Fc receptor is selected from one or more of human FcRn, human FcγR, human FcγRIIB, human FcγRIIC, human FcγRIIIB, human FcγRIIIA, human FcγRIIA, and combinations thereof.

[0081] immunogenicity One advantage of many embodiments of the present invention is that modifications can be used that result in bispecific antibodies that can be easily isolated based on differences in binding to Protein A and that are not or substantially not immunogenic in humans. This property makes such embodiments particularly useful in creating bispecific antibodies for use in human therapy and, for example, in creating immunoadhesins that are not or substantially not immunogenic (when using human binding moieties, i.e., human receptor components and / or human ligands). This property is relevant for bispecific antibodies that have CH3 domains with H95R / Y96F (IMGT numbering) modifications of IgG1, IgG2 and IgG3, as well as further modifications that result in modified positions that reflect the wild-type sequence of the different IgG isotypes. Thus, while modifications associated with a particular IgG isotype are not found in nature, the modified sequences are locally identical to the wild-type sequences of the different IgG isotypes, and the modifications are not expected to be immunogenic or substantially immunogenic. Even if the sequence is not locally identical to any naturally occurring sequence, the modification may not be immunogenic; such modifications may be equally useful. Thus, the minimal point mutation H95R (IMGT numbering), if not immunogenic, may be a suitable embodiment of the invention.

[0082] Thus, bispecific antibodies are provided that have one or more differential modifications of the heavy chain constant domain, including modifications that are not or substantially not immunogenic in humans with respect to the heavy chain constant domain, but that result in differences in the affinity of the heavy chain constant domain for affinity reagents (e.g., Protein A). The modifications include those disclosed herein. In a particular embodiment, the bispecific antibody with differentially modified heavy chains, but that is not or substantially not immunogenic in humans with respect to the CH3 domain, is a human IgG1, IgG2 or IgG4 that comprises a CH3 domain that comprises (or, in another embodiment, consists essentially of) one of the following modifications: H95R or H95R and Y96F (IMGT numbering).

[0083] The bispecific antibodies are expected to be non- or substantially non-immunogenic to humans in which tolerance to human IgG1, IgG2 and IgG4 isoforms has not been broken to any significant extent.

[0084] Specifically, the binding groove of MHC class II molecules accommodates 9-mers that contain the major determinants recognized by the variable loops of T cell receptors, and the FcΔAdp modification is expected to be immunologically "hidden" because any natural sequence whose peptide length is less than 9-mers is unlikely to elicit an immune response. However, peptides longer than 9-mers (usually about 13-17-mers) may be bound by MHC class II, and protruding segments may potentially affect binding. Thus, additional modifications that eliminate longer non-natural sequences (additional modifications to the FcΔAdp modification) may further reduce the potential for immunogenicity. One particular example is the V422I (EU; V82I in IMGT numbering) modification, which extends the minimum non-natural peptide length from 14 to 39 residues in IgG1ΔAdp and from 14 to 43 residues in the similarly defined IgG2ΔAdp. Another example is the modification L445P (EU; L105P in IMGT numbering) in IgG4ΔAdp, which extends the length from 10 to 14 residues.

[0085] Pharmacokinetics The binding site for Protein A overlaps with that for the neonatal Fc receptor FcRN, which is believed to be responsible for giving immunoglobulins a long serum half-life. Thus, given that human IgG3 has a shorter serum half-life (about 7 days) than other IgG subclasses (about 21 days), modifications near the Protein A binding site raise the possibility that the format proposed here may have a shorter serum half-life than IgG1, 2 and 4. Several Fc mutants affecting His435 have been shown to not bind to FcRN and to have a shorter half-life in mice. However, pharmacokinetic analysis shows that the serum half-life of the IgG1ΔA / IgG1 heterodimer is not significantly different from that of the IgG1 homodimer (see Example 2). Thus, the IgG1ΔAdp mutation has the advantage of eliminating Protein A binding while still retaining the long half-life of IgG1.

[0086] Thus, in one embodiment, a bispecific antigen-binding protein is provided that comprises a modification of the CH3 domain as described herein, which exhibits an equivalent pharmacokinetic profile to the same bispecific antigen-binding protein without a modification in the CH3 domain. In one embodiment, a bispecific antibody is provided that comprises an IgG1ΔA / IgG1 heterodimer Fc, which has a serum half-life that is about 1.5-fold, about 2-fold, about 2.5-fold, or about 3-fold longer than an otherwise identical bispecific antibody but comprising an IgG3 CH3 domain or otherwise identical but comprising at least one IgG3 heavy chain. In one embodiment, a bispecific antibody is provided that comprises an IgG1ΔA / IgG1 heterodimer Fc, which exhibits approximately the same serum half-life as a bispecific antibody without the IgG1ΔA modification (i.e., an IgG1 homodimeric bispecific antibody).

[0087] Immunoglobulin heavy chain Immunoglobulin heavy chain variable regions that can be used to generate bispecific antibodies with the desired characteristics (e.g., desired specificity, desired affinity, desired functionality, e.g., blocking, non-blocking, inhibitory, activating, etc.) can be generated using any method known in the art. The desired heavy chain can then be constructed by cloning a nucleic acid sequence containing the variable region into a construct with the desired heavy chain constant region as described herein.

[0088] In one embodiment, the first heavy chain comprises a variable region encoded by a nucleic acid derived from the genome of a mature B cell of a first animal immunized with a first antigen, and the first heavy chain specifically recognizes the first antigen. In a particular embodiment, the second heavy chain comprises a variable region encoded by a nucleic acid derived from the genome of a mature B cell of a second animal immunized with a second antigen, and the second heavy chain specifically recognizes the second antigen.

