METHOD FOR IN VIVO GENERATION OF MULTISPECIFIC ANTIBODIES FROM MONOSPECIFIC ANTIBODIES

Abstract

Herein is reported a method for the generation of multispecific antibodies directly on the cell-surface at the site of action by a half-antibody exchange reaction between two 2/3-IgGs or two 2/3-BiFabs destabilized in one half by asymmetric perturbing mutations fostering the generation of correctly assembled full length bi- or multi-specific antibodies. The method is performed in the absence hinge region disulfide bonds in the starting 2/3-IgGs or 2/3-BiFabs.

Claims

1. A multimeric polypeptide comprising a first polypeptide comprising i) in N- to C-terminal direction a) a first antibody variable domain selected from a pair of an antibody light and heavy chain variable domain specifically binding to a first target, and b) a first human immunoglobulin G CH3 domain, ii) a pair of an antibody light and heavy chain variable domain specifically binding to a second target either N-terminal to the first antibody variable domain or C-terminal to the first CH3 domain, a second polypeptide comprising i) in N- to C-terminal direction a) a second antibody variable domain selected from a pair of an antibody light and heavy chain variable domain specifically binding to a third target, and b) a second human immunoglobulin G CH3 domain, wherein the second antibody variable domain is an antibody light chain variable domain if the first antibody variable domain is an antibody heavy chain variable domain; or the second antibody variable domain is an antibody heavy chain variable domain if the first antibody variable domain is an antibody light chain variable domain, and wherein the second CH3 domain comprises a perturbing mutation selected from the group of mutations consisting of D356K, E357K, K370E and K439E, whereby the first CH3 domain comprises a) the amino acid residue K at position 439 if the perturbing mutations is D356K, or b) the amino acid residue K at position 370 if the perturbing mutations is E357K, or c) the amino acid residue E at position 357 if the perturbing mutations is K370E, or d) the amino acid residue D at position 356 if the perturbing mutations is K439E, and ii) optionally a pair of an antibody light and heavy chain variable domain specifically binding to the second or a fourth target either N-terminal to the second antibody variable domain or C-terminal to the second CH3 domain, whereby all numbering is according to Kabat EU index.

2. A multimeric polypeptide comprising a first polypeptide comprising i) in N- to C-terminal direction a) a first human immunoglobulin G CH3 domain, and b) a first antibody variable domain selected from a pair of an antibody light and heavy chain variable domain specifically binding to a first target, ii) a pair of an antibody light and heavy chain variable domain specifically binding to a second target either N-terminal to the first CH3 domain or C-terminal to the first variable domain, a second polypeptide comprising i) in N- to C-terminal direction a) a second human immunoglobulin G CH3 domain and b) a second antibody variable domain selected from a pair of an antibody light and heavy chain variable domain specifically binding to a third target, wherein the second antibody variable domain is an antibody light chain variable domain if the first antibody variable domain is an antibody heavy chain variable domain; or the second antibody variable domain is an antibody heavy chain variable domain if the first antibody variable domain is an antibody light chain variable domain, and wherein the second CH3 domain comprises a perturbing mutation selected from the group of mutations consisting of D356K, E357K, K370E and K439E, whereby the first CH3 domain comprises a) the amino acid residue K at position 439 if the perturbing mutations is D356K, or b) the amino acid residue K at position 370 if the perturbing mutations is E357K, or c) the amino acid residue E at position 357 if the perturbing mutations is K370E, or d) the amino acid residue D at position 356 if the perturbing mutations is K439E, and ii) optionally a pair of an antibody light and heavy chain variable domain specifically binding to the second or a fourth target either N-terminal to the second CH3 domain or the second variable domain, whereby all numbering is according to Kabat EU index.

3. The multimeric polypeptide according to any one of claims 1 to 2, wherein the first CH3 domain and the second CH3 domain comprise further mutations to foster heterodimer formation between said first CH3 domain and said second CH3 domain and that are different from the perturbing mutation.

4. The multimeric polypeptide according to any one of claims 1 to 3, wherein the first CH3 domain comprises a) the mutation T366W, or b) the mutations T366S/L368A/Y407V, and the second CH3 domain comprises a) the mutations T366S/L368A/Y407V if the first CH3 domain comprises the mutation T366W, or b) the mutation T366W if the first CH3 domain comprises the mutations T366S/L368A/Y407V.

5. The multimeric polypeptide according to any one of claims 1 to 4, wherein the first polypeptide and the second polypeptide are a non-covalent dimer.

6. The multimeric polypeptide according to any one of claims 1 to 5, wherein the first variable domain and the second variable domain form a non-functional binding site.

7. The multimeric polypeptide according to any one of claims 1 to 6, wherein the first and the second polypeptide each comprise the amino acid sequence DKTHTSPPS (SEQ ID NO: 66) or DKTHT (SEQ ID NO: 94) or GGGS (SEQ ID NO: 69) or DKTHGGGGS (SEQ ID NO: 97) N-terminal to each of the first and second variable domains in case the first CH3 domain is located C-terminal to the first variable domain, or N-terminal to each of the first and second CH3 domains in case the first variable domain is located C-terminal to the first and second CH3 domain.

8. The multimeric polypeptide according to any one of claims 1 to 7, wherein i) the first CH3 domain comprises the mutation T366W and the amino acid residue K at position 439, and the second CH3 domain comprises the perturbing mutation D356K and the mutations T366S/L368A/Y407V, or ii) the first CH3 domain comprises the mutation and the amino acid residue K at position 370, and the second CH3 domain comprises the perturbing mutation E357K and the mutations T366S/L368A/Y407V, or iii) the first CH3 domain comprises the mutations T366S/L368A/Y407V and the amino acid residue E at position 357, and the second CH3 domain comprises the perturbing mutation K370E and the mutation T366W, or iv) the first CH3 domain comprises the mutations T366S/L368A/Y407V and the amino acid residue D at position 356, and the second CH3 domain comprises the perturbing mutation K439E and the mutation T366W.

9. The multimeric polypeptide according to any one of claims 1 to 8, wherein the first, second and third target are different.

10. The multimeric polypeptide according to any one of claims 1 to 9, wherein the first target and/or the third target is human CD3.

11. The multimeric polypeptide according to any one of claims 1 to 10, wherein the pair of an antibody light and heavy chain variable domain specifically binding to the second target is selected from the group consisting of Fv, scFc, Fab, scFab, dsscFab, CrossFab, bispecific Fab, sdAb, and VHH.

12. The multimeric polypeptide according to any one of claims 1 to 11, wherein the pair of an antibody light and heavy chain variable domain specifically binding to the fourth target is selected independently of the pair of an antibody light and heavy chain variable domain specifically binding to the second target from the group consisting of Fv, scFc, Fab, scFab, dsscFab, CrossFab, bispecific Fab, sdAb, and VHH.

13. The multimeric polypeptide according to any one of claims 1 to 12, wherein each of the first and the second polypeptide further comprises an immunoglobulin G CH2 domain directly N-terminal to the CH3 domain.

14. The multimeric polypeptide according to any one of claims 1 to 13, wherein the human immunoglobulin G is human IgG1 or human IgG2 or IgG3 or human IgG4.

15. A composition comprising a first multimeric polypeptide according to any one of claims 1 to 14 and a second multimeric polypeptide according to any one of claims 1 to 14, wherein the second CH3 domain of the first multimeric polypeptide comprises the mutation D356K and the second CH3 domain of the second multimeric polypeptide comprises the mutation K439E, or the second CH3 domain of the first multimeric polypeptide comprises the mutation E357K and the second CH3 domain of the second multimeric polypeptide comprises the mutation K370E, and the first antibody variable domain of the first multimeric polypeptide and the first variable domain of the second multimeric polypeptide are a pair of an antibody light chain variable domain and an antibody heavy chain variable domain that specifically bind to the first target, and the second antibody variable domain of the first multimeric polypeptide and the second variable domain of the second multimeric polypeptide are a pair of an antibody light chain variable domain and an antibody heavy chain variable domain that specifically bind to the third target, and the second and fourth target are independently of each other a cell surface antigen.

16. The composition according to claim 15, wherein the first CH3 domain of the first multimeric polypeptide and the second CH3 domain of the second multimeric polypeptide comprise the same mutations to foster heterodimer formation and the second CH3 domain of the first multimeric polypeptide and the first CH3 domain of the second multimeric polypeptide comprise the same mutations to foster heterodimer formation.

17. The composition according to any one of claims 15 to 16, wherein the first CH3 domain of the first polypeptide comprises a) the mutation T366W, or b) the mutations T366S/L368A/Y407V, and the second CH3 domain of the first polypeptide comprises a) the mutations T366S/L368A/Y407V if the first CH3 domain comprises the mutation T366W, or b) the mutation T366W if the first CH3 domain comprises the mutations T366S/L368A/Y407V.

18. The composition according to any one of claims 15 to 17, wherein the first CH3 domain of the first multimeric polypeptide and the second CH3 domain of the second multimeric polypeptide comprise the mutation T366W or the mutations T366S/L368A/Y407V, and the second CH3 domain of the first multimeric polypeptide comprises the mutation D356K and the second CH3 domain of the second multimeric polypeptide comprises the mutation K439E, or the second CH3 domain of the first multimeric polypeptide comprises the mutation E357K and the second CH3 domain of the second multimeric polypeptide comprises the mutation K370E, and the first antibody variable domain of the first multimeric polypeptide and the first variable domain of the second multimeric polypeptide are a pair of an antibody light chain variable domain and an antibody heavy chain variable domain that specifically bind to the first target, and the second antibody variable domain of the first multimeric polypeptide and the second variable domain of the second multimeric polypeptide are a pair of an antibody light chain variable domain and an antibody heavy chain variable domain that specifically bind to the third target, and the second and fourth target are independently of each other a cell surface antigen.

19. The composition according to any one of claims 15 to 18, wherein the first and/or the third target is human CD3.

20. The composition according to any one of claims 15 to 19, wherein the composition is a pharmaceutical composition and optionally further comprises a pharmaceutically acceptable excipient.

21. The multimeric polypeptide according to any one of claims 1 to 14 or the composition according to any one of claims 15 to 20 for use as a medicament.

22. A method of treatment comprising the administration of a multimeric polypeptide according to any one of claims 1 to 14 or a composition according to any one of claims 15 to 20 to a patient in need of such treatment.

Description

DESCRIPTION OF THE FIGURES

[1393] FIG. 1: Design and modular composition of exemplary 2/3-IgGs that can be used in the method according to the current invention.

[1394] FIG. 2: Interactions between knob-cys and hole-cys heavy chains (upper part) and knob-cys heavy chain and MHCFcRP (middle and lower part). The covalent disulfide bond is indicated with a dashed line, attractive interaction pairs are depicted with line between full spheres, repulsive interactions or resulting steric hindrance are indicated with double arrows lines.

[1395] FIGS. 3A, 3B, 3C and 3D: SEC chromatograms of the purified 2/3-IgGs with different MHCFcRPs: shown are SEC profiles of 2/3-IgG preparations following protein A extraction from cell culture supernatants; the main peak of each profile represent the 2/3-IgG; with fluorescein (fluos; anti-fluos) or biotin (bio; anti-bio) binding site/specificities (see Example 2). FIG. 3A: with D356K (hole);

[1396] FIG. 3B: with K370E (knob); FIG. 3C: with E357K (hole);

[1397] FIG. 3D: with K439E (knob).

[1398] FIG. 4: Generation of bsAbs (bispecific antibodies) by exchange reaction according to the current invention exemplified with 2/3-IgGs.

[1399] FIG. 5: TCEP (x molar equivalents in relation to 2/3 input IgGs) is applied to (partially) reduce the hinge-disulfide bonds. SEC differentiates 2/3-IgG starting molecule, generated bsAb and dimeric MHCFcRP. All reactions at different TCEP concentrations were stopped after the same incubation time (triangle: bsAb; cross: 2/3-IgG, diamond: dimeric MHCFcRP).

[1400] FIG. 6: Scheme of removal of undesired non-reacted input molecules and by-products from desired bsAb products if reaction is performed in vitro.

[1401] FIGS. 7A and 7B: FIG. 7A: Scheme of the exchange reaction; FIG. 7B: SDS-page of the NiNTA-purification; NiNTA-bound (upper panel) represents proteins eluted from NiNTA, NiNTA flow through (lower panel) are proteins that do not contain a His-6 or His-8 Tag; n.r.=non-reduced, r.=reduced; M=marker.

