RU2780412C2 - Methods for production of multi-specific and multivalent antibodies - Google Patents

Abstract

FIELD: biotechnology.

SUBSTANCE: method for the production of a mixture of antibodies, containing two monospecific antibodies and one bispecific antibody, is proposed, where all antibodies have a common heavy chain. Two antibodies with different specificity determined by the same variable domain of the heavy chain are isolated. Then, two different light chains and one heavy chain of the specified antibodies are co-expressed in a cell, which leads to the assembly of monospecific and bispecific antibodies. One of the specified light chains contains the Kappa constant domain, and the other contains the Lambda constant domain. A mixture of antibodies is also proposed for isolation of two monospecific antibodies and one bispecific antibody, obtained by the specified method.

EFFECT: invention provides regulation of the ratio of different antibodies in the resulting mixture for maximization of production of a bispecific antibody.

8 cl, 37 dwg, 5 tbl, 15 ex

Description

Related Application

[0001] This application claims U.S. Provisional Application No. 61/374159 filed August 16, 2010, U.S. Provisional Application No. 61/443008 filed February 15, 2011, U.S. Provisional Application No. 61/509260 filed July 19, 2011, the contents of which are incorporated in this document by reference in its entirety.

The technical field to which the invention belongs

[0002] The invention relates to the production of novel bispecific monoclonal antibodies having different specificities for each binding site of an immunoglobulin molecule. The antibodies of the invention are composed of one heavy chain and two different light chains, one of which contains a Kappa constant domain and the other a Lambda constant domain. The invention particularly relates to the isolation of antibodies of different specificity but having a common heavy chain. The invention further relates to the controlled co-expression of two light chains and one heavy chain, resulting in the assembly of monospecific and bispecific antibodies. The invention provides a method for producing a fully human bispecific and bivalent antibody that does not change the sequence and does not include the use of linkers or other sequences with a non-human origin, and also provides a method for producing a mixture of antibodies from two monospecific antibodies and one bispecific antibody. The invention also provides methods for efficiently purifying a bispecific antibody.

Prior Art

[0003] An antibody consists of four polypeptides: two heavy chains and two light chains. The antigen-binding region of an antibody is formed by a light chain variable domain (VL) and a heavy chain variable domain (VH). At one end of these domains, six loops form an antigen-binding site, also referred to as a hypervariable region (CDR). Three CDRs are located in the VH domain (H1, H2 and H3) and three others are located in the VL domain (LI, L2 and L3). During B cell development, a unique region of immunoglobulin is formed by somatic recombination, known as V(D)J recombination. The variable region of an immunoglobulin heavy or light chain is encoded by different gene segments. The heavy chain is encoded by three segments called the variable segment (V), the D-segment (defining diversity) (D), and the connecting (J) segment, while the variability of the light chain is formed by the recombination of only two segments, V and J. A large number of paratopes antibodies can be generated by recombination between one of the many copies of the V, D and J segments that are present in the genome. The V segment encodes CDR1 and CDR2, while CDR3 is obtained through recombination events. During the immune response, additional diversity is introduced into the antigen-binding site through a process called somatic overmutation (SHM). During this process, point mutations are introduced into the heavy and light chain variable genes, and specifically into the CDR-coding regions. This additional variability allows the selection and expansion of B cells expressing antibody variants with increased affinity for the antigen they recognize.

[0004] A large number of immunoglobulins are divalent and monospecific molecules that carry the same specificity on both arms, as they consist of two identical heavy chain polypeptides and two identical light chain polypeptides. However, almost immediately during the development of hybridoma technology, it was found that hybrid hybridomas can be created by merging two hybridomas (Suresh MR et al., Methods Enzymol 1986; 121: 210-228). These 'quadromas' express two different heavy and two different light chains and thus produce many different kinds of antibodies resulting from the random pairing of heavy and light chains. Among these different species, bispecific antibodies (bsAbs) are produced bearing different specificities on each arm. Another naturally occurring exception is the immunoglobulin of the IgG4 isotype, which is able to undergo heavy chain exchange due to less stable dimerization mediated by the hinge region of this isotype (van der Neut Kolfschoten M et al., Science. 2007 317(5844): 1554-7) . Although this exchange appears to occur in vivo, its biological significance remains unclear.

[0005] Monoclonal antibodies have emerged as a successful and attractive class of molecules for therapeutic administration in certain human diseases. However, targeting or neutralizing a single protein is not always sufficient to achieve efficacy in some diseases that limit the therapeutic use of monoclonal antibodies. It is even more obvious that in some cases the neutralization of one component of the biological system is not enough to achieve efficiency. One solution to the problem is the co-administration of several monoclonal antibodies. However, this method is hampered by regulatory aspects if the pooled antibodies have not been previously improved on an individual basis. In addition, combinatorial methods are also expensive from a manufacturing perspective. Accordingly, there is a need to create antibodies and drugs that allow targeted action on multiple antigens with a single molecule.

The essence of the invention

[0006] The invention provides the ability to identify, produce and purify bispecific antibodies that are indistinguishable in sequence from standard antibodies. The invention also provides the possibility of obtaining and purifying a simple mixture of antibodies, consisting of three or more antibodies carrying the same heavy chains. The unmodified nature of the antibodies of the invention provides them with advantageous manufacturing characteristics similar to those of standard monoclonal antibodies.

[0007] The bispecific antibodies of the invention are prepared using the following steps:

- isolated two antibodies with different specificities and shared the same heavy chain variable domain, but with different light chain variable domains. This step was facilitated by using antibody libraries containing fixed heavy chains or by using transgenic animals containing the same VH genes.

the heavy chain variable domain was fused to the heavy chain constant region, the variable domain of one light chain was fused to the Kappa constant domain, and the variable domain of the other light chain was fused to the Lambda constant domain. Preferably, the light chain variable domain fused to the Kappa chain constant domain is of the Kappa type and the variable domain fused to the Lambda constant domain is of the Lambda type. However, the invention also allows the production of hybrid light chains so that two light chain variable domains of the same type can be used to produce the bispecific antibodies of the invention.

the three chains were co-expressed in mammalian cells resulting in the assembly and secretion into the supernatant of a mixture of three antibodies: two monospecific and one bispecific antibody carrying two different light chains. The ratio of different antibodies depends on the relative expression of the chains and their assembly in IgG. The invention provides methods for regulating these ratios and maximizing the production of a bispecific antibody.

- the mixture of antibodies was purified using standard chromatographic methods used to purify antibodies. A mixture of antibodies can be characterized and used as a multi-targeting agent.

The bispecific antibody was purified using a stepwise affinity chromatographic material that specifically binds to the human Kappa and human Lambda constant regions. This purification process is independent of the light chain variable domain sequence and thus is characteristic of all bispecific antibodies of the invention.

isolated bispecific antibody carrying a light chain containing a constant Kappa domain and a light chain containing a constant Lambda domain were characterized using various biochemical and immunological methods.

the bispecific antibody of the invention can be used for therapeutic purposes or as a research or diagnostic reagent.

[0008] The invention relates to a monoclonal antibody having a different specificity at each combined site and comprising two copies of a single heavy chain polypeptide and a first light chain and a second light chain, where the first and second light chains are different.

[0009] In some antibodies, at least the first part of the first light chain is of the Kappa type and at least part of the second light chain is of the Lambda type. In some antibodies, the first light chain includes at least a Kappa constant region. In some antibodies, the first light chain further includes a Kappa variable region. In some antibodies, the first light chain further includes a lambda variable region. In some antibodies, the second light chain includes at least a Lambda constant region. In some antibodies, the second light chain further includes a lambda variable region. In some antibodies, the second light chain further includes a Kappa variable region. In some antibodies, the first light chain includes a Kappa constant region and a Kappa variable region, and the second light chain includes a Lambda constant region and a Lambda variable region.

[0010] In some embodiments, the constant and variable framework sequences are human.

[0011] The invention relates to methods for producing a bispecific antibody by a) isolating an antibody or a portion of an antibody fragment having a specificity determined by a heavy chain variable domain together with a first light chain variable domain; b) isolating an antibody or portion of an antibody fragment having a different specificity defined by the same heavy chain variable domain as the antibody of step a) together with a second light chain variable domain; c) co-expression in a cell: (i) a common heavy chain variable domain fused to an immunoglobulin heavy chain constant region; (ii) a first light chain variable domain fused to either a Kappa-type light chain constant domain or a Lambda-type light chain constant domain; and (iii) a second light chain variable domain fused to a light chain constant domain of a different type than the chain type of the first variable constant domain.

[0012] Some methods also include the additional step d) isolating bispecific antibodies derived from the obtained monospecific antibodies. For example, in some methods, isolation is performed using an affinity chromatography purification step. In some methods, the purification step is carried out using an affinity chromatographic material specific for the Kappa constant domain for a Lambda constant domain specific for both of them, the Kappa constant domain and the Lambda constant domain.

[0013] In some methods, the Kappa light chain variable domain is fused to a Kappa-type constant region. In some methods, the Kappa light chain variable domain is fused to a Lambda-type constant region. In some methods, the lambda light chain variable domain is fused to a Kappa-type constant region. In some methods, the lambda light chain variable domain is fused to a lambda-type constant region.

[0014] In some methods, steps a) and b) are facilitated by the use of antibody libraries having a common heavy chain, and the diversity is limited to light chain variable domains. A heavy chain variable domain that is present in one of these libraries can be based on different variable germline genes and have different sequences in both the CDR and the framework region. In some methods, such libraries are constructed using various types of heavy chain variable domains and can be used to generate antibodies of the invention.

[0015] In some methods, the antibody library is presented on a filamentous bacteriophage, on the surface of yeast, bacteria, or mammalian cells, or used for ribosomes or other type of in vitro display.

[0016] The invention also relates to methods for producing a bispecific antibody that specifically binds to a first antigen and to a second antigen, where the first and second antigens are different, by a) providing a first nucleic acid molecule encoding a first polypeptide comprising a heavy chain variable region an immunoglobulin polypeptide or fragment thereof that binds to the first antigen in association with an immunoglobulin constant region; b) providing a second nucleic acid molecule encoding a second polypeptide comprising a light chain variable region of an immunoglobulin polypeptide, or a fragment thereof, that binds to a first antigen in conjunction with a Kappa-type or Lambda-type first light chain constant region; c) providing a third nucleic acid molecule encoding a third polypeptide comprising an immunoglobulin polypeptide light chain variable region or fragment thereof that binds to a second antigen in conjunction with a Kappa-type or Lambda-type second light chain constant region, wherein the constant domains of the first light chain and the second light chain are of different types; and (d) culturing a host cell comprising the first, second, and third nucleic acid molecules under conditions that allow expression of the first, second, and third polypeptides.

[0017] Some methods also include the additional step of e) recovering the bispecific antibody. For example, in some methods, the bispecific antibody is recovered in step e) using an affinity chromatography purification step. In some methods, the purification step is carried out using an affinity chromatographic material specific for the Kappa constant domain for a Lambda constant domain specific for both of them, the Kappa constant domain and the Lambda constant domain.

[0018] In some methods, the second nucleic acid encodes a Kappa-type light chain variable domain. In some methods, the second nucleic acid encodes a Kappa-type constant region. In some methods, the second nucleic acid encodes a Lambda-type constant region. In some methods, the second nucleic acid encodes a lambda-type light chain variable domain. In some methods, the second nucleic acid encodes a Kappa-type constant region. In some methods, the second nucleic acid encodes a Lambda-type constant region. In some methods, the third nucleic acid encodes a Kappa-type light chain variable domain. In some methods, the third nucleic acid encodes a Kappa-type constant region. In some methods, the third nucleic acid encodes a Lambda-type constant region. In some methods, the third nucleic acid encodes a lambda-type light chain variable domain. In some methods, the third nucleic acid encodes a Kappa-type constant region. In some methods, the third nucleic acid encodes a Lambda-type constant region.

[0019] The invention also relates to a mixture of antibodies, which includes two monospecific antibodies and one bispecific antibody that share a common heavy chain. For example, a bispecific antibody is any of the bispecific antibodies described herein or obtained using the methods described herein. The invention also relates to methods for preparing such a mixture of antibodies by a) isolating an antibody or a portion of an antibody fragment having a specificity determined by the heavy chain variable domain together with the first light chain variable domain; b) isolating an antibody or portion of an antibody fragment having a different specificity defined by the same heavy chain variable domain as the antibody of step a) together with a second light chain variable domain; c) co-expression in a cell: (i) a common heavy chain variable domain fused to an immunoglobulin heavy chain constant region; (ii) a first light chain variable domain fused to either a Kappa-type light chain constant domain or a Lambda-type light chain constant domain; and (iii) a second light chain variable domain fused to a Kappa-type light chain constant domain or fused to a Lambda-type light chain constant domain. Some methods also include the additional step d) isolating the mixture of antibodies obtained in step c) from the cell culture supernatant.

