Method and means for producing Ig-like molecules

Abstract

The present invention relates to the fields of molecular biology, medicine and biological therapeutics, and in particular to the field of therapeutic antibodies used in the treatment of a wide variety of diseases. The present invention provides means and methods for producing one or more Ig-like molecules in a single host cell. Novel CH3 mutations are also provided that enable the production of desired monospecific and / or bispecific Ig-like molecules.

Description

[Technical Field]

[0001] The present invention relates to the fields of molecular biology, medicine and biological therapeutics, and in particular to the field of therapeutic antibodies used in the treatment of a wide variety of diseases. [Background technology]

[0002] Currently, the most commonly used biologics are isolated human or humanized recombinant monoclonal antibodies, which enhance the ability of the body's immune system to neutralize or eliminate cells and / or molecules involved in disease processes or eradicate invading pathogens or infectious agents.

[0003] Monoclonal antibodies bind to a specific site on an antigen, or epitope, and in therapeutic antibodies, the epitope is selected to perform a desired function, such as eliminating tumor cells, inhibiting receptor-ligand interactions, or neutralizing viruses.

[0004] Currently, there are approximately 30 monoclonal antibodies approved by the FDA, which are typically produced in large quantities and undergo extensive testing of their biophysical and biochemical properties to ensure that each batch meets quality standards.

[0005] Despite these advantages, monoclonal antibodies have several disadvantages, some of which are related to their inherent monospecificity and the complexity of diseases.

[0006] Disease processes are often multifactorial in nature. The effects of disease-related factors may be redundant or synergistic, and crosstalk between signaling networks may occur due to increased expression of different receptors. Therefore, inhibiting multiple distinct disease-related factors or pathways can improve the efficacy of treatment. However, due to their inherent monospecificity, monoclonal antibodies can only interfere with a single step in a complex disease process, resulting in insufficient efficacy.

[0007] Aside from the multifaceted nature of disease processes, it has become clear that targeting a single epitope on a single cell, soluble protein, or pathogen is not sufficient to treat disease effectively because the target epitope may no longer be accessible to monoclonal antibodies to exert the desired effect. For example, tumor cells often evade monoclonal antibody therapy by reducing, mutating, or masking target epitopes on growth factor receptors.

[0008] By activating other receptors and / or their ligands, tumor cells activate alternative pathways to continue growing and metastasizing. Similarly, viruses and other pathogens evade monoclonal antibody therapy by mutating, deleting, or masking their target epitopes.

[0009] Monoclonal antibodies that bind to a single epitope do not generate all of the effector mechanisms evoked by polyclonal antibodies, notably opsonization (making the antigen more susceptible to phagocytosis), steric hindrance (antibody-encapsulated antigens are prevented from reaching host cells or mucosal surfaces), toxin neutralization, aggregation, or precipitation (antibodies bound to several water-soluble antigens agglutinate and are subsequently removed), complement activation, and antibody-mediated cytotoxicity (antibodies cause target cell killing by NK cells and neutrophils).

[0010] Polyclonal antibodies suitable for therapeutic applications can be obtained from pooled human serum. These serum-derived therapeutic polyclonal antibodies are used to treat and prevent viral infections such as rabies, cytomegalovirus, and respiratory syncytial virus, to neutralize toxins such as tetanus toxin and botulinum toxin, or to prevent anti-D immunization.

[0011] The broader use of serum-derived polyclonal antibody preparations is hindered by the fact that the plasma source can only be applied to a limited range of targets, such as infectious diseases and toxins. Furthermore, the products are dependent on the donor blood supply in terms of both quantity and compatibility, resulting in considerable batch-to-batch variability. Additionally, screening technology has not kept pace with the constant evolution of viruses, and immunoglobulin preparations carry the risk of transmitting infectious diseases. Finally, plasma-derived immunoglobulins are costly to produce due to the lengthy process of blood collection, screening, and immunoglobulin purification.

[0012] A mixture of monoclonal antibodies can enhance the efficacy of monoclonal antibodies while lacking the limitations associated with serum-derived polyclonal antibodies. In this field, combinations of two human or humanized monoclonal antibodies have been tested in preclinical models and clinical trials (e.g., a mixture of two monoclonal antibodies against the HER2 receptor, a mixture of two antibodies against the EGFR receptor, and two monoclonal antibodies against the rabies virus).

[0013] It has been shown in the art that the combination of two monoclonal antibodies can produce additive or synergistic effects, inducing effector mechanisms not associated with either antibody alone. For example, a mixture of two monoclonal antibodies against EGFR or HER2 has been shown to more potently kill tumor cells. This is due to a combination of activities that include enhanced receptor internalization and enhanced inhibition of downstream receptor signaling pathways, in addition to promoting cytotoxicity by immune system effectors.

[0014] In combination therapies based on two monoclonal antibodies, the component antibodies are produced separately and then mixed at the protein level. The drawback of this approach is the enormous cost of conducting clinical trials for each of the two antibodies and then (partially) repeating the process for the combination. This can make therapies based on antibody combinations cost prohibitive.

[0015] Alternatively, two recombinant cell lines producing the constituent monoclonal antibodies can be mixed in a fermenter and the resulting mixture of antibodies purified as a single sample (WO 2004 / 061104). A drawback of this approach is the difficulty in controlling the composition and therefore the poor reproducibility of the resulting recombinant polyclonal antibody preparation, especially given that such composition changes over time as the cells are cultured.

[0016] Over the past decade, bispecific antibodies have been developed as an alternative to using a combination of two antibodies, where two different immunoglobulin molecules in a mixture bind to different epitopes on the same or different targets. In bispecific antibodies, this is achieved with a single immunoglobulin molecule.

[0017] By binding to two epitopes of the same or different targets, bispecific antibodies can exert effects similar to those of a combination of two antibodies that bind to the same epitope. Furthermore, because bispecific antibodies in IgG format combine two different monovalent binding sites in a single molecule, and mixtures of two IgG antibodies combine two different bivalent binding molecules in a single sample, different effects have also been observed depending on these formats.

[0018] Therefore, from a technical and regulatory perspective, it is less laborious to develop a single bispecific antibody, since production, preclinical, and clinical trials only require a single molecule. Thus, single bispecific antibody-based therapies are facilitated by a less laborious and cost-effective drug discovery process and provide more efficient antibody therapies.

[0019] Bispecific antibodies in the IgG format, consisting of two heavy chains and two light chains, can be produced in a variety of ways. For example, bispecific antibodies can be produced by fusing two antibody-secreting cell lines or by expressing two antibodies in a single cell using recombinant DNA technology. These methods produce multiple types of antibodies because the heavy chains corresponding to each antibody can form monospecific dimers (also called homodimers), which contain two identical heavy chain pairs with the same specificity, or bispecific dimers (also called heterodimers), which contain two different heavy chain pairs with different specificities.

[0020] Furthermore, the light and heavy chains from each antibody can pair randomly, resulting in inappropriate, non-functional combinations. This problem, known as heavy and light chain mispairing, can be solved by selecting antibodies that share a common light chain to be expressed as bispecifics. However, even with a common light chain, expressing two heavy chains and one common light chain in a single cell results in three distinct antibody types: two monospecific "parental" antibodies and a bispecific antibody. Therefore, the desired bispecific antibody must be generated from the resulting antibody mixture.

[0021] Several techniques have been employed to further increase the proportion of bispecific antibodies in mixtures of parental antibodies and bispecific antibodies and to reduce mispaired heavy and light chains, but there is a need for bispecific antibody formats that eliminate or minimize some of the problems mentioned above. Summary of the Invention [Problem to be solved by the invention]

[0022] In summary, the prior art provides a variety of techniques and methods for producing monoclonal antibodies, bispecific antibodies, mixtures of monoclonal antibodies, or mixtures of monospecific and bispecific antibodies that can be applied to treat patients.

[0023] However, as noted above, each of these current technologies and methods has its own drawbacks and limitations. Thus, there is a need to develop improved and / or alternative technologies for producing disease-modifying, multi-molecular targeted, mixture- or bispecific-based biological therapeutics. [Means for solving the problem]

[0024] The present invention provides improved and / or alternative technology methods and means for producing disease-modifying, multi-molecular targeted, mixture or bispecific approach-based biological therapeutics. The invention further provides the resulting products and uses of these methods and means.

[0025] The prior art describes various approaches to promote the formation of specific bispecific antibodies of interest, which can reduce the composition of undesired antibodies in the resulting mixture.

[0026] In antibodies, CH3-CH3 interactions are known to play a major role in Fc dimerization (Ellerson JR., et al., J. Immunol 1976 (116) 510-517 and Deisenhofer J. Biochemistry 1981 (20) 2361-2370). Furthermore, when two CH3 domains interact with each other, they face each other at a protein-protein interface consisting of "contact" residues (also called contact amino acids, interface residues, or interface amino acids).

[0027] Contact amino acids in a first CH3 domain interact with one or more contact amino acids in a second CH3 domain. The contact amino acids are typically within 5.5 Å (preferably 4.5 Å) of each other in the three-dimensional structure of the antibody. The interaction between contact residues in one CH3 domain and contact residues in a different CH3 domain can be mediated, for example, by van der Waals forces, hydrogen bonds, water-mediated hydrogen bonds, salt bridges or other electrostatic forces, attractive interactions between aromatic side chains, disulfide bonds, or other known forces.

[0028] It has been shown that approximately one-third of the contact amino acid side chains at the CH3 domain interface of human IgG1 contribute to the majority of domain folding and association, and it is anticipated that other (adjacent) amino acid residues may also influence interactions at this protein-protein interface.

[0029] Prior art approaches have been employed to interfere with antibody heavy chain dimerization. Specific engineering of the CH3 domain has been applied to promote heterodimerization over homodimerization. Examples of such CH3-CH3 interface engineering are described, for example, in WO 1998 / 050431, Ridgeway et al., 1996, and Merchant et al., 1998, and are also known as the "knob-into-hole" approach. This method introduces mutations that result in complementary protrusions and cavities.

[0030] Generally speaking, this method introduces protuberances at the interface of a first polypeptide and a corresponding cavity into a second polypeptide, with the protuberances positioned in the cavity, thereby promoting heteromultimer formation and preventing homomultimer formation.

[0031] A "protrusion" or "knob" is created by replacing a short amino acid side chain at the interface of a first polypeptide with a longer amino acid side chain (e.g., tyrosine or tryptophan). A complementary "cavity" or "hole" of identical or similar size to the protrusion is created by replacing a long amino acid side chain at the interface of a second polypeptide with a shorter amino acid side chain (e.g., alanine or threonine). Protrusions and cavities can be created by synthetic means, such as modifying a nucleic acid encoding the polypeptide, or by synthesis of the peptide.

[0032] Using the knob-into-hole technique alone, the proportion of the bispecific antibody of interest in a mixture of two parent antibodies and a bispecific antibody can be increased to 87% at best. Merchant et al. succeeded in increasing the proportion of bispecific antibodies in the mixture to 95% by introducing an additional disulfide bond between the two CH3 domains in the CH3-CH3 interface. However, to use such bispecific antibodies as drugs, they must be purified from the homodimer and formulated into a pharmaceutically acceptable diluent or excipient. Purifying the heteromultimer from such a mixture presents significant challenges due to the similar physicochemical properties of the homodimer and heterodimer.

[0033] One object of the present invention is to provide a method for producing bispecific antibodies from a single cell clone, which further improves the proportion of bispecific antibodies in the mixture. According to the present invention, the knob-into-hole technology can be used as one of the means, alone or in combination with other means, to achieve the further improved proportion of bispecific antibodies in the mixture.

[0034] Another example of such CH3-CH3 interface engineering is provided by heterodimeric Fc technology. This technology involves the design of bispecific and asymmetric fusion proteins through the development of strand-exchange engineered domain (SEED) CH3 heterodimers. These SEED·CH3 heterodimers are derivatives of human IgG and IgA, composed of alternating segments of human IgA and IgG CH3 sequences. This results in a pair of complementary human SEED·CH3 heterodimers, termed SEED-bodies (Davis JH. et al., Protein Engineering, Design & Selection 2010(23)195-202; WO 2007 / 110205).

[0035] Yet another approach to producing the desired bispecific antibodies is based on electrostatic manipulation of naturally charged contact residues at the CH3-CH3 interface, as described, for example, in EP 01870459 or US 2010 / 0015133, WO 2007 / 147901, WO 2010 / 129304, Gunasekaran et al (2010) and WO 2009 / 089004.

[0036] The mutations in the heavy chain CH3 domain described in these publications replace naturally charged contact residues with amino acid residues of the opposite charge (charge reversal strategy). This alters the charge polarity at the opposing contact surfaces of the Fc dimer, allowing favorable attractive interactions when electrostatically compatible Fc chains are coexpressed. Thus, the formation of desirable Fc heterodimers is promoted, while the formation of undesirable Fc homodimers is inhibited by repulsive charge interactions.

[0037] Four characteristic pairs of charged residues have been reported to be involved in the interdomain interactions at the CH3-CH3 interface: D356 / K439', E357 / K370', K392 / D399', and D399 / K409' (following the numbering scheme reported by Kabat (1991), where the residues in the first and second chains are separated by a " / " and the prime symbol (') represents the number of the residue in the second chain). Because the CH3-CH3 interface has two-fold symmetry, each characteristic pair of charged residues appears twice in native IgG (i.e., the electrostatic interactions of K439 / D356', K370 / E357', D399 / K392', and K409 / D399' are also present at the interface).

[0038] Taking advantage of this two-fold symmetry, we demonstrated that a single charge inversion, such as K409D in the first chain or D399'K in the second chain, reduced homodimer formation due to identical charge repulsion. This repulsion effect was further enhanced by inverting the opposite charge. Expression of different CH3 domains with different complementary charge inversions was shown to induce heterodimerization, resulting in an increased proportion of bispecific species in the mixture.

[0039] The above-described approach has enabled the percentage of bispecific antibodies produced from a single cell to be between about 76% and about 96%. One object of the present invention is to provide a method for producing bispecific antibodies from a single cell that further improves the percentage of the desired bispecific antibody. According to the present invention, electrostatic manipulation techniques can be used as one of the means to achieve the further improved percentage of the desired (bispecific) antibody, either alone or in combination with other means, such as the knob-into-hole approach.

[0040] In one aspect, the present invention provides a method for producing at least two different Ig-like molecules from a single host cell, each of the two Ig-like molecules comprising two CH3 domains capable of forming an interface, the method comprising providing to the cell: a. a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain, b. a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, c. a third nucleic acid molecule encoding a polypeptide chain comprising a third CH3 domain, and d. a fourth nucleic acid molecule encoding a polypeptide chain comprising a fourth CH3 domain, wherein at least two of the nucleic acid molecules comprise means for selective pairing of a polypeptide comprising the first and second CH3 domains with a polypeptide comprising the third and fourth CH3 domains, and the method further comprises the steps of culturing the host cell to express the at least four nucleic acid molecules and recovering the at least two different Ig-like molecules from the culture.

[0041] For example, it is often desirable to produce multiple (bispecific) antibodies to more efficiently inhibit multiple biological pathways involved in a disease process or pathogen invasion, replication, and / or spread.

[0042] The use of a mixture of multiple bispecific antibodies to treat a particular disease is particularly useful. For example, tumor cells employ many different strategies to develop resistance during treatment with antibodies or small molecule drugs. Resistance can involve multiple cell surface receptors and soluble molecules, and it would be beneficial to develop antibody-based cancer therapies that simultaneously address multiple such disease- and escape-related molecules.

[0043] When two or more target molecules or epitopes associated with such diseases and evasion are involved, mixtures of bispecific antibodies are a novel and interesting therapeutic modality. Such mixtures of bispecific antibodies are preferably produced from a single cell, facilitating the drug discovery process. This makes the drug discovery process less tedious from a regulatory perspective and more cost-effective and feasible from a pharmaceutical and clinical development perspective.

[0044] In a single cell-based approach, it would be desirable to use methods that allow for controlled and efficient production of bispecific antibodies. This would reduce or completely eliminate the need to separate the mixture of desired bispecific IgG molecules from undesired monospecific IgG molecules. In the prior art, monospecific and bispecific antibodies have been produced from single cells (WO 2004 / 009618). However, these mixtures are complex mixtures of several different bispecific and monospecific antibody types.

[0045] A further object of the present invention is to provide methods and means for producing a mixture of defined bispecific antibodies from a single cell. As described below, the method preferably results in a (bispecific) antibody mixture with at least 95%, at least 97%, or greater than 99% dimeric IgG molecules, regardless of the amount of monomeric by-products. Generally speaking, in cells producing multiple, native IgG molecules, half molecules (monomeric by-products) will be present, but they can be easily removed by known size exclusion chromatography.

[0046] In one embodiment, the present invention provides a method for producing a defined mixture of at least two different Ig-like molecules in a single cell instead of a single (bispecific) antibody of interest, with reduced or no formation of undesired dimeric antibody species. The resulting mixture is well-defined, and its composition is controlled by engineered mutations in the CH3 domain. Furthermore, adjusting the expression level and / or different transfection ratios used for expression can affect the composition of the mixture.

[0047] In the methods of the invention, the CH3 domain encoded by the first nucleic acid molecule selectively pairs with the CH3 domain encoded by the second nucleic acid molecule, and the CH3 domain encoded by the third nucleic acid molecule selectively pairs with the CH3 domain encoded by the fourth nucleic acid molecule. The invention further provides a mixture of at least two different Ig-like molecules obtainable by the methods of the invention.

[0048] As used herein, "selective pairing of polypeptides comprising the first and second CH3 domains" means that substantially all of the resulting dimers having a polypeptide comprising the first CH3 domain and / or a polypeptide comprising the second CH3 domain are dimers consisting of a pair of a polypeptide comprising one first CH3 domain and one polypeptide comprising the second CH3 domain.

[0049] Similarly, "selective pairing of polypeptides comprising the third and fourth CH3 domains" means that substantially all of the resulting dimers having a polypeptide comprising a third CH3 domain and / or a polypeptide comprising a fourth CH3 domain are dimers consisting of a pair of a polypeptide comprising one third CH3 domain and one polypeptide comprising a fourth CH3 domain.

[0050] As a result, when a nucleic acid molecule encoding a polypeptide with four different (A, B, C, D) CH3 domains is introduced into a single cell, a mixture of primarily two specific Ig-like molecules is produced instead of a mixture of 10 different Ig-like dimers (AA, AB, AC, AD, BB, BC, BD, CC, CD, and DD).

[0051] As described in more detail below, in a preferred embodiment of the invention, the polypeptide chain comprising the first CH3 domain has the amino acid substitution T366K and the polypeptide chain comprising the second CH3 domain has the amino acid substitution L351D. These amino acid changes are a preferred means for selective pairing of the polypeptide chains comprising the first and second CH3 domains.

[0052] The polypeptide chain comprising the first CH3 domain preferably further comprises the amino acid substitution L351K. The polypeptide chain comprising the second CH3 domain preferably further comprises an amino acid substitution selected from the group consisting of Y349E, Y349D, and L368E, with L368E being most preferred. In yet another preferred embodiment, the polypeptide chain comprising the third CH3 domain comprises the amino acid substitutions E356K and D399K, and the polypeptide chain comprising the fourth CH3 domain comprises the amino acid substitutions K392D and K409D.