[1402] FIG. 8: Bispecific functionality of bsAbs generated by exchange reaction according to the invention. Functionality was assessed by a bridging ELISA that enables detection of simultaneous binding of binding sites of a bispecific antibody. Antigen A coated to the ELISA plate was fluorescein (fluos-BSA, FITC-BSA) and antigen B was biotin (bio-Cy5), which becomes detected by its fluorescence.

[1403] FIG. 9: Exemplary 2/3-IgGs for 2/3-IgG-exchange reaction with binding sites at the C-terminus of the heavy chain.

[1404] FIG. 10: Exemplary 2/3-IgGs for 2/3-IgG-exchange reaction with binding sites at the N-terminus as well as at the C-terminus of the heavy chain.

[1405] FIG. 11: General applicability of the method according to the invention shown by IgG-exchange reaction using starting materials of different binding specificities and formats, exemplified with 2/3-IgGs.

[1406] FIG. 12: Different bsAb format matrix generated via exchange reaction according to the current invention using exemplary 2/3-IgG. The matrix was generated with a fluorescein binding entity and a biocytinamid binding entity. Input molecules and exchange-derived output molecules are shown in FIG. 11. Functionality of generated bsAbs was assessed by bridging ELISA using fluos-BSA as capture antigen and bio-Cy5 to detect bispecific binding functionality. Signals derived from bridging ELISA shows that all formats have bispecific binding efficacy.

[1407] FIG. 13: Matrix for the generation and characterization of bsAb diversity via exchange reaction according to the current invention using a miniaturized high-throughput- and automation-compatible approach.

[1408] FIG. 14: Bispecific antibody formation via exchange reaction according to the method of the current invention with HTS technology. Shown is the signal of an exemplary bridging ELISA showing concentration dependent fluorescence signals that are indicative for bispecific antibody formation. Fluos-bio bridging ELISA, cross: fluos [hole/K370E]+bio [knob/E357K], diamond: bio [hole/K370E]+fluos [knob/E357K]. All other curves: 2/3-IgG input molecules without cognate exchange partners (these do not show bridging signal as only one binding site is present).

[1409] FIG. 15: Scheme of the exchange reaction according to the current invention exemplified with 2/3-IgGs without hinge-region and CH3 domain inter-chain disulfide bonds. This enables chain-exchange reaction in the method according to the current invention without the need to add a reducing agent.

[1410] FIG. 16: The 2/3-IgGs without inter-chain disulfide bridges are secreted into culture supernatants like standard IgGs, purified by standard protein A affinity and size exclusion chromatography, and analyzed by SDS-PAGE confirming the desired 100 kDa 2/3-IgG as expression product. This proves correct assembly of the purified 2/3-IgG-derivatives without inter-chain disulfide bridges as well as absence of undesired dimers and aggregates. Purification of i) anti-bio antibody light chain (SEQ ID NO: 39)+anti-bio antibody heavy chain-knob without hinge region cysteine residues (SEQ ID NO: 57)+MHCFcRP-hole-E357K without hinge regions cysteine residues (SEQ ID NO: 62) (shown on the left) and ii) anti-fluos antibody light chain (SEQ ID NO: 42)+anti-fluos antibody full length heavy chain-hole without hinge region disulfide bonds (SEQ ID NO: 60)+MHCFcRP-knob-K370E without hinge region cysteine residues (SEQ ID NO: 63) (shown on the right).

[1411] FIG. 17: Results of the exchange reaction according to the current invention with starting materials without hinge-region disulfide bonds: 2.5 M concentration of input molecules with purified bsAb as positive control demonstrate successful bsAb generation via chain exchange with monospecific 2/3-IgG input molecules without Fc-region inter-chain disulfide bonds.

[1412] FIG. 18: Design and composition and chain exchange principle of BiFabs to TriFabs.

[1413] FIGS. 19A and 19B: Expression and purification of 2/3-BiFabs:

[1414] FIG. 19A: KappaSelect; FIG. 19B: SEC profiles are exemplarily shown for LeY-proDig (knob)-MHCFcRP(hole).

[1415] FIGS. 20A and 20B: Expression and purification of 2/3-BiFabs: SDS-Page; n.r=non-reduced; r=reduced; L=molecular weight marker. FIG. 20A: LeY-proDig (knob)-MHCFcRP (hole), LeY-proDig (hole)-MHCFcRP (knob), MSLN-proDig (hole)-MHCFcRP (knob);

[1416] FIG. 20B: LeY-proCD3 (knob)-MHCFcRP (hole), LeY-proCD3 (hole)-MHCFcRP (knob), LeY-proCD-AG-2 (knob)-MHCFcRP (hole), LeY-proCD-AG-2 (hole)-MHCFcRP (knob).

[1417] FIG. 21: FACS analysis of cells exposed to the starting molecules and the molecule obtained by exchange reaction according to the current invention.

[1418] FIG. 22: 2/3-BiFab exchange reaction on the cell surface re-arranges and thereby activates the 3rd binding site (Fv) in the stem-region.

[1419] FIG. 23: On-cell conversion of different antigen targeting and different stem-Fv containing TriFab-like prodrugs to cell bound tri-specific TriFabs.

[1420] FIG. 24: Principle of detecting on-cell activation of CD3 binding functionality by CD3-signaling reporter assay.

[1421] FIG. 25: Activation of CD3 binding functionality by 2/3-BiFab exchange on target cells.

[1422] FIG. 26: Principle of on-cell conversion of different antigen targeting 2/3-BiFab prodrugs to fully functional cell bound activated tri-specific TriFabs.

[1423] FIG. 27: On-cell conversion of different antigen targeting 2/3-BiFab prodrugs. LeY and MSLN targeting antibodies on A431-H9 cells. Incubation for six hours at 37 C.

[1424] FIG. 28: TriFab derivatives containing prodrug exchange modules that can rearrange to fully functional trispecific entities.

[1425] FIG. 29: 2/3-BiFab derivatives containing single-chain prodrug exchange modules that can rearrange to fully functional tetraspecific entities.

[1426] FIGS. 30A and 30B: Co-culture of PBMCs from two different donors and MCF7 with anti-LeY-proCD3 2/3-BiFabs. LDH release serves as indicator for cell-mediated killing of the targeted tumor cells.

[1427] FIG. 30A: PBMCs of Donor 1; FIG. 30B: PBMCs of Donor 2.

[1428] FIGS. 31A and 31B: FIG. 31A: Co-culture of PBMCs and A431 cells with anti-EGFR-proCD3 2/3-BiFabs. FIG. 31B: Co-culture of PBMCs and HELA cells with anti-AG-4-proCD3 2/3-BiFabs. LDH release serves as indicator for cell-mediated killing of the targeted tumor cells.

[1429] FIGS. 32A, 32B, 32C, 32D and 32E: Cytokine amounts in supernatant after co-culture of PBMCs and HELA with anti-AG-4-proCD3 2/3-BiFabs at molar concentrations of 4 nM. FIG. 32A: IL-2 amounts; FIG. 32B: IFN amounts; FIG. 32C: Granzyme B amounts; FIG. 32D: TNF amounts; FIG. 32E: legend.

[1430] FIG. 33: Expression level of surface antigens AG-4 and EGFR on HELA cells

[1431] FIG. 34: Dual targeting of AG-4 and EGFR with respective proCD3 2/3-BiFabs and resulting T-cell activation in Jurkat reporter assay.

[1432] FIG. 35: 2/3-BiFab derivatives containing single-chain prodrug exchange modules that can rearrange to fully functional Digoxigenin and Biotin binding entities and can be analyzed by FACS by Dig-Cy5 and Bio-488 binding.

[1433] FIGS. 36A and 36B: FACS analysis of dye binding on cell surface upon TriFab derivative conversion. FIG. 36A: Dig-Cy5; FIG. 36B: Bio-488.

[1434] FIG. 37: 2/3-BiFab derivatives containing single-chain prodrug exchange modules that can rearrange to fully functional CD3 and CD-AG-2 binders.

[1435] FIG. 38: 2/3-BiFab derivatives containing single-chain prodrug exchange modules that can rearrange to fully functional CD3 and CD-AG-2 binders activate T-cells in a CD3 signaling reporter assay.

[1436] FIG. 39: MHCFcRP can be equipped with a variable fragment (VH or VL) and generate upon 2/3-BiFab conversion a CD-AG-2 binding Fab molecule as by-product.

[1437] FIG. 40: FACS analysis CD-AG-2-MHCFcRP by product binding to Jurkat cell surface.

[1438] FIG. 41: N-terminal addition of a targeting Fab entity to MHCFcRP in combination with two proCD3/proCD-AG-2 variable regions within MHCFcRP leads to on cell assembly of two types of trispecific TriFabs that enable CD3 as well as CD-AG-2 binding.

[1439] FIG. 42: T-cell activation capability of trispecific TriFab in a CD3 signaling assay.

[1440] FIG. 43: An alternative 2/3-BiFab format that contains a CH2 domain and thereby an effector function competent Fc-region; the variable domains are each at the C-terminal end of the Fc-region.

[1441] FIG. 44: An alternative 2/3-BiFab format that contains an additional CH2 domain and thereby an effector function competent Fc-region; the variable domain is in between the Fc-region and the targeting Fab or at the N-terminus, respectively.

[1442] FIG. 45: Analytical FcRn affinity chromatography was performed to prove the CH2-dependent binding of the CH2 competent 2/3-BiFab molecules.

[1443] FIGS. 46A and 46B: FACS analysis of Dig-Cy5 binding on cell surface upon application of CH2-containing 2/3-BiFabs. A Dig binding site is just converted upon application of both respective BiFabs. Top to bottom: MCF-7+Dig-Cy5; knob-educt; hole-educt; on-cell shuffling/exchange reaction; product control.

[1444] FIG. 46A: variable domains at the C-terminus of the Fc-region;

[1445] FIG. 46B: variable domains between Fc-region and targeting Fab.

[1446] FIG. 47: The molecular setup of a CH2-containing BiFab for on cell generation of a functional CD3 binding site.

[1447] FIG. 48: A T-cell reporter assay reveals the ability of CH2-containing 2/3-BiFabs to induce a T-cell activation. KN=knob; HL=hole.

[1448] FIGS. 49A, 49B and 49C: Negative Stain Transmission Electron Microscopy (NS-TEM) analysis reveals molecular shape and flexibility of 2/3-BiFabs. FIG. 49A: anti-AG-4-proCD3 (hole)-MHCFcRP (knob); FIG. 49B: AG-3-proCD3 (knob)-MHCFcRP (hole); FIG. 49C: anti-AG-4/CD3/AG-3-antibody.

[1449] FIG. 50: Target-independent shuffling occurs at a low level only at high concentrations (300 nM).

[1450] FIG. 51: By modification of the IgG1 hinge region, i.e. by removal of the disulfide bonds or by shortening the hinge region, different distances between the individual binding sites can be engineered.

EXAMPLES

Example 1

Design and Modular Composition of 2/3-IgGs

General Remarks

[1451] FIG. 1 shows the design and modular composition of the 2/3-IgGs used in the methods according to the current invention. These 2/3-IgGs are composed of three individual chains: one light chain (normally a full length light chain comprising a light chain variable domain and a light chain constant domain), one heavy chain (normally a full length heavy chain comprising a heavy chain variable domain and all heavy chain constant domains including a hinge region) and one heavy chain Fc-region polypeptide (normally a heavy chain Fc-region fragment comprising hinge-CH2-CH3). The variable domains of the light chain and the heavy chain form a functional binding site. The heavy chain (normally derived from the human IgG1 subclass) contains either the knob-cys-mutations or the hole-cys-mutations (the mutations T366W and S354C in the CH3 domain of an antibody heavy chain is denoted as knob-cys-mutations and the mutations T366S, L368A, Y407V, Y349C in the CH3 domain of an antibody heavy chain are denoted as hole-cys-mutations (numbering according to Kabat EU index)) in CH3 to enable the formation of knob-into-hole Fc-region dimers. The heavy chain Fc-region polypeptide is a so called dummy-Fc/MHCFcRP (see below), i.e. an IgG1 derivative that lacks VH and CH1, starts at the N-terminus with at least part of the hinge region sequence and harbors a His6 tag at its C-terminus. In addition, the heavy chain Fc-region polypeptide of the 2/3-IgG contains in its CH3 domains either the knob-mutation or the hole-mutations (the mutation T366W in the CH3 domain of an antibody heavy chain is denoted as knob-mutation and the mutations T366S, L368A, Y407V in the CH3 domain of an antibody heavy chain are denoted as hole-mutations (numbering according to Kabat EU index)). In addition to the knob- or hole-mutation(s) the heavy chain Fc-region polypeptide comprises a destabilizing mutation introducing one (i.e. a single additional) repulsive charge with respect to the wild-type sequence: D356K or E357K or K370E or K439E; SEQ ID NO: 35 to 38; this mutated heavy chain Fc-region polypeptide is denoted as MHCFcRP in the following.

