EA041608B1 - HETERODIMERIC POLYSPECIFIC ANTIBODY FORMAT - Google Patents

Description

The field of technology to which the invention belongs

The present invention relates to a new heterodimeric multispecific format of multiple antibody variable domains, comprising a backbone of two separated pairs of variable domains, where both light chain variable domains and both cognate heavy chain variable domains are located in tandem on two separate protein chains, respectively.

State of the art

Over the past forty years since the creation of the first monoclonal antibodies [R17], antibodies have become an increasingly important class of biomolecules for research, diagnostic and therapeutic purposes.

As therapeutic agents, antibodies are being developed towards more rationally chosen functionalities, thus improving and expanding their inherent properties. Examples include optimization of effector functions by glycoengineering [R18], site-specific, eg crossing the blood-brain barrier [R19], or tunable half-life, eg by enhancing binding to FcRn [R20].

A complementary approach to antibody functionalization is to combine specific binding to different targets in a single molecule to generate bi- or multispecific antibodies or antibody fragments, which provides alternative mechanisms of action such as T-cell retargeting, such as exemplified by the bispecific antibody Blinatumomab or the trispecific antibody Catumaxomab.

Although a large number of different multispecific antibody formats have been developed to date [R21], the current repertoire of bi- and multispecific antibody formats still has significant technical challenges in production and low flexibility, with only a few formats allowing tri- and multispecific binding, and even fewer formats support the formation of heterodimeric proteins.

Previously, various polyspecific formats have been proposed. Conceptually, these formats can be classified into three categories: a) single-stranded polyspecific formats, where the various target-binding domains are all located on the same protein chain expressed from a single gene; b) homo-bi- and homo-multimeric formats, where different target-binding domains are located on identical protein chains, which are assembled using a multimerizing domain, resulting in the formation of bi-/multivalent and possibly also polyspecific complexes; and c) heterodimeric formats where the target-binding domains are located on different protein chains and the assembly of the two protein chains is provided by a heterodimerizing domain.

Heterodimeric polyspecific formats in principle have the advantage that binding domains with different specificities and affinities can be easily tested in different combinations by simply rearranging two heterodimerizing protein chains, thus allowing the search for optimal combinations of specificity and affinity directly in the final format without the need for tedious cloning. .

Such screening in the final product format is required in cases where the binding properties and/or binding strength of different domains must be carefully matched to each other in order to achieve optimal potency of the bispecific protein while minimizing the risk of non-specific effects. In clinical practice, this means optimal efficacy with minimal risk of side effects. Situations where such optimal combinations are required may be, for example, the simultaneous inhibition of two disease-provoking cytokines, which are produced during the disease process at different concentrations. In this situation, the therapeutic bispecific protein should provide effective inhibition of both cytokines at the same therapeutic dose.

Another example where the characteristics of target-binding domains must be matched in a multispecific molecule is in cancer therapy, where a cytotoxic antibody is directed to two targets on the surface of tumor cells. While the two cell surface antibody targets in this situation may be co-expressed exclusively on cancer cells, they may be expressed separately in various healthy tissues. In neoplastic therapy, in order to achieve the best efficacy with a low risk of adverse side effects, the cytotoxic antibody should preferably bind to the cell when both targets are co-expressed, but should not bind to tissues expressing only one of the two targets. To achieve this goal, the affinities of the two target-binding domains must be chosen in such a way that, on the one hand, the affinity of individual domains for their target is too weak to lead to cell lysis, and on the other hand, cooperative avidity as a result of co-binding of the bispecific molecule with both targets on the cancer cell was sufficient to cause cell lysis. Due to geometric constraints resulting from simultaneous binding to different macromolecules immobilized on the cell surface, the combination of domains to achieve maximum cooperative binding is not only a function of affinity, but also of epitopes, and can only be determined by testing different combinations of domains. in the actual product format.

A native IgG-type antibody can be considered as a homodimeric format.

In order to increase the number of binding specificities in a homodimeric antibody format using the classical IgG architecture as a scaffold, additional binding fragments such as single chain Fv [R15], Fv [R16], single domain fragments [e.g. nanobodies: Huang et al., Expert Rev Mol Diagn. 10 (2010):777-85] or alternative scaffolds [eg finomers: Schlatter et al., MAbs. 4 (2012) 497-508] to the amino or carboxy terminus of both heavy and light chains. One advantage of this approach is that bi- and tri-specific constructs can be generated using conventional IgG as the framework domain, allowing most of the manufacturing and modification technologies that have been developed for conventional IgG to be used. However, due to the homodimeric nature of conventional Fc regions, this approach will always result in at least two identical binding domains per molecule and hence bivalent binding to a specific target. This may not always be desirable, in particular: (a) unless cooperative binding to two targets is to produce the desired effect; (b) if the molecular weight is not to be further increased. In addition, this approach often suffers from poor domain stability of the attached binding fragments, making them unsuitable for pharmaceutical use.

The concept of fusion with additional binding domains to increase the number of binding specificities can also be applied to Fab fragments [R14] or other antigen-binding IgG fragments [R23]. Due to the heterodimeric nature of the heavy and light chain Fab, a Fab fragment can be used as a heterodimerizing domain. The Fab fragment, for example, was used to construct the so-called Tribody. In this format, the scFv fragments are fused to the carboxy termini of both the Fab light and heavy chains, resulting in a truly heterodimeric tri-specific molecule. The binding of the light chain and the heavy chain in the Fab occurs mainly through the interaction between CL and CH1, which are additionally connected by a covalent disulfide bond [R2] The problems of this format are: (a) stability limitation by the least stable component, which is most likely the attached scFv ; and (b) limiting the number of binding specificities to a maximum of three targets.

Heterodimeric IgG [R31] have been proposed as an approach to remove the limitations of homodimeric bispecific formats. The simple co-expression of two different monoclonal antibodies in the same cell will, with a very low probability, result in the assembly of heterodimeric bispecific IgGs where two different heavy chains will pair with each other and two different light chains will pair with their respective heavy chains [R24]. However, it can also lead to: a) mismatching between heavy and light chains with different specificities; and b) mixtures of various combinations of heavy chains resulting in mono- and bispecific variants. To eliminate these difficulties, several approaches have been proposed that create an artificial asymmetry of the molecules. The bump-bottom concept [R3, R4] uses a genetically engineered modification of the interaction surface between heavy chains or a heavy chain and a light chain to direct the binding of co-expressed chains into the desired configuration. In another approach, the CrossMab [R5] technique allows selective pairing between the engineered heavy and light chains. A disadvantage of these techniques is that any residual fraction of unpaired molecules is very difficult to separate from the product. Therefore, other techniques focus on the problem of separation by creating different binding properties for mono- and bispecific [R22] binding molecules, and on the other hand, allow yield losses caused by a stochastic distribution of variants.

Another limitation of IgG-based heterodimeric formats is that they must all contain an Fc effector domain. The format in which heterodimerization will be determined by target-binding domains directed to any desired target will allow for more binding specificities/functionality at the same or lower molecular weight. Lower molecular weight molecules penetrate target tissues (eg, solid tumors) more efficiently and thus improve efficacy at the same or lower dose. Such small molecule formats can be further engineered to have a serum half-life comparable to that of IgG by simply adding, for example, a binding domain that interacts with serum albumin.

An alternative approach uses non-antibody class fusion proteins, such as scFv fragments, to confer the desired polyspecificity. Examples of such fusion proteins are Dock-and-Lock [R25], barnase-barstar [R26], jun-fos [R27], TNF [R28] or HSA [R29]. These concepts have in common that they add at least one pair of domains that interact in a heterodimeric manner by combining bi- or polyspecific binding domains. These heterodimerizing domains are not directly involved in target binding, however, they

- 2 041608 increase the molecular weight of the protein similarly to the first constant region (C1) in the format

Tribody. In addition, they may pose a risk of increasing immunogenicity in humans by incorporating animal epitopes and sequences.