[0053] In the method according to the present invention, each polypeptide chain comprising a CH3 domain preferably further comprises a variable region recognizing a target epitope. The variable regions that are part of the polypeptide chain comprising a CH3 domain preferably share a common light chain. In this case, only the VH of the variable regions differ, while the VL of all variable regions are substantially identical.

[0054] Thus, in a preferred embodiment, the method according to the present invention further comprises providing the host cell with a nucleic acid molecule encoding a common light chain. In a particularly preferred embodiment, each of the four variable regions of the polypeptide chain comprising four CH3 domains recognizes a different target epitope. For example, if a first nucleic acid molecule encodes a heavy chain further comprising a variable domain specific for antigen A, a second nucleic acid molecule encodes a heavy chain further comprising a variable domain specific for antigen B, a third nucleic acid molecule encodes a heavy chain further comprising a variable domain specific for antigen C, and a fourth nucleic acid molecule encodes a heavy chain further comprising a variable domain specific for antigen D. A mixture is then produced comprising bispecific Ig-like molecules specific for AB and bispecific Ig-like molecules specific for CD.

[0055] The formation of monospecific antibodies (AA, BB, CC, or DD specific) or bispecific antibodies specific for AC, AD, BC, or BD is reduced or eliminated by means of the selective pairing of polypeptides comprising the first and second CH3 domains with polypeptides comprising the third and fourth CH3 domains. It is of course possible to use additional nucleic acid molecules, such as those encoding polypeptides comprising the fifth and sixth CH3 domains, for the production of defined mixtures containing more than two different Ig-like molecules.

[0056] It should be noted that the ratio of nucleic acids used in the methods of the present invention does not need to be 1:1:1:1. Furthermore, the ratio of the resulting expressed Ig-like molecules does not need to be 1:1. Using known means, it is possible to create an antibody mixture with an optimized ratio. For example, the expression level of the nucleic acid molecules, and thus the ratio of the resulting Ig-like molecules, can be adjusted using different genetic elements such as promoters, enhancers, and repressors, or by controlling the genomic integration site of the copy number of the antibody-encoding DNA construct.

[0057] The means for selective pairing preferably comprises complementary knob-into-hole mutations, disulfide bridges, charge mutations including charge reversal mutations, or a combination thereof. Those skilled in the art will appreciate that the means for selective pairing may be selected from a range of types of mutations. That is, at least four nucleic acid molecules encoding polypeptide chains comprising a CH3 domain are required for charge mutations as a means for selective pairing.

[0058] Furthermore, in certain cases, unengineered wild-type CH3s are also used for the selective pairing of two polypeptide chains comprising wild-type CH3 domains. In a particularly preferred embodiment, the means for selective pairing comprises at least one CH3 mutation selected from Table B, as described herein below.

[0059] A preferred embodiment provides a method according to the invention, wherein all four of said nucleic acid molecules comprise means for selective pairing of polypeptides comprising said first and second CH3 domains with polypeptides comprising said third and fourth CH3 domains, wherein said means for selective pairing of polypeptides comprising said first and second CH3 domains is different from the means for selective pairing of polypeptides comprising said third and fourth CH3 domains.

[0060] One aspect of the present invention provides a method according to the present invention, wherein the means for selective pairing of polypeptides comprising the first and second CH3 domains is different from the means for selective pairing of polypeptides comprising the third and fourth CH3 domains. "Different" here means that the means for selective pairing of polypeptides comprising the first and second CH3 domains is designed to favor selective pairing of the first and second chains, such that interactions between the first and the third and / or fourth CH3 domain polypeptide chains are substantially eliminated. In other words, dimerization of the polypeptide comprising the first CH3 domain with the third or fourth polypeptide is essentially eliminated or nearly eliminated. The polypeptide comprising the third and fourth CH3 domains is either wild-type or is provided with means for selective pairing that are different from the means for selective pairing of the first and second CH3 domains.

[0061] Recent research has focused on producing single bispecific antibodies using, for example, knob-into-hole technology or mutation (inversion) of charged contact amino acids present in the CH3 domain. However, prior to the present invention, the production of a defined mixture of at least two (bispecific) Ig-like molecules without significant co-production of other dimeric by-products has not been feasible.

[0062] The present invention provides an efficient and controlled method for producing mixtures of well-defined Ig-like molecules, including a high percentage of bispecifics in the mixture. In systems where two bispecifics are desired, a percentage of (two) bispecifics of at least 95%, at least 97%, or even higher can be obtained. This translates to no more than 5%, no more than 3%, or even less monospecific, bivalent by-products. Note that the amount of monomeric by-products (i.e., half molecules) is not critical, since these half molecules can be easily separated based on their size.

[0063] In another preferred embodiment of the invention, the variable regions of the polypeptide chains comprising the first and second CH3 domains recognize different target epitopes, while the variable regions of the polypeptide chains comprising the third and fourth CH3 domains recognize the same target epitope. This results in primarily one type of bispecific Ig-like molecule and one type of monospecific Ig-like molecule. For example, if the variable regions of the polypeptide chains comprising the first and second CH3 domains recognize different target epitopes, and if the variable regions of the polypeptide chains comprising the third and fourth CH3 domains both recognize epitopes that are different from the target epitopes recognized by the same first and second CH3 domains, a mixture of Ig-like molecules with AB or CC specificity will be produced.

[0064] The present invention further provides a method wherein the target epitopes recognized by the variable regions of the polypeptide chains comprising the third and fourth CH3 domains are identical but different from the target epitopes recognized by the variable regions of the polypeptide chains comprising the first or second CH3 domains.

[0065] Alternatively, when the variable regions of the polypeptide chains comprising the first and second CH3 domains recognize different target epitopes and the variable regions of the polypeptide chains comprising the third and fourth CH3 domains both recognize the same epitope as the polypeptide chain comprising the first or second CH3 domain, a mixture of Ig-like molecules with specificity for AB and AA, or AB and BB, is produced.

[0066] The methods of the present invention provide that the target epitope recognized by the variable region of the polypeptide chain comprising the third and fourth CH3 domains is identical to the target epitope recognized by the variable region of the polypeptide chain comprising the first or second CH3 domain.

[0067] Another object of the present invention is to provide methods and means for producing defined mixtures of bispecific and monospecific antibodies in single-cell culture. A non-limiting example of such a well-defined mixture is a mixture of an AB-specific bispecific antibody and an AA-specific monospecific antibody. Another example is a mixture of an AB-specific bispecific antibody and a BB-specific monospecific antibody. Yet another example is a mixture of an AB-specific bispecific antibody and a CC-specific monospecific antibody. Again, suitable means and methods are provided for producing a mixture of desired antibodies having at least 90%, preferably 95%, and most preferably at least 97% or even greater than 99% of the desired antibody.

[0068] In another embodiment, the methods of the present invention provide that the variable regions of the polypeptide chains comprising the first and second CH3 domains recognize the same target epitope, while the variable regions of the polypeptide chains comprising the third and fourth CH3 domains recognize a second target epitope that is different from the target epitopes recognized by the first and second variable regions. This results in the production of monospecific Ig-like molecules with predominantly AA or BB specificity. The formation of bispecific Ig-like molecules is reduced or eliminated.

[0069] In some embodiments, producing a mixture of monospecific antibodies in a single cell is preferred over a mixture of bispecific antibodies. For example, when cross-linking of two identical target molecules is desired, or when two targets are too far apart to be bound by a single bispecific antibody. Producing a mixture of monospecific antibodies in a single cell can be advantageous because they can be viewed as a single therapeutic product.

[0070] The therapeutic efficacy and safety of various monospecific antibodies have already been proven in the art, and they have been approved for production. Production of mixtures of monospecific antibodies in single cells could facilitate testing of the efficacy and safety of several such mixtures and reduce the efficiency and cost of regulatory approval and production. However, methods for producing specific mixtures of monospecific antibodies in single cells that result in less than 5% formation of bispecific by-products are currently unavailable. Another object of the present invention is to provide means and methods for producing such mixtures of well-defined homodimers that result in less than 5% formation of bispecific antibodies.

[0071] The method according to the invention is suitable for producing any desired mixture of bispecific and / or monospecific Ig-like molecules. Again, to produce defined mixtures containing more than two different Ig-like molecules, it is possible to use, for example, additional nucleic acid molecules encoding polypeptides comprising the fifth and sixth (and seventh and eighth, etc.) CH3 domains.

[0072] Preferably, the method of the present invention uses at least two CH3 domains that contain a combination of at least one of the mutations provided by the present invention. Through these mutations, a new specific interaction is formed between the two CH3 domains. These mutations of the present invention are described in detail below.

[0073] As used herein, the term "Ig-like molecule" refers to a proteinaceous molecule having at least one immunoglobulin (Ig) domain. The Ig-like molecule comprises a sequence having the function of at least one immunoglobulin CH3 domain, preferably the sequence comprises the CH3 domain of IgG1. A proteinaceous molecule having at least one CH3 domain can further comprise a specific binding moiety.

[0074] The CH3 domains of the present invention comprise a means for selective pairing and are used for the selective pairing of two proteinaceous molecules comprising a CH3 domain to design a desired heterodimeric binding molecule or mixture of binding molecules. The binding moiety introduced into the proteinaceous molecule comprising a CH3 domain can be any binding means, including, but not limited to, single-chain Fvs, single-chain or tandem diabodies (TandAb®), VHHs, Anticalins®, Nanobodies®, BiTEs®, Fabs, ankyrin repeat proteins or DARPINs®, Avimers®, DARTs, TCR-like antibodies, Adnectins®, Affilins®, Trans-bodies®, Affibodies®, TrimerX®, MicroProteins, Fynomers®, Centyrins®, or KALBITOR®.

[0075] In a preferred embodiment, the binding moieties are antibody variable regions (i.e., VH / VL combinations). The variable regions that are part of a polypeptide chain comprising a CH3 domain preferably share a common light chain. In such cases, only the VH of the variable regions differ, while the VL of all variable regions are substantially identical.

[0076] Additionally or alternatively, cytokines, hormones, water-soluble ligands, receptors and / or peptides and other molecules can be introduced into the CH3 domains of the invention.

[0077] In a more preferred embodiment, the Ig-like molecule comprises a full-length Fc main chain. In a most preferred embodiment, the Ig-like molecule is an antibody. Preferably, the variable regions of these antibodies share a common light chain but may differ in the VH region.

[0078] The term "antibody" as used herein refers to a proteinaceous molecule belonging to the immunoglobulin class of proteins, which contains a domain that binds to one or more antigen epitopes, the domain being derived from or having sequence homology with the variable region of an antibody. Known antibodies include several isotypes, such as IgG1, IgG2, IgG3, IgG4, IgA, IgD, IgE, and IgM. Antibodies according to the present invention may be of any isotype or functional derivatives and / or fragments thereof. In a preferred embodiment, an antibody of the IgG isotype is produced as an Ig-like molecule, since IgG antibodies have a longer half-life compared to, for example, antibodies of other isotypes.

[0079] Antibodies produced by the methods of the present invention can have sequences from any source, including murine and human sequences. Antibodies can be composed of sequences from one source, such as from all human antibodies, or they can have sequences from two or more sources, resulting in what are called chimeric or humanized antibodies.

[0080] The closer a therapeutic antibody is to the subject's natural antibody, the better (e.g., a human antibody for a human subject). Antibody binding is expressed in terms of specificity and affinity. Specificity determines which antigen or epitope a binding domain binds to. Affinity is a measure of the strength of binding to a particular antigen or epitope. Specific binding is defined as binding with an affinity (KD) of at least 1×10 M, more preferably 1×10 M, and even more preferably greater than 1×10 M. Typically, therapeutic monoclonal antibodies are used with an affinity of 1×10 M or higher.

[0081] As used herein, the term "antigen" refers to a substance or molecule that, when introduced into the body, triggers the immune system to produce antibodies. Antigens can come from a variety of sources, including pathogens, tumor or other abnormal cells, haptens, or autologous tissue, among others. At the molecular level, antigens are characterized by their ability to bind to the antigen-binding site of an antibody. Mixtures of antigens are also considered "antigens." That is, while tumor cell lysates or virus particles are often considered "antigens" by those skilled in the art, many antigenic determinants are present in such tumor cell lysates or virus particle preparations.

[0082] An antigen has at least one, and often two or more, epitopes. The term "epitope" here refers to the part of an antigen that is recognized by the immune system, specifically antibodies, B cells, or T cells. Epitopes are typically thought to be derived from non-self proteins, although host-derived sequences can also be classified as epitopes.

[0083] The term "CH3 domain" is well known. IgG structure has four chains: two light chains and two heavy chains. Each light chain has two domains: a variable region and a constant region (VL and CL). Each heavy chain has four domains: a variable region (VH) and three constant regions (CH1, CH2, and CH3). The CH2 and CH3 domain regions of the heavy chain are called the Fc (Fragment crystallizable) portion, Fc fragment, Fc main chain, or simply Fc.

[0084] IgG molecules are heterotetramers consisting of two heavy chains and two light chains connected by disulfide bonds (-SS-) at the hinge region. The heavy chains dimerize through interactions at the CH3-CH3 domain interface and at the hinge region. The number of disulfide bonds in the hinge region varies depending on the immunoglobulin subclass (Papadea and Check 1989).

[0085] The Fc fragment of an immunoglobulin is a dimer of two C-terminal constant regions, namely the CH2 and CH3 domains of the heavy chain. Part of its physiological function is interaction with the complement system and with specific receptors on the surface of various cells. It is known that the interaction between the CH3 domains of two individual heavy chains plays an important role in inducing heavy chain dimerization.

[0086] Thus, the CH3 domain plays a primary role in the association of antibody heavy chains, and the interface between the CH3 domains contains more than 20 contact residues from each chain, which play a role in CH3-CH3 interactions (Deisenhofer J., Biochemistry 1981(20)2361-2370; Miller S., J. Mol. Biol. 1990(216)965-973; Padlan, Advances in Protein Chemistry 1996(49)57-133).

[0087] The variant CH3 domains of the invention can be used in conjunction with other antibody domains to generate bispecific or monospecific full-length antibodies. The specificity of the antibody, as determined by the VH / VL combination, generally does not affect the heavy chain dimerization behavior induced by the CH3 domain.

[0088] As used herein, the terms "contact residue," "contact amino acid," "interface residue," and "interface amino acid" generally refer to any amino acid residue present in the CH3 domain that may be involved in interdomain contact. This may be calculated by known techniques, including the calculation of the solvent accessible surface area (ASA) of residues in the CH3 domain in the presence and absence of a second chain, where residues that show a difference in ASA (>1 Å) between the two conditions are identified as contact residues (Lee and Richards J. Mol. Biol. 1971(55)379). Residues identified as contact residues are those at positions 347, 349, 350, 351, 352, 353, 354, 355, 356, 357, 360, 364, 366, 368, 370, 390, 392, 394, 395, 397, 399, 400, 405, 407, 409, and 439 according to the EU numbering system (Table A).

[0089] [Table A]

[0090] The contact residues at the CH3-CH3 interface can be charged or neutral amino acid residues. As used herein, the term "charged amino acid residue" or "charged residue" refers to an amino acid residue having an electrically charged side chain. These can be positively charged side chains, such as those present in arginine (Arg, R), histidine (His, H), and lysine (Lys, K), or negatively charged side chains, such as those present in aspartic acid (Asp, D) and glutamic acid (Glu, E).

[0091] As used herein, the term "neutral amino acid residue" or neutral residue refers to all other amino acids that do not have an electrically charged side chain. These neutral residues include serine (Ser, S), threonine (Thr, T), asparagine (Asn, N), glutamine (GLu, Q), cysteine (Cys, C), glycine (Gly, G), proline (Pro, P), alanine (Ala, A), valine (Val, V), isoleucine (Ile, I), leucine (Leu, L), methionine (Met, M), phenylalanine (Phe, F), tyrosine (Tyr, Y), and tryptophan (Trp, T).

[0092] As used herein, a "CH3-CH3 domain interface" or "CH3 interface," "CH3-CH3 pairing," "domain interface," or simply "interface" refers to the association of two CH3 domains of different polypeptides comprising CH3 domains as a result of an interaction of amino acid residues, i.e., at least one interaction between an amino acid of a first CH3 domain and an amino acid of a second CH3 domain. Such interactions may be, for example, van der Waals forces, hydrogen bonds, water-mediated hydrogen bonds, salt bridges or other electrostatic forces, attractive interactions between aromatic side chains, disulfide bond formation, or other known forces.

[0093] As used herein, said means of selective pairing of a polypeptide comprising a first and second CH3 domain with a polypeptide comprising said third and fourth CH3 domain may be any known means.

[0094] In one embodiment, at least one nucleic acid molecule encodes a CH3 domain containing large amino acid residues (i.e., "knobs" or "protrusions") at the contact residue positions, e.g., R, F, Y, W, I, or L, while at least one other nucleic acid molecule encodes a CH3 domain containing small amino acid residues (i.e., "holes" or "cavities") at the complementary contact residue positions, e.g., G, A, S, T, or V. The resulting CH3 domains are more likely to pair with each other due to the conformation of the contact amino acids. The knob-into-hole technique has been described in detail above.

[0095] In a further embodiment of the invention, at least one nucleic acid molecule encodes a CH3 domain containing an amino acid bearing an opposite charge compared to the wild-type at the contact residue position of a naturally charged residue, i.e., at the position originally occupied by K, H, R, D, or E, while at least one other nucleic acid molecule encodes a CH3 domain containing an amino acid bearing an opposite charge compared to the wild-type at the complementary contact residue position of the naturally charged residue. The resulting engineered CH3 domains are more likely to pair with each other due to the opposite charges of the contact amino acids, but are less likely to pair with other identical CH3 domains due to electrostatic repulsion.

[0096] In one embodiment, the CH3 mutations described in EP 01870459, WO 2009 / 089004, and Gunasekaran et al (2010) are used.

[0097] In one embodiment, the means for selective pairing of the polypeptides comprising the first and second CH3 domains are "knob" and "hole" amino acid residues, and the means for selective pairing of the polypeptides comprising the third and fourth CH3 domains are charge-engineered amino acids. Preferably, the means for selective pairing of both the polypeptides comprising the first and second CH3 domains and the polypeptides comprising the third and fourth CH3 domains are charge-engineered amino acids.

[0098] In one embodiment, the amino acid residues engineered for selective pairing of the polypeptides comprising the first and second CH3 domains are different compared to the amino acid residues engineered for selective pairing of the polypeptides comprising the third and fourth CH3 domains.

[0099] In a particularly preferred embodiment, at least the first and second nucleic acid molecules encode novel mutations in the CH3 domain provided by the present invention. As described in more detail below, the present invention provides novel CH3 mutations that allow for the production of specific bispecific Ig-like molecules of interest without producing significant amounts of undesired (dimeric) by-products. Furthermore, the present invention provides novel CH3 mutations that allow for the production of specific monospecific Ig-like molecules of interest without producing significant amounts of undesired (dimeric) by-products. Therefore, the use of at least one of these CH3 mutations according to the present invention is preferred.