[1452] The heavy chain and the MHCFcRP can form two types of heterodimers depending on the distribution of the knob-into-hole-mutations therein: [1453] i) heavy chain-knob::MHCFcRP-hole, and [1454] ii) heavy chain-hole::MHCFcRP-knob.

[1455] Those heterodimers are, however, somewhat flawed as the complementary Fc-region lacks the additional CH3 cysteine necessary to form inter-chain disulfides to the heavy chain, and also these contain charge mutations without matching heavy chain counterparts.

Example 2

Expression and Purification of 2/3-IgGs

[1456] Expression of 2/3-IgGs was achieved by co-transfection of plasmids encoding light chain, heavy chain (with knob or hole-mutations) and matching MHCFcRP (hole or knob) into mammalian cells (e.g. HEK293) via state of the art technologies.

[1457] In more detail, for example, for the production of the 2/3-IgGs by transient transfection (e.g. in HEK293 cells) expression plasmids based either on a cDNA organization with or without a CMV-Intron A promoter or on a genomic organization with a CMV promoter were applied.

[1458] Beside the antibody expression cassettes, the plasmids contained: [1459] an origin of replication, which allows replication of this plasmid in E. coli, [1460] a B-lactamase gene, which confers ampicillin resistance in E. coli., and [1461] the dihydrofolate reductase gene from Mus musculus as a selectable marker in eukaryotic cells.

[1462] The transcription unit of each antibody gene was composed of the following elements: [1463] unique restriction site(s) at the 5-end [1464] the immediate early enhancer and promoter from the human cytomegalovirus, [1465] followed by the Intron A sequence in the case of the cDNA organization, [1466] a 5-untranslated region of a human antibody gene, [1467] an immunoglobulin heavy chain signal sequence, [1468] the antibody chain either as cDNA or in genomic organization (the immunoglobulin exon-intron organization is maintained), [1469] a 3-non-translated region with a polyadenylation signal sequence, and [1470] unique restriction site(s) at the 3-end.

[1471] The fusion genes comprising the antibody chains were generated by PCR and/or gene synthesis and assembled by known recombinant methods and techniques by connection of the according nucleic acid segments e.g. using unique restriction sites in the respective plasmids. The subcloned nucleic acid sequences were verified by DNA sequencing. For transient transfections larger quantities of the plasmids were prepared by plasmid preparation from transformed E. coli cultures (Nucleobond AX, Macherey-Nagel).

[1472] Standard cell culture techniques were used as described in Current Protocols in Cell Biology (2000), Bonifacino, J. S., Dasso, M., Harford, J. B., Lippincott-Schwartz, J. and Yamada, K. M. (eds.), John Wiley & Sons, Inc.

[1473] The 2/3-IgGs were generated by transient transfection with the respective plasmid using the HEK293-F system (Invitrogen) according to the manufacturer's instruction. Briefly, HEK293-F cells (Invitrogen) growing in suspension either in a shake flask or in a stirred fermenter in serum-free FreeStyle 293 expression medium (Invitrogen) were transfected with the respective expression plasmid and 293fectin or fectin (Invitrogen). For 2 L shake flask (Corning) HEK293-F cells were seeded at a density of 1*10.sup.6 cells/mL in 600 mL and incubated at 120 rpm, 8% CO.sub.2. The day after the cells were transfected at a cell density of approx. 1.5*10.sup.6 cells/mL with ca. 42 mL mix of A) 20 mL Opti-MEM (Invitrogen) with 600 g total plasmid DNA (1 g/mL) and B) 20 ml Opti-MEM+1.2 mL 293 fectin or fectin (2 L/mL). According to the glucose consumption glucose solution was added during the course of the fermentation. Correctly assembled 2/3-IgGs were secreted into culture supernatants like standard IgGs. The supernatant containing the secreted 2/3-IgG was harvested after 5-10 days and 2/3-IgGs were either directly purified from the supernatant or the supernatant was frozen and stored.

[1474] Because 2/3-IgGs contain an Fc-region they were purified by applying standard protein A affinity chromatography: The 2/3-IgGs were purified from cell culture supernatants by affinity chromatography using MabSelectSure-Sepharose (GE Healthcare, Sweden) and Superdex 200 size exclusion (GE Healthcare, Sweden) chromatography.

[1475] Briefly, sterile filtered cell culture supernatants were captured on a MabSelectSuRe resin equilibrated with PBS buffer (10 mM Na.sub.2HPO.sub.4, 1 mM KH.sub.2PO.sub.4, 137 mM NaCl and 2.7 mM KCl, pH 7.4), washed with equilibration buffer and eluted with 25 mM sodium citrate at pH 3.0. The eluted antibody fractions were pooled and neutralized with 2 M Tris, pH 9.0. The antibody pools were further purified by size exclusion chromatography using a Superdex 200 26/60 GL (GE Healthcare, Sweden) column equilibrated with 20 mM histidine, 140 mM NaCl, pH 6.0. The 2/3-IgG containing fractions were pooled, concentrated to the required concentration using Vivaspin ultrafiltration devices (Sartorius Stedim Biotech S.A., France) and stored at 80 C.

[1476] Purity and integrity were analyzed after each purification step by CE-SDS using microfluidic Labchip technology (Caliper Life Science, USA). Protein solution (5 l) was prepared for CE-SDS analysis using the HT Protein Express Reagent Kit according manufacturer's instructions and analyzed on LabChip GXII system using a HT Protein Express Chip. Data were analyzed using LabChip GX Software.

[1477] For example, the following 2/3-IgGs have been produced by co-expression of corresponding L-chain, H-chain and MHCFcRP encoding plasmids:

TABLE-US-00026 anti-fluorescein-2/3- anti-biocytinamid-2/3- IgG-knob-cys+ IgG-hole-cys+ D356K- E357K- K370E- K439E- MHCFcRP hole hole knob knob HEK293 protein A 122 94 129 117 [mg/L] SEC >70 >50 >70 >70 [% yield] Expi- protein A >200 >200 >200 >200 system [mg/L) SEC >90 >90 >80 >80 [% yield]

[1478] The corresponding SEC chromatograms are shown in FIG. 3.

Example 3

Generation of Bispecific Antibodies (bsAbs) by 2/3-IgG-Exchange Reaction

[1479] The 2/3-IgGs that contain a light chain, a heavy chain and MHCFcRP have been generated in two types of KiH heterodimers: full length heavy chain-knob::MHCFcRP-hole and full length heavy chain-hole::MHCFcRP-knob. Both types of 2/3-IgGs are somewhat flawed as the MHCFcRP lacks the additional CH3 cysteine necessary to form inter-chain disulfides to the heavy chain, and the MHCFcRP contains charge mutations without matching counterpart in the full length heavy chain. The modules that make up those flawed heterodimers, however, are capable to rearrange to bispecific heterodimers with matching charges as shown in FIG. 4. The full length heavy chain (knob-cys) of 2/3-IgG A and the full length heavy chain (hole-cys) from 2/3-IgG B form a matching heterodimer. Matching heterodimers are also formed when MHCFcRP (hole-charge) interacts with MHCFcRP (knob-charge). Thus, exchange reactions based on temporary separation of starting heterodimers of two different 2/3-IgGs resulted in products that contain preferentially (charge) matching heterodimers. Exchange reactions therefore converted two monospecific 2/3-IgGs to one bispecific IgG and one MHCFcRP heterodimer:

2/3-IgG(A)-His6(8)+2/3-IgG(B)-His6(8).fwdarw.bsAb(AB)+Fc-His6(8)

[1480] The exchange reaction was initiated by a reduction step (e.g. by applying 2-MEA or TCEP at various concentrations) to break especially the hinge-region inter-chain disulfide bonds. Chain rearrangement occurred spontaneously thereafter.

[1481] Therefore, anti-fluorescein-2/3-IgG and anti-biocytinamid-2/3-IgG input molecules were mixed in equimolar amounts at a protein concentration of 100 g/ml in a total volume of 40 l 1PBS+0.05% Tween 20 with the indicated TCEP concentrations on a 384 well REMP plate (Brooks, #1800030). After centrifugation, plates were sealed and incubated for one hour at 27 C.

[1482] A biotin-fluorescein bridging ELISA was subsequently used to quantify bispecific antibody. Therefore, white Nunc MaxiSorp 384 well plates were coated with 1 g/ml albuming-fluorescein isothiocyanate conjugate (FITC, Sigma, # A9771) and incubated overnight at 4 C. After washing 3 times with 90 l PBST-buffer (PBST, bidest water, 10PBS+0.05% Tween 20) blocking buffer (1PBS, 2% gelatin, 0.1% Tween-20) was added 90 l/well and incubated for one hour at room temperature. After washing 3 times with 90 l PBST-buffer, 25 l of a 1:10 dilution of each exchange reaction was added to each well. After incubation for one hour at room temperature, plates were again washed 3 times with 90 l PBST-buffer. 25 l/well biotin-Cy5 conjugate in 0.5% BSA, 0.025% Tween-20, 1PBS was added to a final concentration of 0.1 g/ml and plates were incubated for one hour at room temperature. After washing 6 times with 90 l PBST-buffer, 25 l 1PBS were added to each well. Cy5 fluorescence was measured at an emission wavelength of 670 nm (excitation at 649 nm) on a Tecan Safire 2 Reader.

[1483] FIG. 5 shows the results of analyses of the redox conditions for generation of bsAbs by 2/3-IgG-exchange. TCEP is applied to (partially) reduce the hinge-disulfide bonds between the heavy chain Fc-region polypeptides, i.e. between the full length half-IgG and the MHCFcRP. Chain exchange can be identified by SEC which differentiates 2/3-IgG input, bsAb output and MHCFcRP by-product. The yield of the exchange reactions depending on the ratio between 2/3-IgG and TCEP are shown in FIG. 5 (for comparison all reactions were analyzed after the same reaction time).

[1484] All 2/3-IgG starting molecules, all non-wanted by-products, as well as all aggregates that were potentially generated during the exchange reaction harbor affinity tags (His6 or His8). The desired bsAb produced in the exchange reaction is the only molecule that does not carry a His-tag. Therefore, a simple NiNTA absorption step was applied to remove all undesired molecules (see FIGS. 6 and 7). The remaining bsAbs (not depleted by NiNTA absorption) were directly applied to screening procedures and analyzed to identify bsAbs with desired functionalities.

Example 4

[1485] Functional Assessment of Bispecific Antibodies (bsAbs) Generated by 2/3-IgG-Exchange Reaction

[1486] Bispecific functionality of bsAbs that were generated as products of 2/3-IgG-exchange reactions was evaluated by bridging-ELISA assays. FIG. 8 shows as an example the binding result for an anti-fluorescein/biocytinamid bispecific antibody generated by an exchange reaction as reported herein. In the reaction biocytinamid (bio)-binding 2/3-IgG and a fluorescein (fluos)-binding 2/3-IgG as starting molecules were employed. The fluos-binding arm of anti-fluos/bio bispecific antibodies bind to fluos-BSA coated ELISA plates. Subsequent exposure to bio-Cy5 generates signals only upon bsAb-mediated capture of bio-Cy5 via the bio-binding arm of the bsAb. Because bridging-mediated signals occur only with bsAbs but not with either monospecific Fluos or Bio binders, no signals were observed when using only 2/3-IgGs in the assay. Because of that and because the exchange reaction does not force molecule aggregation, such bridging ELISA can be performed directly on exchange reaction mixes, without requiring prior NiNTA-mediated depletion of non-bsAb molecules. Signals observed when applying the reaction mix indicated successful generation and presence of functional bsAbs. Signal generation via bridging ELISA was dependent on the amount of input entities used in the exchange reaction.

Example 5

The Exchange Reaction is Functional Independent of Binding Specificities or V-Region Composition of Starting 2/3-IgGs

[1487] A variety of 2/3-IgGs was produced to evaluate if 2/3-IgG production as well as exchange reactions work for different antibodies independent of their binding specificities and V-region composition, as well as for different antibody combinations.

[1488] Therefore, 2/3-IgGs with binding specificities for biocytinamid (bio), digoxigenin (dig), fluorescein (fluos), LeY-carbohydrate (LeY), VEGF and PDGF were used. These were produced by co-transfection of expression plasmids encoding full length light chains, knob- or hole-full length heavy chains and mutated heavy chain Fc-region polypeptides as described above.