[0020] The invention also relates to methods for producing two or more, for example, three or more non-identical antibodies in one recombinant host cell by a) expressing in one recombinant host cell one or more nucleic acid sequences encoding a common immunoglobulin heavy chain and at least two, for example, at least three different immunoglobulin light chains that are capable of pairing with a common immunoglobulin heavy chain to form functional antigen-binding domains and to produce two or more, for example, three or more, non-identical antibodies that include a common heavy chain. Some methods also include the step of collecting cells or otherwise purifying two or more, for example three or more, non-identical antibodies from a recombinant host cell or from a host cell culture. The host cell may be, for example, a mammalian cell. In some methods, non-identical antibodies include monospecific and bispecific antibodies.

[0021] In some methods, non-identical antibodies target different epitopes of the same target antigen. In some methods, non-identical antibodies have different affinities for the same target epitope. In some methods, non-identical antibodies bind to different antigens.

[0022] In some methods, two or more, for example, three or more non-identical antibodies are independently selected from the group consisting of: IgG1, IgG2, IgG3, IgG4, IgA1, IgA2, IgD, IgE, and IgM.

[0023] In some embodiments, two or more, for example, three or more non-identical antibodies contain a modified Fc region that modifies antibody effector functions, such as antigen-dependent cell-mediated cytotoxicity (ADCC), complement-dependent cytotoxicity (CDC), antigen-dependent cellular phagocytosis (ADCP), or modify their pharmacokinetic properties by altering their binding to neonatal Fc receptors.

[0024] In some methods, two or more, for example, three or more different immunoglobulins are represented in the format F(ab')2.

[0025] In some methods, one or more nucleic acid sequences are stably expressed in a host cell.

[0026] In some methods, two or more, for example, three or more, non-identical antibodies are produced by a host cell in vitro.

[0027] Some methods also include the additional steps of selecting at least one host cell by evaluating two or more, for example, three or more non-identical antibodies produced by the recombinant host cell for their ability to bind to the target antigen; a step of culturing the recombinant host cell; and isolating three or more non-identical antibodies. Antibodies can be isolated using any of the methods described herein, or using any other methods known in the art.

[0028] In some methods, different immunoglobulin light chains contain identical constant regions. In some methods, different immunoglobulin light chains contain different constant regions.

Brief description of the drawings

[0029] Figure 1 is a schematic representation of various formats of a bispecific antibody. A. Antibody fragment formats: X-linked Fab, cross-linked Fab fragments; tascFv/BiTE, dual-scFv/bispecific T cell invader; Db, diabody; taDb, double diabody. B. Fc fusion formats: Db-Fc, diatelo-Fc fusion; taDb-Fc hybrid, double diabody-Fc hybrid; taDb-CH3, double diabody-CH3 hybrid; (scFv) ₄ -Fc, tetra-scFv-Fc hybrid; DVD-Ig, double variable domain immunoglobulin. C. IgG formats: "knob-hole" and SEED, a domain constructed by chain exchange; CrossMab, "knob-hole" along with heavy and light chain domain exchange; bsAb, bispecific antibody isolated from quadroma (quadroma); sdAb, single domain antibody.

[0030] Figure 2 is a schematic representation of possible modes of action with the support of bispecific antibodies. 2A. Directed action on two antigens. 2b. Repeated directed exposure of a toxic component or activity to a target cell. 2C. Increased selectivity mediated by avidity.

[0031] Figures 3A-3C are a schematic representation of the structure of various bispecific antibodies of the invention, consisting of two copies of a unique heavy chain polypeptide and two different light chain polypeptides. The localizations and/or locations of the Kappa light chain and the Lambda light chain (or portions thereof) shown in these figures are not intended to be limiting. Those skilled in the art will appreciate that the Kappa light chain and the Lambda light chain (or portions thereof) can also be arranged to mirror the bispecific antibodies shown in Figures 3A-3C. Those of skill in the art will also appreciate that the bispecific antibodies that are presented in full length IgG format in Figures 3A-3C can also be made using other immunoglobulin isotypes or in other immunoglobulin formats such as F(ab')2. 3A. A Kappa variable domain fused to a Kappa constant domain; and a Lambda variable domain fused to a Lambda constant domain. 3b. Kappa variable domains fused to the Kappa constant domain and to the Lambda constant domain. 3C. Lambda variable domains fused to the Kappa constant domain and to the Lambda constant domain.

[0032] Figure 4 is an illustration of an ELISA assay testing clones specific for hCXCL10-NusA or hIL6RC and carrying the same heavy chain variable domain. Each clone was tested against both targets to demonstrate specificity.

[0033] Figures 5A-C are a series of illustrations depicting the three types of libraries used in the Examples, for each type of library, Vκ and Vλ, the libraries were supported separately. Figures 5A and C: Two sets of libraries that contain a fixed VH3-23 variable domain that differs only in their H3 CDR sequence which is H3 below (CDR definition according to IMGT). The light chain combination was altered either by inserting random sequences into CDRL3 of selected light chain variable genes (Figures 5A and 5C) or by capturing naturally rearranged light chain variable domains isolated from human donors, which may include all human variable genes and contain divergences in all 3 CDRs (Figure 5B). Various diversification strategies are illustrated with horizontal lines below the altered portions of the light chain combinations.

[0034] Figures 6A-6B are graphs depicting ELISA results using monospecific IgGλ, and IgGκ selected against hIFNγ and IL6RC, respectively, and carrying a common heavy chain. ELISA formats are shown schematically after each graph. In Figure 6A, INFγ was immobilized on the plate, incubated with anti-IFNγ IgGλ or with anti-IL6RC (i.e., soluble IL-6R/IL-6 receptor complex) IgGκ, and both were detected with antibodies binding human Gκ, or antibodies binding human Cλ associated with horseradish peroxidase. The signal was detected by colorimetry and quantified using a microplate reader.

[0035] Figure 7 is a schematic representation of vectors used for co-expression of one heavy chain and two light chains in mammalian cells. Both vectors contain three promoters to drive gene expression, a glutamine synthetase gene for selection of a stable cell line. In the second vector, pNovI kHλk, expression of the additional Kappa light chain is driven by an internal ribosome entry site (IRES). Various genes and genetic control elements are indicated. hCMV, human cytomegalovirus promoter; SV40, V40 promoter; pA polyadenylation signal; VH, heavy chain variable domain; VK, Kappa light chain variable domain; CK constant domain of the Kappa light chain; VL, Lambda light chain variable domain; CL2, lambda light chain constant domain 2; GS cDNA, glutamine synthetase cDNA; AmpR; selection marker for resistance to ampicillin. The selected number of restriction sites is indicated.

[0036] Figure 8A is a schematic representation of a purification process for bispecific antibodies of the invention. Figure 8B is an illustration depicting the co-expression, purification and SDS-PAGE analysis of the bispecific antibodies of the invention. The gel was stained using simply blue dye. PA: Protein A; K: Kappa selection; λ: Lambda selection; FT: flow column; E: eluted fraction.

[0037] Figure 9 is an illustration of SDS-PAGE analysis of total IgG purified from mammalian cells transfected with vectors allowing different levels of Kappa light chain expression using different IRES elements within the pNovI kHλk vector (lane 1-5) and compared with pNovI kHλ vector. The relative light chain intensities of Kappa and Lambda reveal that the level of expression can be modulated.

[0038] Figure 10A is an illustration depicting IEF gel analysis of purified monospecific IgG (κκ and λλ) and bispecific IgG (κλ). Figure 10B is an illustration of IEX-HPLC analysis of monospecific and bispecific antibodies. Three antibodies were injected independently and their elution profile plotted. The gradient used in the experiment is presented.

[0039] Figure 11 depicts an ELISA assay used to determine the ability of a bispecific antibody to bind to both the target and the Kappa and Lambda light chains in the molecule. Figure 11A is a schematic representation of the ELISA format. Figure 11B is a graph depicting the results of an ELISA with INFγ immobilized on a plate. Figure 11C is a graph depicting the results of an ELISA with IL-6RC immobilized on a plate. IgGk, monospecific antibody to IL6RC; IgGλ, monospecific antibody to IFNγ; IgGkλ, bispecific antibody to IL6RC/INFγ. Secondary detection antibodies, anti-human lambda HRP and anti-human kappa HRP are indicated.

[0040] Figure 12 is an illustration of SPR alysis of bispecific IgGκλ antibodies. In Figure 12A, INFγ was immobilized on the surface of a Biacore chip and an anti-IL6RC monospecific antibody (IgGκ), an anti-INFγ monospecific antibody (IgGλ), and an anti-IL6RC/INFγ bispecific antibody (IgGκλ) were injected onto the surface, followed by injection of IL6RC. In Figure 12B, an anti-IL6RC/INFγ (IgGκλ) bispecific antibody was immobilized on the surface of the chip, and an anti-human Kappa antibody and an anti-human Lambda protein were injected at the same concentration. The experiment was repeated by inverting the order of injection of antibodies with identical results.

[0041] Figure 13 is a schematic representation of a brief description of one method for obtaining bispecific and multispecific antibodies described herein in CHO cells.

[0042] Figure 14A is a graph depicting the growth and antibody production profiles of CHO cell pools at laboratory production level in a conical Erlenmeyer flask. The level of antibody production was determined using the analysis of Protein A-HPLC. VCC means viable cell concentration and Ab means antibody. Figure 14B is a graph depicting a comparison of growth profile and antibody production profile between laboratory and pilot fermentations. The level of antibody production was determined using the analysis of Protein A-HPLC. VCC means viable cell concentration and Ab means antibody.

[0043] Figure 15A and 15B are a series of graphs depicting antibody production in a 96-well plate (96wpl) by cell lines expressing anti-Kappa (KK) monospecific antibody, Lambda (LL) monospecific antibody, and anti-Kappa and anti-Lambda bispecific antibody. (KL) five weeks after transfection in two independent experiments. The level of antibody production was determined using ELISA, where m Ab means monoclonal antibody. Figures 15C and 15D are a series of graphs depicting antibody production in a shake 24-well plate (24wpl) by overnight batch cultures of anti-kappa monospecific antibody (KK), multispecific lambda antibody (LL), and bispecific anti-kappa antibody.

[0044] Figure 16A is an illustration depicting the results of SDS-PAGE analysis of monospecific κ IgG molecules (i.e., monospecific molecules containing Kappa light chains, also referred to herein as "mono κ" molecules), monospecific λ IgG molecules (i.e., monospecific molecules containing Lambda light chains, also referred to as "mono λ" molecules), and κλ antibodies (i.e., antibodies containing both light chains, Kappa and Lambda), through the purification steps described above. Figure 16B is an illustration depicting the results of SDS-PAGE analysis of mono κ, mono λ, and κλ antibodies obtained after the elution steps described above. 16A and 16B, the gel was stained with "simply blue" and E denoted the elution fraction, FT denoted the flow fraction, and MM denoted the molecular weight marker. Figure 16C is an illustration depicting isoelectric focusing (IEF) gel analysis of purified monospecific IgG molecules (κκ and λλ) and bispecific IgG molecule (kλ).

[0045] Figures 17A-D are a series of graphs and illustrations depicting that the methods for making bispecific antibodies of the invention produce antibodies that include both light chains, Kappa and Lambda, and that purified antibodies exhibit bispecificity. Graphs depict the results of ELISA using antibodies to κλ against hIFNγ and IL6RC. ELISA was performed using the indicated anti-Kappa and anti-Lambda detection antibodies. Figures 17A-D illustrate that the Lambda light chain binds to hIFNγ while the Kappa light chain binds to IL6RC. NI-0501 is a control antibody for the Lambda light chain, an anti-hIFNγ antibody, and NI-1201 is a control antibody for the Kappa light chain, an anti-IL6RC antibody.

[0046] Figure 18 is an illustration of an IEF gel analysis for various monospecific and bispecific antibodies, revealing that the difference in pI can vary depending on the variable light chain sequence of the antibody. Lane 1, anti-NusA IgGκ; Lane 2, anti-NusA/anti-IFNγ IgGκλ; Lane 3, anti-IFNγ IgGλ; Lane 4, anti-IL6RC IgGκ; Lane 5, anti-IL6RC/anti-IL6RC IgGκλ; Lane 6, anti-IL6RC IgGλ.

[0047] Figure 19 is a schematic representation of three different hybrid proteins obtained by combining the variable Lambda gene and the constant Kappa gene. The crosslink points differ between different hybrids: at 19A, VLambda crosslinked with CKappa; on 19B, VLambda up to CDR3 crosslinked with VKappa FR4 and CKappa; and at 19C, VLambda and the first four amino acids of CLambda and CKappa excluding the first four amino acids. CDR, hypervariable region; FR, wireframe.