[0089] In one embodiment, the first animal and / or the second animal is a genetically modified animal that comprises an unrearranged human immunoglobulin heavy chain variable region. In one embodiment, the first animal and / or the second animal is a genetically modified animal that comprises an unrearranged human immunoglobulin heavy chain variable region and a human immunoglobulin constant region. In one embodiment, the first animal and / or the second animal is a genetically modified mouse that comprises an unrearranged human immunoglobulin heavy chain variable region.

[0090] Immunoglobulin heavy chain variable region sequences can be obtained by any other method known in the art, such as phage display, and the sequences obtained thereby can be used to generate nucleic acid constructs that can be linked to a nucleic acid encoding any suitable heavy chain, such as a heavy chain in which the CH3 domain has been modified as described herein, placed into an expression construct, and delivered to a cell that can make the heavy chain, for example, in the presence of a suitable light chain.

[0091] Immunoglobulin light chains Bispecific antibodies comprising two heavy chains that recognize two different epitopes (or two different antigens) are more easily isolated if they can pair with the same light chain (i.e., a light chain having the same variable and constant domains). Various methods are known in the art for making light chains that can pair with two heavy chains of differing specificity without interfering or substantially interfering with the selectivity of the heavy chain variable domain and / or its affinity for its target antigen.

[0092] In one approach, a light chain can be selected by examining usage statistics for all light chain variable domains, identifying the light chain most frequently used in human antibodies, and pairing that light chain with two heavy chains of differing specificity.

[0093] In another approach, a light chain can be selected by observing light chain sequences in a phage display library (e.g., a phage display library that contains human light chain variable region sequences, such as a human ScFv library) and selecting the most commonly used light chain variable regions from that library.

[0094] In another approach, light chains can be selected by assaying a phage display library of light chain variable sequences using the heavy chain variable sequences of both heavy chains as probes. Light chains that associate with both heavy chain variable sequences are selected as light chains for those heavy chains, allowing binding and / or activation to both epitopes.

[0095] In another approach, a light chain can be selected by combining a known light chain with the desired heavy chain and assaying the resulting bispecific antibody for binding specificity, affinity and / or activation capacity.

[0096] To the extent that difficulties are encountered in any approach to selecting a light chain (e.g., the light chain interferes with binding of one or both heavy chains to its antigen, or the light chain does not associate well with one or both heavy chains), the light chain can be aligned with the heavy chain's cognate light chain, and modifications are made in the light chain to better match sequence features in common with both heavy chain's cognate light chains. If immunogenicity alterations need to be minimized, based on parameters and methods known in the art for assessing immunogenic potential (i.e., in silico as well as wet assays), modifications preferably result in sequences present in known human light chain sequences, such that proteolytic processing is unlikely to generate T cell epitopes.

[0097] Antibodies and Binding Proteins The compositions and methods are particularly useful in making human bispecific antibodies, i.e., bispecific antibodies comprising human constant and human variable domains. In some embodiments, human antibodies include antibodies having heavy chain variable and heavy chain constant domains derived from human germline immunoglobulin sequences, in some embodiments from somatically mutated human immunoglobulin sequences (e.g., human immunoglobulin sequences produced in animals comprising human immunoglobulin gene sequences). In some embodiments, the human variable and / or constant regions may include amino acid residues not encoded by human germline immunoglobulin sequences, or encoded as a result of recombination and / or splicing, for example in the CDRs, particularly CDR3. Human antibodies are not intended to include antibodies in which CDR sequences derived from the germline of another mammalian species, such as a mouse, have been grafted onto human framework sequences. Such antibodies are referred to as humanized or chimeric antibodies. Human antibodies include antibodies containing mutations introduced in vitro, for example by random or site-directed mutagenesis, which mutations are preferably not immunogenic in humans.

[0098] The above methods and compositions can be used to generate chimeric antibodies, preferably chimeric antibodies that are not or have low immunogenicity in humans. Chimeric antibodies are antibodies in which one of the heavy chain variable region or framework or CDR or heavy chain constant region or heavy chain constant domain is derived from different species (e.g., human and mouse, or human and primate). In some embodiments, chimeric antibodies include antibodies with heavy chain variable region of non-human origin (e.g., mouse) and heavy chain constant region of human origin. In some embodiments, chimeric antibodies include antibodies with heavy chain variable region of human origin and heavy chain constant region of non-human (e.g., mouse) origin. In various embodiments, the region of mouse origin is identical or substantially identical to the sequence of mouse immunoglobulin germline with or without somatic hypermutation. Chimeric antibodies also include antibodies with light chain constant region that is identical or substantially identical to the sequence of human immunoglobulin germline and non-human (e.g., mouse) heavy chain or chimeric human / non-human heavy chain. Chimeric antibodies include antibodies having a light chain constant domain identical or substantially identical to a non-human (e.g., murine) immunoglobulin germline sequence, and a human or chimeric non-human / human heavy chain.

[0099] In some embodiments, the compositions and methods are for generating affinity matured antibodies. In some embodiments, an affinity matured antibody has a higher affinity (e.g., K in the nanomolar or picomolar range) for its target antigen compared to a substantially identical antibody without the alteration. D Affinity matured antibodies contain one or more alterations in one or more CDRs that result in a change in the affinity of the CDR or framework region. Affinity matured antibodies can be generated by any suitable method known in the art, such as by random or site-directed mutagenesis of the CDR and / or framework regions followed by affinity screening, VH domain shuffling, etc.

