Antibodies binding to CD2, and multifunctional molecules comprising such antibodies

Abstract

The present invention relates to an antibody that binds to CD2, or a target-binding fragment or derivative of such antibody, and to bi- or multifunctional constructs comprising such antibody.

Description

[0001] Evotec International GmbH

[0002] Antibodies binding to CD2, and multifunctional molecules comprising such antibodies

[0003] REFERENCE TO SEQUENCE LISTING SUBMITTED AS A COMPLIANT XML 1.0 FORMAT FILE (.xml)

[0004] Pursuant to the EFS-Web legal framework and 37 CFR §§ 1.821-825 (see MPEP § 2442.03(a)), Rule 30 EPC, and § 11 PatV, an electronic sequence listing compliant with WIPO standard ST.26 in the form of an XML 1.0 format file is submitted concurrently with the instant application, and the entire contents of the sequence listing are incorporated herein by reference. For the avoidance of doubt, if discrepancies exist between the sequences mentioned in the specification and the electronic sequence listing, the sequences in the specification shall be deemed to be the correct ones.

[0005] FIELD OF THE INVENTION

[0006] The present application relates to antibodies binding to CD2, and multifunctional molecules comprising such antibodies.

[0007] BACKGROUND

[0008] CD2 (Uniprot ID P06729) is a transmembrane glycoprotein of the immunoglobulin superfamily. Its extracellular domain (amino acids 25-209) consists of two immunoglobulin domains, a membrane distal V-set domain and a membrane proximal C2-set domain, that are connected via a transmembrane domain (amino acids 210-235) to a proline-rich cytoplasmic domain (amino acids 236-351). CD2 is expressed on the cell surface of T cells, NK cells, dendritic cells and thymocytes and plays important roles in T cell activation, adaptive NK cell response and thymocyte development. Via its V-set domain (amino acids 31-126) CD2 binds its primary ligand human CD58 (also known as LFA3). It also binds, but with at least 10-fold lower affinity, human CD48 (also known as SLAMF2). CD2 is unique as a costimulatory receptor because it allows to engage and coactivate both T cells and NK cells, the two complementary cytotoxic effector cell types of the immune system that are key to a successful cancer immunotherapy.

[0009] T cells recognize and kill tumor target cells primarily via binding of their respective T cell receptor (TCR) complexes to antigenic peptides presented by major histocompatibility complex I (MHC-I) molecules on the surface of tumor target cells. This interaction is referred to as signal 1 of T cell activation. In various new cancer immunotherapy approaches, a strong signal 1 is artificially provided: For adoptive T cell therapy, T cells are engineered to express chimeric antigen receptors (CARs), which are artificial TCR molecules that consist of an extracellular binding moiety, being e.g. specific for a tumor antigen, (typically either a single chain variable fragment or a single domain antibody) that is attached via a transmembrane domain to intracellular signaling domains derived from various TCR components. An alternative approach is to recruit endogenous CD3+ T cells via so called CD3-based immune cell engager molecules, which are bispecific antibodies that recognize with one arm the CD3 component of the normal TCR complex on the surface of T cells and with the other arm a tumor selective surface antigen. This latter approach redirects the cytotoxic activity of any endogenous T cell towards tumor cells by providing a strong signal 1 irrespective of the cognate antigenic peptide of its endogenous TCR, while at the same time circumventing human leukocyte antigen (HLA) restriction. However, for an optimal T cell response and to avoid exhaustion of T cells, a second co-stimulatory signal (referred to as signal 2) needs to be provided in addition to signal 1. That is why all clinically approved CAR constructs do not only contain a CD3z signaling domain, which provides signal 1, but also a co-stimulatory signaling domain derived from either the CD28 or the 4-1BB receptor. To also provide co-stimulation in the context of its CD3-based immune cell engagers, Regeneron has recently developed bispecific antibodies that bind CD28 with one arm and a tumor associated antigen (TAA) with the second arm. These molecules can be combined with CD3 -based immune cell engagers directed against the same tumor to provide both signal 1 and 2 for optimal T cell activation. Roche has demonstrated preclinically that combining their CD20-targeted CD3 T cell bispecific with a co-stimulatory CD19-targeted 4- 1BBL bispecific achieved tumor remission in mouse models. Apart from CD28 and 4-1BB, CD2 can also provide an efficient co-stimulatory signal 2 by the interaction with its ligand CD58. The CD2 / CD58 interaction substantially enhances the efficiency of antigen recognition by T cell receptors (TCRs) and has been found to be the primary costimulatory pathway in the, with age increasing, proportion of CD8+ T cells that lack CD28. Moreover, CD2 co-stimulation of CD8 T cells in vitro has been shown to drive a non-exhausted T cell phenotype in the presence of persistent TCR stimulation. It is also important to note that loss or the downregulation of CD58 has been identified as a cancer cell intrinsic mechanism of resistance or escape in tumors of melanoma patients with acquired resistance to therapy with immune checkpoint inhibitors. Thus, co-stimulatory CD2xTAA bispecifics offer great potential to further enhance the therapeutic benefits of treatment with immune checkpoint inhibitors (endogenous signal 1) and with CD3-based immune cell engagers (artificial signal 1).

[0010] CD2xTAA bispecifics have not been published yet and it is not clear which epitope on CD2 works best for a co-stimulatory bi specific construct. Of the 3 well-characterized epitopes in the extracellular domain of CD2, which are referred to as T11.1, T11.2 and T11.3, the former two are located in the V-set domain and antibodies against these epitopes block CD58 binding. The epitope T11.3 has been reported to be part of the C2-set domain and its accessibility is increased upon T cell activation. For a co-stimulatory bispecific construct, it would be preferable to be able to bind to all T cells rather than only activated T cells.

[0011] In contrast to bispecifics that target CD28 or 4-1BB, CD2xTAA bispecifics in principle allow to also engage and coactivate NK cells, because NK cells also express CD2 on their surface. Moreover, simultaneous engagement of CD 16a via the Fc moiety of an antibody (signal 1 for NK cell activation) and of CD2 by its ligand CD58 (signal 2) is known to result in costimulation of the antibody-dependent cellular cytotoxic (ADCC) activity of adaptive NK cells. Therefore, engagement of NK cells by means of a CD2 bispecific can be pursued as a best-in- class approach for T cell co-stimulation or for ADCC. However, the main problem that needs to be overcome when designing such molecules is fratricide, a phenomenon where immune effector cells kill other immune effector cells. Antibodies that can simultaneously engage CD 16 and another NK cell surface target, like e. g. in the case of the CD2 targeting IgGl antibody Siplizumab, have been shown to cause NK cell fratricide by ADCC.

[0012] Hence one object of the present invention is the identification of CD2 binders and a multispecific antibody format that allows to engage T cells and exert a co-stimulatory activity. Another object of the present invention is the identification of CD2 binders and a multispecific antibody format that enhances ADCC.

[0013] Another object of the present invention is the identification of CD2 binders and a multispecific antibody format that activates NK cells without causing major NK cell fratricide.

[0014] These and other objectives are solved by the features of the independent claims. The dependent claims disclose embodiments of the invention which may be preferred under particular circumstances. Likewise, the specification discloses further embodiments of the invention which may be preferred under particular circumstances.

[0015] BRIEF DESCRIPTION OF THE FIGURES

[0016] Figure 1 shows FACS-based sorting of effector memory T cells (TEM).

[0017] Figure 2A shows the criteria used to define hit binders in the flow cytometry screening of hybridoma. “TEM Abs” refers to all antibodies that were obtained in a screen against effector memory T cells.

[0018] Figure 2B shows representative examples of results obtained from the plate-bound antibody TCR co-stimulation assay using antibodies in mouse IgG format.

[0019] Figure 2C shows representative examples of results obtained from the plate-bound antibody TCR co-stimulation assay using antibodies in chimeric human IgG4 format.

[0020] Figure 2D shows confirmation of expression of 18 selected immune checkpoint receptors transiently expressed in HEK293 cells.

[0021] Figure 2E shows CD2 binding flow cytometry data for an isotype control antibody and three anti-CD2 antibodies in chimeric human IgG4 format.

[0022] Figure 3 is a bar graph representation of flow cytometry data showing antibody binding to CD2 knockout, mock-treated and untreated CD4+primary T cells.

[0023] Figure 4A is a bar graph representation of ELISA-based assessment of cross-reactivity to human CD2 and CD2 of cynomolgus monkey (Macaca fascicularis, in the following: “cyno CD2”). Figure 4B is a graphical representation of flow cytometry-based data assessing binding of antibodies to human CD2 and CD2 of mouse (Mus musculus, in the following: “mouse CD2”).