[0048] Figure 20 is an illustration of the analysis of two light chain constructs on the Bionalyzer 2100 system using the Protein 80 chip (Agilent Technologies). The electropherogram corresponding to the gel image is shown.

[0049] Figure 21 is a series of graphs depicting the dose effect results of an ELISA using scFv specific for IFNγ (A) or IL6RC (B) in which the VH domain was either the total initially selected VH (upper curves) or it was other VH domains that allow scFv expression and purification (lower curves).

[0050] Figure 22 is a graph depicting the results obtained for IgGκλ, a bispecific antibody quantification using a sandwich ELISA format. Dose effects were achieved using either purified bispecific antibody alone or mixed with Kappa and Lambda monospecific antibodies in the various ratios indicated to evaluate these molecules in the assay.

Detailed description

[0051] In order to overcome the limitations of therapeutic agents containing monoclonal and monovalent antibodies, which can only target one antigen, or to overcome the limitations of combinations of therapeutic agents containing monovalent antibodies, intensive efforts have been made to obtain formats of bispecific antibodies with targeting multiple antigens. Such antibodies carrying more than one specificity are of interest from a biotechnological point of view and have great potential as therapeutic agents enabling novel therapeutic approaches (Fischer and Léger, Pathobiology 2007; 74:3-14; Morrison SL Nature Biotechnol 2007; 25:1233-1234). Bispecific antibodies are preferred because they allow for multiple targeting, they increase therapeutic potential, they are relevant to the creation of alternative capabilities in biological systems, and they provide new mechanisms of action through capabilities such as repeated targeting and/or increased specificity. It has been found that single therapeutic targets are becoming more and more exhausted, so combinations provided by bispecific antibodies provide new and broad arrays of targets for therapeutic agents and applications.

[0052] Several strategies have been used to generate such bispecific molecules, such as chemical crosslinking of antibody fragments, forced heterodimerization, quadroma technology, crosslinking of antibody fragments via polypeptide linkers, and the use of single domain antibodies. The availability of recombinant DNA technologies results in the production of a variety of antibodies in a bispecific format (see, for example, Ridgway JB et al. (1996) Protein Eng 9: 617-621). Linkers and mutations are often introduced into different regions of an antibody to enhance heterodimer formation or to join different binding components into a single molecule. However, these engineered molecules often have poor performance characteristics, as well as an increased risk of immunogenicity, which limit or prevent their progress towards clinical applications. In addition, previous attempts to develop bispecific formats have been limited due to factors such as poor stability and/or expression. These methods are further described below and the formats discussed are illustrated in Figure 1.

[0053] Chemical Crosslinking . The use of chemical crosslinking reagents to covalently link two antibodies is a substantially efficient method. Antibody fragments derived from their respective antibodies by enzymatic hydrolysis or obtained by recombinant techniques are conjugated using bifunctional reagents (Glennie MJ et al., J Exp Med 1992; 175:217-225). Product homogeneity is a major limitation of this method because bispecific samples must be purified from homodimers and modification steps can alter protein integrity and stability. The many steps involved make the process difficult in terms of production and product homogeneity.

[0054] Quadromes . Quadromas and triomas can be obtained by fusing either two hybridomas or one hybridoma with a B lymphocyte, respectively (Suresh MR et al., Methods Enzymol 1986; 121: 210-228). In this case, the simultaneous expression of two heavy and two light chains results in a random assembly of 10 combinations of antibodies, and the target bsAb represents only a small fraction of the secreted antibodies. bsAb should be purified using a combination of chromatographic methods and this significantly reduces the yield of the product. The main limitation is that quadromas produce rodent-derived bsAbs, which limits their therapeutic potential due to immunogenicity issues.

[0055] Recombinant bispecific antibodies . Most of the bispecific format antibodies were generated by genetic engineering using antibody fragments, such as scFv or Fab fragments, as building blocks connected via polypeptide linkers. Formats based on linked antibody fragments include tandem scFv (BiTE), diabody, and tandem diabody (Kipriyanov SM. Methods Mol Biol 2003; 207:323-333; Korn T et al., Int J Cancer 2002; 100:690-697) . These building blocks can further bind to the Fc region of an immunoglobulin to produce 'IgG-like' molecules. These formats include diatelo-Fc, tandem diatelo-Fc, tandem diatelo-CH3, (scFv) ₄ -Fc and DVD-Ig (Lu D et al., J Immunol Methods 2003; 279: 219-232; Lu D et al. , J Biol Chem 2005; 280: 19665-19672; Lu D et al., J Biol Chem 2004; 279: 2856-2865; Wu C et al., Nat Biotechnol 2007 25:1290-7). A potential limitation of the use of linkers is that the flexible nature of these peptides makes them more susceptible to proteolytic cleavage, potentially leading to poor antibody stability, aggregation, and increased immunogenicity. In addition, these foreign peptides can elicit an immune response against the junction between the protein and the linker, or against the linker itself. In general, bsAbs based on linked building blocks are labor intensive in terms of production, which limits their therapeutic applications.

[0056] An ideal bispecific molecule for human therapy would be indistinguishable from normal IgG. Studied strategies based on forced heterodimerization of two heavy chains. The first 'knob into hole' method developed aims to force the pairing of two different IgG heavy chains by introducing mutations in the CH3 domains to modify the contact surface (Ridgway JB et al., Protein Eng 1996; 9: 617-621). Amino acids with large side chains were added to one chain to create a 'knob'. On the other hand, large amino acids were substituted for amino acids with short side chains to create a 'hole' in another CH3 domain. By co-expression of these two heavy chains, more than 90% heterodimer formation ('knob-hole') was observed compared to homodimer formation ('hole-hole' or 'knob-knob'). A similar concept has been developed using human CH3 domains derived from chain exchange domain formation (SEED) based on human IgG and IgA sequences (Davis JH et al., 2010, PEDS 23:195-202). These designed domains lead to the formation of heterodimeric molecules that can carry two different specificities. These two methods are attractive because they favor the production of heterodimers of interest (up to 95%), but do not completely prevent the formation of homodimers. Thus, there is still a need for a post-purification procedure capable of removing homodimers from heterodimers. Another potential property of these methods is that mutated domains are not fully human and may result in immunogenicity, and may also affect domain stability and aggregation propensity of the molecule. Because these strategies allow for forced pairing of heavy chains, different light chains can randomly pair with any of the heavy chains and result in different antibodies that need to be purified from each other. An improvement regarding the 'knob into hole' method for solving the problem of light chain pairing has recently been described (WO 2009/080253 A1). The method involves exchanging some of the light chain and heavy chain domains in addition to 'knob into hole' mutations. The main advantage of this method is that a bispecific bivalent antibody having two different heavy chain variable domains and two different light chain variable domains can be generated and is referred to as "CrossMab." However, the sequences of this bispecific antibody are not fully human because it contains both mutations in Fc to force heterodimerization and artificial junction points between different immunoglobulin domains. In addition, these modifications result in a reduced level of expression of the bispecific format antibodies compared to the monoconal antibody (Schaefer et al., PNAS 2011; 108:11187-11192).

[0057] Antibodies based on a single domain . Camelid (llama and camel) and cartilaginous fish (nurse shark) immune systems use single V-domains fused to Fc, demonstrating that a single domain can confer high binding affinity to an antigen. Cam, shark, and even human V domains are alternatives to antibodies, but they can also be used to generate bsAbs. They can be reformatted into classic IgGs, in which each arm has the potential to bind to two targets either through the VH domain or through the VL domain. This single domain IgG will have similar biochemical properties to IgG and potentially solves the problems associated with other bsAb formats in terms of production and heterogeneity. However, it is likely that steric hindrance will often prevent simultaneous binding of both antigens on both arms of an antibody.

[0058] Characteristics of the formats of the bispecific antibodies described above are presented in Figure 1. Some of these characteristics of the formats are taken from Fischer and Léger, Pathobiology 2007; 74:3-14; and Morrison S.L. Nature Biotechnol 2007; 25:1233-1234.

[0059] In contrast to these previous formats, the bispecific antibodies, multispecific antibodies, compositions, and methods provided herein overcome such design hurdles. The bispecific antibodies provided herein have a common heavy chain, two light chains - one Kappa (k), one Lambda (λ) - each having a different specificity (ie two light chains, two specificities). Preferably, the bispecific antibodies do not contain any linkers or other modifications, including amino acid mutations. Using the methods presented in this document, get molecules with binding specificity, where the diversity is limited to the VL region. These methods generate bispecific antibodies by controlled co-expression of three chains (one VH chain, two VL chains), and by purifying the bispecific antibody. The bispecific and/or multispecific antibodies described herein exhibit similar affinities for a given target compared to those of monospecific antibodies for the same target. Preferably, the bispecific and/or multispecific antibodies described herein are virtually indistinguishable from standard IgG molecules.

[0060] The methods provided herein also provide a means of making simple mixtures of antibodies from two monospecific antibodies and one bispecific antibody, which are used, for example, for multiple targeting without purifying the bispecific antibody from the mixture.

[0061] Possible modes of action of bispecific antibodies

[0062] Simultaneous inhibition of two targets . By definition, bispecific antibodies have two specificities and thus can inhibit more than one target. These targets may be soluble factors or localized on the cell surface. A number of multiple cytokine targeting formats have been successfully prepared (Wu C et al., Nat Biotechnol 2007 25:1290-7).

[0063] Redirected impact . Because most bispecific antibody formats are capable of binding to two molecules at the same time, they can thus be used as linking molecules to target cytotoxic effector cells or cytotoxic agents to cells involved in disease. This application has been studied in relation to oncology. In some cases, one specificity of the antibody is directed against tumor cell markers such as CD 19, CD20, HER2, carcinoembryonic antigen (CEA). The second arm of an antibody carries in close proximity a toxic component or activity, such as drugs, toxins, cytokines, or an effector cell of the immune system (T cells, NK cells, monocytes, and neutrophils targeting CD3, CD16, CD64, and CD89 , respectively) (Thielemans K, Blood 1996; 87: 4390-4398; Goldstein J et al., J Immunol 1997; 158: 872-879).

[0064] Increased specificity through avidity . In the classical IgG format, antibody binding is directed both by the affinity of each fusion site for its antigen and by the avidity effects afforded by bivalent binding. The avidity effect significantly increases the apparent affinity of the antibody for cell surface markers, since two dissociation events occur for the released antibody. Some of the bispecific formats described above are bivalent (ie, one binding site per target), while others are tetravalent. The latter have four binding sites or more and have the potential to bind each target in a bivalent manner. Divalent bispecific antibodies are used as therapeutic agents that selectively target cell populations that express a combination of cell surface markers. This unique mode of action is generally limited to molecules that can benefit from the avidity component to distinguish between cells expressing both antigens and those expressing only one marker.

[0065] Characteristics of possible modes of action mediated by bispecific antibodies are shown in Figure 2. This characterization was taken from Fischer and Léger, Pathobiology 2007; 74:3-14.

[0066] Characteristics and limitations of bispecific antibody formats

[0067] The key characteristics of these bispecific antibody formats are summarized in Table I. All formats except those based on human domains contain sequences that are not of human origin or contain non-human protein sequences obtained by fusion of different protein domains . Most formats using linkers lead to potential production problems due to domain aggregation. The presence of foreign sequences and unfavorable stability characteristics can potentially significantly increase the risk of immunogenicity. The key difference between the formats is the valency of their binding sites, which is directly related to the ability to mediate redirection or avidity-mediated selective binding. Thus, all formats cannot be capable of all modes of action. Specifically, a single format that is indistinguishable from a fully human immunoglobulin cannot mediate redirection or increased selectivity activities. Thus, there is a need to obtain new bispecific antibodies with favorable properties for the development of therapeutic agents, i.e. those that are indistinguishable from a fully human immunoglobulin molecule, with good manufacturing properties and with the possibility of a wide range of possible modes of action.

[0068] Improved methods for producing bispecific and bivalent antibodies.

[0069] The present invention provides methods for producing bispecific antibodies that are structurally identical to human immunoglobulin. This type of molecule consists of two copies of a unique heavy chain polypeptide, the variable region of the first light chain is fused to the constant Kappa domain, and the variable region of the second light chain is fused to the constant Lambda domain. Each fusion site exhibits a different antigenic specificity, facilitated by both the heavy chain and the light chain. Light chain variable regions may be of the Lambda or Kappa family and are preferably linked to Lambda and Kappa constant domains, respectively. This is preferred in order to avoid the production of unnatural polypeptide compounds. However, it is also possible to obtain the bispecific antibodies of the invention by fusing the Kappa light chain variable domain to the Lambda constant domain for first specificity, and also to fuse the Lambda light chain variable domain to the Kappa constant domain for the second specificity (Figure 3). The bispecific antibodies described herein are also referred to as IgGκλ antibodies or "κλ antibodies", new full length human IgG format bispecific antibodies. This κλ antibody format allows affinity purification of a bispecific antibody that is indistinguishable from a standard IgG molecule with characteristics that are indistinguishable from a standard monoclonal antibody and thus is preferred over previous formats.