In contrast to the interaction between CL and CH1 discussed above, the binding of the paratope-forming domains of VL and VH is generally regarded as weak. However, there are several concepts of heterodimeric antibody fragments that consist solely of antibody variable domains. Approaches such as dianbodies [R6], DART [R10], and tandab [R7, R8], among others, offer an elegant and minimalistic approach to the creation of homo- and heterodimeric bispecific and bi-tetravalent assemblies. The most important limitations of these formatting strategies are: (a) adding additional binding specificities by fusion, e.g. scFv to the amino or carboxy terminus of any diabody chain, or DART can result in the formation of intrachain light and heavy chain variable domain pairs, thus making the two protein chains are very problematic; (b) due to the previously observed weak binding between light and heavy chain variable domain surfaces, these formats suffer from low monomer stability and poor production, such that additional genetic modifications are considered necessary, such as the introduction of interdomain disulfide bonds [R12] to stabilize the binding surface.

To create polyspecific single-chain tandem Fv antibodies, Kipriyanov et al. [R30] proposed a construct containing two protein chains, each of which consists of two separated Fv domains arranged in the following order VL-(linker 1)-VH-(linker 2 )-VL-(linker 3)VH. To create heterodimeric tetraspecific proteins, a heterodimer will be composed of two protein chains with the architecture below. Strand 1: VLA-(linker 1)-VHA-(linker 12)-VLB-(linker 13)-VHC, and strand B: VLD-(linker 1)-VHD-(linker 2)-VLC-(linker 3) -VHB, where the binding of FvB and FvC will cause heterodimerization of the two chains (see Fig. 10A). In order to prevent intra-chain binding that results in a tandem single-stranded Fv(scFv2)-like format, as well as to enhance the heterodimerization of the two monomeric protein chains, truncated linkers at linker 3 positions of up to 10 amino acids have been proposed (EP1293514 A1). The proposed organization of the two separated variable domains with a linker 2 of at least 15 amino acids still allows the second variable domains to bind to the N-terminal domains, resulting in a single chain diaantibody (scDb)-like format consisting of mismatched VH/VL pairs, which, in turn, most likely cannot communicate with their target. In addition, there is also the possibility of forming a heterodimer in which all heavy and light chain variable domains on protein chain 1 would pair with the heavy and light chain variable domains on protein chain 2, respectively, thereby preventing the formation of final scFvs (scFvA and scFvD) and leading to pairing between non-cognate variable domains. Tandem by-products of scFv (scFv2) or scDb-type can be responsible for the very high proportion of protein seen at apparent molecular weight from non-multimerized protein chains [R30].

In theory, the formation of scDb-like structures in the above approach can be further reduced by also shortening the second linker (linker 2) between the two separated variable domains. However, this limits the flexibility of the design, which in many cases can adversely affect the range of available epitopes that allow simultaneous binding of two targets. These geometric constraints are especially problematic when simultaneous binding of two membrane proteins is required.

Most importantly, however, both monomers can form homodimeric moieties (see FIG. 10B), so that statistically up to two thirds of the dimeric products may consist of two homodimers, while only one third will consist of the desired heterodimer.

In general, the production is in great need of heterodimeric multispecific formats that allow simple permutation and subsequent characterization of different binding domains in the final format. The main problems of such formats were: (a) the relatively low efficiency of specific heterodimerization, which leads to a suboptimal product yield; and (b) the need to use either non-target-binding proteins as heterodimerizing domains or engineered heterodimerizing Fc effector domains, which reduces the ability to alter the serum half-life of the protein and limits the flexibility to add new functionality without increasing molecular weight.

Thus, it is desirable that the optimal heterodimeric multispecific format be exclusively composed of target-binding domains and allow the geometry of the molecule to be controlled, for example, by freely varying the length of linkers between different binding domains to overcome the geometrical limitations imposed by interacting partners (targets). A solution to this problem, namely modification of the arrangement of variable domains on monomeric chains, has not yet been shown or proposed in the prior art.

- 3 041608

The essence of the invention

The present invention relates to a novel heterodimeric multispecific format of a plurality of antibody variable domains, comprising a backbone of two separated pairs of variable domains, wherein both light chain variable domains and both cognate heavy chain variable domains are located in tandem on two separate protein chains, respectively, thereby causing heterodimerization of the two protein chains. Up to two additional binding domains, in particular antibody-based binding domains such as scFv fragments, are fused to the amino or carboxy terminus of each protein chain, resulting in a heterodimeric protein capable of specifically binding up to six targets (hexaspecific).

Thus, in a first aspect, the present invention provides a heterodimeric protein comprising a first and a second single chain protein, wherein said first single chain protein comprises a first amino acid sequence consisting of (from N- to C-terminus):

(ia) a first VL domain, (iia) a first polypeptide linker, and (iiiia) a second VL domain; and where the specified second single-chain protein contains a second amino acid sequence consisting of (from N- to C-terminus):

(ib) a first VH domain, (iib) a second polypeptide linker, and (iiib) a second VH domain; and where said first VL domain forms a first cognate pair of variable domains with specificity for a first target antigen with any of said first or said second VH domain, and said second VL domain forms a second cognate pair of variable domains with specificity for a second target antigen with another of said VH domains, and wherein at least one of said first or second single chain protein further comprises (iv) at least one additional domain as a third functional domain that is fused via a third polypeptide linker to said first or said second amino acid sequence, where, optionally, said heterodimeric protein does not contain a cognate pair of first and second immunoglobulin constant domains, where said first immunoglobulin constant domain is contained in said first single-stranded protein, and where said second immunoglobulin constant domain is contained in said nom second single-stranded protein.

In a second aspect, the present invention provides one or two nucleotide sequences encoding said first and second single chain proteins.

In a third aspect, the present invention provides one or two vectors containing said one or two nucleotide sequences.

In a fourth aspect, the present invention relates to a host cell or host cells containing one or two vectors.

In the fourth aspect, the present invention relates to a method for obtaining the first and second single-stranded proteins or heterodimeric protein according to the present invention, including: (i) obtaining a nucleic acid or nucleic acids of the present invention, or a vector or vectors of the present invention, the expression of these nucleic acids or nucleic acids or said vector or vectors, and collecting said first and second single chain proteins or said heterodimeric protein from the expression system; or (ii) obtaining a host cell or host cells of the present invention, culturing said host cell or host cells, and harvesting said first and second single chain proteins or said heterodimeric protein from the cell culture.

In a fifth aspect, the present invention relates to a pharmaceutical composition comprising a heterodimeric protein of the present invention and a pharmaceutically acceptable carrier.

In a sixth aspect, the present invention relates to a heterodimeric protein of the present invention for use in the treatment of a disease, in particular a human disease, more specifically a human disease selected from a cancerous, inflammatory, and autoimmune disease, wherein at least one of said VL- and VH cognate pairs α-domains or said third, fourth, fifth or sixth functional domain is capable of interacting specifically with a target of therapeutic relevance in the respective disease.

In a seventh aspect, the present invention relates to a method of treating a patient suffering from a disease, in particular a human disease, more specifically a human disease selected from a cancerous, inflammatory and autoimmune disease, comprising administering to the subject an effective amount of a heterodimeric protein of the present invention, where at least at least one of said cognate pairs of VL and VH domains or said third, fourth, fifth or sixth functional domain is capable of interacting specifically with a target of therapeutic relevance in the respective disease.

- 4 041608

Specific embodiments of the present invention are set forth in the appended dependent claims.

Blueprints

In FIG. 1 shows a schematic representation of assembly 1 (see example 1).

In FIG. 2 shows a schematic representation of assembly 3 (see example 1).

In FIG. 3 shows a schematic representation of assembly 5 (see example 1).

In FIG. 4 shows a schematic representation of assembly 7 (see example 1).

In FIG. 5 shows size exclusion chromatograms after 1-step purification. (A) Assembly 1; (B) assembly 3; (C) assembly 5; (D) assembly 7.