[0024] Figure 5 shows a representative SPR sensorgram overlay of association and dissociation profiles of proprietary CD2 antibodies and the reference CD2 antibody, Siplizumab.

[0025] Figure 6 shows summary SPR data for Figure 5.

[0026] Figure 7A shows schematic representations of two assay set-ups to study the impact of CD2 antibodies on the CD2-CD58 receptor ligand interaction by flow cytometry.

[0027] Figure 7B shows bar graph representations of flow cytometry data revealing binding intensities of CD58-Fc to CD8+T cells in presence or absence of CD2 antibodies or an isotype control.

[0028] Figure 7C is a graphical representation of flow cytometry-based concentration response analysis data showing the impact of CD2 antibodies on the binding of CD58-Fc to primary CD8+T cells using set-up 2 from Figure 7A.

[0029] Figure 8 shows bar graph representations of results from FACS-based epitope binning assays.

[0030] Figure 9 shows the results of domain mapping experiments using HEK293 cells transfected with human CD2 wt and deletion mutants. Figure 9A shows a schematic representation of the GFP-fusion constructs used, and Figure 9B shows a bar graph representation of flow cytometry data.

[0031] Figure 10 is an alignment of the amino acid sequences of full-length human CD2 (Uniprot code: P06729) and mouse CD2 (Uniprot code: P08920).

[0032] Figure 11A is a schematic representation of full length human CD2 and human / mouse CD2 chimera constructs generated and expressed in HEK293. Figure 1 IB shows bar graph representations of median fluorescence intensity (MFI) data from flow cytometry analysis investigating binding of antibodies to full length human CD2 and human / mouse CD2 chimeric constructs.

[0033] Figure 12 is a schematic drawing of an exemplary format for a tumor targeted CD2 bispecific designed to provide a costimulatory signal to effector T cells.

[0034] Figure 13 is a non-linear regression representation of flow cytometry data showing binding of serially diluted Fc-inactive bispecific antibodies in the IgG-L-scFv format to CD8+T cells and CD2-transfected HEK293 cells.

[0035] Figure 14 is a non-linear regression representation of flow cytometry data showing binding of serially diluted Fc-inactive Trastuzumab-based bi specific antibodies in the IgG-L-scFv format to HER2-positive HCC1954 breast cancer cells.

[0036] Figure 15 shows non-linear regression representations of flow cytometry data showing conjugate formation between CD8+T cells and HCC1954 target cells.

[0037] Figure 16 shows non-linear regression representations of flow cytometry data revealing binding intensities of CD58-Fc to CD8+T cells in the presence of parental anti-CD2 IgG, Trastuzumab, Fc-inactive Trastuzumab-based CD2 bispecifics or control bispecific antibodies.

[0038] Figures 17A and B are non-linear regression representations of anti -tumor T cell cytotoxicity data. Primary human T cells were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) in the presence of Fc-inactive Trastuzumab-based CD2 bispecifics or control bispecific antibodies.

[0039] Figures 18A and B are non-linear regression representations of CD25 upregulation data as marker of T cell activation. Primary human T cells were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) in the presence of Fc-inactive Trastuzumab-based CD2 bispecifics or control bispecific antibodies.

[0040] Figure 19 is a non-linear regression representation of anti -tumor cytotoxicity data. Adherent cell-depleted human PBMCs were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) in the presence of Fc-inactive Trastuzumab-based CD2 bispecifics or control bispecific antibodies.

[0041] Figure 20 is a non-linear regression representation of IFN-gamma release data. Adherent cell- depleted human PBMCs were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) in the presence of Fc-inactive Trastuzumab-based CD2 bispecifics or control bispecific antibodies.

[0042] Figure 21 is a bar graph representation of anti -tumor cytotoxicity data in the absence of signal 1. Adherent cell-depleted human PBMCs were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) in the presence of Fc-inactive Trastuzumab-based CD2 bispecifics or control bispecific antibodies.

[0043] Figures 22A and B show non-linear regression representations of anti-tumor cytotoxicity data cConcentration responses to increasing signal 2 doses in the presence of a fixed signal 1 dose are shown. In Figure 22 A, primary human T cells were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+). In Figure 22B, primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+and HER2+). All co-cultures were done at at an effector-to-target cell ratio of 2: 1 and treated with a fixed concentration of 5 pM EpCAM*CD3 BiTE in the presence of serially diluted trastuzumab-based Fc-inactive CD2 bispecifics or control bispecific molecules.

[0044] Figures 23A and B show non-linear regression representations of CD25 upregulation data. Concentration responses to increasing signal 2 doses in the presence of a fixed signal 1 dose are shown. In Figure 23 A, primary human T cells were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+). In Figure 23B, primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+and HER2+). All co-cultures were done at an effector-to-target cell ratio of 2: 1 and were treated with a fixed concentration of 5 pM EpCAM*CD3 BiTE in the presence of serially diluted trastuzumab-based Fc-inactive CD2 bispecifics or control bispecific molecules. After 48 hours, T cell activation was determined by staining for CD25 and samples were counterstained for CD4+and CD8+T cell subsets.

[0045] Figure 24 is a schematic drawing of an exemplary format for a tumor-targeted CD2 bispecific designed to provide a co-stimulatory signal to NK cells. Figure 25 shows non-linear regression representations of flow cytometry data showing binding of serially diluted bispecific antibodies or parental Trastuzumab antibodyto primary human CD8+T cells, primary human NK cells and CD2-transfected HEK293 cells.

[0046] Figure 26 shows non-linear regression representations of flow cytometry data showing binding of serially diluted bispecific antibodies or parental Trastuzumab antibody to HCC1954 target cells.

[0047] Figure 27 shows non-linear regression representations of flow cytometry data showing conjugate formation between NK cells and HCC1954 target cells or CD8+T cells and HCC1954 target cells.

[0048] Figure 28 is a non-linear regression representation of FACS-based concentration response analysis data showing the impact of Fc-competent CD2 bispecifics on degranulation of CD16+NK-92 in the absence of target cells as a read-out for NK cell fratricide.

[0049] Figure 29 is a non-linear regression representation of NK cell degranulation data in the presence of HCC1954 target cells. CD16+NK-92 cells were co-cultured with CTFR-prelabeled HER2+HCC1954 tumor cells in the presence of Fc-competent Trastuzumab-based CD2 bispecifics or control antibodies.

[0050] Figure 30 is a non-linear regression representation of NK cell-mediated anti-tumor cytotoxicity data. CD 16+NK-92 cells were co-cultured with CTFR-prelabeled HER2+HCC1954 tumor cells in the presence of Fc-competent Trastuzumab-based CD2 bispecifics or control antibodies.

[0051] Figure 31 is a non-linear regression representation of NK cell secretion activity (IFN-gamma release) following co-culture with tumor cells. CD16+NK-92 cells were co-cultured with CTFR-prelabeled HER2+HCC1954 tumor cells in the presence of Fc-competent Trastuzumabbased CD2 bispecifics or control antibodies.

[0052] Figure 32A is a bar graph representation of NK cell degranulation data comparing NK-92 cells with and without CD 16 expression. CD 16+ or CD 16negNK-92 cells were co-cultured with CTFR-prelabelled HER2+HCC1954 tumor cells in the presence of Tra-H-monovalent-16-92- 4_KiH and control molecules.

[0053] Figure 32B is a bar graph representation of NK cell-mediated anti-tumor cytotoxicity data comparing NK-92 cells with and without CD16 expression. CD16+ or CD16negNK-92 cells were co-cultured with CTFR-prelabeled HER2+HCC1954 tumor cells in the presence of Tra- H-monovalent-16-92-4_KiH and control molecules.

[0054] Figure 32C is a bar graph representation of IFN-gamma release comparing the impact of Tra- H-monovalent-16-92-4_KiH and control molecules on cytokine secretion by NK-92 cells with and without CD 16 after 4 hours of co-culture with target cells.

[0055] Figure 33 shows non-linear regression representations of concentration response analysis data showing the impact of Fc-competent CD2 bispecifics, fucosylated and afucosylated, on degranulation and fFNy release of CD16+NK-92 cells over a 4h culture period in the absence of tumor targets.

[0056] Figure 34 shows non-linear regression representations of concentration response analysis data showing the impact of Fc-competent CD2 bispecifics, fucosylated and afucosylated, on degranulation, tumor cell killing and fFNy release in a co-culture of CD16+NK-92 cells with HCC1954 tumor cells over a 4h culture period.