[0070] An essential step of the method is the identification of two Fv regions of an antibody (each consisting of a light chain variable region and a heavy chain variable domain) having different antigenic specificities that share a heavy chain variable domain. Many methods have been described for the production of monoclonal antibodies and fragments thereof. (See, for example, Antibodies: A Laboratory Manual, Harlow E, and Lane D, 1988, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, publication incorporated herein by reference). Fully human antibodies and antibody molecules in which the sequence of both light and heavy chains, including CDR 1 and 2, are derived from human genes. The CDR3 region may be of human origin or may be engineered synthetically. Such antibodies are referred to herein as "human antibodies" or "fully human antibodies". Human monoclonal antibodies can be obtained by using the trioma method (trioma); human B cell hybridoma method (see Kozbor, et al., 1983 Immunol Today 4: 72); and the EBV hybridoma method to produce human monoclonal antibodies (see Cole, et al., 1985 In: Monoclonal Antibodies and Cancer Therapy, Alan R. Liss, Inc., pp. 77-96). Human monoclonal antibodies can be used and can be obtained using human hybridomas (C. Cote, et al., 1983. Proc Natl Acad Sei USA 80: 2026-2030) or by transforming human B cells with Epstein Barr virus in vitro (see Cole, et al., 1985 In: monoclonal Antibodies and Cancer Therapy, Alan R. Liss, Inc., pp. 77-96).

[0071] Monoclonal antibodies are obtained, for example, by immunizing an animal with a target antigen or an immunogenic fragment, derivative, or variant thereof. Alternatively, the animal is immunized with cells transfected with a vector containing a nucleic acid molecule encoding the target antigen such that it is expressed and associated with the surface of the transfected cell. A variety of methods are well known in the art for producing xenogeneic non-human animals. See, for example, US Patent Nos. 6,075,181 and 6,150,584, which are incorporated herein by reference in their entirety.

[0072] Alternatively, antibodies are generated by screening a library that contains antibody sequences or antigen-binding domains for binding to a target antigen. This library is obtained, for example, in bacteriophage as protein or peptide fusions with the bacteriophage coat protein, which is expressed on the surface of the assembled phage particles, and which are encoded by DNA sequences contained in the phage particles (i.e. "phage display library").

[0073] Myeloma/B cell fusion hybridomas are then screened for their reactivity with the target antigen. Monoclonal antibodies are prepared, for example, using hybridoma techniques such as those described in Kohler and Milstein, Nature, 256:495 (1975). In the hybridoma method, a mouse, hamster, or other suitable host animal is typically immunized with an immunizing agent in order to stimulate lymphocytes that produce or are capable of producing antibodies that will specifically bind to the immunizing agent. Alternatively, lymphocytes can be immunized in vitro .

[0074] Although not strictly impossible, it is unlikely to identify different antibodies having the same heavy chain variable domains but directed against different antigens. Indeed, in most cases, the heavy chain contributes more to the antigen-binding surface and is also the most variable in sequence. Specifically, the heavy chain CDR3 is the most distinct CDR in sequence, length and structure. Thus, two antibodies specific for different antigens will almost invariably carry different heavy chain variable domains.

[0075] The method of the invention overcomes this limitation and greatly facilitates the isolation of antibodies having the same heavy chain variable domain by using antibody libraries in which the heavy chain variable domain is the same for all components of the library and thus the diversity is limited to light chain variable domains. . Such libraries are described, for example, in the attached application PCT/US2010/035619, dated May 20, 2010 and published on November 25, 2010 as PCT Publication No. WO 2010/135558, as well as in the attached application PCT/US2010/057780, dated November 23, 2010, each of which is incorporated herein by reference in its entirety. However, since the light chain variable domain is expressed together with the heavy chain variable domain, both domains can contribute to antigen binding. To further facilitate the process, antibody libraries containing the same heavy chain variable domain and differing in the lambda light chain variable domain or the kappa light chain variable domain can be used in parallel for in vitro selection of antibodies against different antigens. This method makes it possible to identify two antibodies having a common heavy chain, but one of the antibodies carries a lambda light chain variable domain and the other a kappa light chain variable domain, which can be used as building blocks to obtain a bispecific antibody in the full length immunoglobulin format of the invention. The bispecific antibodies of the invention can be of different isotypes and their Fc region can be modified to change their binding properties to different Fc receptors, and thus modify the antibody's effector functions as well as its pharmacokinetic properties. Many methods for modifying the Fc region have been described and applied to the antibodies of the invention (see, for example, Strohl, WR Curr Opin Biotechnol 2009(6):685-91; US Pat. . The methods of the invention can also be used to generate bispecific antibodies and mixtures of antibodies in F(ab')2 format that lack an Fc region.

[0076] Another key step of the invention is the optimization of the co-expression of a common heavy chain and two different light chains in the same cell to allow the assembly of a bispecific antibody of the invention. If all polypeptides are expressed at the same level and assemble to form an immunoglobulin molecule in the same way, then the ratio of monospecific antibodies (same light chains) to bispecific antibodies (two different light chains) should be 50%. However, it is likely that different light chains are expressed at different levels and/or are not assembled with the same efficiency. Thus, the methods of the invention also provide a means of modulating the relative expression of different polypeptides to compensate for their intrinsic expression characteristics or different propensity to assemble using a common heavy chain. This modulation can be achieved through the strength of the promoter, the use of internal ribosome entry sites (IRES) that characterize different efficiencies, or through other types of regulatory elements that can act at the transcriptional or translational level, as well as acting on the stability of the mRNA. Various promoters of varying strength may include CMV (cytomegalovirus immediate early promoter); EF1-1α (human elongation factor 1α subunit promoter); Ubc (human ubiquitin C promoter); SV40 (simian virus 40 promoter). Various mammalian and viral IRES have also been described. (See, for example, Hellen CU and Sarnow P. Genes Dev 2001 15 : 1593-612). These IRESs can largely differ in their length and ribosome recruitment efficiency. In addition, further regulation of activity is possible by introducing multiple copies of the IRES (Stephen et al. 2000 Proc Natl Acad Sei USA 97: 1536-1541). Modulation of expression can also be achieved by multiple sequential transfections of cells to increase the copy number of individual genes expressing one or the other light chain and thus modifying their relative expression. The examples provided herein demonstrate that controlling the relative expression of different chains is critical to maximizing the assembly and overall yield of the bispecific antibody.

[0077] Co-expression of the heavy chain and two light chains produces a mixture of three different antibodies in the cell culture supernatant: two monospecific divalent antibodies and one bispecific divalent antibody. The latter should be purified from the mixture to obtain the molecule of interest. The method described herein greatly facilitates this purification procedure by using an affinity chromatographic material that specifically interacts with Kappa or Lambda light chain constant domains, such as CaptureSelect Fab Kappa and CaptureSelect Fab Lambda affinity matrices (BAC BV, Holland). This method of multi-stage affinity chromatographic purification is efficient and generally applicable to the antibodies of the invention. This is in contrast in form to the specific purification methods that have been developed and optimized for each bispecific antibody derived from quadromas or other cell lines expressing mixtures of antibodies. Indeed, if the biochemical characteristics of different antibodies in a mixture are similar, then their separation using a standard chromatographic method, such as ion exchange chromatography, is difficult or impossible.

[0078] The invention also relates to novel means of making simple mixtures of antibodies from two or more monospecific antibodies and one or more bispecific antibodies that have the same heavy chain and can be purified using standard chromatographic techniques used to purify monoclonal antibodies. (See, for example, Lowy, I et al. N Engl J Med 2010; 362:197-205; Goudsmit, J. et al. J Infect Dis. 2006. 193, 796-801). Such simple mixtures can be used as multi-targeting agents for therapeutic applications.

[0079] Successful co-expression, purification and characterization of the heavy chain and two light chains and purification of bispecific antibodies are presented in the Examples. Genes encoding a common heavy chain and two light chains were cloned into a vector containing three promoters. After transient transfection, the supernatant of PEAK cells was collected.

[0080] Co-expression of the three chains resulted in the assembly of three different antibodies: two monospecific and one bispecific antibody. Their theoretical relative ratios should be 1:1:2, provided that the expression level and assembly rate are the same for both light chains. Bispecific antibodies were purified using a three-step affinity chromatography procedure: (1) Protein A: IgG capture (mono- and bi-), (2) Kappaselect: IgG capture containing Kappa light chains, and (3) Lambdaselect: IgG capture containing light lambda chain. Kappaselect and Lambdaselect are affinity chromatography materials developed by BAC, BV and GE Healthcare.

[0081] Purified bispecific antibodies were characterized as described below. The flow and elute fractions from each affinity purification step were analyzed by SDS-PAGE. The results revealed that at each stage, the bispecific antibodies are enriched (Figure 8). The κλ antibody contained equivalent amounts of Kappa and Lambda light chains. The κλ antibody showed an intermediate migration pattern on the isoelectric focusing gel and in ion exchange chromatography compared to monospecific antibodies (Figure 10). Specificity and affinity of κλ antibodies were determined using ELISA and surface plasmon resonance. The methods of the invention allow the identification of antibodies with affinities ranging from subnanomolar to nanomolar without optimization. This is not obvious as the diversity of antibody libraries described herein is limited to the light chain, which contributes less to binding energy in standard antibodies.

[0082] To avoid the requirement of having access to two antibodies having Kappa-type and Lambda-type light chain variable domains, which seemed to be a limitation of the present invention, the methods described herein allow the production of a hybrid light chain in which the Lambda variable domain can be fused to a constant Kappa domain and, conversely, a variable Kappa domain can be fused to a constant Lambda domain, as described in Figure 3. This extends the application of the invention to antibody pairs that share the same type of light chain. As described in the Examples provided herein, the point of crosslinking between the variable and constant domains is important and can affect the purification process of the bispecific antibody.

[0083] An overview of one method for producing bispecific and/or multispecific antibodies of the invention is provided in Figure 13. In some embodiments, methods for producing bispecific and/or multispecific antibodies use a completely serum-free chemically driven process. These methods include the most widely used cell line in the pharmaceutical industry, Chinese Hamster Ovary (CHO) cells. The methods described in this document are used to obtain both semi-stable and stable cell lines. The methods can be used for laboratory production of the bispecific and/or multispecific antibodies of the invention (eg, in an Erlenmeyer flask) and at medium scale (eg, in a 25 L "Wave bag"). The methods are also readily adaptable for larger scale production of bispecific and/or multispecific antibodies and antibody mixtures of the invention.

[0084] Methods for producing bispecific and/or multispecific antibodies of the invention are preferred because they employ versatile purification processes as shown in Figure 8A. Figure 16 shows the purification and integrity testing of a bispecific antibody product purified from a semi-stable cell line. Bispecific antibodies were purified using a three-step affinity chromatography procedure: (i) Protein A purification to capture IgG molecules, including both monospecific and bispecific antibodies; (ii) KappaSelect purification to capture IgGs containing Kappa light chains; (iii) LambdaSelect purification to capture IgG containing Lambda light chains. The flow and elute fractions from each affinity purification step were analyzed by SDS-PAGE. The results demonstrated the removal of each monospecific form (ie monospecific IgG molecules having Kappa light chains and monospecific IgG molecules having Lambda light chains) during the purification process (Figure 16A). Purified κλ-containing antibodies (ie, antibodies having both Kappa and Lambda light chains) contained an equivalent amount of Kappa and Lambda light chains (Figure 16B). Purified κλ-containing antibodies were a sample of intermediate migration on gel with isoelectric focusing compared to two monospecific antibodies (Figure 16C).

[0085] Chemically driven processes for the production of bispecific and/or monospecific antibodies of the invention can be used with either CHO cell pools or established cell lines. Results obtained using chemically driven processes using either pooled or engineered cell lines show productivity and growth characteristics comparable to cells expressing the appropriate Kappa or Lambda containing monospecific antibodies. Thus, the κλ antibody retains both the structure and performance characteristics of classical human IgGs.

[0086] Previous methods for producing a bispecific antibody format, aimed at forcing the production of a homogeneous bispecific molecule using the various design methods described above, were carried out at the expense of productivity, scalability and product stability. The present invention is an excellent method that allows the preparation of a simple mixture of antibodies that have the standard performance and scalability characteristics of monoclonal antibodies and provides a versatile means of purifying a bispecific antibody from a mixture or purifying a mixture of antibodies.