In FIG. 6 shows SDS-PAGE analysis after 1-step purification: panel A: PRO356 (build 1): reducing conditions: lane 4; non-reducing conditions: lane 10; PRO357 (Build 3): Reducing Conditions: Lane 5; non-reducing conditions: lane 11; PRO358 (build 5) recovery conditions: lane 6; non-reducing conditions: lane 12; PRO355 (build 7) recovery conditions: lane 3; non-reducing conditions: lane 9. Panel B: repeat of SDS-PAGE at a lower temperature during sample preparation showing pronounced crosslinking of PRO357 (build 3) under non-reducing conditions.

In FIG. 7 shows protein content after 28 days storage at 37°C (1 g/l) (FIG. 7B) compared to storage at 4°C (FIG. 7A): PRO356 (assembly 1); PR0357 (assembly 3); PRO358 (build 5); PRO355 (build 7).

In FIG. 8 shows the monomer content after 28 days storage at 37°C (1 g/l) (FIG. 8B) compared to storage at 4°C (FIG. 8A): PRO356 (assembly 1); PR0357 (assembly 3); PRO358 (build 5); PRO355 (build 7).

In FIG. 9 shows SDS-PAGE stability analysis of samples after incubation for four weeks at 37°C: PRO356 (assembly 1): reducing conditions: lane 4; non-reducing agents based on polyspecific antibodies obtained by cognate heterodimerization) were immobilized on the sensor chip, and 4 antigens were injected in the indicated sequence. The resulting sensograms show RU shifts consistent with simultaneous interaction with all four antigens of each MATCH format.

In FIG. 12 shows the results of the analysis of the number of binding relatively inactive MATCH molecules. MATCH molecules were pre-incubated with excess TNF (antigen for one of the dimer-forming Fv domains) and the complex was run from the following conditions: lane 10; PRO357 (Build 3): Reducing Conditions: Lane 5; non-reducing conditions: lane 11; PRO358 (build 5) recovery conditions: lane 6; non-reducing conditions: lane 12; PRO355 (build 7) recovery conditions: lane 3; non-reducing conditions: lane 9.

In FIG. 10 shows a schematic representation of polyspecific single-chain tandem Fv antibodies according to Kipriyanov et al. [R30]: VL domains: gray background; VH domains: white background; cognate pairs are indicated using the same shading. (A) Schematic representation of individual chains and heterodimeric product. (B) Schematic representation of potential homodimers.

In FIG. 11 shows the results of an SPR experiment where MATCH molecules (molecules therapeutic by size-exclusion HPLC. The resulting chromatograms were analyzed to calculate the proportion of active (binding) relative to inactive MATCH molecules. The analysis revealed 11.4 to 4.7% inactive protein when superimposed to a stable peak.

Detailed description of the invention

In this invention, the authors present a new format that allows for the quantitative heterodimeric assembly of two protein chains containing multiple antibody variable domains. This format consists of a backbone of two separated pairs of variable domains (two Fv fragments), where both light chain variable domains and both heavy chain variable domains are located on a separate protein chain, thus promoting heterodimerization of the two protein chains. Up to two additional variable domains in scFv format with high intra- and inter-domain stability are fused to the amino and/or carboxyl terminus of each peptide chain, resulting in a heterodimeric protein capable of specifically binding up to six targets (hexaspecific).

Thus, in a first aspect, the present invention provides a heterodimeric protein comprising a first and a second single chain protein, wherein said first single chain protein comprises a first amino acid sequence consisting of (from N- to C-terminus):

(ia) a first VL domain, (iia) a first polypeptide linker, and (iiiia) a second VL domain; and where the specified second single-chain protein contains a second amino acid sequence consisting of (from N- to C-terminus):

(ib) a first VH domain, (iib) a second polypeptide linker, and

- 5 041608 (iiib) second VH domain; and where said first VL domain forms a first cognate pair of variable domains with specificity for a first target antigen with any of said first or said second VH domain, and said second VL domain forms a second cognate pair of variable domains with specificity for a second target antigen with another of said VH domains, and wherein at least one of said first or second single chain protein further comprises (iv) at least one additional domain as a third functional domain that is fused via a third polypeptide linker to said first or said second amino acid sequence.

In this description and the claims that follow, unless the context otherwise requires, the word contain and its variations such as contains and containing, should be understood to mean the inclusion of the specified integer, composition or step, or group of integers or steps, in while any additional integer, composition or step or group of integers, compositions or steps may also optionally be present, including embodiments in which no additional integer, composition or step or group of integers, compositions or steps are present or stages. With respect to such latter embodiments of the present invention, the term containing therefore includes the narrower term consisting of.

Several documents are cited throughout the text of this specification. Each of the documents cited in this specification (including all patents, patent applications, scientific publications, manufacturer's specifications, instructions, GenBank sequence numbers, etc.), regardless of location in the text, is hereby incorporated by reference in its entirety. to the extent possible under the relevant patent laws. Nothing in this document should be construed as an admission that the invention has no right to oppose such disclosure on the basis of prior invention.

In the context of the present invention, the terms VL domain and VH domain refer to the light chain variable domain and the heavy chain variable domain, respectively, of antibodies. In the context of the present invention, the term antibody refers to immunoglobulin molecules and immunologically active regions of immunoglobulin molecules, i.e. to molecules that contain an antigen-binding site that specifically binds to an antigen, ie. comprising sections of antibodies containing at least one antigen-binding antibody fragment.

In the context of the present invention, an antibody or any binding molecule is generally considered to specifically bind to an antigen (in the case of an antibody) or to a cognate binding partner (in the case of a binding molecule in general) if it has a dissociation constant K _D with the specified antigen/cognate binding partner as a target, 100 μM or less, preferably 50 μM or less, preferably 30 μM or less, preferably 20 μM or less, preferably 10 μM or less, preferably 5 μM or less, more preferably 1 μM or less, more preferably 900 nM or less, more preferably 800 nM or less, more preferably 700 nM or less, more preferably 600 nM or less, more preferably 500 nM or less, more preferably 400 nM or less, more preferably 300 nM or less, more preferably 200 nM or less, even more preferably 100 nm or less, even more preferably 90 nm or less, even more e preferably 80 nM or less, even more preferably 70 nM or less, even more preferably 60 nM or less, even more preferably 50 nM or less, even more preferably 40 nM or less, even more preferably 30 nM or less, even more preferably 20 nM or less, and even more preferably 10 nM or less.

In the context of the present invention, the term functional domains refers to a protein domain having a specific function, such as enzymatic activity or specific binding to a cognate ligand, where said protein domain is a structured domain having at least a secondary structural element. Methods for determining the presence of secondary structure in polypeptides or proteins, such as X-ray crystallography, circular dichroism (CD), vibrational circular dichroism (VCD), NMR, or FT-IR, or for predicting the presence of secondary structure in polypeptides, such as PEP-FOLD ( Shen et al., J. Chem. Theor. Comput. 10 (2014) 4745-4758) are well known to those skilled in the art. In specific embodiments, said protein domain is a structured domain having a tertiary structure. In specific embodiments, said protein domain contains at least about 20 amino acid residues (see Heitz et al., Biochemistry 38 (1999) 10615-25), in particular at least about 50 amino acid residues, more specifically at least about 100 amino acid residues.

In the context of the present invention, the term polypeptide linker refers to a linker, consisting of a chain of amino acid residues connected by peptide bonds, which connects two domains, each of which is attached to one end of the linker. In specific cases,

This polypeptide linker has a continuous chain of 2-30 amino acid residues (e.g. 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29 or 30 amino acid residues). In specific embodiments, the polypeptide linker is not a structured polypeptide. As mentioned above, methods for determining the presence of secondary structure in polypeptides, such as X-ray crystallography, circular dichroism (CD), vibrational circular dichroism (VCD), NMR or FT-IR, or for predicting the presence of secondary structure in polypeptides, such as PEP -FOLD (Shen et al., J. Chem. Theor. Comput. 10 (2014) 4745-4758) are well known to those skilled in the art.

The present invention is characterized as follows.

Using antibody variable domains to create a heterodimeric format where both VLs are located on the same protein chain while the corresponding VHs are located on the second protein chain.

The heterodimer creation domain allows the addition of additional functional domains, such as binding domains, to create tri-, tetra-, penta- or hexa-specific molecules.