[0057] Figure 35 shows non-linear regression representation of anti -tumor cytotoxicity mediated by Fc-competent CD2 bispecifics, fucosylated and afucosylated, in a co-culture of primary human PBMCs with CTFR-prelabeled HER2+HCC1954 tumor cells.

[0058] Figure 36 shows an embodiment of the multifunctional construct according to the invention. The construct supports engagement of an NK cell with a tumor target cell. “1” designates signal 1, and “2” designates signal 2.

[0059] Figure 37 shows another embodiment of the multifunctional construct according to the invention. The construct supports engagement of a T cell with a tumor target cell. “1” designates signal 1, and “2” designates signal 2. Figure 38 shows examples of multifunctional constructs according to the invention, with and without Fc domain. The following table explains the different formats:

[0060] Table 1: examples for multifunctional constructs

[0061] Figure 39 shows flow cytometry-based analysis of antigen expression. Wildtype HCC1954 cells and cells of a CD58-knockout HCC1954 pool generated by standard CRISPR-Cas9 methodology were stained with commercially available fluorochrome-labeled antibodies against CD58, HER2 and EpCAM and data collected by flow cytometry.

[0062] Figure 40 is a non-linear regression representation of anti-tumor cytotoxicity data. Macrophage- depleted human PBMCs were co-cultured with either wildtype or CD58-KO HCC1954 breast cancer cells (EpCAM+and HER2+) that were stably transfected with the far-red fluorescent protein mKate2, at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM* CD3 BiTE. Anti -turn or cytotoxicity was monitored over a 4-day period using the Incucyte system.

[0063] Figure 41 shows is a non-linear regression representation of anti -tumor cytotoxicity data. Macrophage-depleted human PBMCs were co-cultured with the CD58-KO HCC1954 breast cancer cells (EpCAM+and HER2+) that were stably transfected with the far-red fluorescent protein mKate2, at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of trastuzumab-based Fc-inactive CD2 bispecifics or control bispecific molecules at a fixed concentration of 50 nM. Anti-tumor cytotoxicity was monitored over a 4-day period using the Incucyte system. Figure 42 shows non-linear regression representation of anti-tumor cytotoxicity data. Macrophage-depleted human PBMCs were co-cultured at an effector-to-target cell ratio of 2: 1 with the Ramos B cell line (CD19+, CD20+and CD22+) that were stably transfected with the far-red fluorescent protein mKate2. Co-cultures were treated with serially diluted Mosunetuzumab (CD20*CD3 T cell engager) in the presence of Epratuzumab-based (anti- CD22) or FMC63-based (anti-CD19) Fc-inactive CD2 bispecifics or control bispecific molecules at a fixed concentration of 50 nM. Anti-tumor cytotoxicity was monitored over a 4- day period using the Incucyte system.

[0064] Figure 43 shows is a schematic drawing of an exemplary alternative format for a tumor-targeted CD2 bispecific designed to provide a costimulatory signal to effector T cells. Here, bivalent binding to CD2 is mediated by the Fab domain of the CD2-targeted IgG. Disulfide stabilized single chain Fv moieties directed against a tumor associated antigen are genetically fused to the C-termini of both heavy chains providing tumor-targeting. The asterisks represent mutations that render the Fc moiety silent to prevent CD 16a engagement.

[0065] Figure 44 shows non-linear regression representations of anti -tumor cytotoxicity data and CD25 upregulation data. Primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of trastuzumab-based Fc-inactive CD2 bispecifics or control bispecific molecules at a fixed concentration of 10 nM. After a 2-day culture, anti-tumor cytotoxicity was determined by LDH release assay.

[0066] Figure 45 are non-linear regression representations of T cell activation data visualized by CD25 upregulation. Primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of trastuzumab-based Fc-inactive CD2 bispecifics or control bispecific molecules at a fixed concentration of 10 nM. After a 2-day culture, T cells activation was determined by staining for CD25 and counterstaining for CD4 and CD8 T cell subsets.

[0067] DETAILED DESCRIPTION OF EMBODIMENTS Before the invention is described in detail, it is to be understood that this invention is not limited to the particular component parts of the devices described or process steps of the methods described, as such devices and methods may vary. It is also to be understood that the terminology used herein is for purposes of describing particular embodiments only and is not intended to be limiting. It must be noted that, as used in the specification and the appended claims, the singular forms "a", "an", and "the" include singular and / or plural referents unless the context clearly dictates otherwise. It is moreover to be understood that, in case parameter ranges are given which are delimited by numeric values, the ranges are deemed to include these limitation values.

[0068] It is further to be understood that embodiments disclosed herein are not meant to be understood as individual embodiments which would not relate to one another. Features discussed with one embodiment are meant to be disclosed also in connection with other embodiments shown herein. If, in one case, a specific feature is not disclosed with one embodiment, but with another, the skilled person would understand that does not necessarily mean that said feature is not meant to be disclosed with said other embodiment. The skilled person would understand that it is the gist of this application to disclose said feature also for the other embodiment, but that just for purposes of clarity and to keep the specification in a manageable volume this has not been done.

[0069] Furthermore, the content of the prior art documents referred to herein is incorporated by reference. This refers, particularly, for prior art documents that disclose standard or routine methods. In that case, the incorporation by reference has mainly the purpose to provide sufficient enabling disclosure and to avoid lengthy repetitions.

[0070] According to a first aspect of the invention, an antibody is provided which binds to CD2, or a target-binding fragment or derivative of such antibody. The antibody a) comprises a set of three heavy chain and three light chain complementarity determining regions (CDR) comprised in the one of the following heavy chain / light variable domain sequence pairs

[0071] SEQ ID NOs 1 and 2;

[0072] SEQ ID NOs 3 and 4;

[0073] SEQ ID NOs 5 and 6; and / or SEQ ID NOs 7 and 8; b) comprises a set of three heavy chain and three light chain complementarity determining regions (CDR) selected from the following sets

[0074] • SEQ ID NOs 9 - 14;

[0075] • SEQ ID NOs 15 - 20;

[0076] • SEQ ID NOs 21 - 26; and / or

[0077] • SEQ ID NOs 27 - 32; c) comprises the set of heavy chain / light chain complementarity determining regions (CDR) of b), with the proviso that at least one of the CDRs has up to 3 amino acid substitutions relative to the respective SEQ ID NOs, and / or d) comprises the set of heavy chain / light chain complementarity determining regions (CDR) of b) or c), with the proviso that at least one of the CDRs has a sequence identity of > 66 % to the respective CDRs comprised in the SEQ ID NOs, wherein the CDRs are embedded in a suitable protein framework, preferably a variable domain framework, so as to be capable to bind to CD2.

[0078] CD2 (cluster of differentiation 2) (GenelD: 914; UniProt identifier: P06729) is a cell adhesion molecule found on the surface of T cells and natural killer (NK) cells. It has also been called T- cell surface antigen Tl l / Leu-5, LFA-2, LFA-3 receptor, erythrocyte receptor and rosette receptor. It interacts with other adhesion molecules, such as lymphocyte function-associated antigen-3 (LFA-3 / CD58), which are expressed on the surfaces of somatic cells, including cancer cells. In addition to its adhesive properties, CD2 also acts as a co-stimulatory molecule on T and NK cells, i.e., when bound by a target cells’ CD58, CD2 releases a costimulatory signal (“signal 2”) that complements a potential signal 1, evoked for example a) in a T cell for example by (i) TCR interaction with pMHC, or by virtue of a CD3 engaging bi- or multispecific antibody that creates an artificial immunological synapse between a cancer cell and a T cell, b) in an NK cell by interaction between the Fey domain of an antibody that has bound to a cancer antigen on a cancer cell and the FcyRIII / CD I 6 protein of an NK cell c) in a CAR T or CAR NK cell by interaction of the chimeric antigen receptor of the CAR T or CAR NK cell with a cancer antigen on a cancer cell

[0079] In one embodiment, the CDRs are determined according to the definition of Kabat, Chothia or MacCallum, preferably wherein the CDRs are determined according to the numbering set forth in Table 2.

[0080] Methods for the production and / or selection of humanized mAbs are known in the art. For example, US6331415 by Genentech describes the production of chimeric antibodies, while US6548640 by Medical Research Council describes CDR grafting techniques and US5859205 by Celltech describes the production of humanized antibodies.