[0100] As used herein, the terms "polypeptide," "polypeptide molecule," or "polypeptide chain" refer to a chain of amino acids covalently linked through peptide bonds. Proteins are generally composed of one or more polypeptide molecules. In all polypeptides, one end, called the amino terminus or N-terminus, has a free amino group. The other end, called the carboxyl terminus or C-terminus, has a free carboxyl group. Polypeptides of the present invention may undergo post-translational modification processes, such as glycosylation. Thus, a polypeptide chain comprising a CH3 domain of the present invention refers to a polypeptide chain comprising at least an Ig CH3 domain, including those that have undergone post-translational modifications.

[0101] As used herein, the term "nucleic acid molecule" is defined as a molecule comprised of a chain of nucleotides, more preferably DNA and / or RNA. In one embodiment, double-stranded RNA is used. In other embodiments, nucleic acid molecules of the invention comprise other types of nucleic acid structures, such as, for example, DNA / RNA helices, peptide nucleic acid (PNA), locked nucleic acid (LNA), and / or ribozymes. Thus, the term "nucleic acid molecule" also encompasses chains comprising non-natural nucleotides, modified nucleotides, and / or non-nucleic acid components that exhibit the same function as natural nucleotides.

[0102] The present invention further provides a method for producing a host cell that produces at least two different Ig-like molecules, the method comprising the step of introducing into the host cell nucleic acid sequences encoding polypeptide chains comprising at least a first, second, third and fourth CH3 domain, wherein at least two of the nucleic acid sequences provide a means for selective pairing of a polypeptide comprising the first and second CH3 domains with a polypeptide comprising the third and fourth CH3 domains, the nucleic acid sequences being introduced sequentially or simultaneously.

[0103] In a further aspect of the invention, a method of generating a host cell for the production of a heterodimeric Ig-like molecule comprises introducing into the host cell nucleic acid sequences encoding polypeptide chains comprising at least first and second CH3 domains, wherein the polypeptide chain comprising the first CH3 domain comprises at least one substitution of a neutral amino acid residue with a positively charged amino acid residue, and the polypeptide chain comprising the second CH3 domain comprises at least one substitution of a neutral amino acid residue with a negatively charged amino acid residue, either sequentially or simultaneously. The method of generating the host cell preferably comprises introducing into the host cell a nucleic acid sequence encoding a common light chain.

[0104] In one aspect further provided herein, in a recombinant host cell comprising nucleic acid sequences encoding polypeptide chains comprising at least first, second, third and fourth CH3 domains, at least two of said nucleic acid molecules comprise means for selective pairing of a polypeptide comprising said first and second CH3 domains with a polypeptide comprising said third and fourth CH3 domains.

[0105] The present invention further provides a recombinant host cell comprising a nucleic acid sequence encoding a polypeptide chain comprising at least a first and a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises at least one substitution of a neutral amino acid residue with a positively charged amino acid residue, and the polypeptide chain comprising the second CH3 domain comprises at least one substitution of a neutral amino acid residue with a negatively charged amino acid residue.

[0106] A recombinant host cell according to the invention preferably comprises a nucleic acid sequence encoding a common light chain.

[0107] A "host cell" of the present invention may be any host cell capable of expressing a recombinant DNA molecule, including bacteria such as Escherichia (e.g., E. coli), Enterobacter, Salmonella, Bacillus, Pseudomonas, Streptomyces, yeasts such as S. cerevisiae, K. lactis, P. pastoris, Candida, or Yarrowia, filamentous fungi such as Neurospora, Aspergillus oryzae, Aspergillus nidulans, and Aspergillus niger, insect cells such as Spodoptera frugiperda SF-9 or SF-21 cells, and preferably murine cells, including CHO (Chinese hamster ovary) cells, BHK cells, SP2 / 0 cells, and NS-0 myeloma cells, primate cells such as COS and Vero cells, MDCK cells, BRL cells, and the like. Included are mammalian cells such as 3A cells, hybridomas, tumor cells, immortalized primary cells, and embryonic retinal cells such as W138, HepG2, HeLa, HEK293, HT1080, or PER.C6.

[0108] When selecting an expression system, mammalian cell expression vectors and hosts are often used to ensure that the antibody is glycosylated appropriately. Human cell lines, preferably PER.C6, are advantageously used to obtain antibodies with glycosylation patterns consistent with those in humans. Conditions for growing or propagating cells (see, Tissue Culture, Academic Press, Kruse and Paterson, editors (1973)) and conditions for expressing recombinant products may differ somewhat. Furthermore, process optimization is typically performed to increase product yield and / or cell growth, respectively, using methods commonly known to those skilled in the art.

[0109] General guidelines, procedures, and practical techniques for maximizing productivity in mammalian cell cultures can be found in Mammalian Cell Biotechnology: a Practical Approach (M. Butler, ed., IRL Press, 1991). Expression of antibodies in recombinant host cells has been extensively described in the literature (e.g., EP 0 120 694, EP 0 314 161, EP 0 481 790, EP 0 523 949, U.S. Pat. No. 4,816,567, WO 00 / 63403). The nucleic acid molecules encoding the light and heavy chains can be extrachromosomal copies and / or stably integrated into the host cell chromosome, the latter being preferred.

[0110] In a further aspect of the invention, there is provided a culture of recombinant host cells according to the invention or a culture of recombinant host cells obtainable or obtainable by a method according to the invention, wherein said culture produces at least either two different Ig-like molecules or a heterodimeric Ig-like molecule.

[0111] When obtaining expression of a sequence of a nucleic acid molecule encoding a polypeptide comprising a CH3 domain, a sequence capable of inducing such expression is known to be functionally related to the sequence of a nucleic acid molecule encoding a polypeptide comprising a CH3 domain. By functionally related, we mean that the nucleic acid sequence encoding a polypeptide comprising a CH3 domain or a precursor thereof is related to the sequence capable of inducing expression in that it is capable of inducing expression of a polypeptide comprising a CH3 domain or a precursor thereof.

[0112] Useful expression vectors are already available, such as the Invitrogen line of commercial pcDNA vectors. When a sequence encoding a polypeptide of interest is properly inserted relative to sequences that control the transcription and translation of the encoded polypeptide, the resulting expression cassette is useful for producing, or expressing, the polypeptide of interest.

[0113] Sequences that drive expression include promoters, enhancers, etc., and combinations thereof. These must be capable of functioning in the host cell, thereby driving the expression of a functionally associated nucleic acid sequence. Promoters can be constitutive or regulatable, and can be obtained from a variety of sources, including viral, prokaryotic, or eukaryotic, or can be artificially designed.

[0114] Expression of a nucleic acid of interest can be driven by the native promoter or its derivatives, or by a completely heterologous promoter. Well-known and commonly used promoters for expression in eukaryotic cells include promoters derived from viruses such as adenovirus (e.g., the E1A promoter), promoters derived from cytomegalovirus (CMV) (e.g., the CMV immediate early (IE) promoter), and promoters derived from simian virus 40 (SV40). Suitable promoters for eukaryotic cells are also available, including the metallothionein (MT) promoter, elongation factor 1 alpha (EF-1α) promoter, actin promoter, immunoglobulin promoter, and heat shock promoters.

[0115] Any promoter or enhancer / promoter capable of inducing expression of a sequence of interest in a host cell is suitable for the present invention. In one embodiment, the sequence capable of inducing expression comprises a CMV promoter region, preferably comprising the -735 to +95 nucleotide region of the CMV immediate early gene enhancer / promoter. As those skilled in the art will recognize, the expression sequence used in the present invention may preferably be combined with elements that stabilize or enhance expression, such as insulators, matrix attachment regions, STAR elements (WO 03 / 004704), etc., which can increase the stability and / or level of expression.

[0116] The production of proteins in recombinant host cells has been extensively described, for example, in Current Protocols in Protein Science, 1995, Coligan JE, Dunn BM, Ploegh HL, Speicher DW, Wingfield PT, ISBN 0-471-11184-8, Bendig, 1988. Culturing cells involves allowing the cells to metabolize and / or grow and / or divide and / or produce the protein of interest. This can be accomplished by methods known to those skilled in the art, including but not limited to providing nutrients to the cells. These methods include growing attached to a surface, growing in suspension, or a combination thereof.

[0117] Some culture conditions can be optimized by known methods to optimize protein production. Culture can be performed, for example, in dishes, roller bottles, or reactors, by batch, fed-batch, continuous, or hollow fiber culture. For large-scale (continuous) recombinant protein production by cell culture, it is known that it is preferable to grow cells in suspension. It is also known that it is preferable to culture cells under conditions free of animal- or human-derived serum or serum components derived from animals or humans. This elimination of additional animal- or human-derived proteins from the culture medium facilitates purification and enhances safety. Meanwhile, the use of a synthetic medium provides optimal reproducibility, making the system highly reliable.

[0118] Ig-like molecules are expressed in host cells and recovered from the cells, or preferably from the cell culture medium, by methods generally known to those skilled in the art. After recovery, these Ig-like molecules can be purified using known methods, including immunoprecipitation, centrifugation, filtration, size exclusion chromatography, affinity chromatography, cation and / or anion exchange chromatography, hydrophobic interaction chromatography, and the like. For antibody mixtures containing IgG molecules, Protein A or Protein G affinity chromatography can be suitably used (see, e.g., U.S. Patents 4,801,687 and 5,151,504).

[0119] The Ig-like molecules and / or mixtures thereof produced by the methods of the present invention preferably share a common light chain. Thus, further provided is a method of the present invention further comprising providing the host cell with a nucleic acid molecule encoding a common light chain. This light chain is capable of pairing with at least two different heavy chains, thereby forming a functional antigen-binding domain. The functional antigen-binding domain is capable of specifically binding to one antigen.

[0120] It is preferable to use a common light chain that can pair with all heavy chains produced by the method of the present invention. This avoids mispairing of incompatible heavy and light chains in the formation of a functional antigen-binding domain. In one embodiment, a common light chain with only one identical amino acid sequence can be used. Alternatively, those skilled in the art will recognize that the term "common" includes functionally equivalent light chains even if they do not have identical amino acid sequences. Many variants of the light chain exist, which contain mutations (deletions, substitutions, additions) that do not substantially affect the formation of a functional binding region. Therefore, such variants can also bind to different heavy chains to form functional antigen-binding domains.

[0121] As used herein, the term "common light chain" refers to a light chain that is identical or has a different amino acid sequence, but which retains the binding specificity of the resulting antibody after pairing with a heavy chain. For example, by introducing and testing conservative amino acid changes and / or amino acid changes in regions that do not contribute, or only partially contribute, to the binding specificity of the heavy chain, it is possible to create or find functionally equivalent, if not identical, light chains.

[0122] The term "shared light chain" includes combinations of specific shared light chains and functionally equivalent variants thereof. A detailed description of the use of shared light chains is provided in WO 2004 / 009618. Preferably, the shared light chain used in the present invention is a germline-like light chain, more preferably a germline-derived light chain, preferably a rearranged human germline κ light chain, and most preferably a rearranged human germline κ light chain IgVκ1-39 / Jκ or IGVκ3-20 / Jκ.

[0123] Alternatively, instead of using a common light chain, and to avoid mispairing of incompatible heavy and light chains, one skilled in the art may opt for a means of forced pairing of heavy and light chains, e.g., as described in WO 2009 / 080251, WO 2009 / 080252 and / or WO 2009 / 080253.

[0124] The present invention provides novel engineered CH3 domains as well as novel combinations of engineered CH3 mutations. Prior to the present invention, charged contact amino acids in the CH3 domain known to be involved in CH3-CH3 pairing were replaced with amino acids of the opposite charge (charge reversal), thereby affecting CH3-CH3 pairing.

[0125] The present invention provides an alternative to this approach because an uncharged or neutral CH3 amino acid is substituted with a charged residue in the wild-type CH3. In this embodiment of the invention, rather than replacing a charged contact amino acid with an amino acid of the opposite charge, an uncharged CH3 amino acid is substituted with a charged one.

[0126] The approach of the present invention not only provides a method for efficiently promoting CH3 domain dimerization, but also has the advantage of creating at least one additional charge-charge interaction at the CH3 interface. This additional charge-charge interaction, in addition to the charge pair present at the CH3-CH3 interface, makes the dimers of the present invention generally more stable compared to wild-type dimers (which are defined as bispecific IgGs (AB) without CH3 engineering, in contrast to the parent homodimers (AA or BB)).

[0127] Surprisingly, it is also possible to further increase the percentage of one or more desired Ig-like molecules in a mixture. As noted above, known methods for preferentially producing bispecific antibodies generally result in the production of undesired dimeric by-products. For example, using knob-into-hole technology, the percentage of desired bispecific antibodies is at most 87%, whereas electrostatic engineering approaches, in which charged contact amino acids are replaced with amino acids of the opposite charge, result in a percentage of 96% (see, e.g., Example 11).

[0128] Quite surprisingly, the present inventors have been able to introduce mutations that further increase the proportion of desired Ig-like molecules in the mixture. For example, Example 17 shows that a method using the mutations of the present invention can produce such a high proportion of desired bispecific antibodies that no dimeric by-products are detectable in the resulting mixture. Although some unpaired half-molecules, in which only one heavy chain is paired with a common light chain, are present in the mixture, these are the result of unequal expression of the heavy chains and can be easily separated from the mixture by size exclusion chromatography.

[0129] Such mutations according to the present invention therefore result in the production of high yields of bispecific Ig-like molecules in a single cell, substantially free of contaminating dimeric by-products, making them particularly suitable for pharmaceutical compositions.

[0130] A preferred embodiment of the present invention provides a method for producing a heterodimeric Ig-like molecule from a single cell, the Ig-like molecule comprising two CH3 domains capable of forming an interface, the method comprising the steps of: providing to the cell: (a) a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and (b) a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises at least one substitution of a neutral amino acid residue with a positively charged amino acid residue, and the polypeptide chain comprising the second CH3 domain comprises at least one substitution of a neutral amino acid residue with a negatively charged amino acid residue, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules and recovering the heterodimeric Ig-like molecule from the culture. Preferably, the method further comprises the step of providing to the host cell a nucleic acid molecule encoding a common light chain, the advantages of which are outlined above.

[0131] The amino acid at position 366 in one CH3 domain and the amino acid at position 351 in the other CH3 domain are reported to form a pair of contact residues at the CH3-CH3 interface, i.e., they are located close enough in the three-dimensional conformation of the resulting Ig-like molecule to interact with each other. Thus, the first CH3 domain preferentially pairs with the second CH3 domain.

[0132] In one embodiment, the threonine (T) at position 366 of the first CH3 domain is replaced with a first charged amino acid, and the leucine (L) at position 351 of the second CH3 domain is replaced with a second charged amino acid, wherein the first and second charged amino acids have opposite charges. If a polypeptide comprising a first CH3 domain bearing a charged residue at position 366 further comprises a variable region with specificity for antigen A, and if a polypeptide comprising a second CH3 domain bearing a charged residue of the opposite charge at position 351 further comprises a variable region with specificity for antigen B, a bispecific Ig-like molecule with predominantly AB specificity is formed.

[0133] In a method further provided by the present invention, the means for selective pairing of polypeptides comprising the first and second CH3 domains or the means for selective pairing of polypeptides comprising the third and fourth CH3 domains is a substitution of threonine at position 366 of the first or third CH3 domain with a first charged amino acid and a substitution of leucine at position 351 of the second or fourth CH3 domain with a second charged amino acid, wherein the first and second charged amino acids have opposite charges.

[0134] One preferred combination of mutations according to the present invention is a substitution of threonine (T) for lysine (K) at position 366 in a polypeptide comprising a first CH3 domain and further comprising a variable region (e.g., specific for A), and a substitution of leucine (L) for aspartic acid (D) at position 351 in a polypeptide comprising a second CH3 domain and further comprising a variable region (e.g., specific for B). This is designated the mutation pair T366K / L351'D.

[0135] As mentioned above, the amino acids at position 366 of one CH3 domain and position 351 of the second CH3 domain are reported to be a pair of contact residues at the CH3-CH3 interface. The lysine introduced at position 366 and the aspartic acid introduced at position 351 have opposite charges, and these amino acids electrostatically attract each other. Therefore, the first CH3 domain selectively attracts the second CH3 domain. Furthermore, Ig-like molecules are predominantly formed by pairs of the first CH3 domain with a lysine at position 366 and the second CH3 domain with an aspartic acid at position 351.

[0136] If a polypeptide comprising a first CH3 domain has specificity for antigen A and if a polypeptide comprising a second CH3 domain has specificity for antigen B, then an "AB"-specific bispecific Ig-like molecule will predominantly be formed. Note that in some embodiments, the variable regions of the polypeptide chains comprising the first and second CH3 domains may both be identical, resulting in the formation of a monospecific Ig-like molecule (e.g., "AA" specificity).

[0137] As mentioned above, one advantage of the mutations of the present invention is that instead of replacing the interactions of the original charged amino acids, novel interactions between newly introduced pairs of charged amino acids are generated, which has not been previously disclosed or suggested.

[0138] One aspect of the present invention provides a method according to the present invention for producing at least two different Ig-like molecules from a single host cell, wherein the polypeptide chain comprising the first CH3 domain comprises the amino acid substitution T366K and the polypeptide chain comprising the second CH3 domain comprises the amino acid substitution L351D.

[0139] One embodiment provides a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, the method comprising providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises the amino acid substitution T366K and the polypeptide chain comprising the second CH3 domain comprises the amino acid substitution L351D, the method further comprising culturing the host cell to express the nucleic acid molecules and recovering the heterodimeric Ig-like molecule from the culture.

[0140] The amino acid mutations described above according to the present invention allow for the production of heterodimeric Ig-like molecules from a single cell, with less than 5%, preferably less than 2%, more preferably less than 1%, or most preferably substantially no homodimeric contamination.

[0141] One embodiment provides a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, and wherein the presence of contaminating homodimers is less than 5%, preferably less than 2%, more preferably less than 1%, and most preferably substantially no homodimer contamination, the method comprising the steps of providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises the amino acid substitution T366K and the polypeptide chain comprising the second CH3 domain comprises the amino acid substitution L351D, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules and recovering the heterodimeric Ig-like molecule from the culture.

[0142] Preferably, in the method of producing at least two different Ig-like molecules according to the invention, or in the method of producing a heterodimeric Ig-like molecule according to the invention, the polypeptide chain comprising the first CH3 domain further comprises the amino acid substitution L351K. Even more preferably, the polypeptide chain comprising the second CH3 domain further comprises an amino acid substitution selected from the group consisting of Y349E, Y349D, and L368E. Most preferably, the polypeptide chain comprising the second CH3 domain further comprises the amino acid substitution L368E.

[0143] Thus, in one preferred embodiment, the T366K / L351'D mutation described above according to the invention can be further combined with a substitution of leucine (L) to glutamic acid (E) at position 368 of the second CH3 domain, which can be designated, for example, as a T366K / L351'D,L368'E mutation (although it can also be designated in other ways, such as T366K / L351D-L368E or T366K / L351D,L368E or T366K-L351D,L368E).