Numerous examples of highly efficient pairing in assemblies of heterodimerizing bases.

A simple solution for combinatorial screening of multiple pools of binding domains that share a common domain to create a heterodimer.

In a specific embodiment, the present invention provides a heterodimeric protein wherein said first or second single chain protein further comprises (v) a fourth functional domain that is fused via a fourth polypeptide bridge to said first or said second amino acid sequence.

In a specific embodiment, the present invention provides a heterodimeric protein wherein said first or second single chain protein further comprises (vi) a fifth functional domain that is fused via a fifth polypeptide bridge to said first or said second amino acid sequence.

In a specific embodiment, the present invention provides a heterodimeric protein wherein said first or second single chain protein further comprises (vii) a sixth functional domain that is fused via a sixth polypeptide bridge to said first or said second amino acid sequence.

In specific embodiments, said heterodimeric protein comprises said third and said fourth functional domains. In such embodiments, said heterodimeric protein is tetravalent, in particular embodiments, said heterodimeric protein is tetraspecific.

In specific embodiments, said heterodimeric protein comprises said third, said fourth, said fifth, and said sixth functional domains. In such embodiments, said heterodimeric protein is hexavalent, in particular embodiments, said heterodimeric protein is hexaspecific.

In specific embodiments, said heterodimeric protein does not contain a first and second immunoglobulin constant domain cognate pair, wherein said first immunoglobulin constant domain is contained in said first single chain protein, and wherein said second immunoglobulin constant domain is contained in said second single chain protein. In specific embodiments, at least one of said first and said second single chain proteins does not contain an immunoglobulin constant domain. In a specific embodiment, both said first and said second single chain proteins do not contain a constant immunoglobulin domain.

In specific embodiments, said heterodimeric protein does not contain a cognate pair of a first protein interaction domain contained in said first single chain protein and a second protein interaction domain contained in said second single chain protein, other than the cognate pairs of (i) said first VL domain and said first VH domain, and (ii) said second VL domain and said second VH domain.

In specific embodiments, said first polypeptide linker consists of 5-20 amino acid residues, in particular 6-15 amino acid residues. In specific embodiments, said polypeptide linker has the sequence (G _m S) _n ; wherein m is independently selected from 2, 3 and 4; and choose from 1, 2, 3, 4 and 5.

In private other embodiments of the invention, the specified first polypeptide linker consists of 11-20 amino acid residues, in particular 11-15 amino acid residues. In specific embodiments, said polypeptide linker has the sequence (GmS)n; where m is independently selected from 2, 3 and 4 and n is selected from 3, 4 and 5.

In particular embodiments of the invention, said second polypeptide linker consists of 5-20 amino acid residues, in particular 6-15 amino acid residues. In specific embodiments, said polypeptide linker has the sequence (G _m S) _n ; and

- 7 041608 m independently selected from 2, 3 and 4; and n is selected from 1, 2, 3, 4 and 5.

In private other embodiments of the invention, the specified second polypeptide linker consists of 11-20 amino acid residues, in particular 11-15 amino acid residues. In specific embodiments, said polypeptide linker has the sequence (G _m S) _n ; where m is independently selected from 2, 3 and 4 and n is selected from 3, 4 and 5.

In particular embodiments of the invention, said third, fourth, fifth and/or sixth polypeptide linkers independently consist of 8-20 amino acid residues, in particular 10-15 amino acid residues. In specific embodiments, said polypeptide linkers independently have the sequence (GmS)n; where m is independently selected from 2, 3 and 4, in particular from 4; and n is selected from 1, 2, 3, 4 and 5, in particular from 2 and 3.

In specific embodiments, said first VL domain (ia) and said first VH domain (ib) form a first cognate pair of variable domains with specificity for the first target antigen, and said second VL domain (iia) and said second VH domain (iib ) form a second cognate pair of variable domains with specificity for the second target antigen. In such an embodiment, said first and said second single chain protein form said heterodimeric protein in a parallel arrangement of said single chain proteins.

In particular such embodiments of the invention, said first polypeptide linker consists of 10-20 amino acid residues, in particular 12-17 amino acid residues, in particular 15 amino acid residues. In specific embodiments of the invention, the specified polypeptide linker has the sequence (G _m S) _n ; where m is independently selected from 2, 3 and 4, in particular from 4; and n is selected from 1, 2, 3, 4 and 5, in particular from 3.

In particular such embodiments of the invention, said second polypeptide linker consists of 10-20 amino acid residues, in particular 12-17 amino acid residues, in particular 15 amino acid residues. In specific embodiments of the invention, the specified polypeptide linker has the sequence (G _m S) _n ; where m is independently selected from 2, 3 and 4, in particular from 4; and n is selected from 1, 2, 3, 4 and 5, in particular from 3.

In particular such embodiments of the invention, said third, fourth, fifth and/or sixth polypeptide linkers independently consist of 10-20 amino acid residues, in particular 12-17 amino acid residues, in particular 15 amino acid residues. In specific embodiments of the invention, the specified polypeptide linker has the sequence (G _m S) _n ; where m is independently selected from 2, 3 and 4, in particular from 4; and n is selected from 1, 2, 3, 4 and 5, in particular from 3.

In particular other embodiments, said first VL domain (ia) and said second VH domain (iib) form a first cognate pair of variable domains with specificity for the first target antigen, and said second VL domain (iia) and said first VH- domain (ib) form a second cognate pair of variable domains with specificity for the second target antigen. In such an embodiment, said first and said second single chain protein form said heterodimeric protein in an antiparallel arrangement of said single chain proteins.

In particular such embodiments of the invention, said first polypeptide linker consists of 5-12 amino acid residues, in particular 5-10 amino acid residues, in particular 6 amino acid residues. In specific embodiments of the invention, the specified polypeptide linker has the sequence (G _m S) _n ; wherein m is independently selected from 2, 3 and 4, in particular from 2; and n is selected from 1, 2, 3, 4 and 5, in particular from 2.

In particular such embodiments of the invention, said second polypeptide linker consists of 5-12 amino acid residues, in particular 6-10 amino acid residues, in particular 8 amino acid residues. In specific embodiments of the invention, the specified polypeptide linker has the sequence (G _m S) _n ; wherein m is independently selected from 2, 3 and 4, in particular from 3; and n is selected from 1, 2, 3, 4 and 5, in particular from 2.

In particular such embodiments of the invention, said third, fourth, fifth and/or sixth polypeptide linkers independently consist of 10-20 amino acid residues, in particular 8-12 amino acid residues, in particular 10 amino acid residues. In specific embodiments of the invention, the specified polypeptide linker has the sequence (G _m S) _n ; where m is independently selected from 2, 3 and 4, in particular from 4; and n is selected from 1, 2, 3, 4 and 5, in particular from 2.

In particular embodiments of the invention, said third, fourth, fifth, and or sixth functional domains are independently selected from a list of binding domains, toxins, enzymes, hormones, signaling proteins, and albumins.

In particular embodiments of the invention, said third, fourth, fifth and/or sixth functional domains are independently selected from binding domains.

In particular such embodiments, binding domains are independently selected from a list of antibody-based binding domains, including, but not limited to, scFv, Fab, and single antibody variable domains, shark VNAR-based single-domain antibodies, and alternative scaffold-based binding domains. , including but limited to ankyrin-based domains, finomers, avimers, anticalins, fibronectins, and binding sites built into antibody constant regions (eg, by F-star technology).

In particular such embodiments, said binding domains are antibody-based binding domains selected from single chain Fv fragments and single antibody variable domains.

In some such embodiments, the order of the variable domains in such a single chain Fv fragment is selected from (N-terminus to C-terminus) VL-(linker)-VH and VH-(linker)-VL. In some embodiments, the order of the variable domains is the same for all single chain Fv fragments contained in the heterodimeric protein. In some embodiments, the three VL domains are connected to each other via said first polypeptide linker and one of said third, fourth, and fifth polypeptide linkers, respectively, e.g. from said first amino acid sequence. In some embodiments, the three VH domains are connected to each other via said second polypeptide linker and one of said third, fourth, and fifth polypeptide linkers, respectively, e.g. the specified second amino acid sequence (see Fig. 1 and 4). Thus, in some embodiments, at least one of said first and said second single chain proteins comprises an amino acid sequence consisting of three VL domains or three VH domains, respectively, connected by two polypeptide linkers.