TABLE-US-00027 Chain MHCFcRPs SEQ ID NO: hole-D356K-His8 35 hole-E357K-His8 36 knob-K370E-His8 37 knob-K439E-His8 38 anti-bio antibody full length light chain 39 anti-bio antibody full length heavy chain-knob-cys 40 anti-bio antibody full length heavy chain-hole-cys 41 anti-fluos antibody full length light chain 42 anti-fluos antibody full length heavy chain-knob-cys 43 anti-fluos antibody full length heavy chain-hole-cys 44 anti-dig antibody full length light chain 45 anti-LeY antibody full length light chain 46 anti-PDGF antibody full length light chain 47 anti-VEGF antibody full length light chain 48 anti-dig antibody VH-CH1 fragment 49 anti-LeY antibody VH-CH1 fragment 50 anti-PDGF antibody VH-CH1 fragment 51 anti-VEGF antibody VH-CH1 fragment 52

[1489] SEQ ID NO: 49-52 describe the VH-CH1 region of 2/3-IgGs with specificities for dig, VEGF, PDGF and LeY. Those were fused to the hinge-CH2-CH3 regions (i.e. replace the bio VH-CH1 regions) of SEQ ID NO: 40 and 41 to generate complete H-chains with desired specificity. The MHCFcRPs applied for generating these molecules are listed as SEQ ID NO: 35-38.

[1490] All of these 2/3-IgGs could be produced and purified to similar yields as for standard IgGs under comparable conditions (see Example 2). Examples for expression of these 2/3-IgGs with different binding specificities are shown in the following Table.

TABLE-US-00028 2/3-IgG = 1/2-IgG-hole-cys + MHCFcRP-knob-E357K anti-dig anti-VEGF anti-PDGF anti-LeY anti-fluos Protein A 76 76 96 81 94 [mg/L] SEC 40-60 >70 >90 >95 >50 [% yield]

[1491] In the exchange-matrix, which was applied to generate bsAbs of different specificity, combinations of 2/3-IgGs with binding specificities for fluorescein, biocytinamid, VEGF, PDGF and digoxigenin in all combinations as shown in the following Table were employed.

TABLE-US-00029 exchange reaction MHCFcRP-knob-E357K between bio fluos Dig VEGF PDGF MHCFcRP- bio bio bio bio bio hole-K370E fluos dig VEGF PDGF fluos fluos fluos fluos fluos bio dig VEGF PDGF dig dig dig dig dig bio fluos VEGF PDGF VEGF VEGF VEGF VEGF VEGF bio fluos Dig PDGF PDGF PDGF PDGF PDGF PDGF bio fluos Dig VEGF

[1492] The chain exchange of starting 2/3-IgGs and generation of bsAbs with desired specificity combinations was monitored by bridging ELISA (see Example 4), wherein plate-coated antigens and signal-generating antigen-conjugates/complexes were applied that match the different bsAb specificity combinations.

[1493] The results of the bridging ELISA applied to assess the functionalities of different bsAb combinations are shown in the following Tables. Only bsAbs that recognize their cognate pair of antigens present as capturing or detection antigen generate signals in the bridging ELISA. Other bsAbs generated in the matrix are negative due to absence of at least one specificity.

TABLE-US-00030 TABLE Bridging ELISA confirms the functionality of bsAbs generated. Shown are the relative signal intensities within one assay at the input molecule concentration 1.3 M. The highest value is set to 100% as a reference. assay biocytinamid-fluorescein capture fluorescein-albumin detection biocytinamid-Cy5 exchange reaction MHCFcRP-hole-K370E between bio fluos dig VEGF3 PDGF MHCFcRP- bio 100% 2.5% 2.5% 1.9% knob-E357K fluos 97.6% 2.5% 1.9% n.a. dig 2.2% 2.5% 2.2% 2.2% VEGF 1.9% 2.2% 2.3% 2.3% PDGF 1.8% n.a. 2.3% 1.9% N.a. = not available.

TABLE-US-00031 assay digoxigenin-fluorescein capture fluorescein-albumin detection digoxygenin-Cy5 exchange reaction MHCFcRP-hole-K370E between bio fluos dig VEGF PD1GF MHCFcRP- bio 1.9% 1.6% 1.4% 1.3% knob-E357K fluos 2.4% 100% 2.8% n.a. dig 2.0% 52.5% 2.0% 1.5% VEGF 1.5% 1.5% 1.5% 1.5% PDGF 1.5% n.a. 1.8% 2.8%

TABLE-US-00032 assay VEGF-biocytinamid capture VEGF detection biocytinamid-Cy5 exchange reaction MHCFcRP-hole-K370E between bio fluos dig VEGF PDGF MHCFcRP- bio 9.0% 9.3% 100% 10.1% knob-E357K fluos 10.2% 9.4% 9.9% n.a. dig 9.0% 9.1% 8.7% 9.9% VEGF 78.3% 9.2% 9.3% 9.5% PDGF 10.5% n.a. 9.2% 10.9%

TABLE-US-00033 assay PDGF-biocytinamid capture PDGF detection biocytinamid-Cy5 exchange reaction MHCFcRP-hole-K370E between bio fluos dig VEGF PDGF MHCFcRP- bio 3.0% 4.1% 4.4% 81.8% knob-E357K fluos 3.2% 3.1% 3.3% n.a. dig 3.3% 3.2% 3.3% 3.4% VEGF 4.0% 3.1% 3.1% 3.2% PDGF 100% n.a. 3.9% 3.2%

TABLE-US-00034 assay digoxigenin-VEGF capture VEGF detection digoxygenin-Cy5 exchange reaction MHCFcRP-hole-K370E between bio fluos dig VEGF PDGF MHCFcRP- bio 7.2% 6.2% 6.4% 6.1% knob-E357K fluos 6.5% 6.3% 6.5% n.a. dig 6.2% 6.7% 59.7%% 7.0% VEGF 6.1% 6.6% 100% 7.0% PDGF 6.0% n.a. 5.9% 6.5%

TABLE-US-00035 assay digoxigenin-PDGF capture PDGF detection digoxygenin-Cy5 exchange reaction MHCFcRP-hole-K370E between bio fluos dig VEGF PDGF MHCFcRP- bio 3.0% 2.9% 2.9% 3.0% knob-E357K fluos 3.7% 3.2% 2.8% n.a. dig 2.9% 3.1% 3.5% 62.3% VEGF 3.1% 3.3% 3.0% 2.9% PDGF 3.7% n.a. 100% 3.8%

[1494] For the VEGF containing bispecific antibodies the same assays have been performed. These also showed only signals above background levels for the respective combinations.

[1495] It can be seen that the exchange reaction according to the current invention is a generally applicable method: exchange reactions lead to functional bsAb independent of binding specificities or V-region composition of the starting molecules.

Example 6

Design, Composition and Generation of Format Variants

[1496] The 2/3-IgG-exchange reaction of Example 4 was expanded to starting molecules that have either one binding site at the C-terminus of the heavy chain, or heavy chains with binding sites at N- as well as C-terminus. For generation of the exchanged bsAbs the exchange driving principle (conversion of flawed input heterodimers to matching output-heterodimers) was kept unaltered. The composition of the MHCFcRPs was also retained as described above.

[1497] FIGS. 1, 9 and 10 show the modular composition of the three 2/3-IgG formats that were applied to generate different bsAb formats. One of the 2/3-IgGs has one Fab arm at the N-terminal position. Another of the 2/3-IgGs has the Fab arm attached via a flexible linker to the C-terminus of the heavy chain (i.e. it starts at the N-terminus with the hinge-region). The third 2/3-IgG has the C-terminal Fab arm as well as the N-terminal Fab arm.

[1498] Expression of these 2/3-IgG variants was achieved by co-transfection of plasmids encoding light chain, heavy chain (knob or hole) and corresponding MHCFcRP (hole or knob) into mammalian cells (e.g. HEK293) (see Example 2).

[1499] Sequences of the full length heavy chains modified used for the generation of the different bsAb formats are as follows:

TABLE-US-00036 chain MHCFcRPs SEQ ID NO: hole-D356K-His8 35 hole-E357K-His8 36 knob-K370E-His8 37 knob-K439E-His8 38 anti-bio antibody full length heavy chain-hole-cys 53 with C-terminal fusion anti-bio antibody full length heavy chain-hole-cys 54 with N- and C-terminal fusion anti-fluos antibody full length heavy chain-hole-cys 55 with C-terminal fusion anti-fluos antibody full length heavy chain-hole-cys 56 with N- and C-terminal fusion

[1500] The 2/3-IgGs are secreted into culture supernatants like standard IgGs and were purified by standard protein A affinity chromatography (see Example 2). Size-exclusion and mass-spec analytics revealed correct assembly of purified 2/3-IgG variants as well as absence of undesired dimers and aggregates. Expression yields of 2/3-IgGs were similar to those observed with standard IgGs in the same expression systems. The respective data is presented in the following Table.

TABLE-US-00037 anti-fluorescein antibody- anti-biocytinamid antibody - knob-cys + hole-cys + MHCFcRP-hole-E357K MHCFcRP-knob-K370E 43 + 36 55 + 36 56 + 36 41 + 37 53 + 37 54 + 37 SEQ ID NO: (N-Fc) (C-Fc) (NC-Fc) (N-Fc) (C-Fc) (NC-Fc) protein A 94 94 75 129 87 75 [mg/L] SEC 55 90 87 40-80 61 63 [% yield]

Example 7

[1501] Characterization of bsAbs with Combined Binding Functionalities in Different Valencies, Stoichiometries and Geometries

[1502] Three different starting molecules (2/3-IgG with N-terminal, C-terminal, N- and C-terminal binding site(s)) can be combined with each other in the method according to the current invention to result in nine different bsAb formats. These differ in valencies, geometries and positions of the individual binding sites. The exchange reaction to generate these different bsAbs was performed under the same conditions as outlined in Example 3.

[1503] All types of input formats are flawed as the MHCFcRP lacks the additional CH3 cysteine necessary to form inter-chain disulfides to the heavy chain and as it contains a repulsive charge mutation (i.e. a charge without matching full length heavy chain counterpart). The heavy chains that make up those flawed heterodimers rearrange to form (charge and disulfide) matching heterodimers in the method according to the current invention. The different types of full length heavy chains (knob-cys with hole-cys) form matching heterodimers. Matching heterodimers are also formed from the MHCFcRP (hole-charge with knob-charge).

[1504] Without being bound by this theory it is assumed that exchange reactions based on temporary separation of flawed heterodimers of two different 2/3-IgGs results in products that contain preferentially perfectly matching heterodimers with matching charges and, if present, cysteine residues for the formation of disulfide bonds. Exchanges therefore convert the monospecific 2/3-IgGs to bispecific IgGs (in different formats), as well as corresponding (variable region free, i.e. non-target binding competent) Fc-region heterodimer.

[1505] For the description of the exchange reactions, the input molecules are termed: [1506] nA or nB for molecules having the Fab arm at the normal N-terminus of the full length heavy chain (H-chain) [1507] cA or cB for molecules having the Fab arm at the C-terminus of the H-chain [1508] ncA or ncB for molecules with Fab at N- as well as C-terminus of the H-chain

[1509] The different format-exchange reactions are as follows:

2/3-IgG(nA)-His-tag+2/3-IgG(nB)-His-tag.fwdarw.bsAb(nAnB)+Fc-His-tag

2/3-IgG(nA)-His-tag+2/3-IgG(cB)-His-tag.fwdarw.bsAb(nAcB)+Fc-His-tag

2/3-IgG(nA)-His-tag+2/3-IgG(ncB)-His-tag.fwdarw.bsAb(nAncB)+Fc-His-tag

2/3-IgG(cA)-His-tag+2/3-IgG(cB)-His-tag.fwdarw.bsAb(cAcB)+Fc-His-tag

2/3-IgG(cA)-His-tag+2/3-IgG(nB)-His-tag.fwdarw.bsAb(cAnB)+Fc-His-tag

2/3-IgG(cA)-His-tag+2/3-IgG(ncB)-His-tag.fwdarw.bsAb(cAncB)+Fc-His-tag

2/3-IgG(ncA)-His-tag+2/3-IgG(nB)-His-tag.fwdarw.bsAb(ncAnB)+Fc-His-tag

2/3-IgG(ncA)-His-tag+2/3-IgG(cB)-His-tag.fwdarw.bsAb(ncAcB)+Fc-His-tag

2/3-IgG(ncA)-His-tag+2/3-IgG(ncB)-His-tag.fwdarw.bsAb(ncAncB)+Fc-His-tag

[1510] Exchange reactions are initiated by a reduction step to break the inter-chain (hinge-region) disulfide bonds, chain rearrangement occurs spontaneously thereafter. All input molecules, all by-products, as well as all aggregates that may potentially form during the exchange reaction harbor affinity tags (e.g. a His6- or His8-tag). The bsAb products of the exchange reaction, however, do not carry the affinity tag and can therefore be separated via affinity (e.g. NiNTA) absorption chromatography. The bsAbs (in different formats) can directly be applied to screening procedures and analyses to identify and to rank the different bsAbs formats with optimal functionality.