[0100] In some embodiments, the antibody is a neutralizing antibody. Neutralizing antibodies include antibodies that can neutralize, inhibit or block the biological activity of an antigen. Neutralizing antibodies include antibodies that, when bound to an antigen, block or reduce the ability of the antigen to act on its natural target in vivo and in vitro. Examples of neutralizing antibodies include antibodies against a ligand of a biological receptor that block the protein ligand from binding to the receptor, or antibodies against a biological receptor that block the binding of the receptor to the ligand, where the receptor undergoes a change inside the cell upon binding of the ligand in the absence of the antibody. Determining whether an antibody is a neutralizing antibody generally requires the implementation of a functional assay that measures the effect of the antibody on the biological activity of the antigen.

[0101] The methods and compositions of the invention are also useful in a variety of applications for antibodies and other binding proteins. A brief description of some useful applications is provided here.

[0102] Bispecific binding proteins can be made with binding specificities for tumor antigens and T cell antigens that target antigens on cells, such as CD20, and also target antigens on T cells, such as CD3. In this way, the bispecific antibody targets cells of interest in the patient (e.g., B cells in lymphoma patients, via CD20 binding) and targets the patient's T cells. The bispecific antibody is designed in various embodiments to activate T cells when bound to CD3, thus linking T cell activation to specific selected tumor cells.

[0103] Bispecific binding proteins can also be made that contain two binding moieties, each of which is directed to a binding partner on the surface of the same cell (i.e., each is directed to a different target). This design is particularly suitable for targeting a particular cell or cell type that expresses both targets on the surface of the same cell. Although the targets may appear individually on other cells, the binding moieties of these binding proteins are selected such that each binding moiety binds to its target with relatively low affinity (e.g., low micromolar or high nanomolar, e.g., KD greater than 100 nanomolar, e.g., 500, 600, 700, 800 nanomolar). Thus, long-term binding to the target is only preferred in situations where the two targets are present in close proximity on the same cell.

[0104] Bispecific binding proteins can be made that contain two binding moieties that bind to the same target, each binding to a different epitope of the same target. This design is particularly suitable for maximizing the probability of successfully blocking a target with a binding protein. Multiple extracellular loops, such as multiple extracellular loops of a transmembrane channel or cell surface receptor, can be targeted by the same bispecific binding molecule.

[0105] Bispecific binding proteins can be made that contain two binding moieties that cluster together and activate negative regulators of immune signaling, resulting in immune suppression. Repression in cis can be achieved when targets are present on the same cell; repression in trans can be achieved when targets are present on different cells. Repression in cis can be achieved, for example, by using a bispecific binding protein with an anti-IgGRIIb binding moiety and an anti-FelD1 binding moiety to downregulate immune responses to FelD1 (IgGRIIb only clusters in the presence of FelD1). Repression in trans can be achieved, for example, by using a bispecific binding protein with an anti-BTLA binding moiety and a binding moiety that specifically binds to a tissue-specific antigen of interest (clustering of inhibitory BTLA molecules occurs only in selected target tissues), which may address autoimmune diseases.

[0106] Bispecific binding proteins can be made to activate multicomponent receptors. In this design, two binding moieties for two components of a receptor bind to the receptor, crosslinking the receptor and activating signal transduction from the receptor. This can be done, for example, with a bispecific binding protein that has a binding moiety that binds to IFNAR1 and a binding moiety that binds to IFNAR2, where the receptors are crosslinked by binding. Such bispecific binding proteins can provide an alternative to interferon treatment.

[0107] Bispecific binding proteins can be created that transport binding moieties across semi-permeable barriers, such as the blood-brain barrier. In this design, one binding moiety binds to a target that can pass a specific selective barrier; the other binding moiety targets a molecule with therapeutic activity, where the target molecule with therapeutic activity cannot normally pass through the barrier. This type of bispecific binding protein is useful for taking therapeutic agents to tissues that they would not otherwise reach. Some examples include targeting the pIGR receptor to transport therapeutic agents to the intestine or lung, or targeting the transferrin receptor to transport therapeutic agents across the blood-brain barrier.

[0108] Bispecific binding proteins can be made that deliver binding moieties to specific cells or cell types. In this design, one binding moiety targets a cell surface protein (e.g., a receptor) that is readily internalized into the cell. The other binding moiety targets an intracellular protein, where binding of the intracellular protein provides a therapeutic effect.

[0109] The bispecific binding protein binds to a surface receptor on a phagocytic immune cell and to a surface molecule on an infectious pathogen (e.g., yeast or bacteria), bringing the infectious pathogen into close proximity with the phagocytic immune cell and facilitating phagocytosis of the pathogen. An example of such a design would be a bispecific antibody that targets CD64 or CD89 molecules and the pathogen.

[0110] Bispecific binding proteins have an antibody variable region as one binding part and a non-Ig part as the second binding part.The antibody variable region achieves targeting, while the non-Ig part is an effector or toxin linked to Fc.In this way, the ligand (e.g., effector or toxin) is delivered to the target that is bound to the antibody variable region.

[0111] A bispecific binding protein has two moieties, each bound to an Ig region (e.g., an Ig sequence containing CH2 and CH3 regions), such that any two protein moieties can be placed near each other in the context of an Fc. Examples of this design include capture, e.g., homodimeric or heterodimeric capture molecules.

[0112] nucleic acid Nucleic acid sequences encoding monoclonal antibodies can be obtained by any suitable method known in the art. Examples of suitable methods for obtaining monoclonal antibodies (and their nucleic acid sequences) include, for example, the hybridoma method (see, for example, Kohler et al. (1975) Nature 256:495-497) or phage antibody libraries (see, for example, Clackson et al. (1991) Nature 352:624-628).