[0081] Humanized antibodies are antibodies in which the complementarity determining regions stem from a parent antibody taken from a non-human species and are grafted into the framework (at least the variable domain) of a human antibody, like e.g. of an IgGl, IgG2 or IgG4. The humanized antibody binds the same target as the parent antibody, but, due to its grafting into a human framework, has reduced immunogenicity (like e.g. HAMA response). For this reason, a humanized antibody is structurally different from its parent (e.g. murine) antibody.

[0082] In humanization, the step of grafting the CDRs into a human framework is often followed by a step of affinity maturation, to reacquire affinity that was lost in the grafting process. This process further modifies the sequence of the human antibody, including its CDRs.

[0083] As used herein, the term “CDR” or “complementarity determining region” is intended to mean the non-contiguous antigen combining sites found within the variable region of both heavy and light chain polypeptides. These particular regions have been described by Kabat et al. (1977), Chothia et al. (1987) and MacCallum et al., (1996) where the definitions include overlapping or subsets of amino acid residues when compared against each other. Nevertheless, application of either definition to refer to a CDR of an antibody or grafted antibodies or variants thereof is intended to be within the scope of the term as defined and used herein. The amino acid residues which encompass the CDRs as defined by each of the above cited references are set forth below in Table 2 as a comparison. Table 2: CDR definitions

[0084] As used herein, the term “framework” when used in reference to an antibody variable domain is entered to mean all amino acid residues outside the CDR regions within the variable domain of an antibody. Therefore, a variable domain framework is between about 100-120 amino acids in length but is intended to reference only those amino acids outside of the CDRs.

[0085] It needs to be mentioned that the term “domain”, as used herein, can refer to subunits of a binding entity (like, e.g., VH- domain and VL domain, both forming a target binding entity), as well as to the entity that comprises two or more of such subunits. In that sense, a scFv fragment, for example, comprising a VH- domain and VL domain, is also called a “domain” or “binding domain” herein.

[0086] As used herein, the term “capable to bind to target X” has to be understood as meaning that respective binding domain binds the target with a KD of 10'4or smaller. KD is the equilibrium dissociation constant, a ratio of koff / kon, between the antibody or fragment and its antigen. KD and affinity are inversely related. The KD value relates to the concentration of antibody or fragment (the amount of antibody or fragment needed for a particular experiment) and so the lower the KD value (lower concentration) and thus the higher the affinity of the binding domain. The following table 3 shows typical KD ranges of monoclonal antibodies

[0087] Table 3: KDand Molar Values

[0088] Preferably, the antibody or fragment has up to 2 amino acid substitutions, and more preferably up to 1 amino acid substitution. Preferably, at least one of the CDRs of the antibody or fragment has a sequence identity of > 67 %; > 68 %; > 69 %; > 70 %; > 71 %; > 72 %; > 73 %; > 74 %; > 75 %; > 76 %; > 77 %; > 78 %; > 79 %; > 80 %; > 81 %; > 82 %; > 83 %; > 84 %; > 85 %; > 86 %; > 87 %; > 88 %; > 89 %; > 90 %; > 91 %; > 92 %; > 93 %; > 94 %; > 95 %; > 96 %; > 97 %; > 98 %; > 99 %, and most preferably 100 % to the respective SEQ ID NO.

[0089] “Percentage of sequence identity” as used herein, is determined by comparing two optimally aligned biosequences (amino acid sequences or polynucleotide sequences) over a comparison window, wherein the portion of the corresponding sequence in the comparison window may comprise additions or deletions (z.e., gaps) as compared to the reference sequence, which does not comprise additions or deletions, for optimal alignment of the two sequences. The percentage is calculated by determining the number of positions at which the identical nucleic acid base or amino acid residue occurs in both sequences to yield the number of matched positions, dividing the number of matched positions by the total number of positions in the window of comparison and multiplying the result by 100 to yield the percentage of sequence identity.

[0090] The terms “identical” or percent “identity,” in the context of two or more nucleic acids or polypeptide sequences, refer to two or more sequences or subsequences that are the same sequences. Two sequences are “substantially identical” if two sequences have a specified percentage of amino acid residues or nucleotides that are the same (z.e., at least 85%, 90%, 95%, 96%, 97%, 98% or 99% sequence identity over a specified region, or, when not specified, over the entire sequence of a reference sequence), when compared and aligned for maximum correspondence over a comparison window, or designated region as measured using one of the following sequence comparison algorithms or by manual alignment and visual inspection. The disclosure provides polypeptides that are substantially identical to the polypeptides exemplified herein. With respect to amino acid sequences, identity or substantial identity can exist over a region that is at least 5, 10, 15 or 20 amino acids in length, optionally at least about 25, 30, 35, 40, 50, 75 or 100 amino acids in length, optionally at least about 150, 200 or 250 amino acids in length, or over the full length of the reference sequence. With respect to shorter amino acid sequences, e.g., amino acid sequences of 20 or fewer amino acids, substantial identity exists when one or two amino acid residues are conservatively substituted, according to the conservative substitutions defined herein. Preferably, at least one of the CDRs has been subject to CDR sequence modification, including

[0091] • affinity maturation, and / or

[0092] • reduction of immunogenicity and / or

[0093] • improvement of manufacturability.

[0094] Affinity maturation is the process by which the affinity of a given antibody is increased in vitro. Like the natural counterpart, in vitro affinity maturation is based on the principles of mutation and selection. It has successfully been used to optimize antibodies, antibody fragments or other peptide molecules like antibody mimetics. Mutations inside the CDRs are introduced. In addition, the genetic diversity can be increased by chain shuffling. Two or three rounds of mutation and selection using display methods like phage display usually results in antibody fragments with affinities in the low nanomolar range. For principles see Eylenstein et al. (2016) or US20050169925A1, the content of which is incorporated herein by reference for enablement purposes.

[0095] Engineered antibodies contain murine-sequence derived CDR regions that have been engrafted, along with any necessary framework back-mutations, into sequence-derived V regions. Hence, the CDRs themselves can cause immunogenic reactions when the humanized antibody is administered to a patient. Methods of reducing immunogenicity caused by CDRs are disclosed in Harding et al. (2010), or US2014227251A1, the content of which is incorporated herein by reference for enablement purposes.

[0096] According to an embodiment of the invention, the CDRs are determined according to the definition of Kabat, Chothia or MacCallum. Preferably the CDRs are determined according to the numbering set forth in Table 2.

[0097] According to embodiments of the invention, the antibody or fragment comprises a) the heavy chain / light chain variable domain (HCVD / LCVD) pairs set forth in the following pairs of SEQ ID NOs:

[0098] • SEQ ID NOs 1 and 2; SEQ ID NOs 3 and 4;

[0099] SEQ ID NOs 5 and 6; and / or

[0100] SEQ ID NOs 7 and 8; b) the heavy chain / light chain variable domains (HCVD / LCVD) pairs of a), with the proviso that

[0101] • the HCVD has a sequence identity of > 80 % to the respective SEQ ID NO, and / or

[0102] • the LCVD has a sequence identity of > 80 % to the respective SEQ ID NO, c) the heavy chain / light chain variable domains (VD) pairs of a) or b), with the proviso that at least one of the HCVD or LCVD has up to 10 amino acid substitutions relative to the respective SEQ ID NO, said antibody or fragment still being capable to bind to CD2.

[0103] The antibody characterized by the VH / VL domains as presented by SEQ ID NOs: 1 and 2 (and the CDRs of SEQ ID NOs: 9 - 14) is called 16-23-7 herein.

[0104] The antibody characterized by the VH / VL domains as presented by SEQ ID NOs: 3 and 4 (and the CDRs of SEQ ID NOs: 15 - 20) is called 16-92-4 herein.

[0105] The antibody characterized by the VH / VL domains as presented by SEQ ID NOs: 5 and 6 (and the CDRs of SEQ ID NOs: 21 - 26) is called 16-92-4_h herein. 16-92-4_h is the humanized variant of 16-92-4.

[0106] The antibody characterized by the VH / VL domains as presented by SEQ ID NOs: 7 and 8 (and the CDRs of SEQ ID NOs: 27 - 32) is called 16-92-4_ho herein. 16-92-4_ho is a modified variant of 16-92-4_h, with two substitutions (Q75K in the VH domain (CDR2) and DIE in the VL domain. These two substitutions comprise reversion to the residues that were in the original mouse-derived sequence of 16-92-4 but were changed in the process of humanization. However, following structural modelling of the humanized variant 16-92-4_h, the position Q75 in VH and DI in the VL were identified as positions that could impact the binding of the humanized variant 16-92-4_h. Hence these positions were changed to original mouse-derived 16-92-4 sequences giving rise of the version 16-92-4_ho.