[1511] The bispecific formats were generated by exchanging the above described input 2/3-IgGs in a 384 well MTP format followed by bridging ELISA to assess functional assembly. Therefore, the exchange partners (2/3-IgG molecule 1 consisting of a full length heavy chain containing the hole-cys-mutations and an MHCFcRP-knob-K370E; 2/3-IgG molecule 2 consisting of a full length heavy chain containing the knob-cys-mutations and a MHCFcRP-hole-E357K) were mixed in equimolar amounts (4 M) in a total volume of 100 l 1PBS+0.05% Tween 20. Protein solutions were diluted in 11 times 1:2 in a 384-deep well plate (Greiner 384 Masterblock). 20 l of each sample from the dilution series were mixed with 20 l of a 0.5 mM TCEP solution to a final protein concentration of 200-0.2 g/ml and 0.25 mM TCEP on a 384 well REMP plate (Brooks, #1800030). After centrifugation, plates were sealed and incubated for one hour at 37 C.

[1512] As control examples, bsAbs containing bio-binding functionality on one side and fluorescein-binding functionality on the other side were used. Functionality of the resulting bsAbs was assessed by biotin-fluorescein bridging ELISA. Therefore, white Nunc MaxiSorp 384 well plates were coated with 1 g/ml albumin-fluorescein isothiocyanate conjugate (Sigma, # A9771) and incubated overnight at 4 C. After washing 3 times with 90 l PBST-buffer (PBST, double distilled water, 10PBS Roche #11666789001+0.05% Tween 20), 90 l/well blocking buffer (1PBS, 2% BSA, 0.1% Tween 20) was added and incubated for one hour at room temperature. After washing 3 times with 90 l PBST-buffer 25 l of a 1:4 dilution of each exchange reaction was added to each well. After incubation for one hour at room temperature, plates were again washed 3 times with 90 l PBST-buffer. 25 l per well biotin-Cy5 conjugate in 0.5% BSA, 0.025% Tween 20, 1PBS was added to a final concentration of 0.1 g/ml and plates were incubated for one hour at room temperature. After washing 6 times with 90 l PBST-buffer, 25 l 1PBS were added to each well. Cy5 fluorescence was measured at an emission wavelength of 670 nm (excitation at 649 nm) on a Tecan Safire 2 Reader.

[1513] Different bsAb formats via exchange of 2/3-IgGs of different formats were generated with one fluorescein binding entity and one biocytinamid binding entity. Input molecules and exchange-derived output molecules are shown in FIG. 11.

[1514] Functionality of generated bsAbs was assessed by bridging ELISA as shown in FIG. 12, using fluos-BSA as capture antigen and bio-Cy5 to detect bispecific bridging binding functionality. All different formats result in a bridging ELISA signal.

[1515] These results show the feasibility to generate different formats using a method according to the current invention via chain exchange reactions in a robust and high-throughput compatible manner.

Example 8

[1516] Generation of Functional bsAbs by 2/3-IgG-Exchange and Screening/Identification of bsAbs with Desired Functionality is Compatible with Miniaturization and High-Throughput as Well as Automation Technologies

[1517] Application of high-throughput and automation technologies is desired and in many instance necessary to handle large numbers of different bsAbsdiffering in binding site sequence and/or format. It has therefore been analyzed if bsAb generation via the 2/3-IgG exchange method according to the current invention, as well as analysis/screening of the functionality, i.e. bispecific binding, of the thereby generated bispecific antibodies, can be miniaturized in order to be compatible with high throughput and automation technologies.

[1518] Therefore, 2/3-IgG exchange reactions were performed and the reaction products were analyzed in miniaturized scale in 348 well plates.

[1519] A matrix screen was set up in 384 well MTP format as follows: The exchange partners (2/3-IgG molecule 1 consisting of a full length heavy chain containing the hole-cys-mutations and an MHCFcRP-knob-K370E; 2/3-IgG molecule 2 consisting of a full length heavy chain containing the knob-cys-mutations and a MHCFcRP-hole-E357K) were mixed in equimolar amounts (4 M) in a total volume of 30 l 1PBS+0.05% Tween 20. Protein solutions were diluted four times 1:3 in a 384-deep well plate (Greiner 384 Masterblock). 20 l of each sample from the dilution series were mixed with 20 l of a 0.5 mM TCEP solution to a final protein concentration of 2 M-0.025 M and 0.25 mM TCEP on a 384 well REMP plate (Brooks, #1800030). After centrifugation, plates were sealed and incubated for one hour at 37 C.

[1520] The functionality of the thereby generated bsAbs was subsequently assessed via bridging ELISA (see above) in a miniaturized high-throughput format: White Nunc MaxiSorp 384 well plates were coated with 1 g/ml albuming fluorescein isothiocyanate conjugate (Sigma, # A9771), 1 g/ml PDGF (CST, #8912) or 1 g/ml VEGF121 and incubated overnight at 4 C. After washing 3 times with 90 l PBST-buffer (PBST, double distilled water, 10PBS+0.05% Tween 20) blocking buffer (1PBS, 2% BSA, 0.1% Tween 20) was added 90 l/well and incubated for one hour at room temperature. After washing 3 times with 90 l PBST-buffer 25 l of a 1:4 dilution of each exchange reaction was added to each well. After incubation for 1 h at room temperature, plates were again washed 3 times with 90 l PBST-buffer. 25 l per well biotin-Cy5 conjugate or dig-Cy5 conjugate in 0.5% BSA, 0.025% Tween 20, 1PBS was added to a final concentration of 0.1 g/ml and plates were incubated for one hour at room temperature. After washing 6 times with 90 l PBST-buffer, 25 l 1PBS were added to each well. Cy5 fluorescence was measured at an emission wavelength of 670 nm (excitation at 649 nm) on a Tecan Safire 2 Reader. The details of the exchange reactions and bridging ELISAs these analyses with 2/3-IgG modules that bind either VEGF or PDGF or dig or bio or fluos are shown in FIG. 13. The results of one exemplary these analysis is shown in FIG. 14 and demonstrates that 2/3-IgG-exchange reactions and subsequent functional analyses can be performed and are compatible with high-throughput and automation technologies.

Example 9

[1521] Generation of bsAbs with Three Binding Sites that Target a First Antigen with One Binding Site and a Further Antigen with the Two Other Binding Sites

[1522] The method according to the current invention can be used for the generation of T-cell bispecific antibodies (TCBs). These can have a format as described before (see e.g. WO 2013/026831). For the TCB-exchange approach, one H-chain (either with knob-cys or with hole-cys as described above) contains a CD3-binding CrossFab-derived entity N-terminal of its hinge, further being extended at the N-terminus by another antibody-derived targeting entity. The exchange reaction is carried out under the same conditions described above and results in a TCB harboring a CD3 binding entity and two additional binding entities. These can bind to a target cell antigen. Those molecules can simultaneously bind to CD3 on T-cells and to an antigen on a target (e.g. tumor) cell and thereby induce killing of target cells.

Example 10

[1523] Design and Generation of 2/3-IgGs without Fe-Region Inter-Chain Disulfide Bonds (in Hinge Region as Well as in CH3 Domain) According to the Current Invention

[1524] Chain exchange with Fc-region (hinge region) disulfide containing 2/3-IgGs requires reduction as initial step to enable chain separation and subsequent assembly of desired bsAbs. To avoid the reduction step and the associated need to remove the reducing agent 2/3-IgGs without hinge region disulfide bonds were generated. The principle is shown in FIG. 15. The cysteine residues in the hinge region responsible for hinge-disulfide formation were removed by mutation to serine. Also the CH3-cysteine at position 354 or 349 that forms the KiH associated disulfide bond has been omitted. The respective amino acid sequences are:

TABLE-US-00038 SEQ ID NO: Chain anti-bio antibody full length heavy chain-knob without 57 hinge-region cysteine residues anti-bio antibody full length heavy chain-hole without 58 hinge-cysteine residues anti-fluos antibody full length heavy chain-knob without 59 hinge-cysteine residues anti-fluos antibody full length heavy chain-hole without 60 hinge-cysteine residues MHCFcRP hole-D356K-His8 without hinge-cysteine residues 61 hole-E357K-His8 without hinge-cysteine residues 62 knob-K370E-His8 without hinge-cysteine residues 63 knob-K439E-His8 without hinge-cysteine residues 64

[1525] Expression of the above 2/3-IgGs was achieved by co-transfection of plasmids encoding light chain, full length heavy chain (knob or hole) and corresponding MHCFcRP (hole or knob) into mammalian cells (e.g. HEK293) (see Example 2). The 2/3-IgGs were secreted into culture supernatants like standard IgGs and were thereafter purified by standard protein A affinity and size exclusion chromatography (see Example 2). Subsequent analytics via size exclusion chromatography and SDS-PAGE the desired 100 kDa 2/3-IgG expression product (FIG. 16). This proves correct assembly of the 2/3-IgG as well as absence of undesired dimers and aggregates. This is surprising as such molecules are not stabilized by disulfides between the Fc-regions (neither hinge region nor CH3 domain). The purification yield of anti-fluos- and anti-bio-2/3-IgGs without Fc-region inter-chain disulfide bonds are presented in the following Table

TABLE-US-00039 anti-fluos antibody light chain anti-bio antibody light chain (SEQ ID NO: 42) + anti-fluos (SEQ ID NO: 39) + anti-bio antibody full length heavy antibody heavy chain-knob chain-hole without hinge without hinge region cysteine region disulfide bonds residues (SEQ ID NO: 57) + (SEQ ID NO: 60) + MHCFcRP-hole-E357K MHCFcRP-knob-K370E without hinge regions without hinge region cysteine residues cysteine residues (SEQ ID NO: 62) (SEQ ID NO: 63) protein A >100 >100 [mg/L] SEC yield >50 >50 [mg/L 100 kDa]

Example 11

[1526] Generation of Functional bsAbs by 2/3-IgG-Exchange Reaction without Reduction in the Method According to the Invention

[1527] The 2/3-IgGs that do not contain Fc-region inter-chain disulfide bonds were subjected to chain exchange reactions as described above (see Example 3), except for omitting the initial reduction step. The 2/3-IgGs either contained fluos- or bio-binding sites and Fc-regions without inter-chain disulfide bonds between the full length heavy chain and MHCFcRP. Composition and production of these 2/3-IgGs was described in Example 10. Following exchange reactions without initiating reduction, a bridging ELISA was performed to demonstrate bispecific functionality of bsAbs. The bridging ELISA comprised the addition of exchange reaction products to immobilized fluos-BSA, followed by wash steps and subsequent addition of bio-Cy5 to probe for presence of the 2.sup.nd binding arm of the bsAb (see previous examples for details of the bridging ELISA). Only correct assembled functional bsAbs can bind by their fluos-binding site to the assay plate, are retained and generate signals by capturing and retaining bio-Cy5. Molecules without bispecificity do not generate signals as they either do not bind to the plate (bio-only binder) or cannot capture the signal generating bio-Cy5 (fluos-only binder). The results of these analyses (performing the exchange reaction in this example at 2.5 M concentration of input molecules with purified bsAb as positive control) are shown in FIG. 17. The results demonstrate successful bsAb generation via chain exchange with monospecific 2/3-IgG input molecules without Fc-region inter-chain disulfide bonds. Productive chain exchange took place without requirement of initial reduction. Thus, removal of inter Fc-region polypeptide disulfide bonds eliminated the necessity of an initial reduction step. The resulting bsAbs are held together by non-covalent Fc-Fc interactions. Elimination of Fc-Fc inter-chain disulfides thus allows for corresponding Fc-region mismatch driven exchange reactions without the need for reduction and thereby allowing in vivo application.