[0113] In various embodiments, the immunoglobulin heavy chain variable domains are derived from nucleic acid sequences of genetically modified or transgenic animals. In some embodiments, the regions are derived from animals that include a human immunoglobulin minilocus. In some embodiments, the regions are derived from mice that include one or more extrachromosomal nucleic acids that include one or more nucleic acids encoding immunoglobulin sequences. In various embodiments, the animals may have one or more human immunoglobulin nucleic acid sequences that are unrearranged. In some embodiments, the animals include human light chain variable region nucleic acid sequences, in some embodiments include human heavy chain variable sequences, in some embodiments include both heavy and light chain variable sequences, and in some embodiments further include human constant region sequences. In certain embodiments, the nucleic acid sequences are derived from mice in which the endogenous mouse heavy and light chain variable gene segments have been replaced with human heavy and light chain variable gene segments.

[0114] In some embodiments, the nucleic acid sequences are obtained from naive B cells or T cells of such animals, hi other embodiments, the nucleic acid sequences are obtained from B cells or T cells of animals immunized with an antigen of interest.

[0115] In various embodiments, the nucleic acid sequence is obtained from the cell by amplifying the nucleic acid sequence with primers, including, for example, a set of degenerate primers that include one or more FR, linking or constant sequences.

[0116] In various embodiments, the immunoglobulin heavy chain variable domain is obtained from the nucleic acid of an animal immunized with an antigen of interest. For example, a non-human transgenic animal or a non-human genetically modified animal is immunized with an antigen of interest (e.g., by exposing the animal to the antigen or a cell bearing the antigen or a nucleic acid encoding an expressible form of the antigen), causing the animal to mount an immune response, isolating immune cells (e.g., B cells) from the animal, optionally immortalizing the cells, and screening the cells to identify reactivity with the antigen and / or to identify and / or isolate nucleic acid sequences encoding immunoglobulin variable regions that can recognize the antigen when placed in the context of an antibody. In some embodiments, the cells are B cells. In some embodiments, the B cells of the immunized animal are used to generate hybridomas, identifying B cells that express an antibody that specifically recognizes an epitope of the antigen, and identifying and / or isolating nucleic acid sequences encoding amino acid sequences of variable regions that recognize the epitope.

[0117] In some embodiments, the nucleic acid is obtained from a human, a non-human primate (e.g., an ape such as a chimpanzee), a monkey (e.g., a cynomologous or rhesus monkey), a rodent (e.g., a mouse, a rat, a hamster), a donkey, a goat, a sheep, etc.

[0118] In some embodiments, the heavy chain comprises sequences obtained from a human cell, for example, human fetal cells are exposed to antigen in vitro and placed into a suitable host animal (e.g., a SCID mouse).

[0119] In some embodiments, the nucleic acid is introduced into the cell using a vector, such as a plasmid, cosmid, retrovirus, adenovirus, adeno-associated virus, plant virus, YAC, BAC, or EBV-derived episome.

[0120] In some embodiments, the nucleic acid is present in an expression vector or expression construct. In some embodiments, the expression vector or expression construct is a vector that includes a promoter operably linked to the nucleic acid sequence of interest so that the nucleic acid sequence of interest can be expressed in an appropriate cell under appropriate conditions. The expression vector or expression construct may include leader sequences, enhancers, promoter elements that enhance transcription or translation, transcription termination sequences, splicing sequences, transcription enhancing introns, IRES elements, marker genes, selection sequences, recombinase recognition sites, homology arms, viral sequences, operators (e.g., prokaryotic operators), and the like. In some embodiments, the expression vector includes elements that allow for inducible expression, such as a prokaryotic operator operably linked to a eukaryotic promoter. In some embodiments, expression is induced upon addition of an expression inducer. In other embodiments, expression is induced upon removal of an expression inhibitor. In some embodiments, expression is induced by a temperature change.

[0121] In some embodiments, one or more heavy chain nucleotide sequences are present on the same vector. In some embodiments, the heavy chain nucleic acid sequences and the light chain nucleic acid sequences herein are present on the same vector. In one embodiment, two heavy chain nucleic acid sequences and one light chain nucleic acid sequence are present on the same vector.

[0122] In some embodiments, the nucleic acid is expressed in a cell that contains one or more nucleic acids from a virus, in certain embodiments, the virus is selected from adenovirus, adeno-associated virus, SV-40, Epstein-Barr virus, retrovirus, lentivirus, baculovirus, coronavirus, herpes simplex virus, poliovirus, Semliki Forest virus, Sindbis virus, and vaccinia virus.

[0123] A host cell is a cell that can be transformed to express a nucleic acid of interest. In various embodiments, transformation involves changing the nucleic acid content of the cell so that the cell contains an exogenous nucleic acid (e.g., a nucleic acid not naturally found in the cell, or one or more additional copies of a nucleic acid that corresponds to a nucleic acid sequence naturally found in the cell). The nucleic acid content of the cell can be altered by any suitable method known in the art, for example, by integrating the nucleic acid into the genome of the cell, or by placing the nucleic acid in the cell in an extrachromosomal or extragenomic form. In some embodiments, the nucleic acid content of the cell can be altered so that the cell transiently expresses the nucleic acid of interest, or the nucleic acid content can be altered so that the cell stably expresses the nucleic acid of interest. In some embodiments, the alteration of the genetic content of the cell is inherited when the cell divides.

[0124] Isolation of bispecific antigen-binding proteins Once a suitable set of modifications had been selected based on the information herein, attempts were made to isolate the bispecific antigen-binding proteins using methods known in the art. Simply applying the published methods did not provide satisfactory separation in all cases.