[0107] A “variable domain” when used in reference to an antibody or a heavy or light chain thereof is intended to mean the portion of an antibody which confers antigen binding onto the molecule and which is not the constant region. The term is intended to include functional fragments thereof which maintain some or all of the binding function of the whole variable region. Variable region binding fragments include, for example, functional fragments such as Fab, F(ab)2, Fv, single chain Fv (scfv) and the like. Such functional fragments are well known to those skilled in the art. Accordingly, the use of these terms in describing functional fragments of a heteromeric variable region is intended to correspond to the definitions well known to those skilled in the art. Such terms are described in, for example, Huston et al., (1993) or Pliickthun and Skerra (1990).

[0108] Preferably, the HCVD and / or LCVD has a sequence identity of > 81 %; > 82 %; > 83 %; > 84 %; > 85 %; > 86 %; > 87 %; > 88 %; > 89 %; > 90 %; > 91 %; > 92 %; > 93 %; > 94 %; > 95 %; > 96 %; > 97 %; > 98 %; > 99 %; or most preferably 100 % to the respective SEQ ID NO.

[0109] According to embodiments of the invention, wherein at least one amino acid substitution as specified above is a conservative amino acid substitution.

[0110] A “conservative amino acid substitution”, as used herein, has a smaller effect on antibody function than a non-conservative substitution. Although there are many ways to classify amino acids, they are often sorted into six main groups on the basis of their structure and the general chemical characteristics of their R groups.

[0111] In some embodiments, a “conservative amino acid substitution” is one in which the amino acid residue is replaced with an amino acid residue having a similar side chain. For example, families of amino acid residues having similar side chains have been defined in the art. These families include amino acids with basic side chains (e.g., lysine, arginine, histidine), acidic side chains (e.g., aspartic acid, glutamic acid), • uncharged polar side chains (e.g., glycine, asparagine, glutamine, serine, threonine, tyrosine, cysteine),

[0112] • nonpolar side chains (e.g., alanine, valine, leucine, isoleucine, proline, phenylalanine, methionine, tryptophan),

[0113] • beta-branched side chains (e.g., threonine, valine, isoleucine) and

[0114] • aromatic side chains (e.g., tyrosine, phenylalanine, tryptophan, histidine).

[0115] Other conserved amino acid substitutions can also occur across amino acid side chain families, such as when substituting an asparagine for aspartic acid in order to modify the charge of a peptide. Conservative changes can further include substitution of chemically homologous nonnatural amino acids (i.e. a synthetic non-natural hydrophobic amino acid in place of leucine, a synthetic non-natural aromatic amino acid in place of tryptophan).

[0116] According to another aspect of the invention, an antibody is provided which binds to CD2, or a target-binding fragment or derivative of such antibody, wherein the antibody or fragment or derivative has a binding affinity of > 50% to CD2 compared to that of the antibody or fragment according to the above description.

[0117] As used herein the term “binding affinity” is intended to mean the strength of a binding interaction and therefore includes both the actual binding affinity as well as the apparent binding affinity. The actual binding affinity is a ratio of the association rate over the disassociation rate. Therefore, conferring or optimizing binding affinity includes altering either or both of these components to achieve the desired level of binding affinity. The apparent affinity can include, for example, the avidity of the interaction. For example, a bivalent heteromeric variable region binding fragment can exhibit altered or optimized binding affinity due to its valency.

[0118] A suitable method for measuring the affinity of a binding agent is through surface plasmon resonance (SPR). This method is based on the phenomenon which occurs when surface plasmon waves are excited at a metal / liquid interface. Light is directed at, and reflected from, the side of the surface not in contact with sample, and SPR causes a reduction in the reflected light intensity at a specific combination of angle and wavelength. Biomolecular binding events cause changes in the refractive index at the surface layer, which are detected as changes in the SPR signal. The binding event can be either binding association or disassociation between a receptorligand pair. The changes in refractive index can be measured essentially instantaneously and therefore allows for determination of the individual components of an affinity constant. More specifically, the method enables accurate measurements of association rates (kon) and disassociation rates (koir).

[0119] Measurements of konand kotr values can be advantageous because they can identify altered variable regions or optimized variable regions that are therapeutically more efficacious. For example, an altered variable region, or heteromeric binding fragment thereof, can be more efficacious because it has, for example, a higher konvalue compared to variable regions and heteromeric binding fragments that exhibit similar binding affinity. Increased efficacy is conferred because molecules with higher konvalues can specifically bind and inhibit their target at a faster rate. Similarly, a molecule of the invention can be more efficacious because it exhibits a lower koff value compared to molecules having similar binding affinity. Increased efficacy observed with molecules having lower koir rates can be observed because, once bound, the molecules are slower to dissociate from their target. Although described with reference to the altered variable regions and optimized variable regions of the invention including, heteromeric variable region binding fragments thereof, the methods described above for measuring associating and disassociation rates are applicable to essentially any antibody or fragment or fragment thereof for identifying more effective binders for therapeutic or diagnostic purposes.

[0120] Another suitable method for measuring the affinity of a binding agent is by FACS / scatchard analysis. See inter alia example 1 for a respective description.

[0121] Methods for measuring the affinity, including association and disassociation rates using surface plasmon resonance are well known in the arts and can be found described in, for example, Jonsson and Malmquist, (1992) and Wu et al. (1998). Moreover, one apparatus well known in the art for measuring binding interactions is a BIAcore 2000 instrument which is commercially available through Pharmacia Biosensor, (Uppsala, Sweden).

[0122] Preferably said target binding affinity is > 51%, > 52%, > 53%, > 54%, > 55%, > 56%, > 57%,

[0123] > 58%, > 59%, > 60%, > 61%, > 62%, > 63%, > 64%, > 65%, > 66%, > 67%, > 68%, > 69%,

[0124] > 70%, > 71%, > 72%, > 73%, > 74%, > 75%, > 76%, > 77%, > 78%, > 79%, > 80%, > 81%,

[0125] > 82%, > 83%, > 84%, > 85%, > 86%, > 87%, > 88%, > 89%, > 90%, > 91%, > 92%, > 93%,

[0126] > 94%, > 95%, > 96%, > 97%, > 98%, and most preferably > 99 % compared to that of the reference binding agent. According to another aspect of the invention, an antibody is provided which binds to CD2, or a target-binding fragment or derivative of such antibody, which competes for binding to CD2 with a) an antibody according to the above description, or b) an antibody selected from clones 16-92-4, 16-92-4_h or 16-92-4_ho.

[0127] According to another aspect of the invention, an antibody or a target-binding fragment or derivative of such antibody is provided that binds to essentially the same, or the same, region or epitope on CD2 as a) an antibody according to the above description, or b) an antibody selected from clones 16-92-4, 16-92-4_h or 16-92-4_ho.

[0128] Clones 16-92-4, 16-92-4_h or 16-92-4_ho are identified in the sequence table herein.

[0129] As used herein, the term “region” shall be understood to mean an extracellular region, a domain, a subdomain, or a secondary structure (e.g. loop), or preferably an epitope.

[0130] As regards the format or structure of such antibody or fragment, the same preferred embodiments as set forth above apply. In one embodiment, said antibody or fragment is a monoclonal antibody, or a target-binding fragment or derivative thereof retaining target binding capacities, or an antibody mimetic.

[0131] As used herein, the term "competes for binding" is used in reference to one of the antibodies defined by the sequences as above, meaning that the actual antibody or fragment has an activity which binds to the same target, or target epitope or domain or subdomain, as does said sequence defined antibody or fragment, and is a variant of the latter. The efficiency (e.g., kinetics or thermodynamics) of binding may be the same as or greater than or less than the efficiency of the latter. For example, the equilibrium binding constant for binding to the substrate may be different for the two antibodies.

[0132] Such competition for binding can be suitably measured with a competitive binding assay. Such assays are disclosed in Finco et al. 2011, the content of which is incorporated herein by reference for enablement purposes, and their meaning for interpretation of a patent claim is 1 disclosed in Deng et al 2018, the content of which is incorporated herein by reference for enablement purposes.

[0133] In order to test for this characteristic, suitable epitope mapping technologies are available, including, inter alia,

[0134] • X-ray co-crystallography and cryogenic electron microscopy (cryo-EM)

[0135] • Array -based oligo-peptide scanning

[0136] • Site-directed mutagenesis mapping

[0137] • High-throughput shotgun mutagenesis epitope mapping

[0138] • Hydrogen-deuterium exchange

[0139] • Cross-linking-coupled mass spectrometry

[0140] These methods are, inter alia, disclosed and discussed in Banik et al (2010), and DeLisser (1999), the content of which is herein incorporated by reference for enablement purposes.