Example 12

[1528] Chain Exchange Reactions are Driven by Partially De-Stabilized Full Length Heavy chainMHCFcRP interfaces

[1529] The driver for conversion of 2/3-IgGs to bsAbs is a designed flawed interface between the full length heavy chain and the MHCFcRP. This artificial repulsive interface is the result of mutations introduced into the knob- or hole-CH3 domains of the MHCFcRP. The MHCFcRP still associate with the corresponding (normal) knob- or hole-partners during expression of 2/3-IgGs (see examples above). Those molecules have sufficient stability to present 2/3-IgGs as well behaved molecules without undesired aggregation tendencies.

[1530] Without being bound by this theory, the exchange reaction according to the current invention leading to bsAbs occurs when two complementary 2/3-IgGs come into close distance and the full length antibody heavy chain::MHCFcRP pairs are partially released next to each other. Re-assembly of the matching, i.e. not charged repulsed, knob-hole full length heavy chains should be favored under such conditions because the full length antibody heavy chain (CH3) interfaces are perfect. Thus, the full length heavy chains of the formed bsAb remain associated with preference over re-formation of the partially imperfect (charge mismatched) 2/3-IgG molecules. Thus, a designed partially de-stabilized (charge repulsed) CH3 interface is a key parameter for successful directed chain exchange reactions.

[1531] Partial de-stabilization of the Fc interface, especially the CH3-CH3 interface, can be achieved by mutating CH3 residues of the MHCFcRP while maintain the interacting residues on the full length antibody heavy chain.

[1532] Exemplary mutations that can be introduced into the CH3 domain of the MHCFcRP affecting the full length antibody heavy chain::MHCFcRP interface are provided in the following Table.

TABLE-US-00040 position perturbing (EU numbering) mutation(s) 345E R 347Q K 349Y W or E 351L F or Y 354S E or V 356D S or A or K 357E S or A or L or F or K 360K S or E 362Q E 364S V or L 366T I 368L F or V 370K E 390N E 392K E or D 394T I 397V Y 399D A or K 400S K 401D R 405F W 407Y W or L or I 409K D or E or I 439K E 441L Y

[1533] Some of the mutations include exchanges that place altered charges into the interface. Charge mutations either weaken or break previously existing stabilizing charge pairs or result in repulsion effects, or in both.

[1534] Similarly, amino acids with differently sized side chains can be introduced to generate steric repulsion effects. Such mutations either weaken or interfere with existing hydrophobic interface interactions or generate steric hindrances, or combine both.

[1535] Mutations that partially de-stabilize via charge and/or steric effects can also be combined with each other.

[1536] Furthermore, a first 2/3-IgG that contains charge and/or steric alterations introduced into its MHCFcRP can be combined with a second 2/3-IgG that contains different charge and/or steric alterations introduced into its MHCFcRP which match those of the MHCFcRP from the first 2/3-IgG.

[1537] The 2/3-IgGs as well as the resulting bsAbs assemble in a manner in which paired CH3 domains harbor knob-mutations on one side and hole-mutations on the other. Therefore, back-mutation to wild-type composition of corresponding knob- or hole-residues of the MHCFcRP generate also interface disturbances. Such combinations of knob- or hole-CH3-domains with wild-type domains are listed in the following Table.

TABLE-US-00041 perturbing backmutation CH3 hole position (EU numbering) 349C* Y 366S T 368A L 407V Y CH3 knob Fc position (EU numbering) 354C* S 366W T

[1538] These backmutations can be applied to partially destabilize the CH3-CH3-interface.

[1539] These backmutations can also be applied in combination with other perturbing mutations incl. those described in the previous Table.

[1540] All partially perturbing individual mutations or combination of mutations as described above can also be chosen in a manner that they partially destabilize the 2/3-IgG, yet stabilize a knob-MHCFcRP::hole-MHCFcRP heterodimer as the 2nd product of the exchange reaction and thereby shifting the reaction equilibrium further to the product side (exchange reaction).

Example 13

[1541] On-Cell Conversion of Monovalent 2/3-IgG Derivatives to Bivalent bsAbs According to the Current Invention

[1542] 2/3-IgG derivatives without inter-chain disulfide bonds between the heavy chain and the MHCFcRP do not require reduction to initiate the exchange reaction. It is therefore possible that exchange may also be achieved under physiological conditions, possibly even when individual 2/3-IgGs are bound to cell surfaces. If the functional Fab arms of 2/3-IgGs bind to cell surfaces, they accumulate on target cells. If two complementary 2/3-IgGs bind to the surface of the same cell, the chain exchange can occur while being bound directly on the surface of said cells. This exchange generates a fully functional bsAb with dual specificity directly on the cell surface.

[1543] To demonstrate this in situ on-cell chain exchange, two complementary 2/3-IgGs that bind either the antigens LeY or Herl are applied to cells that display either high levels of LeY, high levels of Herl, or high levels of both. It can be shown in FACS analyses that individually applied 2/3-BiFabs bind to cells with express their cognate antigen. Co-application (simultaneously or consecutive) of both 2/3-BiFabs results in increased binding only to cells that express both antigens. This indicates successful generation of functional bivalent bsAb products (with avidity-mediated improved binding) by on-cell exchange reactions of monovalent 2/3-BiFabs (prodrugs).

Example 14

[1544] Design & Composition and Functionality of 2/3-BiFabs without Heavy Chain:MHCFcRP Disulfide Bonds According to the Current Invention

[1545] TriFabs are antibody derivatives that harbor bispecific functionalities due to an exchange of the IgG CH2 domains to VH and VL, respectively. The Fc-like stem-region of such molecules is held together by intact KiH CH3 domains. This enables the generation of MHCFcRP containing BiFab analogues with potentially exchange-enabling features. FIG. 18 shows the design and composition of MHCFcRP containing 2/3-IgG-BiFab derivatives. Applying the same general principles as for 2/3-IgGs, engineered 2/3-BiFab analogues are composed of a KiH heavy chain and a MHCFcRP entity harboring a complementary VL or VH domain of an irrelevant antibody instead of the CH2 domain as well as a matching CH3 KiH domain. Thus, the CH2 domains of heavy chain and MHCFcRP in 2/3-IgGs becomes replaced by either a VH or a VL domain. In addition, and as a further difference to the 2/3-IgGs described in Example 1, the heavy chain as well as the MHCFcRP-stem of these 2/3-BiFab derivatives do not harbor cysteines that promote heavy chain:MHCFcRP covalent connections (analogous to 2/3-IgG derivative of Example 10). Because 2/3-BiFabs harbor exchange modules, i.e. CH3 domains, based on the same principle as 2/3-IgGs (without inter-chain-disulfides), exchange reactions can occur in the same manner as described and shown for 2/3-IgGs. The general principle of the 2/3-BiFab associated exchange reaction is shown in FIG. 18. The sequences of the light chains, knob- or hole-heavy-chains, and MHCFcRP hole- or knob-chains applied to produce 2/3-BiFabs are as follows:

TABLE-US-00042 Chain SEQ ID NO: anti-LeY antibody light chain 83 anti-LeY-antibody heavy chain with anti-dig antibody 84 variable domain as CH2 domain-knob anti-LeY-antibody heavy chain with anti-dig antibody 85 variable domain as CH2 domain-hole anti-MSLN-antibody heavy chain with anti-dig antibody 86 variable domain as CH2 domain-hole anti-MSLN antibody light chain (MSLN = mesothelin) 87 anti-LeY-antibody heavy chain with anti-CD3 antibody 88 variable domain as CH2 domain-knob anti-LeY-antibody heavy chain with anti-CD3 antibody 89 variable domain as CH2 domain-hole anti-LeY-antibody heavy chain with anti-CD-AG-2 antibody variable domain as CH2 domain-knob anti-LeY-antibody heavy chain with anti-CD-AG-2 antibody variable domain as CH2 domain-hole MHCFcRP with non-binding variable domain as CH2 domain-hole 90 with non-binding variable domain as CH2 domain-knob 91

Example 15

Expression & Purification of 2/3-BiFabs According to the Invention

[1546] Expression of 2/3-BiFabs was achieved by co-transfection of plasmids encoding the light chain, modified stem-heavy chain (knob or hole) and matching MHCFcRP-stem (hole or knob) into mammalian cells (e.g. HEK293) via state of the art technologies previously described (WO 2016/087416). The 2/3-BiFabs are secreted into culture supernatants like standard IgGs. Due to absence of a functional Fc-region as they lack CH2 domains, 2/3-BiFabs were purified by standard protein L (KappaSelect) affinity chromatography as shown in FIGS. 19 and 20. It is surprising that 2/3-BiFabs can be produced and purified in an effective manner even though they do not possess a functional V region in the stem region (as that is composed of non-matching VH and VL), and though they do not contain inter-chain disulfides for covalent connection of the chains. Size-exclusion and native mass-spec analytics showed correct assembly of purified 2/3-BiFab-derivatives as well as absence of undesired dimers and aggregates. The expression yields of 2/3-BiFabs under non-optimized transient expression conditions are listed in the following Table.

TABLE-US-00043 LeY-proDig LeY-proDig MSLN-proDig LeY-proCD3 LeY-proCD3 LeY-proCD-AG-2 LeY-proCD-AG-2 (knob)- (hole)- (hole)- (knob)- (hole)- (knob)- (hole)- MHCFcRP MHCFcRP- MHCFcRP MHCFcRP MHCFcRP MHCFcRP MHCFcRP 2/3 TriFabs (hole) knob (knob): (hole): (knob) (hole) (knob): HighTrap yield 130 190 135 68 86 70 120 Kappa [mg/L] Select 2/3- 100 150 110 35 75 50 100 BiFab [mg] by- 30 40 25 23 11 20 20 products [mg] SEC purified 90 150 110 35 70 50 100 2/3- BiFab [mg/L]

Example 16

Generation of Functional TriFabs by Reduction-Free Chain Exchange According to the Invention

[1547] Elimination of Fc-Fc inter-chain disulfides as shown above for 2/3-IgGs enables MHCFcRP driven exchange reactions without the need for controlled reduction and re-oxidation, i.e. under physiological conditions. It is, thus, possible to shuffle such molecules under physiological conditions, potentially even when already bound to target cell surfaces.

[1548] Therefore, 2/3-BiFab exchange reactions were performed without reduction as initial trigger of the exchange reaction. Input molecules into these exchange reactions were the LeY-binding 2/3-BiFabs with functional LeY binding arms and split Dig-binding stem region (proDig) as described above. Depletion of unreacted 2/3-BiFabs and MHCFcRP by-products was subsequently achieved via absorption of undesired His8-containing proteins on NiNTA resin. FACS analyses were subsequently applied to assess the binding functionality of the generated TriFab in comparison to the 2/3-BiFab precursor molecules. Therefore, LeY-antigen expressing MCF7 cells were exposed to either individual LeY-binding 2/3-BiFabs or to the TriFab of the exchange reaction. Dig-Cy5 was subsequently added and fluorescence of the cells was assessed. FIG. 21 shows low Dig-Cy5 associated signals with cells that were exposed to either 2/3-BiFabs even though both entities possess intact Fab arms which recognize the cell surface carbohydrate LeY.

[1549] The reason for inability of 2/3-BiFabs to bind the digoxigenylated payload is that they do not harbor a functional Dig-binding Fv in their stem region. The TriFab product of the chain exchange reaction, however, unambiguously displayed Dig-Cy5 associated signals, i.e. Dig-binding functionality. This demonstrates that 2/3-BiFab precursor molecules with inactivated binding functionality of their stem-Fv become converted via chain exchange to fully functional TriFabs.

Example 17

On-Cell Conversion of 2/3-BiFabs Prodrugs to Fully Functional Cell Bound Activated Bi- or Tri-Specific TriFabs According to the Invention

[1550] 2/3-BiFabs are only partially non-binding competent as they comprise one fully functional Fab arm. Only the Fv at the tip of the stem region is non-functional as it is composed of non-complementary VH and VL domains (of different antibodies or containing mutations that interfere with binding to cognate antigens, precursor inactive pro-form of the binding site). If the functional Fab arm binds to cell surfaces, 2/3-BiFabs accumulate on target cells. If two complementary 2/3-BiFabs (both carrying inactivated yet each other complementing stem-Fvs) bind to the surface of the same cell, chain exchange reactions can occur while being bound directly on the surface of said cell. This exchange then generates a fully functional TriFab with at least dual specificity directly on the cell surface (FIG. 22).