[0144] As shown in Example 17, the introduction of this mutation into a polypeptide according to the present invention comprising a first CH3 domain specific for antigen A and a second CH3 domain specific for antigen B allows the production of a particularly high percentage of bispecific Ig-like molecules with dual AB specificity. This pair of mutations allows the production of bispecific antibodies without the formation of appreciable amounts of homodimers.

[0145] One preferred embodiment provides a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, and wherein the presence of contaminating homodimers is less than 5%, preferably less than 2%, more preferably less than 1%, and most preferably substantially free of contaminating homodimers, the method comprising providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises the amino acid substitution T366K and the polypeptide chain comprising the second CH3 domain comprises the amino acid substitutions L351D and L368E, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules and recovering the heterodimeric Ig-like molecule from the culture.

[0146] In another preferred embodiment, the threonine (T) at position 366 in the first CH3 domain is substituted with a lysine (K), the leucine (L) at position 351 in the second CH3 domain is substituted with an aspartic acid (D), and the tyrosine (Y) at position 349 in said second CH3 domain is substituted with a glutamic acid (E), e.g., designated as a T366K / L351'D,Y349'E mutation, as well as, e.g., T366K-L351D:Y349E or T366K / L351D,Y349E or simply T366K / L351DY349E.

[0147] The Y349 residue is a neighboring residue at position 351 that may contribute to dimer interactions. In silico data indicate that Y349E not only destabilizes the single dimer (higher calculated score) but also stabilizes the heterodimer (lower calculated score), and glutamic acid (E) is more preferred than aspartic acid (D) at position 349. Thus, introducing a second amino acid substitution into a polypeptide containing a second CH3 domain that already has an amino acid substitution at position 351 favors heterodimerization.

[0148] A particularly preferred embodiment provides a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, and wherein the Ig-like molecule is less than 5%, more preferably less than 2%, even more preferably less than 1%, and most preferably substantially free of homodimer contamination, the method comprising providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises the amino acid substitution T366K, and the polypeptide chain comprising the second CH3 domain comprises the amino acid substitutions L351D and Y349E, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules, and recovering the heterodimeric Ig-like molecule from the culture.

[0149] In another preferred embodiment, the threonine (T) at position 366 in the first CH3 domain is substituted with a lysine (K), the aspartic acid (D) at position 351 in the second CH3 domain is substituted with a leucine (L), the tyrosine (Y) at position 349 in the second CH3 domain is substituted with a glutamic acid (E), and the leucine (L) at position 368 in the second CH3 domain is substituted with a glutamic acid (E). This is designated as the T366K / L351'D, Y349'E, L368'E mutation. The two residues Y349 and L368 are residues that may contribute to dimer interactions.

[0150] In silico data show that Y349E and L368E not only destabilize the B-chain dimer (higher in silico score) but also stabilize the heterodimer (lower in silico score), with glutamic acid at positions 349 and 368 being preferred over aspartic acid (D). Thus, introducing a second and third amino acid substitution into the B-chain, which already has an amino acid substitution at position 351, further promotes heterodimer formation.

[0151] A particularly preferred embodiment provides a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, and wherein homodimer contamination is less than 5%, more preferably less than 2%, even more preferably less than 1%, and most preferably substantially absent, the method comprising the steps of providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises the amino acid substitution T366K, and the polypeptide chain comprising the second CH3 domain comprises the amino acid substitutions L351D, Y349E, and L368E, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules, and recovering the heterodimeric Ig-like molecule from the culture.

[0152] In another preferred embodiment, the threonine (T) at position 366 in the first CH3 domain is substituted with a lysine (K), the leucine (L) at position 351 in the first CH3 domain is substituted with a lysine (K), the leucine (L) at position 351 in the second CH3 domain is substituted with an aspartic acid (D), and the leucine (L) at position 368 in the second CH3 domain is substituted with a glutamic acid (E). This is designated the T366K,L351K / L351'D,L368'E mutation. As shown in the Examples, this mutation also increases the percentage of the desired (bispecific) antibody. Furthermore, this mutation allows the production of bispecific antibodies without detectable homodimer formation.

[0153] Further provided is a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, and wherein homodimer contamination is less than 5%, more preferably less than 2%, even more preferably less than 1%, and most preferably substantially absent, the method comprising the steps of providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises amino acid substitutions T366K and L351K, and the polypeptide chain comprising the second CH3 domain comprises amino acid substitutions L351D and L368E, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules, and recovering the heterodimeric Ig-like molecule from the culture.

[0154] In another preferred embodiment, the threonine (T) at position 366 in the first CH3 domain is substituted with a lysine (K), the leucine (L) at position 351 in the first CH3 domain is substituted with a lysine (K), the leucine (L) at position 351 in the second CH3 domain is substituted with an aspartic acid (D), the tyrosine (Y) at position 349 in the second CH3 domain is substituted with an aspartic acid (D), and the arginine (R) at position 355 in the second CH3 domain is substituted with an aspartic acid (D). This is designated the T366K, L351K / L351'D, Y349'D, R355'D mutation. This T366K-L351K / L351'D-Y349'D pair is further improved by the R355'D mutation in the B-chain, which increases the in silico score for BB and slightly increases the in silico score for AB.

[0155] Further provided is a method for producing a heterodimeric Ig-like molecule from a single cell, wherein the Ig-like molecule comprises two CH3 domains capable of forming an interface, and wherein the presence of homodimers is less than 5%, more preferably less than 2%, even more preferably less than 1%, and most preferably substantially absent, the method comprising the steps of providing to the cell: a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain; and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain, wherein the polypeptide chain comprising the first CH3 domain comprises amino acid substitutions T366K and L351K, and the polypeptide chain comprising the second CH3 domain comprises amino acid substitutions L351D, Y349D, and R355D, the method further comprising the steps of culturing the host cell to express the two nucleic acid molecules, and recovering the heterodimeric Ig-like molecule from the culture.

[0156] Table B lists mutations that are introduced into the CH3 domain as a preferred means of selective pairing for the generation of heterodimers or homodimers. [Table B]

[0157] A method according to the invention for producing at least two different Ig-like molecules, or a method according to the invention for producing a heterodimeric Ig-like molecule, is provided whereby said means for selective pairing of polypeptides comprising said first and second CH3 domains and / or said means for selective pairing of polypeptides comprising said third and fourth CH3 domains comprises at least one combination of mutations set out in Table B. Preferably, said means for selective pairing of polypeptides comprising said first and second CH3 domains and said means for selective pairing of polypeptides comprising said third and fourth CH3 domains comprise at least two combinations of mutations set out in Table B.

[0158] The novel CH3 mutation combinations provided by the present invention allow the production of a mixture of at least two monospecific Ig-like molecules in a single cell, with less than 5%, preferably more than 2%, more preferably less than 1%, and most preferably substantially no bispecific Ig-like molecule contamination. These mutations according to the present invention are particularly suitable for the production of mixtures of monospecific antibodies. This is particularly useful when a high level of cross-linking of two identical target molecules is desired, when a sufficiently high density of antibody on the target cell is required to activate a specific effector mechanism, such as complement-mediated lysis of tumor cells, or when the two targets are too far apart to be bound by a single bispecific antibody, or to simplify the process of obtaining regulatory approval.

[0159] In such cases, it is often desirable to optimize the production platform for such monospecific antibodies. As shown in Example 10 and insights gained from the present invention, when a polypeptide comprising a first CH3 domain (e.g., having A specificity) is substituted for aspartic acid (D) at position 392, a polypeptide comprising said first CH3 domain is substituted for aspartic acid (K), and a polypeptide comprising said first CH3 domain is substituted for aspartic acid (D), and a polypeptide comprising said first CH3 domain is substituted for aspartic acid (K), and a polypeptide comprising said first CH3 domain is substituted for aspartic acid (D), a mixture of at least two different monospecific Ig-like molecules, including a monospecific Ig-like molecule with AA specificity, can be produced in a single cell, with the formation of bispecific by-products (bispecific Ig-like molecules) reduced to less than 5%, or less than 3%, or even to a level where they are not detected at all.

[0160] Therefore, the above-mentioned combination of mutations (designated K392D, D399K, K409D) is particularly preferred for producing a mixture of monospecific Ig-like molecules. Those skilled in the art will recognize that functional variants, i.e., K392E, D399R, K409E, can achieve similar results. In addition, double variants with D399K and K409D substitutions, or K392D and K409D, D399R and K409E, etc., can achieve similar results.

[0161] The same applies to the combination of mutations where a glutamic acid (E) at position 356 in the polypeptide comprising the first CH3 domain is substituted with a lysine (K), a glutamic acid (E) at position 357 in the polypeptide comprising the first CH3 domain is substituted with a lysine (K), a lysine (K) at position 439 in the polypeptide comprising the first CH3 domain is substituted with an aspartic acid (D), and a lysine (K) at position 370 in the polypeptide comprising the first CH3 domain is substituted with an aspartic acid (D). This combination of mutations (designated E356K, E357K, K439D, K370D) is also particularly preferred for producing mixtures of monospecific Ig-like molecules.

[0162] Those skilled in the art will recognize that functional mutants, i.e., K356R, E357R, K439E, and K370E, can result in similar effects. In addition, triple or double mutants with E356K and K439D, E357K and K370D substitutions, or other functional mutants, may also result in similar effects.

[0163] A further embodiment provides a method for producing at least two different monospecific Ig-like molecules from a single host cell, each of the two Ig-like molecules comprising two CH3 domains capable of forming an interface, the method comprising the steps of providing to the cell a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain having -A specificity and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain having -B specificity, wherein the polypeptide chain comprising the first CH3 domain comprises mutations K392D, D399K, and K409D, and the polypeptide chain comprising the second CH3 domain comprises a wild-type CH3 domain or comprises mutations E356K, E357K, K439D, and K370D, the method further comprising the steps of culturing the host cell to express the nucleic acid molecules, and recovering the at least two different Ig-like molecules from the culture.

[0164] Another embodiment provides a method for producing at least two different monospecific Ig-like molecules from a single host cell, each of the two Ig-like molecules comprising two CH3 domains capable of forming an interface, the method comprising the steps of providing to the cell a first nucleic acid molecule encoding a polypeptide chain comprising a first CH3 domain having -A specificity and a second nucleic acid molecule encoding a polypeptide chain comprising a second CH3 domain having -B specificity, wherein the polypeptide chain comprising the first CH3 domain comprises a wild-type CH3 domain or comprises mutations K392D, D399K, and K409D, and the polypeptide chain comprising the second CH3 domain comprises mutations E356K, E357K, K439D, and K370D, the method further comprising the steps of culturing the host cell to express the nucleic acid molecules and recovering the at least two different Ig-like molecules from the culture.

[0165] As shown in Example 10, two monospecific Ig-like molecules are produced in a single cell and the formation of bispecific Ig-like molecules is substantially undetectable. One skilled in the art can select a third nucleic acid molecule encoding a polypeptide chain comprising a wild-type or engineered CH3 domain and provide it to the host cell such that a mixture of three monospecific antibodies is produced.

[0166] In one aspect of the present invention, a method for producing at least two different Ig-like molecules or a method for producing a heterodimeric Ig-like molecule according to the present invention is provided, wherein each polypeptide chain comprising a CH3 domain further comprises a variable region that recognizes a different target epitope, and the target epitopes are located on the same molecule.

[0167] This allows for more efficient antagonism of the target molecule's (biological) function than when only one epitope is targeted. For example, a heterodimeric Ig-like molecule can simultaneously bind to two epitopes present on a growth factor receptor or a soluble molecule important for tumor cell proliferation. This can effectively inhibit several independent signaling pathways, leading to uncontrolled proliferation. Any combination of at least two Ig-like molecules can simultaneously bind to two, three, or four epitopes present on such growth factor receptors or soluble molecules.

[0168] In one preferred embodiment, the target molecule is a water-soluble molecule, while in another preferred embodiment, the target molecule is a membrane-bound molecule.

[0169] In another embodiment of the invention, in a method for producing at least two different Ig-like molecules or heterodimeric Ig-like molecules according to the invention, each polypeptide chain comprising a CH3 domain further comprises a variable region recognizing a target epitope, the target epitope being located on a different molecule, wherein each different target molecule may be a water-soluble molecule or a membrane-bound molecule.

[0170] In one embodiment, the different target molecules are water-soluble molecules. Alternatively, one target molecule is a water-soluble molecule, while the second target molecule is a membrane-bound molecule. In yet another case, both target molecules are membrane-bound molecules. In one embodiment, the different target molecules are expressed in the same cell, while in another embodiment, the different target molecules are expressed in different cells.

[0171] As non-limiting examples, any heterodimeric Ig-like molecule or any combination of at least two Ig-like molecules is suitable for simultaneously inhibiting multiple membrane-bound receptors, simultaneously neutralizing multiple water-soluble molecules such as cytokines or growth factors to tumor cells, or neutralizing different viral serotypes or strains.

[0172] In one preferred embodiment, the method of the present invention for producing at least two Ig-like molecules or heterodimeric Ig-like molecules provides that at least one of the target epitopes is located on a tumor cell. Alternatively, or in addition, at least one of the target epitopes is located on the surface of an effector cell, which is suitable for recruiting T cells or NK cells to kill tumor cells, for example. For example, at least one Ig-like molecule produced by the method of the present invention can recruit immune effector cells, preferably human immune effector cells, by specifically binding to a target molecule located on the immune effector cell.

[0173] In a further embodiment, the immune effector cells are activated after the Ig-like molecule binds to a target molecule. Induction of effector mechanisms includes, for example, the re-direction of immune-regulated cytotoxicity by the Ig-like molecule produced by the method of the present invention. The Ig-like molecule can bind to molecules that cause cytotoxicity, such as T cell receptors or Fcγ receptors, thereby activating downstream immune effector pathways.

[0174] As used herein, the term "immune effector cells" or "effector cells" refers to a repertoire of natural cell populations of the mammalian immune system that, upon activation, affect the viability of target cells. Immune effector cells include not only lymphoid cells such as natural killer (NK) cells, T cells, including cytotoxic T cells, or B cells, but also myeloid lineage cells such as monocytes, macrophages, dendritic cells, and neutrophilic granulocytes. Thus, the effector cells are preferably NK cells, T cells, B cells, monocytes, macrophages, dendritic cells, or neutrophilic granulocytes.

[0175] Target antigens present on immune effector cells include CD3, CD16, CD25, CD28, CD64, CD89, NKG2D, and NKp46. Also provided are methods for producing at least two different Ig-like molecules or heterodimeric Ig-like molecules according to the invention, wherein the target epitopes are located on CD3, CD16, CD25, CD28, CD64, CD89, NKG2D, or NKp46 molecules. Target cell viability includes the cell's ability to survive, proliferate, and / or interact with other cells.

[0176] One aspect of the present invention provides a method for producing a heterodimeric Ig-like molecule according to the present invention, wherein each polypeptide chain comprising a CH3 domain further comprises a variable region that recognizes a target epitope. In one embodiment, the two variable regions of each polypeptide chain comprising a CH3 domain recognize the same target epitope with different affinities. In another embodiment, the two variable regions of each polypeptide chain comprising a CH3 domain recognize different target epitopes.

[0177] In other embodiments, the different target epitopes are located on the same target molecule, which can be a membrane-bound molecule or a water-soluble molecule. In other embodiments, the different target epitopes are located on different target molecules, which are expressed on the same cell or different cells. Alternatively, the different target molecules can be water-soluble molecules, or one target molecule can be a water-soluble molecule and the second target molecule can be a membrane-bound molecule.

[0178] In a preferred embodiment, at least one target molecule of the heterodimeric Ig-like molecule is located on a tumor cell. In yet another preferred embodiment, at least one target molecule of the heterodimeric Ig-like molecule is located on an effector cell (i.e., an NK cell, a T cell, a B cell, a monocyte, a macrophage, a dendritic cell, or a neutrophil granulocyte, and the target epitope is located on a CD3, CD16, CD25, CD28, CD64, CD89, NKG2D, or NKp46 molecule).

[0179] A preferred embodiment provides a method for producing at least two different Ig-like molecules or a method for producing a heterodimeric Ig-like molecule according to the present invention, wherein, as described above, the at least two different Ig-like molecules are antibodies, most preferably antibodies of the IgG isotype, and even more preferably antibodies of the IgG1 isotype.

[0180] Further provided is an Ig-like molecule, a heterodimeric Ig-like molecule, or a mixture of at least two Ig-like molecules obtainable by a method according to the invention. The (heterodimeric) Ig-like molecule or mixture of Ig-like molecules preferably comprises at least one CH3 mutation listed in Table B. Pharmaceutical compositions according to the invention comprising at least one Ig-like molecule or a mixture of at least two Ig-like molecules, as well as (heterodimeric) Ig-like molecules or mixtures of at least two Ig-like molecules comprising at least one mutation listed in Table B, are also provided.

[0181] In one embodiment, the Ig-like molecule is a bispecific Ig-like molecule, such as a bispecific antibody. In another embodiment, the Ig-like molecule is a monospecific Ig-like molecule, such as a monospecific antibody. A preferred embodiment provides a mixture of at least two different Ig-like molecules obtainable by a method according to the present invention, wherein the at least two different Ig-like molecules bind to different epitopes on the same antigen and / or different epitopes on different antigens.

[0182] Further provided are heterodimeric Ig-like molecules obtainable by the methods of the present invention, said heterodimeric Ig-like molecules binding to different epitopes on the same antigen and / or different epitopes on different antigens, the advantages and preferred uses of which have been described above for the mixtures and antibodies.

[0183] The present invention provides a mixture of at least two different Ig-like molecules obtainable by a method according to the invention, wherein said at least two different Ig-like molecules comprise at least one heterodimeric Ig-like molecule. In one embodiment, two of said at least two different Ig-like molecules are heterodimeric Ig-like molecules.

[0184] Yet another preferred embodiment provides a heterodimeric antibody comprising two CH3 domains, one of which comprises the amino acid substitutions L351D and L368E and the other of which comprises the amino acid substitutions T366K and L351K, which, as described above, are preferred means for selective pairing of the two CH3 domains.

[0185] The amino acid substitutions L351D and L368E in one of the two CH3 domains and the amino acid substitutions T366K and L351K in the other of the two CH3 domains are collectively referred to as the "DEKK combination mutation," "DEKK mutant," "DEKK pair," "DEKK-engineered CH3 domain," "DEKK," or other name referring to DEKK. The CH3 domain carrying the amino acid substitutions L351D and L368E is also referred to as the "DE side," and the CH3 domain carrying the amino acid substitutions T366K and L351K is also referred to as the "KK side."

[0186] Also provided is a pharmaceutical composition comprising a (heterodimeric) Ig-like molecule or a mixture of at least two Ig-like molecules obtainable by any method according to the invention. The (heterodimeric) Ig-like molecule or the at least two Ig-like molecules according to the invention are preferably antibodies. The pharmaceutical composition comprises the (heterodimeric) Ig-like molecule, a mixture comprising monospecific or bispecific Ig-like molecules, or a combination of monospecific and bispecific Ig-like molecules.