[0125] Lindhofer et al. made a bispecific antibody with a heterodimeric heavy chain, with one heavy chain (mouse IgG) that binds Protein A and one heavy chain (rat IgG) that does not bind Protein A, and successfully separated the heterodimer from a quadroma mixture of mouse / mouse and rat / rat dimers using a pH step gradient from neutral to pH 5.8 to elute the rat / mouse heterodimer bispecific antibody, followed by a pH step from 5.8 to 3.5 to elute the mouse / mouse homodimer. (See Lindhofer et al. (1995) Preferential species-restricted heavy / light chain pairing in rat / mouse quadromas: Implications for a single-step purification of bispecific antibodies. J. Immunol. 155(1):219-225).

[0126] The Lindhofer approach failed when applied to separate IgG1 homodimers from IgG1 heterodimers with two IgG1 heavy chains that are identical apart from the fact that one of the IgG1 CH3 domains contains the H435R / Y436F dipeptide modification. The inventors found that in a linear pH gradient, the dipeptide-modified IgG1 elutes at pH 3.9, whereas the IgG1 homodimer elutes at pH 3.7. This pH difference did not seem to be sufficient to adequately separate the heterodimer from the homodimer using the Lindhofer method. This difference was not reproducible in a predictable manner.

[0127] Variations in chromatographic behavior were observed in chromatographic runs using relatively substantial ionic strength, given by the buffer strength required to maintain a particular pH step or pH gradient. However, sufficient separation was not achieved by adding an organic modifier (1-propanol). Somewhat surprisingly, for some chromatographic runs, the separation of homodimeric IgG1 and heterodimeric IgG1 was instead dramatically and unexpectedly improved by adding 0.5-1.0 molar concentrations of an ionic modifier (e.g., NaCl). The addition of the ionic modifier broadened the pH range for elution (1.2 pH units with the ionic modifier compared to 0.2 pH units without the ionic modifier), allowing the successful separation of the two species by pH step gradient. However, in other runs, sufficient separation was achieved using only about 150 mM NaCl concentration (see Example 4). To ensure that sufficient separation can be achieved, in one embodiment, isolation of the bispecific antigen binding protein is performed in the presence of about 0.5 to about 1.0 molar ionic regulator.

[0128] Thus, in one embodiment, a method for separating a bispecific antigen binding protein comprising a heterodimeric IgG with one chain containing a modification as described herein comprises using a pH gradient in the presence of an ionic modifier. In one embodiment, the ionic modifier is present at a concentration sufficient to maximize the pH difference between the elution of the IgG homodimer from the Protein A support and the elution of the IgG heterodimer as described herein (i.e., with CH3 modifications). In a particular embodiment, the ionic modifier is present at a concentration of about 0.5 to about 1.0 molar. In another particular embodiment, the ionic modifier is present at a concentration of about 0.15 to about 0.5 molar.

[0129] In one embodiment, the ionic modifier is a salt. In one embodiment, the ionic modifier is a salt of an alkali metal or alkaline earth metal and a halogen. In a particular embodiment, the salt is an alkali metal or alkaline earth metal chloride salt, such as NaCl, KCl, LiCl, CaCl2, MgCl2. In a particular embodiment, the salt is present at a molar concentration of about 0.5, 0.6, 0.7, 0.8, 0.9 or 1.0.

[0130] In one embodiment, the pH gradient is from about pH 3.9 to about pH 4.5, in another embodiment from about pH 4.0 to about pH 4.4, and in another embodiment from about pH 4.1 to about pH 4.3. In a particular embodiment, the gradient is a linear gradient.

[0131] In one embodiment, the pH gradient is a step gradient. In one embodiment, the method includes applying steps of about pH 3.9, about pH 4.0, about pH 4.1, about pH 4.2, about pH 4.3, or about pH 4.4 to an equilibrated Protein A column (e.g., a column equilibrated with PBS or another suitable buffer or liquid). In a particular embodiment, the steps are at a pH of about 4.2.

[0132] In one embodiment, bispecific antibodies comprising heterodimeric IgG CH3 domains are eluted from the Protein A support in one or more fractions that are substantially free of non-heterodimeric IgG, hi certain embodiments, the eluted bispecific antibody fraction comprises less than about 1%, less than 0.5%, or less than 0.1% by weight of the total protein that is non-heterodimeric antibody. EXAMPLES

[0133] The following examples are proposed to illustrate to those skilled in the art how to make and use the methods and compositions of the present invention, and are not intended to limit the scope of what the inventors consider to be their invention. Efforts have been made to ensure accuracy with respect to numbers used (e.g., amounts, temperatures, etc.), but some experimental error and deviations will be accounted for. Unless otherwise indicated, parts are parts by weight, molecular weight is average molecular weight, temperature is in degrees Celsius, and pressure is at or near atmospheric pressure.

[0134] Example 1: Bispecific IL-4Ra / IL-6Ra antigen-binding proteins Two known antibodies of human IgG1 isotype, one against IL-4Ra and one against IL-6Ra, were found to have light chains that differed by only four amino acids. Co-expression experiments revealed that the light chain of the anti-IL-4Ra antibody could be replaced with the light chain from IL-6Ra and still maintain high affinity binding to IL-4Ra, thus making it possible to create bispecific antibodies using anti-IL-4Ra and anti-IL-6Ra heavy chains and the same light chain. Therefore, the heavy chain of the IL-6Ra antibody was modified to FcΔAdp type (i.e., CH3 dipeptide modification of H95R / Y96F in IMGT exon numbering).