EXAMPLES

EXAMPLE 1 Preparation of Antibody Libraries Having Fixed Heavy Chains

[0087] Antibody libraries in which the heavy chain variable domain is identical for all library components were generated as described below. First, a heavy chain VH3-23 variable domain containing a defined sequence of CDR3 AKSYGAFDY (SEQ ID NO: 1) (CDR nomenclature according to IMGT) and a defined sequence of FR4 was cloned into a pNDS vector using SfiI and XhoI restriction sites to obtain a phagemid vector pNDS_VH fixed. The amino acid sequence of VK FR4 corresponds to the FR4 region encoded by the germline J genes of JK1. The amino acid sequence Vλ, FR4 corresponds to the FR4 region encoded by the germline J genes JL2 and JL3. Two variants of the FR4 Vk sequence were generated with a single amino acid change at position 106 (Arginine or Glycine). In total, 6 genes of the variable domains Kappa (VK1-33, VK1-39, VK3-11, VK3-15, VK3-20, VK4-1) and 5 Lambda genes (Vλ1-44, Vλ1-51, Vλ6-57, Vλ2- 14, Vλ1-40) containing an additional fragment in place of the CDR3 coding sequence into the pNDS_VHfixed vector in order to obtain 17 acceptor vectors into which many genes of synthetic or natural origin could be cloned and libraries with a high degree of diversity could be obtained according to the methods described in the attached application PCT/US2010/035619, dated May 20, 2010, published as W02010/135558, and according to the methods described in Ravn et al. (2010) Nucl Acid Res 38(21):el93, each of which is incorporated herein by reference in its entirety. Process resulting in 17 libraries containing a common VH3-23 domain and different VKappa or VLambda domains located in the light chain hypervariable region 3 (CDRL3 region), containing a total of 6.9×10 ⁹ variants (Figure 5A) . Sequencing of 180 randomly selected transformants revealed that >90% of the clones integrated CDRL3 sequences that were in-frame and thus potentially functional.

EXAMPLE 2 Recovery of Phage Libraries

[0088] Each library was recovered independently according to standard phage display procedures, which are summarized below. A volume of cells and frozen aliquots of the library sufficient to cover at least 10 times the theoretical diversity of the library was added to 500 ml of 2xTYAG (100 μg/ml ampicillin; 2% glucose) and expanded at 37° C. with shaking (240 vol/ min) until an OD600 of 0.3-0.5 is reached. The culture was then superinfected with helper phage MK13KO7 for one hour at 37°C (150 rpm). The medium was then changed by centrifuging the cells at 2000 rpm for 10 minutes, the medium was removed and the pellet was resuspended in 500 ml 2xTY-AK (100 μg/ml ampicillin; 50 μg/ml kanamycin). The culture was then grown overnight at 30° C. (240 rpm). The culture was centrifuged at 4000 rpm for 20 minutes to pellet the cells. The supernatant was collected and 30% (v/v) PEG 8000 (20%)/2.5 M NaCl was added to precipitate the phage particles by incubating the mixture for 1 hour on ice. Phage particles were collected by centrifugation at 10,000 rpm for 30 minutes and resuspended in 10 ml TE buffer (10 mM tris-HCl pH 8.0; 1 mM EDTA). The resuspended solution was centrifuged at 10,000 rpm to remove bacterial debris and the precipitation procedure was repeated. After the final resuspension, the phage was titrated by infecting with E. coli and measuring absorbance at 260 nm. The represented level of scFv on the phage surface was also assessed by Western blot analysis using an anti-c-myc monoclonal antibody. Purified phage from various libraries were kept frozen at -80° C. after adding glycerol to a final concentration of 15% (w/v).

EXAMPLE 3 Phage Display Selection Using Fixed Heavy Chain Libraries

[0089] Liquid phase selections against biotinylated hCXCL10-NusA fusion protein (hCXCL10-NusA) and biotinylated hIL6 receptor complex (hIL6RC) : Aliquots of VH-Vk and VH-Vλ phage libraries (10 ¹¹ -10 ¹² pfu) were maintained separately and blocked using PBS containing 3% (w/v) skimmed milk for one hour at room temperature on a rotary mixer. Blocked phage were then sampled onto streptavidin-coated magnetic microspheres (Dynal M-280) for one hour at room temperature on a rotary mixer. The selected phage was then incubated with in vivo biotinylated hCXCL10-NusA or hIL6RC (100 nM) for two hours at room temperature on a rotary mixer. The microspheres were added to the target and captured using a magnetic rack followed by four washes with PBS/0.1% Tween 20 and 3 washes with PBS. The microspheres were then added directly to 10 ml exponential culture of TG1 cells and incubated for one hour at 37° C. with slow shaking (100 rpm). An aliquot of infected TG1 was serially diluted to titrate selection results. The remaining infected TG1 were centrifuged at 3000 rpm for 15 minutes and resuspended in 0.5 ml 2xTYAG (2xTY medium containing 100 μg/ml ampicillin and 2% glucose) and plated on 2xTYAG Bioassay agar plates. After an overnight incubation at 30° C., 10 ml of 2xTYAG was added to the dishes and the cells were scraped from the surface and transferred to 50 ml polypropylene tubes. 2xTYAG containing 50% glycerol was added to the cell suspension to give a final glycerol concentration of 17%. Aliquots of the selection round were maintained at -80°C.

[0090] Phage Recovery : 100 µl of the cell suspension from previous selection rounds was added to 20 ml of 2xTYAG and grown at 37° C. with shaking (240 rpm) until an OD600 of 0.3-0.5 was reached. The culture was then superinfected with 3.3×10 ¹⁰ helper phage MK13KO7 and incubated for one hour at 37°C (150 rpm). The medium was then changed by centrifuging the cells at 3800 rpm for 10 minutes, the medium was removed and the pellet was resuspended in 20 ml 2xTY-AK (100 μg/ml ampicillin; 50 μg/ml kanamycin). The culture was then grown overnight at 30° C. (240 rpm). The next day, an aliquot of the centrifuged supernatant was used as the starting amount for the next round of selection.

[0091] Recovery of monoclonal phage for ELISA : Single clones were selected into a microplate containing 150 μl of 2xTYAG medium (2% glucose) per well and grown at 37°C (100-120 rpm) for 5-6 hours. Phage the M13KO7 helper was added to each well to give a MOI of 10 (i.e. 10 phages per cell in culture) and incubated at 37°C (100 rpm) for 1 hour. After growing, the plates were centrifuged at 3200 rpm for 10 min. The supernatant was carefully removed, the cells were resuspended in 150 μl of 2xTYAK medium and grown overnight at 30°C (120 rpm). For ELISA, phage were blocked by adding 150 μl of 2x concentration of PBS containing 5% skimmed milk powder, followed by incubation for one hour at room temperature. The plates were then centrifuged for 10 minutes at 3000 rpm and the supernatant containing phage was used for ELISA.

[0092] Phage ELISA : ELISA plates (Maxisorp, NUNC) were coated overnight with 2 μg/ml hCXCL10-NusA in PBS or 2 μg/ml hIL6RC in PBS. The plates were blocked with 3% skim milk/PBS at room temperature for 1 hour. The plates were washed 3 times with PBS 0.05% Tween 20 before transferring the pre-blocked phage containing supernatants and incubating for one hour at room temperature. Each clone was tested against both targets to demonstrate specificity. The plates were washed 3 times with PBS 0.05% Tween 20. 50 μl of 3% skimmed milk in PBS containing (HRP)-conjugated anti-M13 antibody (Amersham, diluted 1:10,000) was added to each well. After incubation at room temperature for 1 h, the plates were washed 5 times with PBS 0.05% Tween 20. Then ELISA was performed by adding 50 μl of TMB (Sigma) and 50 μl of 2N H ₂ SO ₄ to stop the reaction. The absorption intensity was read at 450 nm. Clones specific for hCXCL10-NusA or hIL6RC carrying the same heavy chain variable domain could be identified. (Figure 4).

[0093] Sequencing of Phage Clones : Single clones were grown in 5 ml of 2xTYAG medium (2% glucose) per well and at 37°C (120 rpm) overnight. The next day, the phagemid DNA was purified and used for DNA sequencing using the pNDS1 specific primer: mycseq, 5'-CTCTTCTGAGATGAGTTTTTG. (SEQ ID NO:2).

[0094] Large-scale purification of scFv : 1 ml of 2xTYAG stock culture was inoculated with a single colony from a fresh 2xTYAG agar plate and incubated with shaking (240 rpm) at 37°C for 5 hours. 0.9 ml of this culture was used to inoculate a 400 ml culture of the same medium and grown overnight at 30° C. with vigorous shaking (300 rpm). The next day, the culture was induced by adding 400 μl of 1M IPTG and continued incubation for an additional 3 hours. Cells were collected by centrifugation at 5000 rpm for 10 minutes at 4°C. The pelleted cells were resuspended in 10 ml ice-cold TES buffer supplemented with protease inhibitors as described above. Osmotic shock was achieved by adding 15 ml diluted 1:5 TES buffer and incubating for 1 hour on ice. The culture was centrifuged at 10,000 rpm for 20 minutes at 4°C to pellet the cells. The supernatant was carefully transferred to a fresh tube. Imidazole was added to the supernatant to a final concentration of 10 mM. 1 ml of Ni-NTA resin (Qiagen) equilibrated in PBS was added to each tube and incubated on a rotary mixer at 4°C (20 rpm) for 1 hour. The tubes were centrifuged at 2000 rpm for 5 minutes and the supernatant was carefully removed. The precipitated resin was resuspended in 10 ml of cold (4° C.) wash buffer 1 (50 mM NaH ₂ PO ₄ 300 mM NaCl, 10 mM imidazole, pH to 8). The suspension was added to a "polyprep" column (Biorad). 8 ml of cold wash buffer 2 (50 mM NaH ₂ PO ₄ , 300 mM NaCl, 20 mM imidazole, pH to 8) was used to wash the column by gravity. scFv was eluted from the column with 2 ml of elution buffer (50 mM NaH ₂ PO ₄ 300 mM NaCl, 250 mM imidazole, pH up to 8). Fractions were analyzed by measuring absorbance at 280 nm and protein containing fractions were collected on a NAP5 desalting column (Amersham) equilibrated with PBS before buffer exchange. scFv in PBS was analyzed by SDS-PAGE and quantified by absorbance at 280 nm. Purified scFvs were aliquoted and stored at -20°C and at 4°C. Purified scFvs were used in ELISA to confirm specific binding to the target against which they were selected.

EXAMPLE 4: Additional selections using different fixed VH libraries

[0095] Antibody libraries were also generated by capturing naturally rearranged light chain variants and cloning them against the VH domain described in Example 1. In this case, the entire light chain variable gene region was amplified from human cDNA using primers that correspond to 5' and 3' region of human rearranged variable regions, and cloned into the pNDS_VHfixed vector described in Example 1. Another set of libraries were prepared as described in Example 1, but using a fixed VH3-23 domain containing a different ARGDDVS CDRH3 sequence (SEQ ID NO :3). The libraries described above are shown schematically in Figure 5. These fixed VH libraries were used against a panel of target proteins using the selection and screening methodology described in Example 2 and 3. Selection was performed using the following targets: hCXCL10-NusA, IL-6RC , CD47, CD16, CD8 and hIFNγ. Candidates identified that have been shown to be specific to their targets are listed in Table II. These results demonstrate that antibodies can be made that bind to different targets and share a common heavy chain, and that diversity restricted to the light chain is sufficient to confer antigen specificity. Candidates can be generated from libraries containing the VH3-23 domain with different CDRH3 sequences or having VL variants that differ as a result of different strategies. Thus, the results demonstrate that the method is not limited to a particular VH sequence or a particular strategy for producing light chain variable domain differences.