[0187] The pharmaceutical compositions of the present invention also include a pharmaceutically acceptable carrier. As used herein, such a "pharmaceutically acceptable carrier" includes any and all solvents, salts, dispersion media, coatings, antibacterial and antifungal agents, isotonic and absorption delaying agents, and similar physiologically compatible substances. Depending on the route of administration (e.g., intravenous, subcutaneous, intraarticular, etc.), the Ig-like molecules may be coated with a material to protect them from the action of acids and other natural conditions that may inactivate the Ig-like molecules.

[0188] In one aspect, provided is a pharmaceutical composition comprising a mixture of at least two Ig-like molecules obtained by any method according to the invention, wherein said at least two different Ig-like molecules are produced by a recombinant host cell according to the invention. Additionally, provided is a pharmaceutical composition comprising a heterodimeric Ig-like molecule obtained by any method according to the invention, wherein said heterodimeric Ig-like molecule is produced by a recombinant host cell according to the invention.

[0189] In addition to nucleic acid molecules encoding a polypeptide chain comprising a CH3 domain comprising at least one mutation described in Table B, recombinant host cells comprising at least one nucleic acid molecule encoding a polypeptide chain comprising a CH3 domain comprising at least one mutation described in Table B are also provided.

[0190] The present invention is further illustrated by the following examples, which are not intended to limit the invention and are presented merely to clarify the invention. [Brief explanation of the drawings]

[0191] [Figure 1] Figure 1A) is a schematic diagram of the construct vector MV1057. The stuffer region is the region into which the VH region of an antibody is cloned. Figure 1B) is a schematic diagram of the phage display vector MV1043. [Figure 2] This is the amino acid sequence of wild-type IgG1 Fc present in construct vector MV1057 (EU numbering system applied). [Figure 3] Nucleotide and amino acid sequences of the VH region used for cloning into the various constructs. [Figure 4A] Mass spectrum data of transfection A. [Figure 4B] Mass spectrum data of transfection G. [Figure 4C] Mass spectrum data of transfection H. [Figure 5A] Mass spectrum data of transfection M. [Figure 5B] Mass spectrum data of transfection U. [Figure 6] Mass spectrum data of transfection O. [Figure 7A] Inhibition of homodimerization by substituting neutral amino acids with charged amino acids. [Figure 7B] Inhibition of homodimerization by substituting neutral amino acids with charged amino acids. [Figure 8] Figure 8(A) shows the native mass spectrometry (MS) spectrum of the transfected sample ZO(T366K / L351'D). Figure 8(B) shows the convoluted MS spectrum of the transfected sample ZO(T366K / L351'D). The second / major peak is that of the bispecific molecule. [Figure 9] HADDOCK scores for experimentally validated mutation pairs. [Figure 10] Illustrated diagrams of CH3-CH3 interface interactions: Figure 10A) for K409D:K392D / D399'K:E356'K, Figure 10B) for D399K:E356K / D399'K:E356'K, and Figure 10C) for K409D:K392D / K409'D:K392'D. [Figure 11] HADDOCK scores of various 366 / 351' charge variants. [Figure 12] Illustrated diagrams of CH3-CH3 interface interactions. Figure 12A) is for L351D / L351'D, and Figure 12B) is for L351D:S354A:R355D / L351'D:S354'A:R355'D. [Figure 13] HADDOCK scores for additional charge mutations near position L351. [Figure 14] HADDOCK scores for additional charge mutations near position T366 in strand A and position L351 in strand B. [Figure 15] This is a diagram of the CH3-CH3 interface interactions. [Figure 16] HADDOCK scores for variants near T366 / L351. [Figure 17] HADDOCK scores for additional variants near T366 / L351. [Figure 18]Example of nMS spectra of bispecific IgG obtained after co-expression of constructs T366K,L351K with constructs L351D (left panel) or L351D,Y349E (right panel), scaled based on the monovalent full-length IgG (half not shown). [Figure 19A] Mass spectrometry results showing the relative abundance of AA, AB, BB, A, and B (total of all species is 100%). [Figure 19B] The data is the same as in FIG. 19A, but AB is omitted to make it easier to see the appearance of the undesirable AA, BB, A, and B. [Figure 20] Results of thermostability assay. Squares: wild type, triangles: charge-reversed pair E356K:D399K / K392D:K409D, circles: combinations of mutated CH3s shown above each graph. [Figure 21] Results from 10 freeze-thaw experiments. 1122 = primary parent antibody BB; 1337 = secondary parent antibody AA; wild type = AA,AB,BB; CR = bispecific antibody with charge-reversed pair E356K:D399K / K392D:K409D; 3-6 and 9-12 = bispecific molecule combinations 3-6 and 9-12 in Table 15. [Figure 22] Serum stability was measured by ELISA using fibrinogen as the coated antigen. Figure 22A) shows the ELISA data when the IgG sample was diluted to 0.5 μg / ml. Figure 22B) shows the ELISA data when the IgG sample was diluted to 0.05 μg / ml. Results were normalized to 100% at T=0 (day). 1337 = second parent antibody AA; wild type = AA, AB, BB; CR = bispecific antibody with charge-reversed pair E356K:D399K / K392D:K409D; 3-6 and 9-12 = bispecific molecule combinations 3-6 and 9-12 in Table 15. [Figure 23A] nMS results for experiments with transfection ratios ranging from 1:5 to 5:1. Figure 23A shows the DEKK combination mutant, with "A" specificity on the DE side and "B" specificity on the KK side. [Figure 23B] nMS results for experiments with transfection ratios ranging from 1:5 to 5:1. Figure 23B shows the DEKK combination mutants, with the "C" specificity on the DE side and the "B" specificity on the KK side. [Figure 23C] nMS results for experiments with transfection ratios ranging from 1:5 to 5:1. Figure 23C shows charge reversal mutation combinations, with "A" specificity on the E356K:D399K side and "B" specificity on the K392D:K409D side. [Figure 24] nMS results for transfections #1-11 from Table 20. [Figure 25] HADDOCK scores of dimers of different CH3-engineered vectors. Grey bars: desirable species AB and CD; black bars: undesirable species AA, BB, CC, DD, AC, BC, AD, BD. [Figure 26] SDS-PAGE of transfections #1-11 from Table 20. Control samples DE / KK, DE / DE, and KK / KK are included. [Figure 27A] nMS of transfection #9. [Figure 27B] nMS of transfection #11. [Figure 28A] This is nMS of gel filtered sample 1516:1516. [Figure 28B] This is nMS of gel filtered sample 1337:1337. [Figure 28C] nMS of gel filtered sample 1516:1337. [Figure 29] Blood concentrations of the DEKK-engineered antibody and its two parent antibodies (pK study) DETAILED DESCRIPTION OF THE INVENTION

[0192] Example 1: Amino acid substitutions to generate a variety of different CH3 domains To selectively promote or inhibit pairing of Ig-like molecules with CH3 domains and obtain a wide variety of Ig-like molecules with different CH3 domains, we introduced many amino acid substitutions known to promote heterodimer formation, as well as many other amino acid substitutions that have not previously been reported or tested but were selected to promote homodimer formation, into a construct vector (construct vector MV1057; Figure 1A).

[0193] Construct vector MV1057 contains the nucleic acid sequence encoding the Fc region of a normal wild-type IgG1, as depicted in Figure 2. Table 1 lists the amino acid substitutions introduced into this wild-type Fc, resulting in a series of seven constructs. All constructs were generated by Geneart. Constructs 1, 2, and 3, or their alternatives, have previously been reported to promote heterodimerization (European Patent Application Publication No. 01870459, International Publication No. WO 2009 / 089004). Constructs 6 and 7 have also been reported (WO 98 / 50431). Constructs 4 and 5 are new and are designed to promote homodimerization.

[0194] [Table 1]

[0195] Example 2: Cloning VH into constructs with CH3 mutations Several antibody VH regions with known specificities and known binding ability to the human IGKV1-39 light chain were used for cloning into these constructs. As mentioned above, all CH3 variants can be used with other antibody domains to create full-length antibodies with either bispecific or monospecificity. The antibody specificity determined by the VH / VL combination does not affect the heavy chain dimerization behavior induced by the CH3 domain. Model VH / VL combinations, in which all light chains are based on human germline IGKV1-39 and the VHs are varied, were used throughout this study.

[0196] Figure 3 shows the complete sequences and specificities of the antibody VH regions used throughout this study. The MF codes are internal designations for various VHs at Merus. For example, VH MF1337 has specificity for tetanus toxoid, MF1025 for porcine thyroglobulin, and MF1122 for bovine fibrinogen.

[0197] The VH region of the phage display vector MV1043 (Figure 1B) was digested with the restriction enzymes SfiI and BstEII (New England Biolabs, cat# R0123L and R0162L, according to the manufacturer's instructions) to excise the VH fragment from the vector. The vector MV1057 was digested with SfiI and BstEII by standard methods (according to the manufacturer's instructions).

[0198] The fragment and vector are gel purified (Promega / cat# V3125 / according to the manufacturer's instructions), and the excised vector and VH gene insert are isolated. Both are joined by ligation, and the ligated nucleic acid is then transformed into E. coli DH5α (Invitrogen / cat# 12297-016) (according to the manufacturer's instructions). The next day, single colonies are picked and sequenced to identify the vector with the appropriate insertion.

[0199] Example 3: Transfection and expression of full-length IgG in HEK293T cells The various plasmids encoding the recloned VH variants, as well as the common light chain, human IGKV1-39, were transfected into HEK293T cells using standard procedures (de Kruif et al. Biotech Bioeng. 2010) to allow IgG expression. After transfection, the IgG expression levels in the supernatant were measured using a ForteBIO Octet-QK system. This system is based on Bio-Layer Interferometry (BLI), allowing real-time quantification and dynamic characterization of biomolecular interactions. For details, see 1418274598331_0.com. If expression levels exceeding 5 μg / ml were measured, the IgG was purified using Protein A affinity purification.

[0200] Example 4: Purification of IgG Culture supernatants were purified using a Protein A column (GE Healthcare, cat# 11-0034-95, according to the manufacturer's instructions) and then eluted in 0.1 M citrate buffer, pH 3.0. The column was immediately neutralized with an equal volume of 1.0 M Tris-HCl, pH 8.0, or directly rebuffered in PBS using a desalting column. Alternatively, IgG could be purified using Protein A beads (Sepharose beads CL-4B, GE Healthcare, cat# 170780-01).

[0201] Example 5: Antigen-specific ELISA Antigen-specific ELISA is performed to assess the binding activity of the antigen. Antigen capture ELISA is performed to demonstrate the binding activity of bispecific antibodies. A biotinylated second antigen is used to detect the complex (de Kruif et al. Biotech Bioeng. 2010).

[0202] Example 6: SDS-PAGE The purified IgG mixture was analyzed by SDS-PAGE (NuPAGE® 4-12% bis-tris gel / Invitrogen / cat# NP0323BOX) under reducing and non-reducing conditions using standard procedures, and the proteins in the gel were stained with colloidal blue reagent (PageBlue™ protein staining solution / Fermentas / cat# RO571).

[0203] Example 7: Enzymatic deglycosylation of IgG1 Because IgG glycosylation is heterogeneous, deglycosylation was performed to obtain a single product with a distinct mass suitable for mass spectrometry analysis. One unit of N-glycosidase F (PNGase F; Roche Diagnostics, Mannheim, Germany) was added to 10 μg of IgG1 and incubated overnight at 37°C. To remove the original purification buffer (0.1 M citrate buffer pH 3.0 / 1.0 M Tris-HCl pH 8.0), buffer exchange was performed using a 10 kDa MWCO centrifugal filter column (Millipore) and rebuffered with PBS. To remove the separated glycan chains, a similar buffer exchange procedure was performed, with the buffer exchanged into 150 mM ammonium acetate, pH 7.5. The filter was then rinsed with 200 μl of 150 mM ammonium acetate, pH 7.5, for 12 minutes at 11,000 rpm and 4°C. After rinsing, 50 μl of deglycosylated IgG was applied to the filter, and 450 μl of 150 mM ammonium acetate, pH 7.5, was added. This was followed by another centrifugation step of 12 minutes at 11,000 rpm at 4°C. Fresh 150 mM pH 7.5 ammonium acetate buffer was added each time to a total volume of 500 μl, for a total of five centrifugations. After the final centrifugation step, approximately 25 μl of the remaining buffer-exchanged deglycosylated IgG1 was collected, transferred to an Eppendorf tube, and prepared for mass spectrometry analysis.

[0204] Example 8: Native Mass Analysis Mass spectrometry was used to identify different IgG species in purified IgG mixtures and determine the ratios of these IgG species present. Briefly, 2–3 μl of 150 mM ammonium acetate, pH 7.5, containing 1 μM IgG was injected into a gold-plated borosilicate capillary tube (using a Sutter P-97 puller [Sutter Instruments Co., Novato, CA, USA] and an Edwards Scancoat six sputtering system [Edwards Laboratories, Milpitas, CA, USA]) and analyzed on an LCT1 mass analyzer (Waters Corp., Milford, MA, USA) optimized for high-mass detection (Tahallah et al., RCM 2001). A capillary voltage of 1300 V and a sampling cone voltage of 200 V were used; however, these settings were adjusted if higher resolution signal-to-noise ratios were required. To facilitate collisional cooling, the source backing pressure was increased to approximately 7.5 mbar. To measure denatured IgG1, the protein was sprayed at a concentration of 1 μM in 5% formic acid.

[0205] Example 9: Data Processing and Quantification The acquired spectra were processed using MassLynx 4.1 software (Waters Corp., Milford, MA, USA). Minimal smoothing was used, and the spectra were centered. Using each series of charge states, the mass of the species was calculated. For each charge state, the corresponding intensity was assigned by MassLynx and summed. This approach allows for relative quantification of all species in a sample. Alternatively, peak quantification can be performed using the well-known area-under-the-curve (AUC) method. All analyses were repeated three times to calculate the standard deviation of the IgG mass and its relative abundance.

[0206] Example 10: Mixture of two or three monospecific antibodies from a single cell The VH regions of several antibodies with known specificities and known human IGKV1-39 light chain binding abilities (Figure 3) were recloned into wild-type construct vector MV1057 or construct 4 or construct 5 in Table 1, resulting in vectors I–III (Table 2). The resulting vectors I, II, and III, each containing nucleic acid sequences encoding Ig heavy chains with different CH3 regions and different VH specificities and a common human light chain, were then transfected into cells. Transfections were performed singly to observe the formation of intact monospecific antibodies, or in combination with one or two other construct vectors to obtain a mixture of two or three monospecific antibodies. Table 3 lists the transfections and results.

[0207] [Table 2]

[0208] [Table 3]

[0209] Transfections A, G, and H yielded bivalent, monospecific AA, BB, or CC antibodies that only formed homodimers from cells transfected with either vector I, II, or III (Figure 4). This was predicted and previously demonstrated for transfection A. For the first time, we report the homodimerization of CH3-engineered Ig heavy chains containing triple amino acid substitutions (i.e., K392D, D399K, K409D) or quadruple amino acid substitutions (i.e., E356K, E357K, K439D, K370D) (transfections G and H).

[0210] Next, co-expression experiments of the two vectors in single cells were performed. Interestingly, as shown by transfections M and N, co-expression of wild-type and CH3-engineered Ig heavy chains with a common light chain in single cells yielded a mixture of the two monospecific antibodies, without the presence of undesired bispecific antibodies, and with only 4–5% of contaminating "other molecules" present in the mixture. "Other molecules" are defined as all molecules that do not have the mass of intact IgG, including half-molecules consisting of a single heavy and light chain pair. Importantly, the "other" category does not include bispecific products.

[0211] In transfection M, equal proportions of vector DNA were transfected, resulting in an AA:BB ratio of approximately 1:1. However, in transfection N, the AA:CC ratio was approximately 10:1. Therefore, this transfection was repeated (transfection U) with adjusted DNA ratios. In fact, when vector DNA I:III was 1:5, the AA:CC ratio of the antibody products in the mixture was approximately 1:1. Therefore, as demonstrated by transfections M and U, two distinct, substantially pure monospecific antibodies can be expressed in a single cell without undesired by-products (i.e., no excessive presence of AC or half-molecules A or C) (Figure 5). The novel CH3 modifications in constructs 4 and 5 are substantially different from the wild-type CH3 to prevent heterodimerization between wild-type and construct 4 or between wild-type and construct 5. This is advantageous for the mass production of monospecific antibody mixtures from a single cell.

[0212] Similar to these results, two different CH3 engineered Ig heavy chains (constructs 4 and 5) are predicted to result in a mixture of only two different monospecific antibodies, without the presence of additional undesired species. If the CH3 modifications in construct 4 and construct 5 are substantially different, it is inferred that heterodimerization will not occur. In that case, co-expression of the CH3 engineered heavy chains of constructs 4 and 5 with the wild-type CH3 heavy chain in a single cell will result in only three monospecific antibodies.

[0213] Indeed, this result was observed. It was found that expressing three different Ig heavy chains designed to form homodimers rather than heterodimers with a common light chain in a single cell yielded a mixture of three pure monospecific antibodies without any other contaminants in the mixture (transfection O) (Figure 6). As can be clearly seen from Table 3, even when the same proportion of vector DNA was used in transfection O, the antibody AA:BB:CC ratio was not 1:1:1. Transfection with a different ratio of vector DNA (1:1:10, transfection V) demonstrated that the AA:BB:CC ratio in the mixture could be manipulated to the desired ratio. These experiments demonstrate that two or three substantially pure monospecific antibodies can be produced in a single cell without undesired by-products. This is advantageous for the large-scale production of mixtures of monospecific antibodies for therapeutic use.

[0214] Example 11: Mixture of two bispecific antibodies from a single cell The production of a single bispecific antibody using a CH3-engineered heavy chain has been reported elsewhere. Here, this experiment was designed to investigate the feasibility of producing a mixture of two different bispecific antibodies from a single cell. The VH regions of antibodies with known specificity and binding ability to the known human IGKV1-39 light chain (Figure 3) were recloned into vectors containing constructs 1-3 or 6-7 from Table 1, resulting in vectors IV-X (Table 4). Vectors IV-X, each containing nucleic acid sequences encoding Ig heavy chains with different CH3 regions and VH specificities and a common human light chain, were then transfected into cells. The transfections were performed singly to demonstrate the interference with intact monospecific antibody formation, or in combination with other construct vectors to obtain bispecific antibodies or mixtures of two bispecific antibodies. Table 5 lists the transfections and results.

[0215] [Table 4]

[0216] [Table 5]

[0217] It has been shown that the CH3-engineered Ig heavy chains encoded by constructs 1 and 2 retain the ability to form homodimers when expressed in single cells (WO 2009 / 089004). However, a further report in WO 2009 / 089004 showed that CH3 domains engineered with triple charge pair mutations, such as construct 3, no longer form homodimers when expressed alone. In this study, these findings were only partially confirmed. In fact, transfections B, C, and D showed the presence of full-length IgG in addition to a high proportion of unpaired half molecules, indicating homodimerization of the CH3 domains encoded by constructs 1 and 2. Transfections E and F also resulted in the production of full-length IgG in addition to unpaired half molecules, indicating that the triple charge mutation in construct 3 does not completely impair homodimerization. It was further shown that the “knob” and “hole” CH3 mutants of constructs 6 and 7 also formed homodimers (18% for “knob-knob” homodimers and 42% for “hole-hole” homodimers).