[0135] Anti-IL-6Ra light chain was then co-expressed with anti-IL4Ra / Fc and anti-IL6Ra / FcΔAdp heavy chains in CHO cells, and conditioned media from these cells were subjected to Protein A chromatography. Cell supernatants containing a mixture of homodimers and heterodimers were loaded onto a Protein A column, followed by elution with a pH step gradient created by varying the combination of two buffers (A: 100 mM Na citrate, 150 mM NaCl, pH 6.0 and B: 100 mM Na citrate, 150 mM NaCl, pH 3.0; see FIG. 4) to give three phases at pH 6.0, pH 4.2, and pH 3.0, respectively. In FIG. 4, IL-4R represents anti-IL-4Ra and IL-6RΔ represents anti-IL-6Ra(IgG1ΔAdp). The indicated column fractions were assayed for binding to IL-6Ra and IL-4Ra proteins (see FIG. 5). Step elution resulted in one peak eluting at pH 4.2 and a second peak at pH 3.0 (Figure 4). BIACORE™ analysis showed that, as expected, the flow-through material was able to bind soluble IL-6Ra, but not IL-4Ra (Figure 5). The fraction corresponding to the pH 4.2 peak was able to bind approximately equal amounts of IL-6Ra and IL-4Ra, consistent with a heterodimer. The peak eluting at pH 3.0 was able to bind only IL-4Ra, but not IL-6Ra, corresponding to the expected anti-IL-4Ra homodimer. This demonstrated that heterodimeric bispecific antibodies can be efficiently isolated using Protein A chromatography with a simple pH step gradient.

[0136] Example 2: Pharmacokinetics of FcΔAdp protein To test whether the FcΔAdp modification affected the pharmacokinetics of molecules containing heterodimeric Fc / FcΔAdp, the purified heterodimeric species of anti-IL-4Ra / anti-IL-6Ra described above were injected into mice and human immunoglobulin concentrations in serum were measured over a 28-day period ( FIG. 6 ; Table 1 ). The serum half-life of the heterodimer was approximately 10 days, which was similar to that of the wild-type. This demonstrated that the FcΔAdp modification had no detectable effect on serum half-life.

[0137] [Table 1] Example 3: Bispecific CD20 / CD3 antigen-binding proteins It was found that the heavy chain of a known anti-human CD20 antibody was still able to bind to CD20 when co-expressed with the light chain of a known (activating) anti-human CD3 antibody. The anti-CD3 light chain was then co-expressed with anti-CD20 / Fc heavy chain, anti-CD3 / FcΔAdp heavy chain or both heavy chains. The resulting mixed population of homodimers and heterodimers was then used in a bioassay to measure their ability to kill target cells expressing CD20 (Figure 7). Briefly, 2x10 7 6 x 10 human PBMC cells 7 The cells were activated with 0.5×10 CD3×CD28 beads (Invitrogen) for 72 hours. Then, 30 units of IL-2 (R&D Systems) were added and the cells were incubated for an additional 24 hours. The cells were then cultured at 0.5×10 6 / mL and 30 U of IL-2 was added. The cells were then incubated for another 48 hours and used for bioassay. On the day of bioassay, 2 × 10 6Target cells (Raji) at 10 / mL were labeled with 8 μM calcein-AM (Invitrogen) for 30 min. Washed target cells were added to activated hPBMC cells at a 1:10 ratio of target:effector cells (total of 220,000 cells per well) in a total volume of 200 microliters with the indicated amount of antibody-containing supernatant. Cells were incubated for 2 h, supernatants were collected, and fluorescence was quantified. Cytotoxicity was measured by calculating the ratio of specific fluorescence to maximum fluorescence. Both CD20 antibody alone (with anti-CD3 light chain) and anti-CD3 antibody failed to induce killing of target cells; even mixing the two reagents was ineffective. However, significant killing was observed when all three components were co-expressed, suggesting that the effect was due to the heterodimeric bispecific species. The EC for this effect was calculated based on the estimated amount of bispecific antibody in the supernatant of transiently transfected CHO cells. 50 was estimated to be approximately 15 pM.

[0138] Example 4: Cell killing specificity using purified bispecific CD20 / CD3 antigen-binding proteins CHO cell supernatants from transfections as described in Example 3 were subjected to Protein A affinity chromatography using a step gradient for elution. The step gradient was created by varying the combination of two buffers (A: 20 mM Na citrate, 1 M NaCl, pH 5.2; B: 20 ​​mM Na citrate, 1 M NaCl, pH 2.7) to give three phases at pH 5.2, pH 4.2 and pH 2.8, respectively. Protein from the peak eluting at pH 4.2 was then used in a cell killing assay as described in Example 3. An EC of 3 pM was obtained. 50Killing of target cells was observed at 100-fold higher than that of the control (Figure 8). An additional cytotoxicity assay was performed to examine the target specificity of the observed killing. In this experiment, labeled target cells expressing CD20 (Raji) or not bearing CD20 (293) were incubated with activated human PBMCs. Each target cell type was added to the assay alone or in combination with other types of unlabeled target cells. In all cases, target cells expressing CD20 had an EC 50 CD20-negative cell lines were specifically killed by IFN-γ, whereas CD20-negative cell lines were not killed.