[1551] To experimentally demonstrate in situ on-cell chain exchange, we applied the individual 2/3-BiFab modules as well as those that were subjected to biochemical chain shuffling (described in examples above) to MCF7 cells followed by FACS analyses. MCF7 cells carry the LeY antigen on their cell surface and hence bind the individual 2/3-BiFabs with their functional Fab arms. FIG. 23 shows that individually applied 2/3-BiFabs bind to MCF7 cells, but are not capable to capture the 2nd target Dig-Cy5 (FIG. 23, rows 2 and 3). This reflects absence of functional stem-Fv in the 2/3-BiFab and hence on cells that bind only one 2/3-BiFab. Simultaneous application of both (complementary) 2/3-BiFabs however, enables to capture and retain Dig-Cy5 on the surface of MCF7 cells (FIG. 23, row 4). This indicates successful chain rearrangement/exchange and generation of functional TriFabs with Dig-binding functionality of the stem-Fv. Successful chain exchange of the stem region and recovery of the functional stem-Fv (Dig-Cy5 associated targeted FACS signals) was also observed upon consecutive application of the complementary 2/3-BiFabs. Application of the first entity to enable cell binding followed by extensive washing to remove unbound molecules and subsequent application of the 2nd entity also generates FACS signals on antigen positive cells. This confirms that successful exchange reaction (either by simultaneous or sequential application) can occur under physiological conditions on the surface of target cells.

Example 18

On-Cell Chain Conversion of Targeted Anti-CD3-Prodrug 2/3-BiFabs to Fully Functional Bispecific Antibodies According to the Invention

[1552] To demonstrate that on-cell conversion of 2/3-BiFab prodrugs is a general principle that can be applied for different binding specificities, 2/3-BiFabs were generated that contain VH or VL domains of CD3-binding antibodies. T-cell bispecific antibodies (also termed T-cell recruiters) are proteins that combine binding entities that recognize antigen on the surface of tumor cells with CD3-binding functionalities. Such molecules bind to tumor cells via their tumor-antigen-binding entities as well as to T-cells (via CD3-binding functionality). That in turn generates (activation) signals and processes which ultimately results in tumor cell lysis/death mediated by the antibody binding-induced T-cell attack.

[1553] 2/3-BiFabs with anti-CD3-prodrug functionality, i.e. the CD3 binding site is located in the stem region and, thus, not binding competent in the 2/3-BiFab educts, were designed as described above (see Examples 14 and 15). The light chains, knob- or hole-heavy chains and matching MHCFcRPs were co-expressed to generate the 2/3-BiFab LeY-proCD3(hole)-MHCFcRP(knob) and the 2/3-BiFab LeY-proCD3(knob)-MHCFcRP(hole). 2/3-BiFabs were purified as described above (see Table in Example 15) and subjected to on-cell chain exchange reaction.

[1554] To experimentally demonstrate on-cell chain exchange of 2/3-BiFabs with anti-CD3-prodrug functionality, these were then applied to MCF7 cells. Presence or absence of CD3-binding functionality was subsequently determined via cell-based reporter assays that generate signals upon CD3-receptor binding (Promega T-cell Activation Bioassay (NFAT), cat. # J1621, FIG. 24).

[1555] MCF7 cells carry the LeY antigen on their cell surface and hence the individual 2/3-BiFabs can bind with their functional Fab arms. In FIG. 25 it can be seen that individually applied 2/3-BiFabs bind to MCF7 cells but generate no/low CD3 reporter signals. This reflects absence of CD3-binding functionality in these molecules. Simultaneous application of both (complementary) 2/3-BiFabs however generated significant signals that reflect efficient CD3-binding. This indicates successful on-cell/in vivo chain rearrangement/exchange and generation of functional TriFabs with CD3-binding functionality on the cells.

[1556] Successful chain exchange of the stem region and recovery of the functional stem-Fv (CD3-signals) was also observed upon consecutive application of the complementary 2/3-BiFabs. Application of the first entity to enable cell binding followed by extensive washing to remove unbound molecules and subsequent application of the 2nd entity also generated signals on antigen positive cells. This confirmed that successful exchange and activation of targeted anti-CD3 prodrug molecules reaction can occur under physiological conditions on the surface of target cells.

Example 19

On-Cell Conversion of Different Antigen Targeting 2/3-BiFab Prodrugs to Fully Functional Cell Bound Activated Tri-Specific TriFabs According to the Invention

[1557] 2/3-BiFab mediated chain exchange and subsequent activation of prodrug-like antibody derivatives can be combined with dual antigen binding principles. This enhances the prodrug-activation specificity as depicted in FIG. 26: pairs of 2/3-BiFabs that upon chain exchange generate functional TriFabs can be generated which comprise stem-Fv complementing functionalities of the same specificity (e.g. CD3 (=CD-antigen 1) or CD-AG-2 (=CD-antigen 2) or other binders that benefit from targeted prodrug approaches), but comprise cell surface binding Fab arms of differing specificities. Thereby, productive chain exchange and recovery of stem-Fv functionality occurs only on cells that express both antigens. In such settings, exchange reaction generates on the cell surface trispecific TriFabs with one Fab arm derived from each 2/3-BiFab and the stem-Fv recovered from complementing VH and VL from both. High specificity of prodrug-activation is provided because the stem-Fv functionality is absent in the individual 2/3-BiFab or on cells that bind only one 2/3-BiFab. Only cells that carry target antigens for both complementary 2/3-BiFabs (in sufficient density) enable chain exchange and thereby re-creation of the functional stem-Fv region. On-cell generation of the bispecific binding functionality also contributes to avidity as it converts monovalent to bivalent cell surface binders. Thus, it increases and stabilizes the binding of the TriFab derivative on the surface of the target cell (avidity-mediated improvement as shown also for disulfide-lacking 2/3-IgGs).

[1558] Thus, cell surface concentration is higher and thereby additional specificity is gained.

[1559] FIG. 27 shows the experimental results for on-cell conversion of different antigen targeting 2/3-BiFab prodrugs to fully functional, cell bound, activated, tri-specific TriFabs. Complementary 2/3-BiFabs binding either the cell surface antigen LeY or the cell surface antigen mesothelin (MSLN) were produced and combined for 6 hours on a cell line that simultaneously express both antigens. The (inactivated) stem-Fv of both constructs harbored VH or VL of a Dig-binding antibody. FACS analyses indicated lack of relevant Dig-binding activity upon application of only the LeY-binding or only the mesothelin-binding 2/3-BiFab (FIG. 27, columns 2 and 3). Co-application of both, however, lead to generation of Dig-binding cell surface associated functionalities as indicated by increased fluorescence of these cells (FIG. 27, column 4).

Example 20

On-Cell Conversion of Different Antigen Targeting TriFab Prodrugs to Tri-Specific TriFabs According to the Invention

[1560] 2/3-BiFab derivatives as starting molecules comprise as MHCFcRP a modified stem-Fv region without Fab arm at their N-terminus. Attachment of Fab arms to these entities and expression thereof in combination with heavy chains generates exchange-enabled 2/3-TriFab-derivatives as shown in FIG. 28. Those molecules have their MHCFcRPs altered to potentially antigen-binding competent chains whichupon finding their corresponding partnerexchange to functional trispecific TriFabs.

[1561] Thus, 2/3-TriFab derivatives with cell surface target binding specificity A can be generated that harbor a non-functional stem-Fv composed of VH with specificity X and VL of specificity Y. Correspondingly, complementary 2/3-TriFab derivatives can be generated with cell surface target binding specificity B harbor a non-functional stem-Fv composed of VH with specificity Y and VL of specificity X. On-cell chain exchange of such molecules generates two types of trispecific TriFabs both carrying (avidity-enhanced) bispecific cell binding Fab arms (both A+B). One of those TriFabs harbors fully active stem-Fv functionality of the first specificity, the other TriFab contains the second stem Fv functionality (either fully active X TriFab or fully active Y TriFab). Applying such TriFab-prodrug pairs, for example, should enable the simultaneous conversion of inactive to active CD3 as well as CD-AG-2 binders (or of other binder pairs that benefit from avidity-enhanced specific prodrug activation) selectively on only those cells that simultaneously express two defined (cancer) surface antigens in sufficient density.

Example 21

Tri- or Tetraspecific 2/3-BiFab Prodrug Derivatives According to the Invention

[1562] 2/3-BiFab derivatives were designed and can be generated as starting molecules for the exchange reaction as reported herein, wherein the MHCFcRPs are covalently conjugated to the C-terminus of the heavy chain via a peptidic linker, such as e.g. a (G4S)6 linker. This generates a non-functional entity resembling a single-chain stem-module of TriFabs as shown in FIG. 29. Attachment of Fab arms to these single-chain stem modules generates exchange-enabled TriFab-derivatives. Two of these can exchange to functional tri- or tetraspecific antibody derivatives as shown in FIG. 29.

TABLE-US-00044 Chain SEQ ID NO: single chain anti-LeY-Dig TriFab heavy 92 chain (knob) - Bio VL (hole) single chain anti-LeY-Dig TriFab heavy 93 chain (hole) - Bio VH (knob)

[1563] The LeY binding region can be exchanged with sequences that enable binding to other antigens to generates antibody-prodrugs that bind different antigens on cells and rearrange in the same manner as described in FIG. 28.

Example 22

On-Cell Chain Conversion of Targeted Anti-CD3-Prodrug 2/3-BiFabs According to the Invention Enables Effective T-Cell-Mediated Killing of Tumor Cells

[1564] This example demonstrates that on-cell conversion of 2/3-BiFab prodrugs to CD3-binding TriFabs enables T-cell mediated killing of the targeted tumor cells: LeY-tumor antigen binding 2/3-BiFabs were generated that contain either VH or VL domains of CD3-binding antibodies. Design and generation of those 2/3-BiFab prodrugs is described in the previous examples. The VH and VL sequences of the CD3-binder of these 2/3-BiFab prodrugs are described in US 2015/0166661 A1.

[1565] To demonstrate that those molecules induce cell-mediated killing upon simultaneous binding, they were applied at different concentrations to LeY positive MCF7 cells. Therefore, MCF7 cells were seeded out in 96 well plates and incubated overnight, followed by exposing those cells to 2/3-BiFab anti-LeY-proCD3 (knob)-MHCFcRP (hole) and 2/3-BiFab anti-LeY-proCD3 (hole)-MHCFcRP (knob). These components were either added individually/sequentially or in combination. To assess T-cell mediated killing, PBMCs from whole blood of healthy donors (isolated via state-of-the-art Ficoll purification) were added in a 5:1 ratio. Cultures were then maintained at 37 C., 5% CO.sub.2 for 48 hours, followed by assessment of the degree of tumor cell lysis (applying state-of-the art LDH release assays).

[1566] The results of these assays are shown in FIG. 30.

[1567] MCF7 cells carry the LeY antigen on their cell surface and hence the individual 2/3-BiFabs can bind with their functional Fab arms. Individually applied 2/3-BiFab, i.e. administration of either anti-LeY-proCD3 (knob)-MHCFcRP (hole) or 2/3-BiFab anti-LeY-proCD3 (hole)-MHCFcRP (knob) does not result in relevant T-cell-mediated killing (no release of LDH) even at high molar concentrations. This reflects absence of CD3-binding functionality (required for T-cell mediated cell lysis) of these molecules.

[1568] In contrast, simultaneous application of both (complementary) 2/3-BiFabs resulted in significant LDH release. This reflects significant tumor cell lysis already at quite low concentrations. The reason for this is successful chain rearrangement/exchange on tumor cell surfaces and generation of functional TriFabs with CD3-binding functionality. Those functional CD3-binding TriFabs then recruit and engage T-cells whichin turnleads to targeted tumor cell lysis.

Example 23

On-Cell Chain Conversion of AG-4 and EGFR-Targeted Anti-CD3-Prodrug 2/3-BiFabs According to the Invention Enables Effective T-Cell-Mediated Killing of Tumor Cells and Strong Cytokine Release

[1569] Analogously to the experimental setup described in Example 22, AG-4 expressing HELA cells and EGFR expressing A431 cells were targeted with respective 2/3-BiFabs.

[1570] Administration of either anti-AG-4-proCD3 (knob)-MHCFcRP (hole) or 2/3-BiFab anti-AG-4-proCD3 (hole)-MHCFcRP (knob) to HELA cells does not result in relevant cell-mediated killing (no release of LDH). This reflects absence of CD3-binding functionality (required for T-cell mediated cell lysis) of these molecules.

[1571] Additionally, administration of either anti-EGFR-proCD3 (knob)-MHCFcRP (hole) or 2/3-BiFab anti-EGFR-proCD3 (hole)-MHCFcRP (knob) to A431 cells does not result in relevant cell-mediated killing (no release of LDH).