[0218] To prevent or minimize unwanted by-products (homodimers) when co-expressing a second CH3 mutant for heterodimer formation, CH3 mutants that completely inhibit homodimerization when expressed alone are preferred.

[0219] Interestingly, this experiment demonstrated for the first time that a mixture of bispecific antibodies could be expressed in a single cell, with virtually no homodimers in the mixture. Transfections K and L indeed yielded the predicted bispecific species BC + AB (38% + 47% in transfection K, 16% + 60% in transfection L). A relatively high percentage of undesired half molecules was observed in both transfections (15% half molecules A + half molecules C in transfection K, 24% half molecules A + half molecules C in transfection L). The relatively high percentage of half molecules still present was due to the low amount of heavy chain in vector IV, resulting from the unbalanced expression of the matched heavy chain. Therefore, in transfections S and T, the ratio of vector DNA was adjusted to 2:1:1 and transfections were repeated. This resulted in a pure bispecific IgG mixture with equal amounts of the matched IgG heavy chains, no IgG half molecules, and only 3% homodimeric BB. Ideally, this low percentage of monospecific product contamination should be reduced to essentially zero. It is therefore desirable to find additional CH3 variants that will result in mixtures of bispecific antibodies that minimize the presence of contaminating monospecific antibodies.

[0220] This study demonstrates, for the first time, that it is possible to produce a substantially pure mixture of two bispecific antibodies recognizing three different target epitopes in a single cell, while minimizing the presence of monospecific antibodies in the mixture.

[0221] Example 12: Various Mixtures Having shown that it is technically feasible to produce a mixture of two bispecific antibodies recognizing three epitopes, or a mixture of two or three monospecific antibodies from a single cell, we next explored the feasibility of controlled production of various other mixtures.

[0222] A fourth antibody VH region with known specificity and known binding ability to the human IGKV1-39 light chain was recloned into a vector containing constructs 1-3 or 7 in Table 1, resulting in vectors I', II', III', or X' (where "'" indicates a different specificity compared to the corresponding vector number). Vectors I'-III', X', and IV-IX, each containing nucleic acid sequences encoding Ig heavy chains with different CH3 regions and different VH specificities and a common human light chain, were then transfected into cells. The transfections were performed in combination with other construct vectors to obtain various mixtures of bispecific and / or monospecific antibodies. The various mixtures that could be obtained include a mixture of two bispecific antibodies recognizing four epitopes, two bispecific and one monospecific antibody, or one bispecific and one monospecific antibody from a single cell. Table 6 lists the transfections and predicted results.

[0223] [Table 6]

[0224] Although the production of all mixtures is theoretically feasible, other studies have shown that stability issues have hindered the mass production of previous knob-into-hole mutants, and therefore the mixtures resulting from transfections ZA, ZB, ZL, ZM, and ZN are expected to be problematic when moved to mass production.

[0225] Thus, the set of constructs in Table 1 may not be capable of producing all theoretical mixtures in large-scale production from single cells. This is because knob-into-hole mutants are reported to be unstable, and it cannot be excluded that the CH3 domains containing the "knobs" or "holes" may dimerize with either the charged mutants or the wild-type CH3 domain. Therefore, it is desirable to design new CH3 mutants for co-expression in single cells that are engineered to selectively form only homodimers or heterodimers and not homodimerize or heterodimerize with constructs 1–5 in Table 1.

[0226] Example 13: Identification of novel charge pair mutations The goal of this study was to engineer the CH3 region of IgG so that expression of a mixture of different IgG heavy chains in a single cell would result in the production of only heterodimers or only homodimers. Here, the new engineered CH3 domain would not homodimerize or heterodimerize with known engineered or wild-type CH3 domains. Therefore, as a first step in identifying new engineered CH3 domains that meet this criterion, we investigated the replacement of many contact residues in the interface of the IgG CH3 domain, either individually or in groups, with modifications that would result in electrostatic repulsion between identical heavy chains—i.e., reduced homodimer formation. The goal was to obtain a list of residues that, when replaced with charged residues, would result in repulsion between identical heavy chains, so that these mutations could be used to induce homo- and / or heterodimer formation when different IgG heavy chains are expressed together. This would ensure the stability of the resulting full-length IgG and ensure high production rates.

[0227] In further exploration, the identified mutations could be used to generate bispecific antibodies or mixtures of bispecific and monospecific antibodies by engineering compatible pairs of CH3 residues in one or more IgG heavy chain-CH3 regions. Furthermore, newly identified pairs of charge mutations could be combined with existing pairs and expressed in cells so that multiple nucleic acid molecules encoding different heavy chains all have different but complementary CH3 mutations. This allows for selective generation of mixtures of monospecific or bispecific antibodies, or mixtures of defined monospecific and bispecific antibodies. The residues tested in this study are previously identified contact residues (Deisenhofer J., 1981; Miller S., 1990; Padlan, 1996; Gunasekaran, 2010). The rationale for this approach is that repulsive charges are introduced at the contact residues of each potential pair.

[0228] The samples were then analyzed by SDS-PAGE under non-reducing conditions to identify pairs that reduced dimer formation by looking for a band at approximately 72 kD. All resulting pairs were screened for single mutations or combinations with other single mutations, since it was unclear whether the repulsive electrostatic interactions of a single mismatched pair would be sufficient to obtain enough half-molecules for detection by this method. These mutations were also used in combination.

[0229] Amino acid substitutions were introduced into the construct vector MV1057 using Geneart according to Table 7. The constructs were expressed by transfection into HEK293T cells using standard procedures. IgG expression levels were measured using Octet. If production failed twice, the mutation was considered to be inhibiting expression, and the mutation was not further investigated.

[0230] [Table 7]

[0231] Supernatants containing ≥5 μg / ml IgG were analyzed by SDS-PAGE and purified using Protein A. Proteins were stained with colloidal blue reagent. The homodimer was visualized as a band of approximately 150 kD. A smaller band of approximately 75 kD indicated the presence of half molecules (see negative controls: K392D, K409D). The blot is shown in Figure 7.

[0232] The SDS-PAGE results were analyzed and scored, as shown in the rightmost column of Table 7. Many residues, including Q347, S354, Y349, L351, K360, T366, T394, and V397, are promising and should be further tested in combination. The selection here took into account both high scores for inhibiting homodimer formation and the availability of contact residues that could be modified without causing problems in relation to other non-complementary charges. For example, residues F405 and Y407 are known to have multiple interactions at the CH3-CH3 interface, including interactions with already charged residues. Introducing multiple charge mutations within these interacting residues (see Table A) can be problematic.

[0233] To test further combinatorial mutations, new constructs were made in vector MV1057 (Table 8), and antibody VH regions with known specificity and known binding ability to the human IGKV1-39 light chain were used to reclone into vectors containing these new constructs (see Table 9). Table 10 lists the transfection results.

[0234] [Table 8]

[0235] [Table 9]

[0236] [Table 10]

[0237] Combinations of CH3 mutants were expressed and analyzed by SDS-PAGE (data not shown) and native mass spectrometry (MS). The results are summarized in Table 10. ZO transfection yielded the highest percentage of heterodimers in the mixture (69% AC). Interestingly, AA homodimers were absent in ZO transfection, while CC homodimers were present in a small proportion (7%). Mass spectrometry analysis revealed that the remaining protein in the mixture consisted of half A molecules, likely the result of imbalanced expression of the A and C heavy chains. Raw MS data from the transfected ZO sample are shown in Figure 8. Surprisingly, while transfected ZO yielded significant amounts of bispecific product, its opposite charge pair, transfected ZP (L351K / T366'D vs. T366K / L351'D for ZO), did not yield similar results; only 52% bispecific product was observed, with significant amounts of the two homodimers present (30% AA and 13% CC). The explanation for this is that the negatively charged D is structurally very similar to T, so the repulsion of T366D with itself is not strong enough, and therefore T366D can still form homodimers, which was indeed observed.

[0238] It is expected that minor variants of the newly discovered T366K / L351'D pair (e.g., by testing all variants including the new construct T366R and L351E) will result in a similar proportion of bispecific antibodies (BsAbs).

[0239] Example 14: Design of new CH3 mutations to induce efficient heterodimerization by HADDOCK As described in Example 13, the newly discovered charge pair T366K / L351'D increases the proportion of heterodimers in the mixture (69%) while also "contaminating" the mixture with a small proportion of undesired CC homodimers (7%) (L351D / L351'D) and a significant proportion of half-molecule A (24%). In this example, an in silico approach is used to gain further insight into the amino acid residues involved in interactions at the CH3 interface, test opposing complementary substitutions in the CH3 region, and discover a novel CH3 pair containing complementary substitutions that inhibit efficient homodimer formation of the two heavy chains while further increasing efficient heterodimerization.

[0240] HADDOCK (High Ambiguity Driven protein-protein DOCKing) is an information-driven, flexible docking approach for modeling biomolecular complexes. Unlike first-principles docking methods, HADDOCK encodes identified or predicted protein contacts within ambiguous interaction restraints (AIRs) to guide the docking process (de Vries et al., 2010). Input to the HADDOCK web server consists of protein structure files, such as crystal structures, NMR structure clusters, or structural models. After docking or refinement, HADDOCK returns a so-called HADDOCK score, which is a weighted average of van der Waals energy, electrostatic energy, surface area covered, and desolvation energy. The HADDOCK score is often difficult to translate directly to experimental data and is interpreted as an indicator of binding energy or affinity. Additionally, HADDOCK provides structure files of the "top four" structures from the docking calculations, which can be downloaded and visualized, allowing detailed analysis of the interactions of individual residues.

[0241] In this example, the interactions between the CH3 domains of the IgG1 heavy chain are studied. The high-resolution crystal structure of the Fc portion of IgG (structure 1L6X) is used as the starting structure (http: / / www.rcsb.org / pdb / explore / explore.do?structureId=1l6x; Idusogie, EE et al., JI 2000(164)4178-4184).

[0242] In Example 13, it was found that co-expression of vectors XIII and XVI resulted in the formation of contaminating CC homodimers (Table 10). HADDOCK was used to explore additional mutations in addition to T366K / L351'D that disrupt homodimerization.

[0243] The output of HADDOCK scoring is the calculated energy, the HADDOCK score (a weighted average of several energies), and four structure files corresponding to the four lowest energy structures found by the program. The HADDOCK score is used to compare different structures. The other energies are used only to get an indication of what is happening in the structure (e.g., better electrostatic interactions, less covered surface, higher van der Waals energy). The lower the HADDOCK score, the better. For each mutation pair, scores are calculated for the AA, AB, and BB dimers.

[0244] The set of mutation pairs in Example 12 was run in HADDOCK to see if the experimental data correlated with the calculated energies. All theoretical energies are listed in Table 11 and visualized in Figure 9.

[0245] [Table 11]

[0246] For the two wild-type CH3 domains, the HADDOCK scores are identical for AA, AB, and BB because the CH3 regions of A and B are identical. In most other cases, the AB pair had the lowest score, as expected. For the T366K / L351D pair, the BB score was slightly better than the AB score (-210.6 vs. -212.5). However, this difference is within the calculation error range. Using HADDOCK, the heterodimer structures of these pairs were visualized. For example, construct combinations 1-2, 1-1, and 2-2 are displayed in Figure 10. These visualizations clearly show the formation of salt bridges in the heterodimer (Figure 10A, left panel) and electrostatic repulsion between residues on the same chain (Figure 10B and C, center and right panels). The higher HADDOCK score for the homodimer is explained by electrostatic repulsion of the mutated contact residues. These residues bend away from each other and do not interact with residues on the other chain, resulting in reduced affinity.

[0247] Table 11 and Figure 9 confirm the observations in Example 13. The T366K / L351'D AC heterodimer and the L351D / L351'D CC homodimer have similar energies, explaining the presence of both heterodimers and homodimers in the mixture. On the other hand, the T366K / T366'K AA homodimer is barely detectable in the mixture, and the T366K half molecule A is present. Table 11 and Figure 9 show that the HADDOCK score of the T366K / T366'K AA homodimer is higher than that of the AC heterodimer. Therefore, the formation of this homodimer is energetically less favorable.

[0248] Example 15: 366 / 351 Mutation In Example 13, it is hypothesized that alternatives to the T366K / L351'D mutation charge pair can be designed to achieve similar results in terms of the proportion of bispecific antibodies in the mixture. These alternatives may include substitutions of T366R, T366D, T366E, L351E, L351K, and L351R. The proportion of L351D / L351'D CC homodimers may be reduced by generating the 366 / 351 pair of mutants. All possible mutation pairs were run in HADDOCK, and the resulting scores are shown in Table 12 and visualized in Figure 11.

[0249] [Table 12]

[0250] Looking at the HADDOCK scores, we observed that some mutations had similar "patterns" when compared to T366K / L351'D. In many variants, AA homodimers were found to have higher HADDOCK scores than AB heterodimers, while BB homodimers were equally favored. Residue 351 is known to be "neighboring" the same residue in the other chain. That is, residue 351 in chain A pairs with residue 351 in chain B at the CH3-CH3 interface. When the BB dimer forms, there is little negative influence from the same charge. Looking at the L351D / L351'D structure, the aspartates bend away from each other, and there is a stabilizing influence from at least the naturally occurring arginine at position 355, as well as a stabilizing negative charge from the naturally occurring serine at position 354 (see Figure 12A). Mutation of these residues (S354A and R355D) provides little improvement. Figure 12B clearly shows that the main chain hydrogen of A354 is responsible for stabilizing the homodimer. From this series, the T366R / L351'E pair appears to be the most favorable, with the lowest HADDOCK score for the bispecific molecule.

[0251] Example 16: Mutations around T366K / L351'D In this series of HADDOCK analyses, the T366K / L351'D or T366K / L351'E pair was used as the starting structure. To identify additional mutations that could further improve the predicted value of the dual specificity ratio of these A and B chains, mutations were added to the B chain and HADDOCK scores and energies were calculated. When examining the CH3 domain structure using a viewer for visualizing protein structures at the molecular level (YASARA, www.yasara.org), distances between individual residues can be calculated. During this analysis, two residues, Y349 and L368, were observed to be adjacent residues that could positively or negatively contribute to dimer interactions. In this example, the effects of these mutations—in addition to the L351 mutation—on the dimer formation of homo- and heterodimers were examined (see Figure 13). Both residues increase the stability of the heterodimer (lower HADDOCK score) and destabilize the B-B dimer (higher HADDOCK score). Glutamic acid (E) at positions 349 and 368 was suggested to be more preferred than aspartic acid (D). Therefore, the introduction of a second amino acid mutation in the B chain, which already has an amino acid substitution at position 351, was suggested to favor heterodimerization.

[0252] In the next set of HADDOCK analyses, the T366K / L351'D pair was again used as the starting structure. Following the substitutions in the B chain (i.e., Y349D / E and L368E) that further increased heterodimerization, additional mutations were added to the A chain, which already contained the T366K substitution. As shown in Figure 14, there are several mutations that appear to favor the formation of bispecific heterodimers. In the T366K-L351K / L351'D-Y349'D pair, all four mutated residues are involved in heterodimer pairing. This is not the case for T366K-L351K / L351'E-L368'E, where K351 is not directly involved in binding. However, the HADDOCK score of the latter heterodimer was −228.9, significantly lower than the −214.2 of T366K / L351'E-L368'E. This is explained by the hydrogen bond interaction at K at position 351 (see Figure 15). The T366K-L351K / L351'D-Y349'D pair can be further improved by the R355'D mutation in the B chain, which increases the HADDOCK score for BB but also slightly increases the HADDOCK score for AB. Thus, compared to the single mutation of T366K in the A chain, the additional L351K lowers the AB score, while leaving the AA and BB scores unchanged. Theoretically, this will result in a higher amount of bispecific heterodimers in the sample.

[0253] As is evident from Figure 11, substituting R rather than K at position 366 is expected to be more effective in inducing heterodimerization. Therefore, we repeated several HADDOCK analyses shown in Figure 13, this time substituting T366R rather than T366K in the A chain. The combination of the double mutations R366 in the A chain and R366 in the B chain was shown to be unfavorable (Figure 16). This is because the large size of this residue interferes with other interface interactions, even if all salt bridges with R366 are present in the structure. Furthermore, the HADDOCK score of the AA homodimer is lower for R366 than for K366. This also contributes to unfavorable heterodimer formation. Therefore, we discontinued further HADDOCK analyses using R366 at the interface.

[0254] A total of 14 best-performing pairs were selected from HADDOCK predictions (see Table 13 and Figure 17). Some pairs contain the R355D substitution to remove the stabilizing influence of the naturally occurring R355 on the L351 / L351'D interaction.

[0255] [Table 13]

[0256] Example 17: In vitro expression of bispecific molecules using CH3 mutations based on HADDOCK predictions As suggested by the analysis in Example 16, several CH3 mutants with additional mutations around the T366K / L351'D pair can produce mixtures with a higher percentage of bispecific components and a lower percentage of homodimeric components. These best-performing pairs were selected for production and further analysis. In addition, constructs T366R and L351E were also generated. A list of the constructs generated and those used to reclone antibody VH regions with known specificity and binding ability to the known human IGKV1-39 light chain is shown in Table 14.

[0257] The expression levels of IgG containing individual constructs were reported in Example 13 above and were repeated for the constructs listed in Table 14. The goal was to assess which constructs homodimerize in the absence of a compatible heterodimerization partner. Ideally, a high percentage of half molecules and a low percentage of homodimers would be formed. As controls, constructs containing previously reported charge mutations and constructs containing previously reported knob-in-hole mutations were also used to express whole IgG in recombinant cells. Protein A-purified supernatants were analyzed by SDS-PAGE, and the results, along with scores, are shown in Table 14.

[0258] [Table 14]

[0259] The results of coexpression of a common light chain with two different heavy chains that retained the amino acid substitutions of the constructs shown in Table 14 or the amino acid substitutions of the previous constructs are shown in Table 15. Expression of two different heavy chains, each with the amino acid substitutions T366K and L351'D:L368'E, resulted in approximately 87% bispecific heterodimer AB in the mixture (combination number 3 in Table 15), with no homodimers AA or BB. Approximately 12% of half molecules containing the T366K substitution (half molecule A) were observed. Furthermore, the percentage of bispecific heterodimer AB was found to increase when an additional amino acid substitution, L351K, was introduced into the first heavy chain. For example, co-expression of two different heavy chains with the amino acid substitutions T366K:L351K and L351'D:L368'E, respectively, resulted in approximately 92% bispecific heterodimer AB, while AA and BB homodimers were virtually absent from the mixture (combination number 12 in Table 15).

[0260] The combination of 10 and 11 also resulted in a favorable distribution of heterodimers with a high percentage and virtually no homodimers. The absence of homodimers is advantageous because the fraction containing intact IgG molecules is composed exclusively of heterodimers AB. For purification and subsequent therapeutic applications, half-molecules can be removed by standard approaches such as size-exclusion chromatography. Known charge mutants and knob-into-hole mutants do not remove "contaminating" homodimeric antibodies. However, applying these newly identified charge mutants to bispecific antibody production processes would be advantageous.