In some other embodiments, the variable domain of any such antibody-based binding domain that is directly linked via an appropriate linker to the N- and/or C-terminus of said first or second amino acid sequence is: (a) a VH domain when fused to said a first amino acid sequence, and (b) a VL domain when fused to said second amino acid sequence. Thus, the VH domain is fused to the N- and/or C-terminus of the VL-linker-VL core region, and the VL domain is fused to the N- and/or C-terminus of the VH-linker-VH core region (see, for example, FIG. .3).

In specific embodiments, said third, fourth, fifth and/or sixth binding domains are single chain Fv fragments.

In particular such embodiments, the polypeptide linker connecting the variable domains of said single chain Fv fragments consists of 15-25 amino acid residues, in particular 20 amino acid residues. In specific embodiments of the invention, said polypeptide linker has the sequence (GGGGS) _n , n being selected from 3, 4 and 5, in particular from 4.

In specific embodiments, at least one of said antibody variable domains contains CDR regions derived from a parental rabbit antibody.

In specific embodiments, at least one of said antibody variable domains comprises human antibody framework regions.

In particular such embodiments of the invention, at least one of these VL domains contains: (i) framework regions I-III from human Vk; (ii) the CDR domains of CDR1, CDR2 and CDR3; and (iii) a framework region IV which is selected from:

a) a human Vλ germline sequence for frame IV, in particular a Υλ germline sequence selected from the list of SEQ ID NOs 16-22 according to WO 2014/206561;

b) a VX-based sequence that is (bi) a Υλ consensus sequence from human Υλ germline sequences for framework IV, in particular SEQ ID NO 17 according to WO 2014/206561; or (bii) a Υλ consensus sequence from rearranged human Υλ sequences for framework IV, in particular a VX consensus sequence selected from the list of SEQ ID NOs 16 and 17 according to WO 2014/206561; And

c) a VX-based sequence that has one or two mutations, in particular one mutation, compared to the nearest human Υλ germline sequence for the IV framework.

In some embodiments, a cognate pair of one of said first and said second VL and VH domains is specific for an antigen selected from a list of cancer targets; and a target present on immune effector cells such as CD3.

In particular such embodiments, said third, fourth, fifth and/or sixth binding domains are single chain Fv fragments with a specificity for a target selected from the list cancer target; and a target present on immune effector cells such as CD3.

In the context of this application, the term target refers to a cognate binding partner

- 9 041608 binding domain, such as an antigen of an antibody that specifically binds to such a binding domain.

In specific embodiments, said target is a cancer target, in particular an antigen or epitope that is present on the surface of one or more types of tumor cells or tumor-associated cells at an increased concentration and/or in a different steric configuration compared to the surface of non-tumor cells. In particular, said cancer target is present on the surface of one or more types of tumor cells or tumor stromal cells, but not on the surface of non-tumor cells.

In other specific embodiments, said target is an antigen or epitope that is predominantly expressed on cells involved in autoimmune diseases. In other embodiments of the invention, the specified antigen or epitope is predominantly expressed in cells involved in an inflammatory disease.

In specific embodiments, said target is a target present on immune effector cells. In specific embodiments, said target is CD3.

In some embodiments, said first and said second single chain protein are selected from the following list, where VLA, VLB, VHA, and VHB correspond to said first and second VL and VH domains, and VLC, VLD, VLE, VLF, VHC, VHD, VHE and VHF are part of single-stranded fragments with a linker corresponding to the specified third, fourth, fifth and/or sixth functional domains, respectively, linked via third, fourth, fifth and/or sixth linkers (LINKER3, LINKER4, LINKER5 and LINKER6) to the main domain (highlighted in bold); all constructs are written in the direction from the N-terminus to the C-terminus.

A (parallel arrangement 6Fv):

Circuit 1: VLC-(linker)-VHC-(LINKERS )-VLA-(LINKER1)-VLB(LINKER4)-VLD-(linker)-VHD

Circuit 2: VLE-(linker)-VHE-(LINKERS)-VHA-(LINKER2)-VHB(LINKERS)-VLF-(linker)-VHF

B (anti-parallel arrangement 6Fv):

Circuit 1: VLC-(linker)-VHC-(LINKERS )-VLA-(LINKER1)-VLB(LINKER4)-VLD-(linker)-VHD

Circuit 2: VLE-(linker)-VHE-(LINKER52)-VHB-(LINKER2)-VHA(LINKERS)-VLF-(linker)-VHF

Cl (anti-parallel arrangement 4 Fv) (see figure 1):

Circuit 1: VLC-(linker)-VHC-(LINKERS )-VLA-(LINKER1)-VLB

Circuit 2: VLD-(linker)-VHD-(LINKER4)-VHB-(LINKER2)-VHA

C2 (anti-parallel arrangement 4 Fv) (see fig.3):

Circuit 1: VLC-(linker)-VHC-(LINKERS )-VLA-(LINKER1)-VLB

Circuit 2: VHB-(LINKER2)-VHA-(LINKER4)-VLD-(linker)-VHD

NW (anti-parallel arrangement 4 Fv):

Circuit 1: VLA-(LINKER1)-VLB-(LINKERS )-VLC-(linker)-VHC

Circuit 2: VLD-(linker)-VHD-(LINKER4)-VHB-(LINKER2)-VHA

C4 (anti-parallel arrangement 4 Fv):

Circuit 1: VLA-(LINKER1)-VLB-(LINKERS )-VLC-(linker)-VHC

Circuit 2: VHB-(LINKER2)-VHA-(LINKER4)-VLD-(linker)-VHD

DI (parallel arrangement 4 Fv) (see figure 4):

Circuit 1: VLC-(linker)-VHC-(LINKERS )-VLA-(LINKER1)-VLB

Circuit 2: VLD-(linker)-VHD-(LINKER4)-VHA-(LINKER2)-VHB

D2 (parallel arrangement 4 Fv):

Circuit 1: VLC-(linker)-VHC-(LINKERS )-VLA-(LINKER1)-VLB

Circuit 2: VHA-(LINKER2)-VHB-(LINKER4)-VLD-(linker)-VHD

D3 (parallel arrangement 4 Fv):

Circuit 1: VLA-(LINKER1)-VLB-(LINKERS)-VLC-(linker)-VHC

Circuit 2: VLD-(linker)-VHD-(LINKER4)-VHA-(LINKER2)-VHB

D4 (parallel arrangement 4 Fv):

Circuit 1: VLA-(LINKER1)-VLB-(LINKERS)-VLC-(linker)-VHC

Circuit 2: VHA-(LINKER2)-VHB-(LINKER4)-VLD-(linker)-VHD

In this format, localization of two separated VHB and VHC heavy chain variable domains on one protein chain and two corresponding VLB and VLC light chain variable domains on another protein chain (VH-VH/VL-VL) prevents pairing between domains on the same chain,

- 10 041608 resulting in inactive single-stranded diabody-like (scDb) structures, as would be the case if the VH-VL/VH-VL orientation of a conventional diabody was used to provide heterodimerization, similar to the design proposed by Kupriyanov et al. In contrast, the VH-VH/VL-VL orientation allows the formation of exclusively heterodimeric proteins capable of specifically binding 2 to 6 targets.

There is a theoretical possibility that pairing between the VH/VL binding target A and B domains of the VHA-VHB/VLA-VLB core domain will result in an inactive core domain due to mispairing of VHA to VLB and VHB to VLA, resulting in pairing VHA-VLB and VHB-VLA. Unexpectedly and surprisingly, such inactive variants have not yet been discovered. Without being bound to any particular hypothesis, dimerization can be shifted towards cognate pairing due to more efficient packing of the CDRs of cognate pairs, as opposed to potential packing interference occurring in mismatched pairs.