Table II
A number of independent clones identified against a panel of targets using libraries containing a fixed VH sequence with two different CDRH3 sequences. NA: no selection was made. Targets: Total Fixed CDRH3 SEQ ID:1 Fixed CDRH3 SEQ ID:3 κ λ κ λ κ λ hCD16 5 one four - - one hCD8 5 12 2 ten 3 2 hCD47 21 9 16 four 5 5 IL6RC 17 fourteen eight 7 9 7 IFNγ - 5 - 5 - - NusA-CXCL10 2 four 2 four NA NA

EXAMPLE 5 Fixed VH Candidates Converting to IgG and Transient Expression in Mammalian Cells

[0096] After screening, scFv candidates were converted to IgG and expressed upon transient transfection of PEAK cells. Received several IgGλ (n=5) and IgGκ (n=9), having a common heavy chain, and specific to IFNγ and IL6RC, respectively, and having different light chain sequences, as indicated below. The VH and VL sequences of the selected scFvs were amplified with specific oligonucleotides and cloned into an expression vector containing heavy and light chain constant regions, and the resulting constructs were verified by sequencing. Expression vectors were transfected into mammalian cells using Fugene 6 transfection reagent (Roche, Basel, Switzerland). Briefly, Peak cells were cultured in 6-well plates at a concentration of 6×10 ⁵ cells per well in 2 ml culture medium containing fetal bovine serum. Expression vectors encoding candidate VH and VL sequences were co-transfected into cells using Fugene 6 transfection reagent according to the manufacturer's instructions. One day after transfection, the culture medium was removed and 3 ml of fresh serum-free medium was added to the cells and cultured for three days at 37°C. After a three day culture period, the supernatant was harvested to obtain IgG purified on protein G-Sepharose 4B fast flow columns (Sigma, St. Louis, MO) according to the manufacturer's instructions. Briefly, supernatants from transfected cells were incubated at 4° C. in the presence of ImmunoPure (G) IgG binding buffer (Pierce, Rockford, IL). Samples were then passed through Protein G-Sepharose 4B fast flow columns and then IgG purified using elution buffer. The eluted IgG fraction was then dialyzed against PBS and the IgG content was quantified by absorbance at 280 nm. The purity and integrity of the IgG was confirmed by SDS-PAGE. The IgG was then tested for binding to IFNγ to the IL-6/IL-6R receptor complex, referred to herein as IL6RC, by ELISA. Biotinylated IFNγ and IL6RC were immobilized on streptavidin-coated microplates (Streptawell, Roche) and the plates were then blocked with PBS supplemented with 2% BSA at room temperature for 1 hour. The plates were washed 3 times with PBS 0.05% Tween 20 before addition. purified anti-IFNγ IgGλ or anti-IL6RC IgGk to the wells coated with each target. After one hour of incubation at room temperature and washing of bound plates, antibodies were detected using horseradish peroxidase bound anti-human Ck or anti-human Cλ antibodies. An ELISA was then carried out by adding 50 μl of TMB (Sigma) and 50 μl of 2N H ₂ SO ₄ to stop the reaction. The absorption intensity was read at 450 nm. The results reveal that the binding specificity of scFvs isolated from fixed VH libraries was maintained in IgG format and demonstrated that two IgGs having the same heavy chain but different light chains can be specific for different targets (Figure 6).

[0097] The heavy and light chain amino acid sequences of anti-IFNγ IgGλ or anti-IL6RC IgGk are as follows:

Anti-IL6RC VKappa light chain

EIVLTQSPATLSLSPGERATLSCRASQSVSSYLAWYQQKPGQAPRLLIYDASNRATGIPARFSG SGSGTDFTLTISSLEPEDFAVYYCQQWLPTTPATFGQGTKVEIKRTVAAPSVFIFPPSDEQLKS GTASVVCLLNNFYPREAKVQWKVDNALQSGNSQESVTEQDSKDSTYQSLSSTLTLSKADYEKHKV YACEVTHQGLSS4

Anti-IFNγ VLambda light chain

NFMLTQPHSVSESPGKTVTISCTRSSGSIASNYVQWYQQRPGSSPTTVIYEDNQRPSGVPDRFS GSIDSSSNSASLTISGLKTEDEADYYCQSQSWDGNHIVFGGGTKLTVLGQPKAAPSVTLFPPSS EELQANKATLVCLISDFYPGAVTVAWKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWK SHRSYQVTHE5

General heavy chain

EVQLLESGGGLVQPGGSLRLSCAASGFTFSSYAMSWVRQAPGKGLEWVSAISGSGGSTYYADSV KGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAKSYGAFDYWGQGTLVTVSSASTKGPSVFPLA PSSKSTSGGTAALGCLVKDYFPEPVTVSWNSGALTSGVHTFPAVLQSSGLYSLSSVVTVPSSSL GTQTYICNVNHKPSNTKVDKRVEPKSCDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTP EVTCVVVDVSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKC KVSNKALPAPIEKTISKAKGQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNG QPENNYKTTPPVLDSDGSFFLYSKLTVDKSRWQQGNVFSCSVMHEALHNHYTQKSLSLSPG (SEQ ID NO: 6).

EXAMPLE 6: Co-expression of one heavy chain and two light chains in mammalian cells

[0098] Simultaneous expression of one heavy chain and two light chains in one cell can result in the assembly of three different antibodies (Figure 8A). Simultaneous expression can be achieved in various ways, such as by transfecting multiple vectors expressing one of the chains to co-express them, or by using vectors that direct the expression of multiple genes. The pNovi kHλ vector was designed to allow co-expression of one heavy chain, one Kappa light chain, and one Lambda light chain (Figure 7A). Expression of the three genes is driven by human cytomegalovirus (hCMV) promoters, and the vectors also contain the glutamine synthetase (GS) gene which allows selection and the establishment of stable cell lines. Another vector, pNovi kHλk, was also constructed in which a second copy of the Kappa light chain was inserted after the Lambda light chain and their expression was driven by an internal ribosome entry site (IRES). Using this bicistronic construct, it was possible to increase the relative expression of the Kappa light chain. The two vectors are schematically represented in Figure 7.

EXAMPLE 7: Transient co-expression of one heavy chain and two light chains in mammalian cells and purification of total IgG.

[0099] The VH and VL gene of anti-IFNγ IgGλ or anti-IL6RC IgGk was cloned into the pNovi kHλ vector described in Example 5 for transient expression in mammalian cells. Peak cells were cultured in 6-well plates at a concentration of 6×10 ⁵ cells per well in 2 ml culture medium containing fetal bovine serum. 2 μg of plasmid DNA was transfected into cells using the TransIT-LT1 transfection reagent (Mirus) according to the manufacturer's instructions. One day after transfection, the culture medium was removed and 3 ml of fresh serum-free medium was added to the cells and cultured for five days at 37°C. After a five day culture period, the supernatant was collected, centrifuged at 4000 rpm and applied to MabSelect Sure PA resin (GE Healthcare) according to the manufacturer's instructions. Total IgG was eluted under acidic conditions followed by neutralization and buffer exchange with PBS. The IgG content was quantified by absorbance at 280 nm and analyzed on SDS-PAGE. The presence of two bands corresponding to two light chains and one band corresponding to a heavy chain revealed that three chains could be expressed simultaneously in one cell (Figure 8B).

EXAMPLE 8 Purification of Lambda and Kappa Light Chain Bispecific Antibodies

[00100] Co-expression of a heavy chain and two light chains in the same cells can result in the assembly of three different antibodies: two monospecific antibodies carrying the same light chain on both arms, and a bispecific antibody carrying different light chains associated with heavy chain on each shoulder. The latter can be separated from monospecific antibodies by chromatographic techniques that take advantage of the different biochemical properties of monospecific and bispecific antibodies. To develop a general purification approach that is applicable to the production of all bispecific antibodies, we applied the affinity chromatography approach, schematically depicted in Figure 8A. Culture supernatants from transfected cells were applied to a MabSelect Sure PA resin column (GE Healthcare) and total IgG was purified as described in Example 7. Total IgG was then applied to a column containing CaptureSelect Fab Kappa affinity matrix (BAC BV, Holland) equilibrated with 10 volumes of PBS. The column was then washed with PBS, 5-10 column volumes. Immunoglobulin molecules bearing the Kappa light chain were eluted from the column by loading 5 column volumes of 0.1 M glycine pH 2.0 and the fractions were collected and neutralized. Fractions containing antibody were pooled on a PD10 desalting column (Amersham) equilibrated with PBS before buffer exchange. The antibody was then applied to a second column containing CaptureSelect Fab Lambda affinity matrix equilibrated with 10 volumes of PBS. The column was then washed with PBS, 5-10 column volumes. Immunoglobulin molecules carrying only the Kappa light chain were bound to the column and detected in the eluate. Antibodies carrying the Lambda light chain were eluted from the column by applying 5 column volumes of 0.1M glycine pH 3.0 and fractions were collected. Fractions containing the bispecific antibody were pooled on a PD10 desalting column (Amersham) equilibrated with PBS before buffer exchange. Immunoglobulin molecules carrying the Kappa light chain were eluted from the column by loading 5 column volumes of 0.1 M glycine pH 2.0 and the fractions were collected and neutralized. Fractions containing antibody were pooled on a PD10 desalting column (Amersham) equilibrated with PBS before buffer exchange. The antibody was then applied to a second column containing CaptureSelect Fab Lambda affinity matrix equilibrated with 10 volumes of PBS. The column was then washed with PBS, 5-10 column volumes. Immunoglobulin molecules carrying only the Kappa light chain were bound to the column and detected in the eluate. Antibodies carrying the Lambda light chain were eluted from the column by applying 5 column volumes of 0.1M glycine pH 3.0 and fractions were collected. Fractions containing the bispecific antibody were pooled on a PD10 desalting column (Amersham) equilibrated with PBS before buffer exchange. The flow-through fraction and eluted fraction of each purification step were analyzed by SDS-PAGE and it was shown that antibodies carrying Lambda light chains were found in the flow-through fraction of the CaptureSelect Fab Kappa affinity matrix and conversely that antibodies carrying Kappa light chains were detected in the flow-through fraction of the CaptureSelect Fab Lambda affinity matrix (Figure 8B). As expected, the intensity of the bands corresponding to the two light chains was equivalent in the final eluted fraction. Thus, this three-step method allows the purification of bispecific antibodies carrying Kappa and Lambda light chains. The final recovery of the bispecific antibody carrying the Kappa and Lambda light chains was approximately 10-21% in various experiments. The Protein A eluted fraction also revealed that more Lambda light chain than Kappa was assembled to purified IgG, suggesting that increased Kappa light chain expression enhances assembly and recovery of the bispecific antibody.

EXAMPLE 9 Modulation of Light Chain Expression to Optimize Bispecific Antibody Assembly

[00101] To increase expression of the Kappa light chain, the common VH gene, the anti-IFNγ antibody VLambda gene, and two copies of the anti-IL6RC VKappa were cloned into the pNovi kHλk vector described in Example 6. Expression of the second copy of the VKappa light chain is driven by IRES. Various IRES elements were tested to achieve different levels of VKappa light chain expression. This led to the creation of five independent vectors: pNovi kНλκ 1-5. These vectors were used for transient transfections into mammalian cells as described in Example 7 and total IgG was purified from the supernatant by protein A affinity purification. in IgG compared with expression using the vector pNovi kHλ, (Figure 9). Total IgG fractions obtained after protein A purification for each of these constructs were further purified in two consecutive steps with CaptureSelect Fab Kappa and CaptureSelect Fab Lambda as described in Example 8. The bispecific antibody yield was 20% for pNovi kHλ, and was ranging from 33% to 41% for pNovi kHλk constructs, indicating that increased Kappa light chain expression results in improved assembly of the bispecific antibody. To further confirm these results, co-transfections were performed with pNovi kHλ and a vector expressing monospecific anti-IL6RC IgGk. Thus, the relative levels of the Kappa light chain as well as the yield of the bispecific antibody were also increased, indicating that several approaches could be taken to correct for the relative light chain expression.

EXAMPLE 10 Description of Purified Bispecific Antibody

Purified bispecific antibodies, isolated as described in the Examples above, were described using various methods.

[00102] Isoelectric focusing (IEF) . Purified bispecific antibodies isolated as described in Example 9 were analyzed using "IsoGel Agarose IEF" plates pH 3-10 (Lonza) and compared with anti-IFNγ IgGλ and anti-IL6RC IgGk monospecific antibodies. After focusing, the gel was placed in a fixative solution for 30 min, washed with water for 5 min, and then dried at RT. The gel was stained with Coomassie for 15 minutes, briefly rinsed with water and decolorizing solution 2 x 15 minutes. Finally, the gel was dried at CT before imaging. The results shown in Figure 10A indicate that IgGkλ, a bispecific antibody, has a different and intermediate pI compared to the pI of the two monospecific antibodies, as predicted by the antibody format and theoretical calculations. Staining also demonstrated that the bispecific antibody is highly pure following the three step purification process described in Example 8.

[00103] Ion exchange chromatography (IEX) . 50 μg of purified monospecific and bispecific antibodies were analyzed by ion exchange high performance liquid chromatography (IEX-HPLC) (HPLC Waters e2695/detector 2489) on an Agilent Bio Mab, NP5, (Agilent) column: mobile phases were: A: Na ₂ HPO ₄ /NaH ₂ PO ₄ 10 mM, pH 6.5; B: Na ₂ HPO ₄ /NaH ₂ PO ₄ 10 mM, NaCl 100 mM, pH 6.5; at a flow of 0.8 ml/min and a 20%-60% gradient for 133 minutes. Three different antibodies have different retention periods with a peak corresponding to a bispecific antibody having an intermediate profile compared to monospecific antibodies (Figure 10B).

[00104] ELISA . Monospecific anti-IFNγ IgGλ, anti-IL6RC IgGk and bispecific IgGkλ antibodies were tested by ELISA for binding to IFNγ and IL6RC. The results shown in Figure 11 indicate that the IgGkλ bispecific antibody is able to bind both targets and can be detected by both human CK-binding secondary antibodies and human Cλ-binding secondary antibodies.