[0141] According to one embodiment of the invention, the antibody or fragment is a monoclonal antibody, or a target-binding fragment or derivative thereof retaining target binding capacities.

[0142] According to one embodiment of the invention, the antibody or fragment according is in at least one of the formats selected from the group consisting of: IgG, scFv, Fab, or (Fab)2.

[0143] As used herein, the term “monoclonal antibody (mAb)” shall refer to an antibody composition having a homogenous antibody population, i.e., a homogeneous population consisting of a whole immunoglobulin, or a fragment or derivative thereof retaining target binding capacities.

[0144] Particularly preferred, such antibody is an IgG antibody, or a fragment or derivative thereof retaining target binding capacities. Immunoglobulin G (IgG) is a type of antibody. Representing approximately 75% of serum antibodies in humans, IgG is the most common type of antibody found in blood circulation. IgG molecules are created and released by plasma B cells. Each IgG has two antigen binding sites.

[0145] IgG antibodies are large molecules with a molecular weight of about 150 kDa made of four peptide chains. It contains two identical class y heavy chains of about 50 kDa and two identical light chains of about 25 kDa, thus a tetrameric quaternary structure. The two heavy chains are linked to each other and to a light chain each by disulfide bonds. The resulting tetramer has two identical halves, which together form the Y-like shape. Each end of the fork contains an identical antigen binding site. The Fc regions of IgGs bear a highly conserved N-glycosylation site. The N-glycans attached to this site are predominantly core-fucosylated diantennary structures of the complex type. In addition, small amounts of these N-glycans also bear bisecting GlcNAc and a-2,6-linked sialic acid residues.

[0146] There are four IgG subclasses (IgGl, 2, 3, and 4) in humans, named in order of their abundance in serum (IgGl being the most abundant).

[0147] As used herein, the term “fragment” shall refer to fragments of such antibody retaining target binding capacities, e.g.

[0148] • a CDR (complementarity determining region)

[0149] • a hypervariable region,

[0150] • a variable domain (Fv)

[0151] • an IgG or IgM heavy chain (consisting of VH, CHI, hinge, CH2 and CH3 regions)

[0152] • an IgG or IgM light chain (consisting of VL and CL regions), and / or

[0153] • a Fab and / or F(ab)2.

[0154] As used herein, the term “derivative” shall refer to protein constructs being structurally different from, but still having some structural relationship to, the common antibody concept, e.g., scFv, Fab and / or F(ab)2, as well as bi-, tri- or higher specific antibody constructs, and further retaining target binding capacities. All these items are explained below.

[0155] Other antibody derivatives known to the skilled person are Diabodies, Camelid Antibodies, Nanobodies, Domain Antibodies, bivalent homodimers with two chains consisting of scFvs, IgAs (two IgG structures joined by a J chain and a secretory component), shark antibodies, antibodies consisting of new world primate framework plus non-new world primate CDR, dimerized constructs comprising CH3+VL+VH, and antibody conjugates (e.g. antibody or fragments or derivatives linked to a toxin, a cytokine, a radioisotope or a label). These types are well described in the literature and can be used by the skilled person on the basis of the present disclosure, without adding further inventive activity. Methods for the production of a hybridoma cell are disclosed in Kohler & Milstein (1975).

[0156] Methods for the production and / or selection of fully human mAbs are known in the art. These can involve the use of a transgenic animal which is immunized with the respective protein or peptide, or the use of a suitable display technique, like yeast display, phage display, B-cell display or ribosome display, where antibodies from a library are screened against CD2 in a stationary phase.

[0157] In vitro antibody libraries are, among others, disclosed in US6300064 by MorphoSys and US6248516 by MRC / Scripps / Stratagene. Phage Display techniques are for example disclosed in US5223409 by Dyax. Transgenic mammal platforms are for example described in EP1480515A2 by Taconic Artemis.

[0158] IgG, IgM, scFv, Fab and / or F(ab)2 are antibody formats well known to the skilled person. Related enabling techniques are available from the respective textbooks.

[0159] As used herein, the term “Fab” relates to an IgG / IgM fragment comprising the antigen binding region, said fragment being composed of one constant and one variable domain from each heavy and light chain of the antibody.

[0160] As used herein, the term “F(ab)2” relates to an IgG / IgM fragment consisting of two Fab fragments connected to one another by disulfide bonds.

[0161] As used herein, the term “scFv” relates to a single-chain variable fragment being a fusion of the variable regions of the heavy and light chains of immunoglobulins, linked together with a short linker, usually serine (S) or glycine (G). This chimeric molecule retains the specificity of the original immunoglobulin, despite removal of the constant regions and the introduction of a linker peptide.

[0162] Modified antibody formats are for example bi- or trispecific antibody constructs, antibodybased fusion proteins, immunoconjugates and the like. These types are well described in the literature and can be used by the skilled person on the basis of the present disclosure, with adding further inventive activity. In this context, it is surprising that the antibody 16-23-7, when bound to CD2, blocks CD58 binding to CD2 expressed by CD8+T cells (ICso= 0.448 nM)

[0163] Contrary thereto, the antibody 16-92-4 has the opposite effect and increases, stabilizes and / or enhances CD2-mediated binding of CD58 to human CD8+T cells (ECso= 1.173 nM).

[0164] Without being bound to theory, 16-23-7 appears to bind to an epitope on CD2 that seems to overlap with the binding site for CD58, and / or binding of 16-23-7 sterically impedes the interaction of CD2 with CD58. Instead, 16-92-4 seems to bind to a different epitope on CD2 that does not overlap with the binding site for CD58. CD2 antibody 16-92-4 seems to stabilize a conformation of CD2 that facilitates or even favors interaction with CD58.

[0165] As will be discussed herein, the two antibodies belong to different epitope bins and bind to separate sub-domains of the human CD2 molecule. Their different binding epitopes and binding orientations could represent the molecular basis for the observed differences in mediating effective costimulation in the T cell engager format, and in avoiding fratricide in the NK cell engager format.

[0166] According to another aspect of the invention a bi- or multifunctional molecule is provided, which comprises at least: a) a first binding domain capable of binding to the CD2 antigen, or to a subdomain of CD2, and b) a second binding domain capable of specifically binding to a target antigen other than the CD2 antigen.

[0167] As used herein, the terms “bifunctional” and “bispecific” are being used synonymously. Still, a multifunctional molecule can be bispecific or bifunctional, while the use of the terms “bifunctional” and “bispecific” does not exclude that the respective molecule has further functionalities, thus rendering it trispecific / trifunctional or even higher.

[0168] As discussed above, CD2 is a costimulatory protein expressed on T cells and NK cells. It binds to LFA3, also known as CD58, present on B cells, T cells, monocytes and granulocytes. Contrary to other co-stimulatory receptors, CD2 is highly expressed in tumor-infiltrating lymphocytes. Co-stimulation via CD2 may hence offer an approach to address the problem that CD3 T cell engaging molecules that only provide a “signal 1” to T cells have so far only demonstrated limited success in the treatment of solid tumors due to their strong immune suppressive microenvironment.

[0169] Further, while almost all CD28-positive immune cells are also CD2-positive, CD28-negative immune cells exist (e.g., a substantial proportion of CD8+ T cells) which are CD2-positive nonetheless. It has also been shown that the proportion of CD8+ T cells which are CD2-positive yet CD28 -negative, increases with the aging of the immune system (Leitner et al 2015).

[0170] As a result, the inventors of the present invention have surprisingly found that CD2 engagement provides a universal co-activatory signal for bispecific immune cell engagers. Hence, recruiting CD2 expressing tumor infiltrating lymphocytes provides an option to engage and activate T and NK immune effector cells in the tumor microenvironment.

[0171] Further, CD2 engagement can serve as co-stimulating signal even in CD28-negative CD8+T cells. This is important because CD28 expression in human T cells decreases with age and following T cell activation (Eck et al 1997).

[0172] In adaptive NK cells CD2 provides co-stimulation to the Fc / CD16 response, whereas CD2 blockade caused e.g. by antibodies is inhibitory forNK cells.

[0173] According to one embodiment of the bi- or multifunctional molecule, the second binding domain binds to a cancer-associated target on a cancer target cell surface or to an immune checkpoint agonist on a cancer cell.