In order to further enhance heterodimerization towards active pairing in the main VH-VH/VL-VL domain, bulge-hollow techniques or similar techniques can be applied to a single VL/VH pair (or reciprocally applied to both VL/VH pairs) main domain VH-VH/VL-VL. Thus, in some embodiments, the formation of active pairs in the main VH-VH/VL-VL domain of said heterodimeric protein is further enhanced by a technology selected from: bulge-hollow technology and interstrand cysteine bridges.

In a second aspect, the present invention provides one or two nucleotide sequences encoding said first and second single chain proteins.

In a third aspect, the present invention provides one or two vectors containing said one or two nucleotide sequences.

In a fourth aspect, the present invention relates to a host cell or host cells containing one or two vectors.

In the fourth aspect, the present invention relates to a method for obtaining the first and second single-stranded proteins or heterodimeric protein according to the present invention, including: (i) obtaining a nucleic acid or nucleic acids of the present invention, or a vector or vectors of the present invention, the expression of these nucleic acids or nucleic acids or said vector or vectors, and collecting said first and second single chain proteins or said heterodimeric protein from the expression system; or (ii) obtaining a host cell or host cells of the present invention, culturing said host cell or host cells, and harvesting said first and second single chain proteins or said heterodimeric protein from the cell culture.

In a fifth aspect, the present invention relates to a pharmaceutical composition comprising a heterodimeric protein of the present invention and a pharmaceutically acceptable carrier.

In a sixth aspect, the present invention relates to a heterodimeric protein of the present invention for use in the treatment of a disease selected from a cancerous, inflammatory, and autoimmune disease, wherein at least one of said VL and VH domain cognate pairs or said third, fourth, fifth, or of the sixth functional domain is able to specifically interact with a target of therapeutic significance in the corresponding disease.

In a seventh aspect, the present invention provides a method for treating a patient suffering from a disease selected from a cancerous, inflammatory, and autoimmune disease, comprising administering to the subject an effective amount of a heterodimeric protein of the present invention, wherein at least one of said VL- and VH cognate pairs α-domains or said third, fourth, fifth or sixth functional domain is capable of interacting specifically with a target of therapeutic relevance in the respective disease.

- 11 041608

Literature.

Rl. Skerra, A. and Pluckthun, A. (1988). Assembly of a functional immunoglobulin Fv fragment in Escherichia coli.

Science 240, 1038-1041.

R2. Rbthlisberger et al., (2005). Domain interactions in the Fab fragment: A comparative evaluation of the single-chain Fv and Fab format engineered with variable domains of different stability. J Mol Biol 347, 773-789.

R3. Ridgway et al., 1996. 'Knobs-into-holes' engineering of antibody CH3 domains for heavy chain heterodimerization. ProteinEng. 9, 617-621.

R4. Zhu (1997) Remodeling domain interfaces to enhance heterodimer formation. Protein Sci. 6, 781-788

R5. Schaefer, W., et al., 2011b. Immunoglobulin domain crossover as a generic approach for the production of bispecific IgG antibodies. Proc. Natl. Acad. sci. U.S.A. 108, 11187-11192.

R6. Holliger et al. Diabodies: small bivalent and bispecific antibody fragments. Proc. Natl. Acad. sci. U.S.A. 90, 6444-6448.

R7. Arndt et al., 1999. Abispecific diabody that mediates natural killer cell cytotoxicity against xeno-transplanted human Hodgkin's tumors. Blood 94, 2562-2568.

R8. Kipriyanov et al., 1999. Bispecific tandem diabody for tumor therapy with improved antigen binding and pharmacokinetics. J. Mol. Biol. 293, 41-56.

R9. Alt et al., 1999. Novel tetravalent and bispecific IgGlike antibody molecules combining single-chain diabodies with the immunoglobulin gammal Fc or CH3 region. FEBS Lett. 454, 9094.

R10. Johnson et al., 2010. Effector cell recruitment with novel Fv-based dual-affinity re-targeting protein leads to potent tumor cytolysis and in vivo B-cell depletion. J. Mol. Biol.399, 436-449.

Rll. De Jonge et al., (1995) Production and characterization of bispecific single-chain antibody fragments. Mol. Immunol. 32, 1405-1412.

R12. Reiter et al., (1994) Engineering interchain disulfide bonds into conserved framework regions of Fv fragments: improved biochemical characteristics of recombinant immunotoxins containing disulfide-stabilized Fv. ProteinEng. 7, 697-704.

R13. Pack, P., and Pluckthun, A. (1992). Miniantibodies: Use of amphipathic helices to produce functional, flexibly linked dimeric Fv fragments with high avidity in Escherichia coli. Biochemistry 31, 1579-1584.

R14. Schoonjans et al., Fab chains as an efficient heterodimerization scaffold for the production of recombinant bispecific and trispecific antibody derivatives. J Immunol. 2000 Dec 15,-165(12):7050-7 .

R15. Orcutt et al., 2009. A modular IgG-scFv bispecific antibody topology. ProteinEng. Des. Sei. 23, 221-228.

R16. Wu, C. et al., 2007. Simultaneous targeting of multiple disease mediators by a dual-variable-domain immunoglobulin. Nat.Biotechnol. 25, 1290-1297.

R17. mAbs; Kohler & Milstein, Nature. 256 (1975) 495-7

R18. Umana et al., 1999. Engineered glycoforms of an antineuroblastoma IgGl with optimized antibody-dependent

- 12 041608 cellular cytotoxic activity. Nat. Biotechnol. 17, 176-180

R19. Yu, Y. J. et al. sci. Trans. Med. 3, 84ra44 (2011).

R20. Hinton P. R. et al., 2004. Engineered human IgG antibodies with longer serum half-lives in primates. J Biol Chern. 279(8):6213-6.

R21. Spiess et al., 2015. Alternative molecular formats and therapeutic applications for bispecific antibodies. Mol Immunol. 2015 Jan 27.

R22. Davis et al., 2013. Readily isolated bispecific antibodies with native immunoglobulin format. U.S. Patent 8,586,713. Regeneron Pharmaceuticals Inc.

R23. Shahied LS, et al., Bispecific minibodies targeting

HER2/neu and CD16 exhibit improved tumor lysis when placed in a divalent tumor antigen binding format. J Biol Chern. 2004 Dec 24,-279(52):53907-14. Epub 2004 Oct 7.

R24. Milstein.C and Cuello.A.C. (1983) Nature, 305, 537-54

R25. Chang et al., The dock and lock method: a novel platform technology for building multivalent, multifunctional structures of defined composition with retained bioactivity.

Clinic Cancer Res. 2007 Sep 15,-13(18 Pt 2) : 558 6s-5591s.

R26. Deyev et al., (2003). Design of multivalent complexes using the barnase barstar module. Nature biotechnology, 21(12), 1486-1492.

R27. Pack, P., and Pluckthun, A. (1992). miniantibodies:

Use of amphipathic helices to produce functional, flexibly linked dimeric Fv fragments with high avidity in Escherichia coli. Biochemistry 31, 1579-1584.

R28. Halin et al. (2003). Synergistic therapeutic effects of a tumor targeting antibody fragment, fused to interleukin 12 and to tumor necrosis factor a. Cancer research, 63(12), 32023210.

R29. D. Muller et al., Improved pharmacokinetics of recombinant bispecific antibody molecules by fusion to human serum albumin J. Biol. Chern., 282 (2007), pp. 12650-12660

R30. EP1293514

R31. Milstein C, and Cuello AC (1983) Hybrid hybridomas and their use in immunohistochemistry. Nature 305:537-540

Examples

Example 1. Construction of polyspecific formats.