[00105] Surface Plasmon Resonance (SPR) . The binding affinity and kinetics of monospecific anti-IFNγ IgGλ, anti-IL6RC IgGk and bispecific IgGkλ antibodies were described on the Biacore 2000 instrument (Biacore AB, Uppsala, Sweden). 200RU goat anti-human IgG polyclonal antibodies (ahIgG; Biacore) were immobilized with the EDC/NHS chemistry on a CM5 Biacore chip. This surface was used to capture monospecific or bispecific human IgG. The surface was refreshed after each cycle with an injection of 10 mM glycine pH 1.5 at 30 μl/min, for 30 sec, followed by 1 min stabilization in HBS-EP buffer. Data were matched according to a 1:1 Langmuir model and K _on , K _off and K _D values were determined (Table III). Similar affinity values were obtained for monospecific antibodies and bispecific IgGkλ antibodies. The data indicate that the two antigen-binding active sites of the antibody bind hIFNγ and IL6RC in a similar manner in a monospecific and bispecific format.

[00106] The ability of bispecific IgGkλ to interact with both targets was simultaneously tested using SPR. Biotinylated IFNγ was immobilized on a streptavidin-coated CM5 Biacore chip. Monospecific and bispecific antibodies were injected onto this surface followed by injection of IL6RC. The sensorgram shown in Figure 11A demonstrates that the IgGkλ bispecific antibody was able to bind immobilized IFNγ and was able to simultaneously capture IL6RC. The SPR was also used to evaluate the relative amounts of Kappa and Lambda light chains in the purified bispecific antibody. IgGkλ was directly immobilized via amine coupling on the surface of a CM5 Biacore chip, and human CKappa-binding antibody was injected followed by human CLambda-binding antibody at the same concentration. Equivalent responses were obtained as anti-light chain antibodies, indicating that, as predicted by the format, equivalent amounts of Kappa and Lambda light chains are present in the IgGkλ bispecific antibody (Figure 12B).

Table III
Analysis of the kinetics of binding to IFNγ and IL6RC for monospecific and bispecific IgGkλ antibodies was measured on the Biacore 2000 system. Sample being analyzed ligand K _D (M) k _on (1/Ms) k _off (1/s) IFNγ IgGkλ 1.84E-10 9.19E+05 1.69E-04 IgGλλ 1.96E-10 6.08E+05 1.19E-04 IL6RC IgGkλ 2.72E-07 7.44E+03 2.02E-03 IgGkk 2.66E-07 8.08E+03 2.15E-03

EXAMPLE 11 Production of bispecific IgGkλ antibodies

[00107] Expression of IgGkλ, a bispecific antibody, has also been carried out in Chinese Hamster Ovary (CHO) cells, which are widely used for the production of monoclonal antibodies. In the example presented here, both semi-stable pools of transfected CHO cells and stable CHO cell lines were established to generate bispecific IgGkλ antibodies. In the study presented in this case, stable CHO lines were obtained and grown using a chemically defined animal-free manufacturing process (CDACF). The complete process is depicted in Figure 13.

[00108] CHO pools . CHO cells were electroporated with a linearized vector pNovi kHλk encoding IgGkλ, an anti-IFNγ/anti-IL6RC bispecific antibody described in the examples above, and plasmids expressing monospecific anti-IFNγ IgGλ and anti-IL6RC IgGk. After electroporation, pools of transfected cells were grown under 10 days of overgrowth conditions with no feeding. Cells transfected with a bispecific construct present similar growth profiles compared to cells transfected with vectors for the expression of monospecific antibodies. In addition, the productivity was also comparable and reached the typical range of antibody productivity: between 100-200 mg/l for crops, overgrown pools with no top dressing (Figure 14A). Scaling from a small-scale yield in an Erlenmeyer flask (100 ml) to a medium-scale yield in a 25 L Wave Bag was successfully performed (Figure 14B). These results indicate that similar growth and productivity curves are obtained during the expression of the bispecific IgGkλ antibody in the CHO cell line and the corresponding monospecific antibodies.

[00109] Stable CHO lines . Recombinant cell lines producing the bi-specific antibody were generated by electroporation of CHO cells with the pNovi cHkλ vector. After transfection, recombinant cell lines were selected by cell culture dilution in the presence of a final concentration of 50 μM methionine sulfoximine (MSX). After 6 weeks of incubation, colonies of the recombinant cell lines were screened for total IgG productivity using FastELISA® (R&D Biotech) (Figure 15A-B). Selected cell lines were grown in cell culture medium containing 25 μM MSX, transferred to 24-well microplates, and screened for productivity and growth habit in suspension culture (Figure 15C-D). The results revealed that IgGkλ, a bispecific antibody, can be produced under stirred batch conditions of growth at a level comparable to a cell line expressing standard monospecific antibodies. Cell lines with the highest yield were selected and processed in 50 ml batch overgrown cultures in shake flasks for a maximum of 10 days. Protein A HPLC was used to quantify total IgG in the supernatant. Total IgG from the supernatant of the 10 highest producing cell lines were purified by "MabSelect SuRE" chromatography using 1 ml HiTrap prefilled columns (GE Healthcare). The relative amounts of monospecific and bispecific antibodies in total purified IgG were assessed by IEX-HPLC as described in Example 10. the amount of IgGkλ (22 and 25%). The results of the 10 CHO cell lines are summarized in Table IV.

Table IV Total antibody titer (mg/ml) Total antibody count after MabSelect purification
SuRr (mg) Relative amount of bispecific antibody from IEX-HPLC (%) Total bispecific antibody (mg) Cell line 1 0.35 10.73 37 3.97 Cell line 2 0.32 9.54 25 2.37 Cell line 3 0.31 10.02 37 3.72 Cell line 4 0.42 10.01 40 4.00 Cell line 5 0.38 11.59 38 4.39 Cell line 6 0.43 11.10 43 4.75 Cell line 7 0.49 12.67 42 5.32 Cell line 8 0.33 8.96 22 2.01 Cell line 9 0.38 9.94 42 4.18 Cell line 10 0.39 10.98 42 4.61

[00110] Purification and characterization of an IgGkλ bispecific antibody expressed in CHO . The supernatant from a pool of CHO cells transfected with bispecific and monospecific constructs was used for purification. Monospecific anti-IFNγ IgGλ and anti-IL6RC IgGk were purified by protein A affinity chromatography and desalted in PBS, while bispecific IgGkλ antibodies were purified by the three-step affinity chromatography process described in Example 8. also the final purified samples were analyzed by SDS-PAGE and IEF (Figure 16). IgGkλ specificity was monitored after each purification step by ELISA (Figure 17). The results indicate that the purification process was robust and comparable to CHO expression and produced a high purity IgGkλ bispecific antibody.

EXAMPLE 12: Additional examples of bispecific IgGkλ antibodies

[00111] Other examples of bispecific IgGkλ antibodies were isolated, expressed and purified as described in the Examples above. These include anti-NusA/anti-IFNγ IgGkλ and anti-IL6RC/anti-IL6RC IgGkλ bispecific antibodies in which both combined sites bind IL6RC but carry a Kappa and Lambda light chain. These purified bispecific antibodies were analyzed by IEF along with their respective monospecific equivalents (Figure 18). The results suggest that the pI of a bispecific antibody is always midway between the pI of monospecific antibodies, but that these differences can vary significantly depending on the sequence of the light chain variable domain. This example illustrates that purification of a bispecific antibody based on charge differences can be difficult if two light chains have the same biochemical properties and highlights the advantage of the affinity purification approach of the invention.

EXAMPLE 13 Generation of IgGkλ Bispecific Antibodies with Two Kappa or Two Lambda Variable Domains

[00112] Bispecific antibodies can be generated using two variable domains of the same type (Lambda and Kappa). Because the affinity purification steps require the presence of the Kappa constant and Lambda constant light chain domains, any light chain variable domain can in principle be combined with these constant domains to produce hybrid light chains, as shown in Figure 3. This was demonstrated using two antibodies directed against IFNγ and IL6RC isolated from the fixed VH antibody libraries described in Example 1. In this case, both libraries share the same VH and have a lambda light chain variable domain. The VLamda domain of an anti-IFNγ antibody was fused to a Lambda constant domain, while the VLamda domain of an anti-IL6RC antibody was fused to a Kappa constant domain. In the case of the latter, three different constructs were obtained, including different join points between the VLamda domains and the CKappa domain. In one construct (hybrid 1), the fusion point was at the end of framework region 4 (FR4) of the VLambda domain and included the entire constant Kappa domain (Figure 19A). In another construct (hybrid 2), the FR4 lambda region was changed to the FR4 Kappa region (Figure 19B). In the third construct (hybrid 3), the first 4 amino acids of the Kapp constant domain were replaced with 4 amino acids of the Lambda constant domain (Figure 19C). These various Lambda-Kappa light chain hybrids were cloned together with the conventional heavy chain and the Lambda light chain into the pNovi kHλ vector. Mammalian cell transfection, protein expression and three steps of IgGkλ affinity purification were performed as described in the Examples above. Analysis of the eluted fractions indicates that the hybrid 3 light chain does not bind to the Kappa selective resin and therefore does not allow efficient purification of the bispecific antibody. Hybrid 1 and hybrid 2 enable efficient bispecific purification of IgGkλ. These purified IgGkλ bearing the hybrid light chain were analyzed on an Agilent Bionalyzer 2100 using a Protein 80 chip (Agilent Technologies) and the electropherogram peaks corresponding to the light chains were shown to be equivalent (Figure 20). The results suggest that the bispecific antibodies of the invention can be generated with two antibodies having a common heavy chain and two light chains having the same type of variable domains (Kappa or Lambda).

[00113]

EXAMPLE 14 VH and VL of the described antibodies are involved in antigen binding

[00114] To test the contribution of total VH to antigen binding, different scFv light chains bearing the total VH selected in Examples 1 and 4 were combined with two different VHs that differ from the original total VH. Various scFvs can be expressed and purified as described in Example 3 and tested for binding to the target against which they were generated. The examples presented in Figure 21 demonstrate that IFNγ-specific scFv and IL6RC-specific scFv can only bind to their respective targets when combined with the VH domain they were originally selected for, and no binding was observed when combined with the two other VH domains that allow their expression and purification. The results indicate that despite the fact that the variability is exclusive to the VL domain, both total VH and VL contribute to antigen binding.

EXAMPLE 15: Development of an ELISA for IgGkλ, quantitation with a bispecific antibody

[00115] IgGkλ ELISA Development, Bispecific Antibody Quantification Bispecific antibodies have been developed for sandwich ELISA to quantify IgGkλ. Maxisorp 96-well plates (Nunc) were coated with 10 μg/ml mouse anti-human Lambda antibody (Southern Biotech) and incubated at 4° C. overnight. After washing (PBS Tween 20 at 0.05%, 3 washes), the plate was blocked with PBS-BSA 3% (Sigma) for 2 hours at room temperature. The purified IgGkλ standard was serially diluted in PBS-BSA 1% from 500 ng/mL to 1 ng/mL to obtain a good linearity range for sample quantitation. Depending on their origin, the samples were diluted to fall within the quantitation range as follows: untreated supernatant CHO 1/1500; Protein A, purified 1/15000. The samples were then serially diluted 1:2. After washing the plates, 50 μl of each prepared dilution was added in duplicate and incubated for 1 hour at room temperature. The plates were washed again and incubated for 1 hour at room temperature with 50 μl of 1:2000 anti-human Kappa antibody (Southern Biotech). After the last wash (PBST 0.05%, 5 washes), the reaction was developed with 50 µl of TMB substrate (Sigma) and stopped after 15 min by the addition of 50 µl of H ₂ SO ₄ (Sodium hydroxide 2N). Absorbance at 450 nm was recorded using an accurate microplate reader (Epoch, Witec). Since monospecific antibodies can affect the assay by binding used to coat the capture antibody, experiments were performed with the addition of increasing amounts of monospecific IgGk and monospecific IgGλ anti-IgGkλ bispecific standard. Various ratios were tested: (50% IgGkλ bispecific, 25% monospecific IgGk, 25% monospecific IgGλ); (67% IgGkλ, bispecific; 33% monospecific IgGλ); (50% IgGkλ bispecific, 50% monospecific IgGλ); (25% IgGkλ, bispecific, 75% monospecific IgGλ); (50% IgGkλ bispecific, 50% monospecific IgGk).