[0139] Example 5: Protein A separation of differentially modified hIgG2 Fc Differentially modified heterodimeric human IgG2 Fc / ΔAdpFc and unmodified human homodimeric IgG2 Fc / Fc were first concentrated by a bind-and-wash process over a Protein A column (rProteinA FF, GE). To further separate hIgG2 Fc / ΔAdpFc from hIgG2 Fc / Fc, a step gradient elution was performed using a SMART™ system (GE) as follows. The solvent system consisted of solvent A (PBS, 1×), solvent B (20 mM sodium citrate and 1 M NaCl, pH 5.5) and solvent C (20 mM sodium citrate and 1 M NaCl, pH 2.5). Elution started with an isocratic elution with 100% A for the first 20 min, followed by a quick switch to 100% B after 20 min. A linear gradient was then started towards 33.5% C and 66.5% B over the next 10 min; the concentration of C was maintained at 33.5% for 20 min until the first peak (Fc / ΔAdpFc) was completely eluted. This was followed by a linear gradient from 33.5% C to 100% C for 30 min. The flow rate was maintained at 250 microliters / min and the chromatogram was detected at 280 nm by a UV detector. hIgG2 Fc / ΔAdpFc eluted at pH 4.5, while hIgG2 eluted at pH 3.5.

[0140] Example 6: Protein A separation of differentially modified hIgG4 Fc Differentially modified heterodimeric human IgG4 (Fc / ΔAdpFc) and unmodified homodimeric IgG4 (Fc / Fc) were first concentrated by a bind-wash process over a Protein A column (rProteinA FF, GE). To further separate hIgG4 Fc / ΔAdpFc from hIgG4 Fc / Fc, a step gradient elution was performed using a SMART™ system (GE) as follows. The solvent system consisted of solvent A (PBS, 1×), solvent B (20 mM sodium citrate and 1 M NaCl, pH 5.1) and solvent C (20 mM sodium citrate and 1 M NaCl, pH 2.8). Elution started with an isocratic elution with 100% A for the first 20 min, followed by a quick switch to 100% B after 20 min. A linear gradient was then started towards 50% C and 50% B over the next 10 min; the concentration of C was kept at 50% for 20 min until the first peak (Fc / ΔAdpFc) was completely eluted. This was followed by a linear gradient from 50% C to 100% C for 30 min. The flow rate was kept at 250 microliters / min and the chromatogram was detected at 280 nm by a UV detector. hIgG4 Fc / ΔdpFc eluted at pH 4, while the homodimer eluted during the gradient from pH 4 to pH 2.8.

[0141] Example 7: Protein A separation of differentially engineered hIgG1 CD3xCD20 Differentially engineered heterodimeric anti-hCD3xCD20 IgG1(Fc / ΔAdpFc) and unmodified homodimeric anti-hCD20 were separated on a 1 mL rProtein AFF (GE Biosciences) column as follows: The solvent system consisted of Buffer A1 (PBS 1x), Buffer A2 (20 mM sodium citrate and 1M NaCl pH 5.1), Buffer B (20 mM sodium citrate and 1M NaCl pH 5.1), and Buffer C (20 mM sodium citrate and 1M NaCl pH 5.1). The mixed sample was allowed to bind and washed with PBS and buffer A2. A step was used to reach pH 4.2, which resulted in the bispecific CD3* ×CD20 IgG1(Fc / ΔAdpFc) was eluted, followed by homodimeric anti-hCD20 IgG1 with a linear gradient from pH 4.2 to pH 2.8.

[0142] Example 8: Binding affinity of modified CH3 to Fc receptors The binding affinity of a human IgG1 isotype bispecific antibody with ΔAdp modifications (H435R and Y436F, EU numbering) to various human Fc receptors was tested in a Biacore™ steady-state equilibrium binding assay.

[0143] Briefly, carboxymethylated dextran (CM5) chips with amine-coupled anti-penta-his mAb (Qiagen) were used to capture various constructs of human Fc receptors. The following his-tagged Fc receptor ectodomains were coupled to the surface of various CM5 chips coated with anti-penta-his: FcγRI, FcγRIIA (R131 polymorph), FcγRIIB, and FcγRIIIB (each obtained from R&D Systems); and RcγRIIA (H131 polymorph), FcγRIIIA (V176 polymorph), and RcγRIIIA (F176 polymorph) (each manufactured by Regeneron). Antibodies were passed over the surfaces at three concentrations (25 nM, 50 nM, and 100 nM) for the high affinity receptor FcγRI ectodomain, and at concentrations between 5 micromolar and 39 nanomolar for the low affinity FcγR receptor ectodomain, and the association and dissociation rate constants (k a and k d ) values ​​and used to determine the equilibrium dissociation constants (K D ) was calculated. Binding studies were performed at room temperature using HBS-T buffer at pH 7.2. K D The K for anti-CD3Δdp antibody was measured. DThe values ​​revealed no significant differences in binding to any of the Fc receptors tested compared to the unmodified hIgG1 isotype antibody (Table 2).

[0144] [Table 2] Example 9: Pharmacokinetics of bispecific hIgG1ΔAdp in hFcRn mice The pharmacokinetic clearance rates of bispecific anti-hCD3 / hCD20 IgG1ΔAdp antibodies and their antibody-related controls (anti-hCD3 IgG and anti-hCD3 IgGΔAdp homodimer) were measured in wild-type (WT) mice and mice homozygous for replacement of mouse FcRn with the hFcRn gene (hFcRn mice). Wild-type and hFcRn mice were from crossbred strains with backgrounds including C57BL6 (75%) and 129Sv (25%). Cohorts included 4 each of WT or hFcRn mice, except for one cohort of WT mice receiving an IgG1 isotype-matched control antibody (this cohort included 3 mice). Mice were administered 1 mg / kg of isotype-matched (hIgG1) control, anti-hCD3×CD20IgG1ΔAdp bispecific, anti-hCD3IgG1 or anti-hCD3IgG1ΔAdp homodimer. All test substances were administered subcutaneously. Bleeds were collected at 0 hours, 6 hours, 1 day, 2 days, 3 days, 4 days, 7 days, 10 days, 14 days, 21 days, and 30 days.