Four well-characterized variable domains targeting human tumor necrosis factor (TNF) alpha, human interleukin-5 receptor (IL5R ), the human CD3 epsilon chain (CD3), and the interleukin-23 receptor (IL23R), respectively. Based on the known binding characteristics of the respective variable domains in the scFv format, the activity and thus the correct association of the VL/VH cognate pairs was evaluated in the context of multispecific molecules. The corresponding variable domains at the periphery of the molecule were located either at the amino or carboxy terminus of each protein chain as single chain Fv fragments (scFv), or were located in the main heterodimerizing domain. Unlike peripheral scFv fragments, for which VL and VH are located on the same protein chain, the cognate variable domains VL and VH of the main domain are located on two different protein chains. In the examples below, the target-binding domains located in the two main domains bind to CD3 or TNF, respectively. IL23R or IL5R binding variable domains were used for peripheral scFv modules that were fused to the N- or C-terminus of the main domain using a flexible 10 or 15 amino acid linker.

In order to explore different assembly options for the core heterodimer of the present invention, parallel as well as anti-parallel orientations of cognate pairs of variable domains were designed simultaneously, each having one or two additional scFv modules connected to the N or C terminus of the core domain.

In the anti-parallel arrangement, the main domain was designed in the VHA orientation

- 13 041608

VHB/VLB-VLA from the N-terminus to the C-terminus of each protein chain (protein chains 1 to 9). In one embodiment, the tetraspecific format is formed by the N-terminal fusion of one scFv module to each of two protein chains (1+2 protein chain constructs). The corresponding trispecific format contains the scFv module fused to only one of the two protein chains (constructs 1+5). In order to investigate the possible effects of stabilizing core domain assembly by genetically engineered introduction of disulfide bridges, the two formats described above were also made with a C-terminal cysteine, resulting in cross-linking of cognate Fvs in the core domain of each protein chain. The corresponding heterodimeric formats consist of 3+4 protein chains for the tetraspecific format and 4+6 protein chains for the trispecific format. In an anti-parallel arrangement, the scFv module located on the strand containing the VH tandem in the main domain was fused at the C-terminus instead of the N-terminus and was fused to the protein strand containing the assembled scFv module at the N-terminus, resulting in a tetraspecific format (protein chains 1+7), or with a protein chain containing only the core domain, resulting in a trispecific format (protein chains 5+7).

In a parallel arrangement, the core domain was designed in the VHA-VHB/VLAVLB orientation from the N-terminus to the C-terminus of each protein chain. A tetraspecific format with both scFv modules fused to the N-terminal side of the core domains was generated by co-expression of 9+10 protein chains. The corresponding trispecific assembly with the scFv module exclusively on the strand containing the tandem VH was generated by co-expression of the 10+11 protein chains.

To create the structures described in Table. 1, the amino acid sequences for the Fv domains and linkers were translated back into the corresponding nucleotide sequences, which were synthesized de novo.

The coding sequences were assembled and cloned using standard molecular biology techniques (see, e.g., Sambrook, J., et al., Molecular Cloning: A Laboratory Manual) into an appropriate expression vector (e.g., pcDNA3.1, Invitrogen) for secretion of the recombinant squirrel.

Example 2 Expression and purification.

Expression of the polyspecific format assemblies was performed by co-transfection of the constructs into a suspension cell line (eg, CHO-S Freestyle™, Invitrogen) using a transient gene expression protocol (Freestyle™ MAX system). The combination of co-expressed expression vectors to create assemblies of polyspecific formats is described in table. 2. After culturing for several days, the supernatant of cells secreting antibody fragments was collected for purification. Proteins were adsorbed onto a suitable affinity resin (eg Capto L., GE Healthcare), washed extensively and eluted with a pH shift. The eluted proteins were neutralized and buffer exchanged to give purified pools. Proteins were analyzed by size exclusion high performance liquid chromatography (SE-HPLC) (Table 3 and Fig. 5) and sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 6) to assess purity and by UV visible spectroscopy to assess protein content. The protein concentration was adjusted to the required level and a stability analysis was performed.

Using a one-step affinity chromatography procedure, all constructs could be eluted in a high purity and monomeric fraction (FIGS. 5 and 6), confirming efficient and correct pairing of cognate variable domains. In addition, on SDS-PAGE under non-reducing conditions, the PRO357 fraction migrated almost quantitatively at a covalently bound heterodimer size (about 106 kDa), maintaining proper inter-chain bonds of MATCH and demonstrating high efficiency and almost complete formation of an inter-chain disulfide bond. Due to structural limitations, the formation of this disulfide bond between mismatched variable domains is very unlikely. Thus, this result suggests that heterodimerization occurred almost exclusively between cognate pairs of variable domains.

Example 3 Evaluation of storage stability.

Efficient dimerization of MATCH chains was further demonstrated by the remarkable homogeneity of protein content in protein L-purified samples. The protein was analyzed over four weeks of storage at 4 and 37°C for oligomerization by SEHPLC and degradation by SDS-PAGE (see Fig. 7-9). Prior to the start of the study, the concentration of the samples was adjusted to 1 g/l and the time point t0 was noted. The monomer content was quantified by splitting the samples on Shodex KW-402.5-4F (Showa Denko) and evaluating the obtained chromatograms. To calculate the relative percentage of protein monomer, the area of the monomeric peak was divided by the total area of the peaks that could not be assigned to the sample matrix. Protein degradation was assessed by SDS-PAGE analysis on Any kD™ Mini-Protean TGX gels (Bio-Rad Laboratories) and stained with Coomassie brilliant blue. The protein concentration was monitored at different time points using UV-visible spectroscopy on an Infinite M200 Pro spectrophotometer equipped with

- 14 041608 Nanoquant plate (Tecan Group Ltd.).

Example 4 Thermal Denaturation

The median transition point for thermal denaturation of the tested constructs was determined by differential scanning fluorometry (DSF), essentially as described by Niesen (Niesen et al., Nat Protoc. 2 (2007) 2212-21). DSF analysis is performed on a quantitative PCR instrument (eg, MX3005p, Agilent Technologies). Samples were diluted in buffer (citrate-phosphate buffer, pH 6.4, 0.25 M NaCl) containing a final concentration of 5x Sypro orange in a total volume of 25 μl. Measurements of the samples were carried out in triplicate, and the mode of linear temperature change 25-96°C was programmed. The fluorescent signal was recorded and the raw data was analyzed using GraphPad Prism (GraphPad Software Inc.).

Example 5 Affinity Determination.

Binding affinities of individual target-binding domains in single-chain Fv(scFv) format, as well as purified heterodimeric tetraspecific constructs, with recombinant target proteins - human IL-5 receptor (IL5R), extracellular domain of human IL-23 receptor (IL23R), monochain gamma human CD3 epsilon (CD3) were measured by surface plasmon resonance (SPR) using the MASS-1 SPR instrument (Sierra Sensors). For affinity measurements (performed in running HEPES buffer: 0.01 M HEPES, 0.15 M NaCl, 0.05% Tween) target proteins - extracellular domain of heterodimeric single chain human CD3γδ (produced in the authors laboratory), human IL5R (R&D Systems), human IL23R (Trenzyme) and human TNF (Peprotech) were immobilized at 100-250 RU on a sensor chip (SPR-2 Affinity Sensor High Capacity Amine, Sierra Sensor) using buffer systems optimized for each individual target using standard amino linkage procedure. For human TNFa (TNF), a standard amine sensor was used. Two-fold serial dilutions of the purified heterodimeric tetraspecific constructs ranging from 90 to 0.703 nM were injected into the flow cell over 3 min (20 µl/min) and allowed to dissociate for 720 s. After each injection cycle, surfaces were regenerated with a 45 second second injection of 10 mM glycine-HCl, pH 1.5. Affinity was calculated by fitting sensorgrams for at least six concentrations such that the mean chi-square was less than 10% of R _max . No serial dilutions were made for TNF, only measurements at a single concentration of 90 nM. The background is subtracted from the data twice (reference channel and control cycle are subtracted).

The affinity of the heterodimeric tetraspecific binding constructs for each of the four targets is generally very close to the affinity of the individual binding domains (scFv) used in the tetraspecific format, including those CDRs whose immune reactivity is presumably dependent on proper dimerization (i.e., those , which are represented by forming Fv-dimers, which target TNFa and CD3ε, respectively). This demonstrates the full functionality of each variable domain in the tetraspecific constructs and validates the assembly of the variable domain cognate pairs.