[0261] Additionally, the T366K / L351'D:L368'E and T366K:L351K / L351'D:L368'E charge pairs offer additional advantages over the previously described E356K:D399K / K392'D:K409'D and E356K:D399K / K392'D:K409'D:K439'D charge reversal pairs. Specifically, while previously described charge variants are based on reversal of existing charges within the CH3-CH3 interface, the newly identified charge variants add an additional charge pair (charge-charge interaction) at the CH3-CH3 interface. The introduction of an additional charge pair at the CH3-CH3 interface further increases the stability of the interface, thereby further increasing the stability of the intact antibody. The same is true for the mutations used in combinations 4, 5, 6, 9, 10, and 11. This leads to a favorable proportion of bispecific heterodimers with very low proportions of AA and BB homodimers in the mixture.

[0262] [Table 15]

[0263] Native MS was performed on all dual-specific samples. The resulting graphs were analyzed to derive the relative ratios of the species of interest using two methods: peak height and peak area. While the peak area method is the more scientifically correct analysis method, since all previous analyses were performed using peak height in other studies, both methods were included for comparison. The difference between the two methods was within the measurement error range, so only the peak area values were used for subsequent measurements.

[0264] Two typical spectra are shown in Figure 18. A summary of the results is shown graphically in Figure 19. Values can be found in Table 15. In about half of the samples, the total monospecific IgG contamination was less than 5%, and in only three cases it exceeded 10%. With wild-type IgG, on the other hand, about 50% monospecific IgG would be expected to be found in the mixture.

[0265] Ten combinations of two different heavy chains were selected for further analysis from Table 15. These 10 combinations include combinations 1, 2, 3, 4, 5, 6, 9, 10, 11, and 12 (Table 15). These 10 combinations were selected not only based on the low percentage of homodimers present in the nMS-derived mixtures, but also on their overall physicochemical properties, including product yield, SDS-PAGE, and the number of mutations in the CH3 domain.

[0266] Example 18: Analysis of IgG stability In this study, we further analyze the stability of the Fc region of IgG molecules for a series of CH3 mutation pairs that yield a high percentage of bispecific heterodimers in the intact IgG fraction, and very low amounts (<5%) of the parent IgG. Mutated CH3 domains used to promote heavy chain heterodimerization can have unexpected destabilizing effects on the Fc region of IgG, which can result in undesirable properties such as reduced in vivo half-life, reduced effector function, and / or increased immunogenicity.

[0267] The newly identified charge pairs were compared to wild-type bispecific molecules and bispecific molecules containing previously identified charge mutations (chain A with construct 1 and chain B with construct 2). All bispecific molecules in this study contain identical heavy and light chain variable regions, ensuring that any observed effects are due to mutations in the Fc portion of the molecules and not due to differences in the variable regions.

[0268] A series of stability studies will be performed on these bispecific molecules, including spectroscopic (UV-Vis light absorption, fluorescence, and light scattering) and microscopic (light and fluorescence microscopy with Nile Red staining) analyses that provide information on the aggregation state of the CH3 variants.

[0269] UV-Vis absorption spectra were recorded at 25°C on a double-beam, two-monochromator Cary 300 Bio spectrophotometer. Spectra were monitored between 250 and 400 nm with a 1 cm path length. Absorbance at 320 nm and longer wavelengths provided information on the aggregation state of IgG.

[0270] The intrinsic fluorescence spectrum is monitored at 25°C using a FluoroMax fluorometer. The fluorescence method is optimized accordingly. Fluorescence emission provides information about conformational and aggregation properties.

[0271] 90° light scattering spectra are monitored at 25°C using a FluoroMax fluorometer, performing a synchronous scan (λem = λex) from 400 nm to 750 nm with an integration time of 0.01 seconds. The excitation and emission slits are optimized accordingly. For example, right-angle light scattering can distinguish the presence of 5% dimers in an IgG sample.

[0272] For fluorescence microscopy with Nile Red staining, Nile Red in ethanol is added to the sample immediately before measurement. The sample is loaded onto a microscope slide and analyzed by fluorescence microscopy. Particles are counted; the smallest particle size observed by fluorescence microscopy is approximately 0.5 μm.

[0273] Stresses on proteins, such as temperature, pH, mechanical stress, or exposure to denaturants, can cause conformational changes (e.g., unfolding) and / or aggregation. Previously, charge-engineered bispecific antibodies have been reported to have lower melting temperatures of modified CH3 (Gunasekaran 2010). Therefore, these studies aim to distinguish the novel charge variants of the present invention from previously known charge variants.

[0274] A thermostability study was performed using an Octet™ Protein A biosensor and FcRn binding to IgG. To examine the thermostability of CH3-engineered IgG, samples were incubated at 4, 50, 55, 60, 65, 70, and 75°C for 1 hour at a concentration of 100 μg / ml (PBS). After this, the samples were slowly cooled to 25°C in 15 minutes, held at this temperature for 2 hours, and then stored at 4°C until the next day. Precipitated antibodies were removed by centrifugation, and the total IgG concentration of water-soluble antibodies was determined by Octet™ Protein A biosensor (1 / 10 dilution in PBS).

[0275] Assays measuring the binding of CH3-engineered IgG to FcRn have been investigated using Octet. Protein L biosensors are used to bind IgG light chains to the sensor and then incubate with FcRn in solution. Alternatively, Anti-Penta-HIS biosensors can be used to bind His-tagged FcRn proteins and then incubate with the IgG of interest. These methods can be more sensitive than Protein A biosensors and may be used for thermal stability studies.

[0276] All samples were also subjected to serum stability analysis. Briefly, the (engineered) IgG samples were incubated in human serum at 37°C, while the control samples were kept at 4°C. After 1, 2, 3, and 4 weeks, the samples were centrifuged and the precipitated IgG was removed. The samples were then added to an antigen-specific ELISA to determine the relative amount of functional IgG. A purified control antibody was acutely injected into human serum and used as a reference group.

[0277] Example 19: Stability analysis In previous experiments, co-expression of two different heavy chains with CH3 mutations and a common light chain resulted in a high percentage of bispecific antibodies (Example 17).

[0278] Eight combinations of two different heavy chains were selected from Table 15 for further analysis. These eight combinations included combinations 3, 4, 5, 6, 9, 10, 11, and 12 (Table 15). This study focused on the stability of the IgG Fc region, and these eight combinations were analyzed. Control groups included wild-type bispecific molecules (i.e., no CH3 mutations) and / or previously reported CH3 charge mutations. Note that for wild-type bispecific molecules, two heavy chains and a common light chain were coexpressed without a means to selectively induce heterodimerization. Thus, these "wild-type bispecific molecules" represent a mixture of AA, AB, and BB. All bispecific molecules in this study retained the same heavy and light chain combinations, ensuring that any observed effects were due to mutations in the Fc region of the molecules and not changes in the Fab region.

[0279] It is hypothesized that the mutation pairs used to promote heterodimer pairing of two different heavy chains may be associated with unexpected structural or other destabilizing effects on the Fc region of IgG, which may subsequently result in undesirable problems that could hinder further clinical development, such as reduced in vivo half-life, reduced effector function, and / or increased immunogenicity due to the presence of these mutations.

[0280] (thermal stability) Stress, such as increasing or decreasing temperature, can cause conformational changes (e.g., unfolding) and / or aggregation of proteins. To investigate the thermal stability of CH3-engineered IgGs, combinations 3-6 and 9-12 (Table 15) were incubated with wild-type bispecific molecules and bispecific molecules derived from constructs 1 and 2 (E356K:D399K / K392D':K409D', also known as "charge-reversed" pairs) at 4, 60, 62.5, 65, 67.5, 70, and 72.5°C for 1 h at 100 μg / ml (PBS). After this, the samples were slowly cooled to 25°C over 15 min and held at this temperature for 2 h before being stored at 4°C until the next day. The precipitated antibodies were removed by centrifugation (18,000 rpm; 4°C, 20 minutes), and the total IgG concentration of water-soluble antibodies was determined by Octet using a protein A biosensor (1 / 10 diluted in PBS).

[0281] The results are shown in Figure 20. The control CH3-engineered bispecific antibody (charge-reversed combination E356K:D399K / K392D':K409D' (triangles)) exhibits decreased thermal stability compared to the wild-type bispecific molecule (squares). Bispecific molecules from combinations 3-6 and 9-12 (diamonds) also demonstrated decreased thermal stability compared to the wild-type. Notably, however, three combinations demonstrated improved stability compared to the control CH3-engineered bispecific antibody. Bispecific molecules from combinations 9, 10, and 11 were significantly more stable than the other CH3-engineered (charge-reversed) bispecific molecules and were as stable as the wild-type bispecific molecule at the highest temperature measurement point.

[0282] (Freeze-thaw stability) To investigate the stability of CH3-engineered IgGs during repeated freeze-thaw cycles, bispecific molecules from combinations 3-6 and 9-12 (Table 15), wild-type bispecific molecules, and bispecific molecules obtained using constructs 1 and 2 (E356K:D399K / K392D':K409D' combinations (charge-reversed pairs)) were subjected to 10 freeze-thaw cycles. For each cycle, the samples were placed at -80°C for at least 15 minutes until completely frozen, and then thawed at room temperature. Once completely thawed, the freeze-thaw cycle was repeated again. After 10 freeze-thaw cycles, precipitated antibody was removed by centrifugation (18,000 rpm; 4°C, 20 minutes), and the total IgG concentration of water-soluble antibody was determined using a Protein A biosensor (1 / 10 dilution in PBS) by Octet. The freeze-thaw stability test was repeated three times.

[0283] The results are shown in Figure 21. The control charge-reversed CH3-engineered bispecific antibody appeared to have slightly reduced stability compared to the wild-type bispecific molecule. In contrast, the bispecific molecules from combinations 3, 4, and 9 appeared to have slightly improved stability compared to the wild-type bispecific molecule. From these results, it was concluded that the harsh conditions of freeze-thaw cycling do not pose significant stability issues for the CH3-engineered variants.

[0284] In vitro serum stability: To investigate the stability of CH3-engineered IgG in serum maintained at 37°C, bispecific molecules from combinations 3-6 and 9-12 (Table 15) were incubated with wild-type and charge-reversed bispecific molecules in 10% human serum at 37°C. Control samples were kept at 4°C. After 1, 2, or 5 days, precipitated antibody was removed by centrifugation. Samples were then added to a fibrinogen-specific ELISA to derive the relative amount of functional IgG. A purified control antibody was acutely injected into human serum and used as a reference group.

[0285] Fibrinogen ELISA data showed that all samples were very stable in 10% human serum at 37°C for 5 days. The bispecific molecules from combinations 4 and 5 with lower IgG concentrations were slightly less stable, especially at T=1 and T=2, but the differences were minimal at the end of the experiment (Figure 22).

[0286] Example 20: Further stability tests A series of further analytical methods were used to assess the stability of the mutant IgGs. Bispecific molecules from combinations 3-6 and 9-12 (Table 15), wild-type bispecific molecules (AA, AB, BB), the individual parental antibodies (AA and BB) and bispecific molecules obtained using constructs 1 and 2 (E356K:D399K / K392D':K409D' combination (charge-reversed pair)) were used as samples for these stability assays.

[0287] All IgGs were diluted to 0.2 mg / ml and subjected to several stress conditions (50°C for 2 days, 40°C for 2 weeks, and five freeze-thaw cycles) to allow for differentiation between different samples. Note that these high stress levels destabilized one of the parent antibodies used in all bispecific molecules (the BB parent, carrying two 1122 Fabs). At 50°C for 2 days, aggregation of this protein was detected by UV absorption. This suggests that these stress conditions may not be able to distinguish between the instabilities of the Fab and CH3 in bispecific molecules. Therefore, data from the 50°C incubation should be treated with caution.

[0288] The results are summarized in Table 16. The analytical methods used included: Nile Red fluorescence microscopy ("Nile Red Particles" in Table 16), to observe the amount of particles larger than 0.5 μm after adding Nile Red dye; UV spectroscopy at 350 nm ("UV350nm"), where changes in absorbance at wavelengths longer than 320 nm provide information about the protein's aggregation state; 90° light scattering at 400 nm ("LS400nm"), a sensitive technique for observing changes in protein aggregation, e.g., the difference between IgG monomers and dimers; autofluorescence, where the fluorescence wavelength maximum and intensity of aromatic residues in proteins change depending on the environment (e.g., unfolding); 1,8-ANS fluorescence spectroscopy, where 1,8-ANS binds to cationic groups through electrostatic interactions via ion pair formation, and changes in protein structure and / or conformation are detected.

[0289] (UV-Vis spectroscopy) UV-Vis absorption spectra were measured at 25°C using different Varian quartz cuvettes (e.g., a 1.0 cm black low-volume Hellma cuvette and a 0.2 cm x 1.0 cm clear Hellma cuvette) on a Cary 300 Bio spectrophotometer with a double beam and two monochromators. Using a 1.0 cm path length, the spectrum was monitored between 220 and 450 nm. Absorbance around 280 nm provides information about the protein concentration. The region between 320 and 450 nm provides information about the aggregation state of the sample.

[0290] (90° light scattering) A 90° light scattering spectroscopy method was developed to study protein aggregation. The method was performed similarly to that described in Capelle 2005 and Demeule 2007a. 90° light scattering spectra were monitored at 25°C using a FluoroMax fluorometer (Spex, Instruments SA, Inc. UK) with a synchronous scan from 400 nm to 750 nm (λem = λex) using an integration time of 0.01 seconds. Different slit settings were tested to obtain optimal conditions. Once optimized, all measurements were performed using that slit setting.

[0291] (steady-state fluorescence emission) The fluorescence emission of tryptophan, tyrosine, and phenylalanine residues provides information about the local environment of these fluorophores. Changes or differences in hydrophilicity and / or rigidity are measured. Generally, a more hydrophobic and rigid environment leads to an increase in fluorescence intensity and a blue shift in the emission maximum. Autofluorescence spectroscopy provides information about the current state of the protein and monitors changes in its physical and chemical properties. More detailed information about tyrosine and tryptophan can be found in Lakowicz's book (Lakowicz, 2006).

[0292] Fluorescence emission and excitation spectra were recorded at 25°C in different quartz cuvettes. Samples were excited at different wavelengths. The integration time and slit settings were optimized. After optimization, the integration time and slit settings were applied to all samples.

[0293] Fluorescence microscopy with Nile Red staining. Nile Red staining was developed to visualize protein aggregation. The staining was performed as described by Demeule et al., 2007b.

[0294] Microscopic observations were performed using a Leica DM RXE microscope (Leica Microsystems GmbH, Wetzlar, Germany) equipped with a mercury lamp. Images were acquired using a Sony NEX-5 camera and its firmware. Objectives were 10x, 20x, and 40x. A fixed distance of 0.1 mm between the slide and the cover slip was used for microscopic studies. The size of the 4x4 grid was 1mm x 1mm, corresponding to 0.1 μl.

[0295] (1,8-ANS fluorescence spectroscopy) 8-Anilino-1-naphthalene-8-sulfonic acid (1,8-ANS) is an uncharged, hydrophobic, fluorescent small molecule (molecular weight 299.34 Da) that can be used to study both membrane surfaces and proteins.

[0296] 1,8-ANS is virtually non-fluorescent in water and only emits appreciable fluorescence when bound to membranes (quantum yield ~0.25) or proteins (quantum yield ~0.7). This property makes 1,8-ANS a sensitive indicator of protein folding, conformational changes, and other processes that occur when the probe is exposed to water. Information about 1,8-ANS is available at Molecular Probes' website at www.probes.com.

[0297] The fluorescence emission spectra of 1,8-ANS were recorded using a FluoroMax fluorometer. Direct comparison of 1,8-ANS fluorescence between IgGs was not performed because each IgG has a different number of 1,8-ANS binding sites. In principle, the lower the 1,8-ANS fluorescence, the fewer 1,8-ANS molecules bound to the antibody. Stress-induced changes in 1,8-ANS fluorescence intensity and emission wavelength were evaluated.

[0298] [Table 16]

[0299] Taken together, these data demonstrate that the various IgG samples are remarkably stable. Severe stress conditions (e.g., 50°C for 2 days) are required to produce measurable differences between the samples tested. Under these conditions, samples from combinations 9 and 10 appear more prone to aggregation than the other samples.

[0300] The most significant factors that differentiate the stability of proteins are freeze-thaw cycles and elevated temperatures. Taking into account the very severe stress factor of incubation at 50°C, the two mutants T366K / L351E,Y349E (combination #4) and T366K,L351K / L351D,Y349E (combination #11) are the most stable proteins in this group, closely followed by T366K,L351K / L351D,Y349D (combination #10) and T366K,L351K / L351D,L368E (combination #12).

[0301] Example 21: Experiments for native MS at varying ratios; transfection ratios from 1:5 to 5:1 To gain a better understanding of the behavior of CH3-mutated IgGs in mixtures when transfected at biased ratios, we performed more detailed ratio studies, especially for the T366K:L351K / L351D':L368E' combination (hereafter referred to as KK / DE or DEKK).

[0302] The VH regions of previously used antibodies with known specificity and binding ability to the known common light chain, human IGKV1-39, were used to reclone into constructs 1, 2, 68, and 69, resulting in vector IV (Table 17). Vector IV, each containing nucleic acid sequences encoding Ig heavy chains with different CH3 regions and different antigen specificities and the common human light chain, was then transfected into cells at different transfection ratios shown in Table 18. The results are shown in Figure 23.

[0303] [Table 17]

[0304] [Table 18]

[0305] As shown in Figures 23A and B, in DEKK combination mutations, when an excess of A or C is present (A or C on the "DE side," B on the "KK side"), AB or BC is formed, but in all cases the excess A or C exists as a mixture of both homodimers and half molecules. However, when an excess of B is present (B on the "KK side," A or C on the "DE side"), a clear difference is seen: AB or BC is still formed, but the excess B is essentially not present as a homodimer, and only half molecules are formed.

[0306] It should be noted that the percentages are again measured by peak height. Peaks detected at 2% or less are below the threshold that the nMS technique used here can accurately measure. Therefore, measurements below 2% were considered to be within the noise level of the analysis and were ignored.

[0307] It is noteworthy that excess A:B results in a high proportion of only the B half molecule. In particular, at A:B ratios of 1:3 and 1:5, a high proportion of B half molecules was observed without the presence of BB homodimers (Figures 23A and 23B), indicating that the CH3 mutations on the KK side make homodimerization less likely. The absence of homodimers is a crucial advantage, since the "KK side" of the DEKK combination was chosen to add specificity, potentially resulting in known adverse effects when present as homodimers (e.g., cMET or CD3 antibodies are known to have undesirable side effects when present in pharmaceutical compositions as bivalent homodimers).

[0308] The observation of different DE:KK ratios contrasts with the control charge-reversal CH3 mutations in Vectors IV and V. As shown in Figure 23C, in the E356K:D399K / K392D':K409D' combination mutations, when A is present in excess (A is the "K392D:K409D side"), the excess A exists in all cases as a mixture of both homodimers and half molecules. When B is present in excess (B is the "E356K:D399K side"), the excess B exists in all cases as a mixture of both homodimers and half molecules. Even at higher ratios of 1:3 and 1:5, homodimers are present, but half-molecules of B are not observed. This indicates that the E356K:D399K side does not favor homodimers as much as the KK side of the DEKK combination.