[1572] In contrast, simultaneous application of both (complementary) 2/3-BiFabs resulted in significant LDH release in both target cell setups. This reflects significant tumor cell lysis already at quite low concentrations (FIG. 31). The reason for this is successful chain exchange between the 2/3-BiFabs on tumor cell surfaces and generation of functional TriFabs with CD3-binding functionality.

[1573] Moreover, FIG. 32 shows the amounts of secreted cytokines at concentrations of 4 nM for AG-4 targeting on HELA cells. Significant more amounts of IL-2, IFN-, Granzyme B and TNF are present in the setting where both anti-AG-4-proCD3 (knob)-MHCFcRP (hole) and 2/3-BiFab anti-AG-4-proCD3 (hole)-MHCFcRP (knob) were applied reflecting a strong immune response also on cytokine level (FIG. 32).

Example 24

[1574] Dual Targeting of EGFR and AG-4 on HELA Cells Enables On-Cell Chain conversion of anti-CD3-prodrug 2/3-BiFabs according to the invention and Effective T-Cell Activation

[1575] Dual targeting was evaluated in a functional assay using reporter cell line that generates signals upon CD3-receptor binding (Promega T-cell Activation Bioassay (NFAT), cat. # J1621).

[1576] HELA cells that express both cell surface antigen 4 (AG-4) and epidermal growth factor (EGFR) (FIG. 33) were treated with either 2/3-BiFab targeting AG-4 or EGFR. The (inactivated) stem-Fv of both constructs harbored VH or VL of a CD3-binding antibody. The results indicated lack of relevant CD3-binding activity upon application of only the EGFR-proCD3 (knob)-MHCFcRP (hole) or only the AG-4-proCD3 (hole)-MHCFcRP (knob). Co-application of both, however, leads to significant binding to CD3 and activation of the reporter cell line (FIG. 34). As control EGFR-proCD3 (hole)-MHCFcRP (knob) and AG-3-proCD3 (knob)-MHCFcRP (hole) molecules were analyzed. In regard to antigen expression HELA cells are AG-3 negative, hence only the EGFR targeted 2/3-BiFabs are able to bind. This control serves as another prove that the shuffling reaction takes place on the cell surface and only when both 2/3-BiFabs are bound to the cell surface. In case the conversion would have been efficient even at low concentrations in media, the conversion product EGFR/AG-3/CD3 TriFab would have been able to bind the cell surface via the EGFR-binding entity and, thus, induce a CD3-mediated activationwhich was not detected.

Example 25

[1577] Trispecific 2/3 Fab Prodrug Derivatives with Single-Chain Stem Motive Undergo On-Cell Conversion According to the Invention, Thereby Generate Two Additional Binding Sites and Strongly Activate T-Cells

[1578] 2/3-BiFab derivatives were designed and were generated as starting molecules for the exchange reaction as reported herein, wherein the MHCFcRPs were covalently conjugated to the C-terminus of the heavy chain via a peptidic linker, such as e.g. a (G4S)6 linker (SEQ ID NO: 816). This generated a non-functional entity resembling a single-chain stem-module of TriFabs as shown in FIG. 29. Attachment of Fab arms to these single-chain stem modules generated exchange-enabled TriFab-derivatives. Two of these were shown to exchange to functional tri-specific antibody derivatives as shown in FIG. 29.

TABLE-US-00045 Chain SEQ ID NO: single chain anti-LeY-Dig TriFab heavy 92 chain (knob) - Bio VL (hole) single chain anti-LeY-Dig TriFab heavy 93 chain (hole) - Bio VH (knob)

[1579] FIG. 35 depicts the setup for the binding studies via flow cytometry. The obtained results are shown in FIG. 36. Whereas, either of the constructs alone did not lead to binding of Dig-Cy5 or Bio-488 (FIG. 36A: row 2 and 3; FIG. 36B: row 2 and 3), the co-application of both, however, lead to generation of Dig and Bio-binding cell surface associated functionalities as indicated by increased fluorescence of these cells (FIG. 36A: row 4; FIG. 36B: row 4).

[1580] Accordingly, trispecific antibodies were generated with CD3 and CD-AG-2 binding entities by combining single chain anti-LeY-CD3-TriFab heavy chain (knob)-CD-AG-2-VL (hole) and single chain anti-LeY-CD3-TriFab heavy chain (hole)-CD-AG-2-VH (knob) (FIG. 37). The ability of these molecules to activate T-cells was proven by using (Promega T-cell Activation Bioassay (NFAT), cat. # J1621) and is shown in FIG. 38.

Example 26

[1581] Fab-Shaped MHCFcRP Dimers with an Additional Binding Site as by-Products of the On-Cell Chain Conversion of Targeted Anti-CD3-Prodrug 2/3-BiFabs According to the Invention

[1582] As depicted in FIG. 18, the MHCFcRP by-products with specificity 4 can be non-binding. However, by adding VH or the respective VL of a functional binding entity into the MHCFcRP, a functional, i.e. binding competent, Fab molecule was generated during the exchange reaction. To show the binding functionality of the MHCFcRP by-products, LeY-targeting 2/3-BiFabs pairs carrying CD-AG-2 MHCFcRP were added either alone or in combination to media. In addition, LeY-negative Jurkat cells that express CD-AG-2 were added. The whole setup is shown in FIG. 39. The MHCFcRP by-products were detected with PE-anti His6 antibody. 2/3-BiFabs (FIG. 40, row 2 and 3) alone did not lead to a His6 binding of the detection antibody due to the fact that the Jurkat cells are LeY negative and the 2/3-BiFab could not bind the cell surface. Combining the two 2/3-BiFabs lead to the generation of MHCFcRP by-products which were able to bind to CD-AG-2 and could be detected via an anti-His6 antibody by flow cytometry (see FIG. 40, row 4). Non-reacted 2/3-BiFabs were not able to bind the cell, because of absent LeY expression.

Example 27

[1583] MHCFcRP with an Additional N-Terminal Fused Fab as Targeting Entity Allow Bivalent Cell Targeting, Mediate On-Cell Conversion and the Production of Trivalent MHCFcRP by-Products According to the Invention

[1584] 2/3-BiFab derivatives as starting molecules comprise as MHCFcRP a modified stem-Fv region without Fab arm at their N-terminus. Attachment of Fab arms to these entities and expression thereof in combination with heavy chains generates exchange-enabled 2/3-TriFab-derivatives as shown in FIG. 41. Those molecules have their MHCFcRPs altered to LeY-binding competent chains whichupon finding their corresponding partnerexchange to functional trispecific TriFabs.

[1585] Thus, 2/3-TriFab derivatives with LeY target binding specificity can be generated that harbor a non-functional stem-Fv composed of VH with CD3 specificity and VL of CD-AG-2 specificity, or vice versa. On-cell chain exchange of such molecules generates two types of trispecific TriFabs both carrying (avidity-enhanced) bispecific LeY Fab arms. One of those TriFabs harbors fully active stem-Fv functionality of the CD3 specificity, the other TriFab contains the CD-AG-2 stem Fv functionality. Applying such TriFab-prodrug pairs to LeY-expressing MCF7 cells, for example, enable the simultaneous conversion of inactive to active CD3 as well as CD-AG-2 binders.

[1586] Results are depicted in FIG. 42. The results indicated lack of relevant T-cell activating entities upon application of only the LeY-proCD3 (knob)-LeY-proCD-AG-2-MHCFcRP(hole) or LeY-proCD3 (hole)-LeY-proCD-AG-2-MHCFcRP(knob). Co-application of both, however, leads to a significant T-cell activation.

Example 28

Alternative 2/3-BiFab Derivatives Remain CH2-Dependent FcRn Binding Intact and Show On-Cell Chain Conversion According to the Invention

[1587] 2/3-BiFab derivatives as starting molecules comprise as MHCFcRP a modified stem-Fv region. To keep the CH2 domain from conventional IgG and hence the ability to bind FcRn for prolonged half-life, the variable fragment can be attached at the C-terminal end of a CH2-CH3 dimer as depicted in FIG. 43. The partially destabilizing mutations in the CH3 domain are the same as in 2/3-BiFabs. Alternatively, the variable fragment can be introduced in-between the Fc-part (CH2+CH3) and the Fab part as shown in FIG. 44. All antibodies were expressed as described earlier in good yields (72 mg/L to 161 mg/L). FIG. 45 confirms the ability of the CH2 containing formats to display a higher binding to FcRn revealing an extended serum half-life compared to initial 2/3-BiFabs (the analytical FcRn affinity chromatography was performed as described in Schlothauer, T., et al., MAbs 5 (2013) 576-586).

[1588] The ability of CH2 containing 2/3-BiFabs to confer on-cell chain conversion was analyzed by flow cytometry as described above. Two educt molecules were sequentially applied (with two times wash in-between) to MCF7 cells. Upon chain exchange a functional anti digoxigenin binding site was assembled and hence able to bind digoxigenylated Cy5 on the cell surface, which was detected by FACS. Both molecule classes according to FIGS. 43 and 44, successfully underwent chain exchange reaction according to the invention as depicted in FIG. 46.

Example 29

[1589] CH2 Domain Containing 2/3-BiFabs are Able to Convert into Functional CD3-Binding Sites in a Method According to the Invention and Thereby Mediate T-Cell Activation

[1590] CH2-containing 2/3-BiFabs with the variable fragment attached at the C-terminal end (FIG. 47) were analyzed for their ability to generate a functional CD3 binder using the Jurkat activation assay as outlined above. MCF7 cells served as target cells, LeY as target antigen. Whereas the 2/3-BiFabs did not induce a significant Jurkat activation, the combination of both resulted in a dose-dependent activation and light emission in the reporter system (RLU) (FIG. 48).

Example 30

[1591] Transmission Electron Microscopy Confirms the Proposed Structure of 2/3-BiFabs and the Corresponding Product and Reveals them as being Highly Flexible

[1592] To analyze the shape and structure of the 2/3-BiFabs and the respective chain exchange product obtained in a method according to the invention Negative Stain Transmission Electron Microscopy (NS-TEM) was performed. Results (FIG. 49) reveal a rigid intradomain character but high interdomain flexibility.

[1593] Grid Preparation: Freshly thawed samples are diluted in D-PBS to a concentration of about 5 mg/ml. 4 l of the diluted sample was adsorbed to glow discharged 400 mesh carbon coated parlodion copper grids, washed with 3 drops of water, incubated with 3 l of TMV containing solution, further washed with 2 drops of water and finally stained with 2 drops of uranyl acetate 2%.

[1594] Transmission electron microscopy: Samples were imaged using a Tecnai12 transmission electron microscope (FEI, Eindhoven, The Netherlands) operating at 120 kV. Electron micrographs were recorded on a 20482048 pixel charge-coupled device camera (Veleta Gloor Instruments) at a nominal magnification of 195,000 yielding a final pixel size of 0.296 nm on the specimen level. Alternatively, samples were imaged using a FEI Tecnai G2 Spirit TEM (FEI, Eindhoven, The Netherlands) operating at 80 kV. Electron micrographs were then recorded on a 20482048 pixel charge-coupled device camera (Veleta Soft Imaging Systems) at a nominal magnification of 135,000 yielding a final pixel size of 0.33 nm on the specimen level.

[1595] Image processing: Reference-free alignment was performed on manually selected particles from recorded images using the EMAN2 image processing package (see e.g. G. Tang, L., et al., J. Struct. Biol. 157 (2007) 38-46). The extracted particles were aligned and classified by multivariate statistical analysis yielding 2D class averages. Additionally, when class averaging was not possible, raw images of particles were also manually stained for clarity using Photoshop.

Example 31

Target-Unrelated Chain Conversion Occurs Inefficiently Only at High Concentrations

[1596] To address whether chain conversion also occurs at high concentrations in media, 2/3-BiFabs were applied to HELA cells co-cultured with PBMC. In the first setup, anti-AG-4-proCD3 (hole)-MHCFcRP (knob) and EGFR-proCD3 (knob)-MHCFcRP(hole) were applied (on-cell conversion takes place since both targets are expressed on HELA cells). In the second setup anti-AG-4-proCD3(hole)-MHCFcRP(knob) and AG-3-proCD3(knob)-MHCFcRP(hole) were applied (AG-3 is not expressed on HELA cells, thus on-cell conversion should not be possible). However, the percentage of killing at concentrations of 300 nM reveals, that conversion in media and monovalent binding to the HELA cell surface via the AG-4 binding site is rarely occurring and mediating killing of target cells only to a low percentage (FIG. 50).