In addition, each of the three polyspecific formats was able to bind all four target antigens simultaneously, apparently regardless of the order of interaction with the antigen, as demonstrated by SPR analysis of the immobilized MATCH protein (Fig. 11).

It is important to recognize that while these data indicate correct assembly between MATCH chains, they do not necessarily indicate the absence of non-cognate binding of variable domains, in particular inverted pairing between MATCH chains, which produces chimeric CDR pairs. It has been suggested that CDR pairing affects the efficiency of VLVH association, and our SE-HPLC, SDS-PAGE, and SPR data seem to suggest that cognate pairing in MATCH strands is highly favored. However, in an attempt to assess the degree of MATCH chain flip pairing, we performed SE-HPLC analysis of antibodies and antigen-antibody complexes after incubation of MATCH proteins with a molar equivalent of trimeric TNFa (i.e., a 3-fold excess of the TNFa epitope). Using this assay on the parental anti-TNFa scFv (data not shown), SE-HPLC chromatograms showed discrete peaks corresponding to three different antigen-antibody complex populations reflecting the unequal sizes of TNFa complexes with 1, 2 and 3 scFv. In addition, a peak was observed that is consistent with the presence of residual uncomplexed TNFa in solution, while uncomplexed scFv was completely absent from solution, confirming the use of this method to identify inactive anti-TNFa antibody.

Separation of MATCH protein and MATCH-antigen complexes was less effective due to the large molecular weight of polyspecific molecules. However, the authors' results (FIG. 12) also clearly showed the presence of three populations of MATCH-TNFa complexes and a residual non-complexed TNFa. In addition, the formation of shoulders at the peak of the 1xMATCH:TNFa complex suggests the presence of an inactive but dimeric MATCH protein. To estimate the proportion of inactive MATCH protein in solution, the peaks were deconvoluted using PeakFit v.1.2 software, assuming a Gaussian distribution for each peak, and plotted to optimize the degree of agreement (FIG. 12). This analysis showed that the proportion of inactive MATCH protein ranged from 4.7 to

11.4% (PRO357<PRO356<PRO355) of the total MATCH protein content.

Table 1

Constructions

proteins Linker 1 Main Domain Fv 1 Linker 2 Basic Fv 2 domain Linker 3 1 scFv (a!L23R) GGGGSGGGGS VL (aTNF'a) GGSGGS VL (aCD3) 2 scFv (aIL5R) GGGGSGGCGS VH (aCD3) GGSGGS VH (aTNFa) 3 scFv (aIL23R) GGGGSGGGGS VL (aTNFa) GGSGGS VL (aCD3) GSC 4 scFv (aIL5R) GGGGSGGGGS VH (aCD3) GGSGGS VH (aTNFa) GSC 5 VL {a™ Fa) GGSGGS VL (aCD3) 6 VL (aTNFa) GGSGGS VL (aCD3) GSC 7 VH (aCD3) GGGSGGGS VH (aTNFa) GGGGSGGGGS SCFv (alLSR) IN VL (aTNFa) GGSGGS VT, (aCD3) 9 scFv (aTL23R) GGGGSGGGGSGGGGS VL (aTNFa) GGGGSGGGGSGGGGS VL (aCD3) 10 scFv (a!L5R) GGGGSGGGGSGGGGS VH (aCD3) GGGGSGGGGSGGGGS VH (aTNFa) eleven VL (aTNFa) GGGGSGGGGSGGGGS VL (aCD3)

table 2

Assemblies of polyspecific formats

Squirrel ID (Numab) Assembly Protein chain 1 protein chain PRO356 1 (see Fig. 1) 1 2 PRO469 2 1 5 PRO357 3 (see fig. 2) 3 ⁴ ( PRO470 4 4 6 PRO358 5 (see Fig. 3) 1 7 PRO471 6 5 7 PRO353 7 (see Fig. 4) 9 10 PRO468 e 10 eleven

Table 3

Size exclusion chromatograms after 1-step purification

Squirrel ID (internal) Build ID Content monomer Figure PR0356 Assembly 1 93.9 5A PR0357 Assembly 3 94.4 5V PR0358 Build 5 93.9 5C PR0355 Assembly 7 90.4 5D

Table 4

Mean denaturation point for proteins determined by differential scanning fluorometry

Squirrel ID (interior) Build ID Tt [°C] PRO356 1 67.99 PRO469 2 67.24 PRO357 3 71.27 PRO410 4 70.34 ^: PRO358 5 68.51: PRO471 6 67.98 PRO355 7 67.33 PRO468 .......... 8 66, 67

Claims (5)

Table 5 Affinity of heterodimeric tetraspecific constructs Protein ID I Affinity I Affinity..........[Affinity I Affinity.........I scFvs PR0355 to IL5R [M] 2.32E-10 1.03E- 10 to CD3 M] 8 _? 57E-09 2.01E-08 to IL23R [M] 1.50E-10 6.54E-10 to TNF [M] 2.02E-10 3.3OE-10 PRO356 1.26E-10 7.14Е-09 2.41Ε-Ί; 2, 01E-10 PR0357 1.28E-10 b,69E-09 3.58E-10 1.81Е-10 PR0358 2.12E-10 5.60E-09 4.14Е-10 2.11E-10 CLAIM 1. An antibody-based heterodimeric construct containing at least the first and second cognate pairs of variable domains with specificity for the first and second target antigens, respectively, containing the first and second single-stranded proteins, where the specified first single-stranded protein contains the first amino acid sequence consisting of ( from N- to C-terminus): (ia) a first VL domain, (iia) a first polypeptide linker, and (iiiia) a second VL domain; and where the specified second single-chain protein contains a second amino acid sequence consisting of (from N- to C-terminus): (ib) a first VH domain, (iib) a second polypeptide linker, and (iiib) a second VH domain; and wherein said first VL domain forms a first cognate pair of variable domains with specificity for the first target antigen with any of said first or said second VH domain and said second VL domain forms a second cognate pair of variable domains with specificity for the second target antigen with the other of said VH domains, and wherein at least one of said first or second single chain protein further comprises (iv) at least one additional domain as a third functional domain that is fused via a third polypeptide linker to said first or said second amino acid sequence, where the specified heterodimeric construct does not contain: neither (i) a cognate pair of the first and second immunoglobulin constant domain, where the specified first immunoglobulin constant domain is contained in the specified first single-stranded protein, and where the specified second immunoglobulin constant domain is contained in the specified second single-stranded protein, nor (ii) any additional pair of heterodimerizing domains other than said first and second VL and VH domains (ia-iiiia) and (ib-iiib), in which one heterodimerizing domain of said additional pair of heterodimerizing domains is located in said the first single chain protein, and the other heterodimerizing domain is located in the second single chain protein. 2. Heterodimeric construction according to claim 1, additionally containing: (v) a fourth functional domain that is fused via a fourth polypeptide bridge to said first or said second amino acid sequence; (vi) fourth and fifth functional domains that are fused via fourth and fifth polypeptide linkers, respectively, to said first and/or said second amino acid sequence; or (vi) fourth, fifth, and sixth functional domains that are fused via fourth, fifth, and sixth polypeptide linkers, respectively, to said first and said second amino acid sequence. 3. Heterodimeric construct according to claim 1 or 2, wherein said first polypeptide linker consists of 5-20 amino acid residues, in particular 6-15 amino acid residues. 4. A heterodimeric construct according to any one of claims 1 to 3, wherein: (a) said first VL domain (ia) and said first VH domain (ib) form a first cognate pair of variable domains with specificity for the first target antigen, and said second VL domain (iiia) and said second VH domain (iiib) form a second cognate pair of variable domains with specificity for a second target antigen; or (b) said first VL domain (ia) and said second VH domain (iiib) form a first cognate pair of variable domains with specificity for the first target antigen, and said second VL domain (iiiia) and said first VH domain (ib) form a second cognate pair of variable domains with specificity for the second target antigen. 5. A heterodimeric construct according to any one of claims 1 to 3, wherein said third, fourth, fifth and/or sixth functional domains are independently selected from a list of binding domains, toxins, -