[0174] According to embodiments of the invention, such “cancer-associated target” is a receptor, a protein, a protein complex or another macromolecular structure that is present on the surface of a cancer cell. In the context of the present invention, the term “tumor associated antigen (TAA)” is used synonymously therewith. Such “cancer associated antigen” may be characteristic for a cancer cell, for example because it is a) exclusively presented on cancer cells, yet not on healthy cells, b) presented in higher abundancy or surface density on cancer cells than on healthy cells c) presented on cancer cells, yet not on post-embryonic healthy cells d) presented on cancer cells and healthy cells of reproductive organs (like testis, ovary and trophoblasts), yet not on healthy somatic cells (“cancer-testis antigens”) e) presented on B lymphocytes

[0175] In some embodiments, such “cancer-associated antigen” is a complex of a major histocompatibility complex (MHC) and a peptide presented by the latter. Such complex is also abbreviated as pMHC. Preferably, such presented peptide is a tumor associated peptide (TUMAP), or a neoantigen.

[0176] In some embodiments, such “cancer-associated antigen” is selected from the group consisting of AXL, BCMA, CA9, CD5, CD7, CD10, CD19, CD20, CD22, CD30, CD33, CD37, CD38, CD44, CD52, CD56, CD72, CD70, CD79a (mb-1), CD79b (B29), CD117 (also known as c- kit), CD99, CD123 (also known as IL-3R), CD133, CD135 (also known as flt3), CD138, CD147, CD174, CD276, CLEC12A, CSPG4, c-Met, CEA, EBV-related antigens, GPC3, GUCY2C, EPCAM, EPHA2, EGFR, EGFRvIII, DLL3, DLL4, FAP, FOLH1, folate receptor alpha, GD2, GPC3, GPRC5D, IL1RAP, HER2, HER3, HER4, HPV-related antigens, KDR, L1CAM, LILRA3, LILRB4, MAGE antigens, Nectin-4, Mesothelin, MS4A1, MUC-1, NCAM- 1, NY-ESO-1, PDCD1, PSCA, PSMA, PTK7, SDC1, SLAMF7 (also known as CS1), SIRP- alpha, SSTR2, TEM1, Tissue Factor, TNFRSF8, TNFRSF17, ULBP1, ULBP2, VEGFR2, WT1, Claudin 1, Claudin 6, Claudin 18.2, ROR-1 or R0R2. This list is not limiting though.

[0177] According to another embodiment, such surface antigen on a target cell is an antigen associated with a non-cancer disease, including, but not limited to, autoimmune diseases, for instance, but not limited to systemic lupus erythematosus (SLE) caused by overproduction of autoantibodies by autoreactive B lymphocytes. In some embodiments, such non cancer antigen is selected from the group consisting of CD10, CD19, CD20, CD22, CD52, CD79a, and / or CD79b. This list is not limiting though. As used herein the term “immune checkpoint agonist” relates to an agent that interacts with immune checkpoints on the surface of a T cell or NK cell and triggers intracellular signalling pathways. Such agonists include, inter alia, agonists interacting with at least one immune checkpoint selected from the group consisting of PD1, 0X40, TIM3, CTLA4 and / or LAG3. Embodiments wherein the second binding domain binds to such immune checkpoint agonist on a cancer cell support reduction of immune escape by the cancer cells.

[0178] In one embodiment, the immune checkpoint agonist the second binding domain binds to is PD- L1 expressed on a cancer target cell surface, by binding an immune checkpoint agonist on a cancer cell, one of the major immune escape mechanisms of cancer cells can be interrupted, namely the presentation of immune checkpoints like e.g. PD-L1. Cancer cells present the latter and engagement thereof with PD1 on the surface of e.g. T cells then inhibits T cell mediated immune attacks.

[0179] According to one embodiment of the bi- or multifunctional molecule, the first binding domain capable of binding the CD2 antigen is an anti-CD2 antibody, or a target binding fragment thereof.

[0180] According to one embodiment of the bi- or multifunctional molecule, the second binding domain is capable of specifically binding to a target antigen other than the CD2 antigen, and is an antibody, or a target binding fragment thereof, or a T cell receptor (TCR), or a target binding fragment thereof.

[0181] As discussed above, at the priority date, skilled persons were able, using hybridoma technologies, transgenic mammals comprising a human immune repertoire, genetic engineering and library technologies (e.g. phage display, yeast display), to make antibodies against almost every conceivable protein target, provided the latter is well characterized. Hence, the description of the target is sufficiently enabling to make an antibody thereagainst.

[0182] Binding domains capable of binding pMHC are for example T cell receptors (TCRs) and soluble forms thereof, and so-called TCR mimicking antibodies (TCRmAbs). The latter, as well as technologies to produce them are e.g. disclosed in Duan and Ho (2021), the content of which is incorporated herein by reference for enablement purposes. For TCRs and TCRmAbs, the same enablement principles apply as to “regular” antibodies. TCRs, too, can be obtained by screening of a respective library, or can be obtained from a donor and then optionally be affinity matured.

[0183] Soluble Augments of TCRs are often provided as lieterodimers comprising of a Va-Ca chain and a VfkCp chain, stabilized by a disulfide bridge acironiplislied by two doneddn cysteine residues in the respective C domains, or as a scTCR comprising a ¥a domain and a Vf> domain fused by a peptide linker.

[0184] In embodiments of the bi- or multifunctional molecule according to the invention, both the antibody or fragment representing the first binding domain (anti-CD2) as well as the antibody or TCR or respective fragment representing the second binding domain, can adopt different formats, with or without Fc domain.

[0185] As discussed below, in formats with an Fc domain, the latter can be Fc silenced, can have increased effector function, or can have increased affinity to FcRn - depending on the respective application. In this case, the bi- or multifunctional molecule can adopt the form of an IgG-antibody, yet with a silenced Fc domain, or a soluble T cell receptor fused to a silenced Fc domain, or a so-called scFv-Fc construct with a silenced Fc domain. See Figure 38 and respective discussion.

[0186] Further, as discussed below, instead of using an (unsilenced) Fc domain or an Fc domain that has increased effector function, a third binding domain binding to CD 16 can be applied.

[0187] Formats without a third binding domain, in particular without a Fc domain or without an antiCD 16 binder, can also be used. Such molecule can for example, be provided in the form of a bispecific construct comprising two different scFv fragments fused to one another by a peptide linker (such format would be similar to the well-known BiTE format (Baeuerle et al 2009), yet with an anti-CD2 scFv instead of an anti-CD3 scFv, or in the form of a bispecific nanobody comprising two different single-domain antibodies (Harmsen et al 2007) derived e.g. from camelids or sharks. See Figure 36 and 37 and respective discussion.

[0188] According to one embodiment of the bi- or multifunctional molecule, the second binding domain capable of specifically binding to a target antigen other than the CD2 antigen is a molecule that comprises a soluble T cell receptor, or a target binding fragment thereof. According to one embodiment of the bi- or multifunctional molecule, the anti-CD2 binding moiety is derived from an antibody, that upon binding to CD2, increases, stabilizes and / or enhances binding between CD2 and CD58.

[0189] The inventors have surprisingly found that these characteristics apply for the anti CD2 antibody 16-92-4, which seems to stabilize a conformation of CD2 that facilitates or even favors interaction with CD58, as disclosed elsewhere herein.

[0190] The stabilizing effect of 16-92-4 with regard to the binding between CD2 and CD58 is particularly advantageous in the context of the present invention, in that the endogenous costimulatory effect of CD58 presented by the target cell on CD2 on the T cell, or NK cell may be stabilized, made more sustainable and may be even enhanced.

[0191] According to one embodiment of the bi- or multifunctional molecule, the anti-CD2 antibody, upon binding to CD2, reduces or inhibits the binding between CD2 and CD58.

[0192] The inventors have surprisingly found that these characteristics apply for the anti CD2 antibody 16-23-7, which appears to bind to an epitope on CD2 that seems to overlap with the binding site for CD58, and / or binding of 16-23-7 sterically impedes the interaction of CD2 with CD58, as disclosed elsewhere herein. Still, the binding of such antibody to CD2, even though interfering with the interaction thereof with CD58, provides a respective stimulatory signal.

[0193] According to one embodiment of the bi- or multifunctional molecule, the anti-CD2 antibody, or the target binding fragment thereof, comprises the CDR sequences or the VL / VH domains of one of the antibodies 16-92-4 and / or 16-23-7.