[0309] In summary, the DEKK combination mutations have a distinct advantage over charge-reversal CH3 mutations in that one of the chains in the heterodimer does not form a homodimer.

[0310] Example 22: Various mixtures using DEKK combinations Mutation of the DEKK combination was shown to induce the formation of highly pure bispecific IgG molecules ("AB"). We next explored the feasibility of controlled production of more complex antibody mixtures, such as "AB and AA" or "AB and AC" mixtures, from a single cell. Previously used Fab types were incorporated into vectors containing either the "DE construct" or the "KK construct." To demonstrate the versatility of the technology, various combinations of these vectors were coexpressed to generate mixtures. The Fab types MF1337 (tetanus toxin), MF1122 (fibrinogen), and MF1025 (thyroglobulin) were selected for their overall stable behavior, good expression levels, and mass differences between the IgGs containing these Fabs (see Table 19).

[0311] [Table 19]

[0312] [Table 20]

[0313] SDS-PAGE analysis revealed that most samples consisted predominantly of full-length IgG, with a small proportion of half-molecules present in some cases. Furthermore, many samples showed two bands at approximately 150 kDa on gels under nonreducing conditions, reflecting the presence of two distinct IgG species in the samples. Even on gels under reducing conditions, two heavy chain bands were visible in some samples (data not shown).

[0314] Native MS was performed on all samples, and the percentage of observed species was calculated based on peak height (percentage of "observed species" in Table 20). The results are shown in Figure 24. Two major peaks corresponding to the expected species were observed in all eight samples in which the three heavy chains were coexpressed. Small amounts of contaminating DE-DE homodimers were observed in two of these samples (transfections 2 and 4) and in transfection 11. Half molecules were detected in very low amounts (less than 2%) in most samples. However, as mentioned above, this is not a problem because they can be easily separated from the full-length IgG portion.

[0315] After nMS, it was found that a different species was predicted for sample 11 than the species corresponding to the observed IgG mass, which was concluded to be a transfection error: it was apparent that 1337-KK, instead of 1122-KK, was co-expressed with 1025-DE in sample 11.

[0316] To confirm the functional presence of the desired specificity, IgG samples were further tested by sandwich ELISA. ELISA plates were coated with fibrinogen or thioglobulin, and detection was performed with fluorescein-labeled thioglobulin or tetanus toxoid. The detection antigen was labeled with fluorescein (Pierce NHS-fluorescin Antibody Labeling kit, cat. #53029) according to the manufacturer's instructions. The fluorescein-labeled antigen was then detected with an FITC-conjugated anti-fluorescein antibody (Roche diagnostics, cat. #11426346910).

[0317] A summary of the results of the dual-specific ELISA (OD450 values) is shown in Table 21. The gray cells indicate the expected species for each transfection. Generally, the experimental results matched the expected results, with exceptions noted in italics or bold. For transfections 1-3, species BC (transfections #1 and #2) or AC (transfection #3) are presumed "negative" cells, but they exhibit significant background signal. Previous studies have shown that dual-specific ELISAs can suffer from high background levels. These background levels may be due to half molecules that may be present in the samples. Of note, the results of the dual-specific ELISA confirmed that an error had occurred in transfection #11; species AC (bold values) was detected, but not BC.

[0318] [Table 21]

[0319] Example 23: Improved Mixture of Two Bispecific Antibodies Recognizing Four Distinct Epitopes (AB and CD) from a Single Cell. In Example 12, we hypothesized that the mixture from transfection ZA or ZB might be problematic when moving to large-scale production. This is because knob-into-hole mutants have been reported to be unstable, and dimerization of the CH3 domain with the "knob" or "hole"-engineered CH3 domain cannot be ruled out. As shown in the above examples, novel charge pair mutants have been found that selectively induce heterodimerization with virtually no homodimer formation. Polypeptide chains with CH3 domains containing these novel charge pair mutants can be expressed in cells together with previously known polypeptide chains or SEED bodies containing charge-engineered CH3 domains, resulting in the selective formation of only two bispecific molecules.

[0320] As is evident from the above examples, DEKK combinatorial mutagenesis is excellent for producing one bispecific molecule (AB) or two bispecific molecules (AB and AC) in clonal cells where heavy chain dimerization is driven by the CH3 domain. However, having only one set of vectors available for complementary CH3 mutagenesis limits the types of mixtures that can be produced. If a second, "orthogonal" set of vectors were available for use in combination with DEKK, it would be possible to produce more complex IgG and / or bispecific molecules, such as "AB and CD" or "AB and CC" mixtures.

[0321] An important requirement when combining two sets of vectors is that the heavy chains expressed from the two different sets of CH3-engineered vectors do not form "crossover" dimers, where heavy chains produced by one set of vectors dimerize with heavy chains expressed by the other set of vectors to form full-length IgG.

[0322] To test whether such "crossover" dimers could form, in silico analysis was performed using HADDOCK to gain insight into pairing between the wild-type CH3 domain and a CH3 domain containing DE- or KK-mutations. Similarly, we analyzed pairing between the wild-type CH3 domain and a CH3 domain containing E356K, D399K, or K392D, K409D mutations, between the wild-type CH3 domain and a CH3 domain containing knob-into-hole mutations, and any combination of the above. A list of the CH3 mutation combinations analyzed by HADDOCK is shown in Table 22. The resulting HADDOCK scores are summarized in Figure 25.

[0323] [Table 22]

[0324] Based on these HADDOCK predictions, the most successful combination appears to be the CH3 of the DEKK combination with the CH3 of the charge-reversal combination, as shown in Figure 25. This combination allows the formation of two desired bispecific molecules (AB and CD) without contaminating by-products (especially AC, AD, BC, and BD) when co-expressed in a single cell.

[0325] As can be seen in Figure 25, these undesirable bispecific species AC, AD, BC, and BD have relatively high HADDOCK scores, while the desirable AB and CD have the lowest HADDOCK scores. Of course, when either DEKK or charge-reversed CH3 combinations are inserted into constructs that retain the same specificity (e.g., "C" on the DE side, "C" on the KK side, "A" on the E356K,D399K side, and "B" on the E356K,D399K side, or "A" on the DE side, "B" on the KK side, "C" on the E356K,D399K side, and "C" on the E356K,D399K side), CC and AB are predominantly produced when co-expressed in cells.

[0326] In contrast, when we look at the predictions for coexpression of DEKK and wild-type, the HADDOCK scores for AC and AD are lower than the HADDOCK score for CD. This indicates that AC and AD are highly likely to be contaminated when attempting to produce a mixture of AB and CD by coexpressing a vector encoding the DEKK CH3 combination with a vector encoding the wild-type CH3.

[0327] Finally, when either the DEKK or charge-reversal mutants were coexpressed with knob-into-hole mutants, the undesired dual-specificity mutants had relatively low HADDOCK scores, indicating a high probability of producing these undesired species upon coexpression.

[0328] Therefore, it was concluded that the combination of the CH3 of the DEKK combination with the CH3 of the charge-reversal combination (E356K, D399K / K392'D, K409D') is ideally preferred for obtaining substantially pure mixtures of "AB and CD" and / or "AB and CC" antibodies.

[0329] To put these results into practice, a mixture of two bispecific molecules recognizing four targets / epitopes (AB and CD) and a mixture of one bispecific and one monospecific molecule recognizing three targets / epitopes (AB and CC) were generated. In these mixtures, all four different VHs can be paired with the common light chain IGVK1-39, but each individual VH / VL combination has a different specificity.

[0330] To allow native MS analysis, the mass difference between the (predicted) species must be sufficient, i.e., greater than 190 Da. Four individual VHs were selected so that the predicted species upon co-expression have mass values that allow them to be identified and separated by nMS. Furthermore, the four selected VHs have a mass difference large enough to identify not only the two desired species but also most of the contaminants that may occur in the mixture. The list of selected VHs is shown in Table 23.

[0331] [Table 23]

[0332] As shown in Table 24, four different VHs were cloned into vectors containing "DE" or "KK" constructs or charge-reversal constructs, and some were co-expressed. As before, all vectors also contained nucleic acids encoding the common light chain, IGKV1-39. As previously shown, when combining two sets of vectors, an important requirement is that heavy chains expressed from two different sets of CH3-engineered vectors do not form "crossover" dimers. "Crossover" refers to the dimerization of heavy chains produced from one set of vectors with heavy chains expressed by the other set of vectors to form full-length IgG. Control transfections were performed to test the possibility of "crossover" dimer formation between heavy chains containing charge-reversal mutations and heavy chains containing DE or KK mutations.

[0333] [Table 24]

[0334] Table 25 provides a further summary of the expected species masses and possible contaminants present in transfection numbers 9-11 of Table 24.

[0335] [Table 25]

[0336] All purified protein samples from transfections #1-11 were analyzed by SDS-PAGE, including three control samples (Figure 26). Additionally, MS analysis was performed on protein samples from transfections #9-11 to identify all species in the samples.

[0337] As can be seen in Figure 26, transfections #3 and #4 resulted in the expected mismatch between the "KK" construct and either "E356K:D399K" or "K392D:K409D." The amount of half molecules in the protein samples from these transfections exceeded that of the full-length IgG molecule.

[0338] Transfections #7 and #8 resulted in approximately equal amounts of half molecules and full-length IgG in the protein samples. However, it was not possible to determine from SDS-PAGE whether this full-length IgG represented the DE / DE dimer, the DE / E356K:D399K dimer, or the DE / K392D:K409D dimer. Notably, virtually no half molecules were observed in samples from transfections #9-11.

[0339] Figure 27 shows the results of nMS analysis of transfections #9 and #11. The percentages of expected and contaminating species were calculated by peak height. In transfection #9, the expected species "AB and CD" represented 97% of the mixture (30% AB and 67% CD), while only about 3% of the contaminating BD was present (Figure 27A). In transfection #11, the expected species "AB and CC" represented 94% of the mixture (33% AB and 61% CC), while only 6% of the contaminating BC (4.1%) and AC (1.8%) were present (Figure 27B).

[0340] These data demonstrate that when a second, "orthogonal" set of vectors is used in combination with DEKK, it becomes possible to produce more complex mixtures of IgG and / or bispecific molecules, such as "AB and CD" or "AB and CC" mixtures. The use of charge-reversal and DEKK constructs in combination results in very limited amounts of "crossover" dimer formation. Adjusting the transfection ratio is expected to further reduce these low-level contaminating by-products.

[0341] Example 24: Single-dose pharmacokinetic study in mice To investigate the pharmacokinetic (pK) behavior of bispecific antibodies carrying DEKK mutation combinations in the CH3 region, this study measured and compared the pK parameters of three different IgG batches: 1) the wild-type anti-tetanus toxin parent antibody 1337:1337 (two MF1337 Fabs in a wild-type Fc backbone), 2) the wild-type anti-tetanus toxin parent antibody 1516:1516 (two MF1516 Fabs in a wild-type Fc backbone), and 3) the CH3-engineered bispecific anti-tetanus toxin antibody 1516:1337 carrying DEKK mutation combinations in the Fc region (MF1516 Fab on the DE side and MF1337 Fab on the KK side).

[0342] The DEKK-bispecific antibody products were selected so that the parent antibodies had specificities of 1337:1337 and 1516:1516. This was because, based on previous studies, there was no known pre-treatment serological response to these antibodies in some mouse strains. Pre-treatment serological responses would have complicated the interpretation of study results. Furthermore, there was sufficient mass difference between the parent antibodies to allow differentiation between 1337:1337 (wild-type Fc), 1516:1337 (DEKK Fc), and 1516:1516 (wild-type Fc) by nMS.

[0343] Three IgG batches were prepared as previously described, except that the DNA used for transfection was prepared using an endotoxin-free maxiprep kit to minimize endotoxin levels. The batches were then tested for protein concentration, aggregation, endotoxin levels, and percentage of bispecific product. The IgG batches were shown to meet the acceptance criteria for use in subsequent pK studies: after gel filtration, the IgG concentration was greater than 0.3 mg / ml, aggregation levels were less than 5%, endotoxin levels were 3 EU / mg protein, and the DEKK batches contained greater than 90% bispecific IgG.

[0344] Native mass spectrometry analysis of the samples after gel filtration revealed a high proportion of the expected species. A small amount of DE:DE homodimer, estimated at approximately 2%, was detected in sample 1516:1337 (Figure 28). It was concluded that the three IgG batches met the criteria for use in pK studies.

[0345] To compare pK parameters among the three batches, three groups of female C57BL / 6J mice (Harlan, The Netherlands) were administered 1 mg / kg of human IgG (5 ml / kg immunoglobulin solution / kg body weight). Animals were 7–8 weeks old and weighed approximately 18–20 grams at the time of administration. Blood samples were collected pre-administration, 15 and 60 minutes post-administration, and 2, 4, 8, 24, 48, 96, 168, 268, and 336 hours post-administration. Serum samples were prepared and stored below -20°C until analysis. Each group consisted of three subgroups of four mice, i.e., 12 mice per group. Samples were collected from each mouse at six time points.

[0346] Mice welfare was maintained in accordance with Directive 86 / 609 / EEC (the general principles governing the use of animals in experiments of the European Communities) and Dutch legislation (The Experiments on Animals Act, 1997). This study also complied with the Standards for Humane Care and Use of Laboratory Animals, issued by the Laboratory Animal Welfare Division of the National Institutes of Health, US, identification number 45859-01 (expiry date: April 30, 2015).

[0347] Mice in group 1 received the monospecific IgG antibody 1516:1516 full-length (triangles), mice in group 2 received the monospecific IgG antibody 1337:1337 full-length (squares), and mice in group 3 received the bispecific IgG antibody 1516:1337 full-length with a DEKK-engineered CH3 region (1516 on the DE side and 1337 on the KK side) (diamonds).

[0348] Quantitative analysis of monoclonal human antibodies in mouse serum was performed by ELISA using a quantitative human IgG ELISA (ZeptoMetrix, NY USA; ELISA kit No. 0801182). Briefly, the ELISA assay is based on the following principle: a 96-well ELISA plate is coated with anti-human IgG, and human monoclonal antibodies bind to the anti-human IgG. The bound antibodies were then visualized using horseradish peroxidase (HRP)-conjugated polyclonal anti-human IgG antibodies.

[0349] The optical density (OD) of each well is directly proportional to the amount of antibody in the serum sample. The results are shown in Figure 29. We observed that the full-length bispecific IgGs carrying the DEKK mutation combinations and the parental monospecific antibodies were remarkably similar. The CH3 mutations present in the DEKK-bispecific antibodies do not alter the stability or half-life, and the DEKK mutants behave like wild-type IgG.

[0350] (References) TIFF0007723694000028.tif185169

Claims

1. A heterodimeric antibody comprising two IgG CH3 domains, one of the two CH3 domains comprises an aspartic acid (D) at amino acid 351 according to the EU numbering system and the other of the two CH3 domains comprises a lysine (K) at amino acid 366 according to the EU numbering system; or one of the two CH3 domains comprises an aspartic acid (D) at amino acid 351 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 351 according to the EU numbering system; or one of the two CH3 domains comprises a glutamic acid (E) at amino acid 368 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 366 according to the EU numbering system; or A heterodimeric antibody, wherein one of the two CH3 domains contains a glutamic acid (E) at amino acid 368 based on the EU numbering system, and the other of the two CH3 domains contains a lysine (K) at amino acid 351 based on the EU numbering system.

2. The heterodimeric antibody of claim 1 further comprising a common light chain.

3. The heterodimeric antibody of claim 1, which is a bispecific antibody.

4. The heterodimeric antibody of claim 1 , which comprises a variable region that recognizes a target epitope.

5. The heterodimeric antibody of claim 4 , which comprises two variable regions that recognize different target epitopes.

6. The heterodimeric antibody of claim 5 , wherein the different target epitopes are located on the same target molecule.

7. The heterodimeric antibody of claim 6 , wherein the target molecule is a soluble molecule or a membrane-bound molecule.

8. The heterodimeric antibody of claim 5 , wherein the different target epitopes are located on different target molecules.

9. The heterodimeric antibody of claim 8 , wherein the different target molecules are expressed in the same cell or in different cells.

10. The heterodimeric antibody of claim 8 , wherein the different target molecules are soluble molecules.

11. The heterodimeric antibody of claim 8 , wherein one target molecule is a soluble molecule while the second target molecule is a membrane-bound molecule.

12. The heterodimeric antibody of claim 5 , wherein at least one of the target epitopes is located on a tumor cell and / or at least one of the target epitopes is located on an effector cell.

13. The heterodimeric antibody of claim 12, wherein the effector cell is a NK cell, a T cell, a B cell, a monocyte, a macrophage, a dendritic cell, or a neutrophil granulocyte.

14. The heterodimeric antibody of claim 4, wherein the target epitope is located on a CD3, CD16, CD25, CD28, CD64, CD89, NKG2D, or NKp46 molecule.

15. The heterodimeric antibody of claim 1 , which is human IgG or human IgG1.

16. The heterodimeric antibody of claim 2 , wherein the common light chain is a germline light chain.

17. The heterodimeric antibody of claim 2, wherein the common light chain is a rearranged human germline kappa light chain IgVκ1-39*01 / IGJκ1*01.

18. A recombinant host cell comprising a nucleic acid sequence encoding at least two IgG CH3 domains, one of the two CH3 domains comprises an aspartic acid (D) at amino acid 351 according to the EU numbering system and the other of the two CH3 domains comprises a lysine (K) at amino acid 366 according to the EU numbering system; or one of the two CH3 domains comprises an aspartic acid (D) at amino acid 351 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 351 according to the EU numbering system; or one of the two CH3 domains comprises a glutamic acid (E) at amino acid 368 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 366 according to the EU numbering system; or A recombinant host cell, wherein one of the two CH3 domains comprises a glutamic acid (E) at amino acid 368 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 351 according to the EU numbering system.

19. 20. The recombinant host cell of claim 18, further comprising a nucleic acid sequence encoding a common light chain.

20. A pharmaceutical composition comprising the heterodimeric antibody of claim 1 and a pharmaceutically acceptable carrier.

21. The pharmaceutical composition of claim 20, wherein the heterodimeric antibody is produced by a recombinant host cell of claim 18.

22. 1. A method of producing a host cell for producing a heterodimeric antibody, said method comprising the step of introducing into said host cell a nucleic acid sequence encoding at least two CH3 domains comprising a polypeptide chain; one of the two CH3 domains comprises an aspartic acid (D) at amino acid 351 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 351 according to the EU numbering system; or one of the two CH3 domains comprises a glutamic acid (E) at amino acid 368 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 366 according to the EU numbering system; or one of the two CH3 domains comprises a glutamic acid (E) at amino acid 368 according to the EU numbering system, and the other of the two CH3 domains comprises a lysine (K) at amino acid 351 according to the EU numbering system; The nucleic acid sequences are introduced sequentially or simultaneously.

23. 23. The method of claim 22, further comprising introducing into the host cell a nucleic acid sequence encoding a common light chain.

24. A recombinant host cell according to claim 18, or a culture of a recombinant host cell obtained by the method of claim 22, which produces a heterodimeric antibody.

Images