[00116] The results presented in Figure 22 indicate that the assay is not significantly affected by monospecific antibodies and that it can therefore be used to quantify bispecific IgGkλ antibody in complex samples such as cell culture supernatants. ELISA was used to quantify the bispecific IgGkλ antibody in the supernatant from stable CHO cell lines, and then total IgG was purified from the same supernatants by protein A affinity chromatography. by HPLC with protein A or by absorbance at 280 nm and summarized in Table V. Concentrations obtained by ELISA corresponded to 30-40% of total IgG, the proportion of expected bispecific antibodies. The data suggest that this assay can be used to quantify IgGkλ, a bispecific antibody in cell culture supernatant, and it facilitates the screening of stable cell lines for their IgGkλ specific antibody productivity.

Table V
Quantification of IgGkλ bispecific antibody of various stable CHO cell lines by raw supernatants or total IgG by ELISA, and comparison with Protein A HPLC and A280 total IgG quantification results. cell line CHO supernatants Protein A purified total IgG Prot. A HPLC ELISA A280 nm ELISA #eleven 0.3 mg/ml 0.18 mg/ml 4.38 mg/ml 2.3 mg/ml #fourteen 0.42 mg/ml 0.234 mg/ml 4.13 mg/ml 3.3 mg/ml #twenty 0.38 mg/ml 0.18 mg/ml 4.48 mg/ml 1.7 mg/ml 10E4 0.39 mg/ml 0.19 mg/ml 4.81 mg/ml 3 mg/ml

Other embodiments

[00117] While the invention has been described in conjunction with a detailed description, the foregoing description is intended to be illustrative and not to limit the scope of the invention, which is defined by the scope of the appended claims. Other aspects, advantages and modifications are within the scope of the following claims.

--->

SEQUENCE LIST

<110> Fischer, Nicolas

Magistrelli, Giovanni

Gueneau, Frank

Ravn, Ulla

Elson, Greg

<120> METHODS FOR PRODUCING MULTI-SPECIFIC AND MULTIVALENT

ANTIBODIES

<130> 23135-423001US

<140> PCT/IB2011/002664

<141> 2011-08-16

<150> US 61/374159

<151> 2010-08-16

<150> US 61/443008

<151> 2011-02-15

<150> US 61/509260

<151> 2011-07-19

<160> 6

<170>PatentIn version 3.5

<210> 1

<211> 9

<212> PROTEIN

<213> Artificial sequence

<220>

<223> chemically synthesized

<400> 1

Ala Lys Ser Tyr Gly Ala Phe Asp Tyr

fifteen

<210> 2

<211> 21

<212> DNA

<213> Artificial sequence

<220>

<223> chemically synthesized

<400> 2

ctcttctgag atgagttttt g 21

<210> 3

<211> 7

<212> PROTEIN

<213> Artificial sequence

<220>

<223> chemically synthesized

<400> 3

Ala Arg Gly Asp Asp Val Ser

fifteen

<210> 4

<211> 215

<212> PROTEIN

<213> Artificial sequence

<220>

<223> chemically synthesized

<400> 4

Glu Ile Val Leu Thr Gln Ser Pro Ala Thr Leu Ser Leu Ser Pro Gly

1 5 10 15

Glu Arg Ala Thr Leu Ser Cys Arg Ala Ser Gln Ser Val Ser Ser Tyr

20 25 30

Leu Ala Trp Tyr Gln Gln Lys Pro Gly Gln Ala Pro Arg Leu Leu Ile

35 40 45

Tyr Asp Ala Ser Asn Arg Ala Thr Gly Ile Pro Ala Arg Phe Ser Gly

50 55 60

Ser Gly Ser Gly Thr Asp Phe Thr Leu Thr Ile Ser Ser Leu Glu Pro

65 70 75 80

Glu Asp Phe Ala Val Tyr Tyr Cys Gln Gln Trp Leu Pro Thr Thr Pro

85 90 95

Ala Thr Phe Gly Gln Gly Thr Lys Val Glu Ile Lys Arg Thr Val Ala

100 105 110

Ala Pro Ser Val Phe Ile Phe Pro Ser Asp Glu Gln Leu Lys Ser

115 120 125

Gly Thr Ala Ser Val Val Cys Leu Leu Asn Asn Phe Tyr Pro Arg Glu

130 135 140

Ala Lys Val Gln Trp Lys Val Asp Asn Ala Leu Gln Ser Gly Asn Ser

145 150 155 160

Gln Glu Ser Val Thr Glu Gln Asp Ser Lys Asp Ser Thr Tyr Ser Leu

165 170 175

Ser Ser Thr Leu Thr Leu Ser Lys Ala Asp Tyr Glu Lys His Lys Val

180 185 190

Tyr Ala Cys Glu Val Thr His Gln Gly Leu Ser Ser Pro Val Thr Lys

195 200 205

Ser Phe Asn Arg Gly Glu Cys

210 215

<210> 5

<211> 218

<212> PROTEIN

<213> Artificial sequence

<220>

<223> chemically synthesized

<400> 5

Asn Phe Met Leu Thr Gln Pro His Ser Val Ser Glu Ser Pro Gly Lys

1 5 10 15

Thr Val Thr Ile Ser Cys Thr Arg Ser Ser Gly Ser Ile Ala Ser Asn

20 25 30

Tyr Val Gln Trp Tyr Gln Gln Arg Pro Gly Ser Ser Pro Thr Thr Val

35 40 45

Ile Tyr Glu Asp Asn Gln Arg Pro Ser Gly Val Pro Asp Arg Phe Ser

50 55 60

Gly Ser Ile Asp Ser Ser Ser Asn Ser Ala Ser Leu Thr Ile Ser Gly

65 70 75 80

Leu Lys Thr Glu Asp Glu Ala Asp Tyr Tyr Cys Gln Ser Gln Ser Trp

85 90 95

Asp Gly Asn His Ile Val Phe Gly Gly Gly Thr Lys Leu Thr Val Leu

100 105 110

Gly Gln Pro Lys Ala Ala Pro Ser Val Thr Leu Phe Pro Pro Ser Ser

115 120 125

Glu Glu Leu Gln Ala Asn Lys Ala Thr Leu Val Cys Leu Ile Ser Asp

130 135 140

Phe Tyr Pro Gly Ala Val Thr Val Ala Trp Lys Ala Asp Ser Ser Pro

145 150 155 160

Val Lys Ala Gly Val Glu Thr Thr Thr Pro Ser Lys Gln Ser Asn Asn

165 170 175

Lys Tyr Ala Ala Ser Ser Tyr Leu Ser Leu Thr Pro Glu Gln Trp Lys

180 185 190

Ser His Arg Ser Tyr Ser Cys Gln Val Thr His Glu Gly Ser Thr Val

195 200 205

Glu Lys Thr Val Ala Pro Thr Glu Cys Ser

210 215

<210> 6

<211> 445

<212> PROTEIN

<213> Artificial sequence

<220>

<223> chemically synthesized

<400> 6

Glu Val Gln Leu Leu Glu Ser Gly Gly Gly Leu Val Gln Pro Gly Gly

1 5 10 15

Ser Leu Arg Leu Ser Cys Ala Ala Ser Gly Phe Thr Phe Ser Ser Tyr

20 25 30

Ala Met Ser Trp Val Arg Gln Ala Pro Gly Lys Gly Leu Glu Trp Val

35 40 45

Ser Ala Ile Ser Gly Ser Gly Gly Ser Thr Tyr Tyr Ala Asp Ser Val

50 55 60

Lys Gly Arg Phe Thr Ile Ser Arg Asp Asn Ser Lys Asn Thr Leu Tyr

65 70 75 80

Leu Gln Met Asn Ser Leu Arg Ala Glu Asp Thr Ala Val Tyr Tyr Cys

85 90 95

Ala Lys Ser Tyr Gly Ala Phe Asp Tyr Trp Gly Gln Gly Thr Leu Val

100 105 110

Thr Val Ser Ser Ala Ser Thr Lys Gly Pro Ser Val Phe Pro Leu Ala

115 120 125

Pro Ser Ser Lys Ser Thr Ser Gly Gly Thr Ala Ala Leu Gly Cys Leu

130 135 140

Val Lys Asp Tyr Phe Pro Glu Pro Val Thr Val Ser Trp Asn Ser Gly

145 150 155 160

Ala Leu Thr Ser Gly Val His Thr Phe Pro Ala Val Leu Gln Ser Ser

165 170 175

Gly Leu Tyr Ser Leu Ser Ser Val Val Thr Val Pro Ser Ser Ser Leu

180 185 190

Gly Thr Gln Thr Tyr Ile Cys Asn Val Asn His Lys Pro Ser Asn Thr

195 200 205

Lys Val Asp Lys Arg Val Glu Pro Lys Ser Cys Asp Lys Thr His Thr

210 215 220

Cys Pro Pro Cys Pro Ala Pro Glu Leu Leu Gly Gly Pro Ser Val Phe

225 230 235 240

Leu Phe Pro Pro Lys Pro Lys Asp Thr Leu Met Ile Ser Arg Thr Pro

245 250 255

Glu Val Thr Cys Val Val Val Asp Val Ser His Glu Asp Pro Glu Val

260 265 270

Lys Phe Asn Trp Tyr Val Asp Gly Val Glu Val His Asn Ala Lys Thr

275 280 285

Lys Pro Arg Glu Glu Gln Tyr Asn Ser Thr Tyr Arg Val Val Ser Val

290 295 300

Leu Thr Val Leu His Gln Asp Trp Leu Asn Gly Lys Glu Tyr Lys Cys

305 310 315 320

Lys Val Ser Asn Lys Ala Leu Pro Ala Pro Ile Glu Lys Thr Ile Ser

325 330 335

Lys Ala Lys Gly Gln Pro Arg Glu Pro Gln Val Tyr Thr Leu Pro Pro

340 345 350

Ser Arg Glu Glu Met Thr Lys Asn Gln Val Ser Leu Thr Cys Leu Val

355 360 365

Lys Gly Phe Tyr Pro Ser Asp Ile Ala Val Glu Trp Glu Ser Asn Gly

370 375 380

Gln Pro Glu Asn Asn Tyr Lys Thr Thr Pro Pro Val Leu Asp Ser Asp

385 390 395 400

Gly Ser Phe Phe Leu Tyr Ser Lys Leu Thr Val Asp Lys Ser Arg Trp

405 410 415

Gln Gln Gly Asn Val Phe Ser Cys Ser Val Met His Glu Ala Leu His

420 425 430

Asn His Tyr Thr Gln Lys Ser Leu Ser Leu Ser Pro Gly

435 440 445

<---

Claims (23)

1. A method for producing a mixture of antibodies containing two monospecific antibodies and one bispecific antibody, where all these antibodies have a common heavy chain, where the method includes: a. isolating an antibody or portion of an antibody fragment whose specificity is determined by the heavy chain variable domain combined with the first light chain variable domain; b. isolating an antibody or portion of an antibody fragment having a different specificity defined by the same heavy chain variable domain as the antibody of step a) in combination with a second light chain variable domain; c. co-expression in the cell: i. a common heavy chain variable domain fused to an immunoglobulin heavy chain constant region; ii. a first light chain variable domain fused to a kappa light chain constant domain; and iii. a second light chain variable domain fused to a lambda light chain constant domain, thereby obtaining an antibody mixture containing two monospecific antibodies and one bispecific antibody, where all these antibodies share a common heavy chain, where the bispecific antibody has different specificities in the first antigen binding site and the second antigen binding site and consists of two copies of a common heavy chain polypeptide and a first light chain and a second light chain, where the first and second light chains are different from each other, and wherein the first attachment site specifically binds to the first antigen and the second attachment site specifically binds to the second antigen. 2. The method according to claim 1, where: (a) in step (c) (ii) the kappa light chain variable domain is fused to the kappa light chain constant domain; (b) in step (c) (ii) the lambda light chain variable domain is fused to the kappa light chain constant domain; (c) in step (c) (iii) the kappa light chain variable domain is fused to the lambda light chain constant domain; or (d) in step (c) (iii) the lambda light chain variable domain is fused to the lambda light chain constant domain. 3. The method of claim 1, further comprising the step (d) of purifying the antibody mixture obtained in step (c) from the cell culture supernatant. 4. The method of claim 3, wherein step (d) of purifying the antibody mixture is an affinity chromatography step. 5. The method of claim 4, wherein the purification step is an affinity chromatography purification step performed with an affinity chromatography medium specific for the kappa constant domain specific for the lambda constant domain or a medium specific for both the kappa constant domain and for lambda constant domain. 6. An antibody mixture for isolating two monospecific antibodies and one bispecific antibody, containing two monospecific antibodies and one bispecific antibody, where all these antibodies have a common heavy chain, obtained by the method according to p. 1. 7. The method of claim 1 wherein steps (a) and (b) are facilitated by using antibody libraries having a common heavy chain and diversity in the light chain variable domains. 8. The method of claim 7, wherein the antibody library uses a filamentous bacteriophage display, the surface of yeast, bacteria, or mammalian cells, or uses ribosome libraries or other in vitro libraries.

Images