[0145] Serum levels of human antibodies were measured by sandwich ELISA. Briefly, goat polyclonal anti-human IgG (Fc specific) antibody (Jackson ImmunoResearch) was coated on 96-well plates at a concentration of 1 microgram / mL and incubated overnight at 4°C. After blocking the plate with BSA, serum samples in 6-point serial dilutions and control standards of each antibody in 12-point serial dilutions were added to the plate and incubated at room temperature for 1 hour. After washing to remove unbound antibodies, captured human antibodies were detected using the same goat polyclonal anti-human IgG (Fc specific) antibody (Jackson ImmunoResearch) conjugated with horseradish peroxidase (HRP) and developed with standard colorimetric tetramethylbenzidine (TMB) substrate according to the manufacturer's recommendations. The absorbance at 450 nm was recorded on a plate reader, and the concentration of hIgG in serum samples was calculated using a standard calibration curve made on the sample plate.

[0146] No significant differences were observed in the serum half-life of the four IgG1 antibodies over the 30 days tested. In particular, no significant differences were observed between the IgG1 antibodies with the ΔAdp modification and the wild-type IgG1 antibodies. No differences were observed between the antibodies when using wild-type (mFcRn) mice or mice with humanized FcRn (hFcRn). As expected, the hFcRn mice showed a slightly faster clearance than the wild-type mice. The results are shown in Table 3.

[0147] [Table 3] Example 10: Large scale isolation in low salt buffer Bispecific CD3xCD20ΔAdp antibodies were isolated from bulk cultures according to the present invention. Briefly, CHO-K1 cell lines expressing bispecific anti-hCD3xCD20ΔAdp (CD3 heavy chain modified) antibodies were cultured in 11 liter bioreactors. Cells producing the bispecific antibodies grew to a density of about 8.25×106 cells / mL, yielding about 250-350 mg antibody / L. In contrast, control anti-hCD3 antibodies were obtained at about 100-150 mg / L.

[0148] Antibodies were isolated on MabSelect SuRe™ resin (GE) (20 cm bed height, 1 cm internal diameter) equilibrated with 10 mM sodium phosphate, 0.5 M NaCl, pH 7.2, the clarified cell culture was loaded at 19 g / L, and the column was washed with 3 column volumes of 10 mM sodium phosphate, 0.5 M NaCl, pH 7.2, followed by 2 column volumes of 20 mM sodium phosphate, pH 7.2 (no NaCl). Antibodies were eluted with 40 mM acetate, pH 3.0.

[0149] The monospecific anti-CD30 antibody eluted at pH 3.6, whereas the bispecific anti-hCD3xCD20DAdp eluted at pH 4.4.

[0150] Example 11: Selective Protein A elution of mouse heterodimers at near-neutral pH CHO-K1 cells were transiently transfected with expression constructs for human IFNAR1 (hIFNAR1) and human IFNAR2 (hIFNAR2) extracellular domains fused to wild-type or mutant (TTTK or PTTK) mIgG2a Fc. The ratio of hIFNAR1-mFc to hIFNAR2-mFc was kept 1:1 by transfecting the cells with equal amounts of the two expression plasmids. The culture medium was harvested 4 days after transfection and subjected to Protein A purification using a 0.2 mL NAb Protein A Plus™ spin column (Thermo Scientific / Pierce). Briefly, the column was equilibrated with 1×PBS, pH 7.2. 1 ml of CHO-K1 culture medium was incubated with Protein A resin for 10 min at room temperature. The column was then washed three times with 1×PBS, pH 7.2. Bound proteins were eluted with 20 mM sodium citrate buffer containing 1 M NaCl. Elution was performed three times with 0.4 mL of elution buffer with decreasing pH. Proteins in different fractions were detected by Western blot analysis.

[0151] The results show that it is possible to separate wild type and star mutant mIgG2a heterodimers from wild type mIgG2a homodimers when using pH gradient elution (see fraction E1 in both gels in Figure 9).

[0152] Example 12: Preferential heterodimer formation of mIgG2a mutants versus isotype heterodimers DNA plasmids were constructed for mammalian expression of the extracellular domains of human type I interferon receptors (hIFNAR1 and hIFNAR2) with C-terminal mouse Fc (mIgG2a or mIgG1). Site-directed mutagenesis was used to introduce mutations within the mIgG2a sequence. These mutants are: TTT=M252T, S254T, S256T; TTTK=M252T, S254T, S256T, I258K; PTTTK=I247P, M252T, S254T, S256T, I258K; RF=H435R, H436F. CHO-K1 cells were transiently transfected with the above expression constructs. The ratio of IFNAR1-mFc to IFNAR2-mFc was kept at 4:1 by transfecting cells with four times more hIFNAR1-mFc expression plasmid (hIFNAR1-mIgG2a) than hIFNAR2-mFc (mIgG1 or mutant mIgG2a). The culture medium was harvested 4 days after transfection, and mFc protein was detected by Western blot analysis.

[0153] The results show that heterodimer formation between mIgG2a and mIgG1 is much less efficient than that between wild-type mIgG2a and mIgG2a mutants (compare lane 1 with lanes 2-5 in FIG. 10). In this experiment, an excess of wild-type IgG2a constructs was maintained by using a 4:1 ratio of IFNAR1:IFNAR2 constructs.

Claims

[Claim 1] A method or composition as described in the specification.

Images