[0194] According to one embodiment of the bi- or multifunctional molecule, the anti-CD2 antibody is an antibody according to the above description. Such antibody comprises the sequences as shown in the following table:

[0195] Table 4: Sequences of the antibodies of the above description

[0196] According to one embodiment of the bi- or multifunctional molecule, the molecule further comprises a third domain capable of interacting with, or binding to, CD 16. This embodiment is also called „Trifunctional NK cell costimulator” (TriNCo).

[0197] FcyRIIIa (also called CD 16) is a cluster of differentiation molecule found on the surface of inter alia natural killer cells. Hence, antibodies comprising such domain are able to engage NK cells, in order to have the latter carry out effector functions on the target cell.

[0198] According to one embodiment of the bi- or multifunctional molecule, said domain capable of interacting with, or binding to, CD 16 is at least one selected from the group consisting of

[0199] • an anti-CD16 antibody or a target binding fragment thereof, or

[0200] • an antibody Fc domain

[0201] Technically, a molecule which has a) a first binding domain capable of binding the CD2 antigen, b) a second binding domain capable of specifically binding to a target antigen other than the CD2 antigen, and c) a domain capable of interacting with, or binding to, CD 16 (the latter being, preferably, an Fc domain or an anti CD 16 antibody). is a trifunctional molecule and hence multifunctional.

[0202] The typical bispecific T cell engagers come in a double scFv format with one domain binding CD3 and the other one binding a cancer antigen, and are devoid of an antibody Fc domain - hence, lacking the ability to interact with CD 16. Bispecific NK cell engagers have one domain binding CD16 and the other one binding a cancer antigen. For this reason, such molecules are truly bifunctional. Preferably, the antibody Fc domain is part of the second binding domain capable of specifically binding to a target antigen other than the CD2 antigen, in case if said binding domain is an antibody or a target-binding fragment thereof.

[0203] Antibody Fc domains mediate, inter alia, antibody dependent cellular cytotoxicity (ADCC). NK-cell mediated ADCC is mainly triggered by interaction between the antibody Fc domain and the FcyR Illa receptor (CD 16), on the NK cells.

[0204] According to one embodiment of the bi- or multifunctional molecule the antibody Fc domain is modified to exhibit increased effector function.

[0205] Ways to increase effector function are for example disclosed in van der Horst et al (2020), the content of which is incorporated herein by reference for enablement purposes.

[0206] Effector functions as referred to above include not only ADCC, but also antibody-dependent cellular phagocytosis (ADCP), and complement-dependent cytotoxicity (CDC).

[0207] Increased effector function of an antibody Fc domain can be accomplished in different ways. Enhanced ADCC and ADCP can be accomplished by different measures, namely glycoengineering to enhance FcyR affinity, site-directed mutagenesis to enhance FcyR affinity and / or Fc multimerization

[0208] 1. In Fc glycosylation, fucose restricts the number of conformations recognized by the FcyRIIIa N-glycan and inhibits direct carbohydrate-carbohydrate interactions with the receptor glycan. Therefore, afucosylation is an effective approach to increase the potency of e. g.

[0209] 2. IgGl antibodies to induce ADCC. To some extent, Fc galactosylation also modulates FcyRIIIa binding.

[0210] Afucosylation of Fc can be accomplished by antibody expression under conditions that suppress fucosylation (without a prior Fc sequence mutation). CHO culture media used for antibody production do not contain fucose. Hence nearly all fucose attached to the produced antibodies are generated de novo by the CHO cells. Supplementing culture media during antibody production with fucose analogues such as 2-Deoxy-2-fluoro-L-fucose inhibits enzymes involved in the generation and transfer of fucose groups to glycoproteins (Okeley et al 2013). Other approaches involve inactivation or knock-out of the FUT8 gene in the expressing cell line, or using cells lines which have lower endogenous expression thereof (like e.g. rat hybridoma YB2 / 0 cells), inactivation or knock-out of the Golgi GDP -fucose transporter (GFT) gene (Slc35cl inactivation or knock-out of the endogenous GDP-mannose 4,6-dehydratase (GMD) (like e.g. in CHO Lecl3 cells, and overexpression of P-1, 4-mannosyl-gly coprotein 4- P-7V-acetylglucosaminyltransferase (GnT-III), which catalyzes the formation of a bisecting GlcNAc by attaching a GlcNAc in pi,4 linkage to the P-linked mannose of the tri-mannosyl core of A-glycans, yet is normally not expressed in CHO cells.

[0211] 2. FcyRs interact with residues Leu234-Ser239 on the lower hinge and residues Asp265- Glu269, and Asn297-Thr299 on the CH2 domain. Alanine scanning in the CH2 and CH3 domains has revealed that several mutations could enhance binding to FcyRIIIa, with the most potent mutations combined in S298A / E333A / K334A for enhanced ADCC.

[0212] S239D / I332E mutations can also enhance FcyRIIIa binding and ADCC. P247I / A339Q have also been shown to enhance binding to the lower-affinity FcyRIIIa. Applying mutations to Fc regions asymmetrically can maximize the FcyR binding affinity. In addition, afucosylation of the heterodimeric antibodies further improves the FcyRIIIa bindings.

[0213] Further, Fc substitutions L235V, F243L, R292P, Y300L, and P396L as e.g. used in Margetuximab, which is an Fc engineered variant of the anti-Her2 antibody Trastuzumab, increase ADCC e.g. in IgGl isotype antibodies.

[0214] All of these substitutions, alone or in combination, can be used in the context of the bispecific antibody of the present invention, to confer increased ADCC to the Fc domain thereof.

[0215] Different than ADCC, ADCP induction is highly dependent on the balance of binding to the activating receptors versus the inhibitory receptor FcyRIIb. The activating FcyRIIa shares 90% similarities with the inhibitory FcyRIIb. Increasing FcyRIIa binding while simultaneously decreasing FcyRIIb can be achieved by mutations F243L / R292P / Y300L / V305I / P396L. In another study, in which ADCC could be enhanced by S239D / I332E mutations, a third mutation (A330L) was necessary to improve ADCP because the sole S239D / I332E mutation also resulted in increased binding to FcyRIIb. G236A selectively enhances FcyRIIa binding when added to S239D / I332E and S239D / A330L / I332E. 3. Avidity modulation is a less established but more straightforward approach to increase ADCC and ADCP. Fc duplication (or tandem-Fc) or multiplication, whereby multiple Fes are linked within one IgGl molecule, has been shown to augment FcyR binding avidity and increase ADCC and ADCP.

[0216] Enhanced CDC can be accomplished by different measures, namely modified glycosylation, site directed mutagenesis to increase Clq binding, antibody hexamerization and / or cross- isotype antibodies.

[0217] 1. Molecular interactions between galactose and amino acid residues on the CH2 domains have increased Clq binding affinity, and further Fc glycosylation modulates Fc / Fc interactions and thereby affects not the affinity but the avidity of Clq binding.

[0218] 2. Residues D270, K322, P329, and P331 of the CH2 domain are critical for the interaction with Clq. There are basically two main interaction sites: residues 266-272 and 294-300 on one CH2 domain and residues 325-331 on the other. Further, mutations in residues located on or in proximity to these binding sites significantly affect Clq binding: The double mutant K326WZE333S and triple mutant S267E / H268E / S324T enhanced Clq binding and CDC. Further, mutations in the upper hinge region can enhance Clq binding and CDC.

[0219] 3. Antibody hexamerization facilitates Clq binding by modulating avidity. Introducing the specific point mutations E345R and E430G at the Fc and CH2-CH3 interface stimulates the Fc / Fc interactions between antibodies and facilitates the natural concept of antibody hexamerization, leading to superior Clq binding and enhanced CDC.

[0220] 4. Cross-isotype antibodies can be generated by replacing the CH2 and CH3 domains of an IgGl antibody with the corresponding regions of an IgG3 antibody, and hence increases the CDC response.

[0221] As discussed above, in another embodiment of the bi- or multifunctional molecule, the domain capable of interacting with, or binding to, CD 16 is an anti-CD16 antibody or a target binding fragment thereof. Like antibody Fc domains, anti-CD16 antibodies can be used, in bi-or multifunctional antibodies, to trigger NK-cell mediated responses, including, inter alia, ADCC and CDC. For this reason, similar to the well-known BiTE format (“bispecific T cell engager”), which comes in a heterodimeric scFv format, wherein one scFv binds a cancer target like e.g. CD19, and the other one binds to CD3, so as to engage with T cells and establish an artificial immunological synapse between T cell and cancer cell, bispecific constructs have been developed in which one part binds to a cancer target and the other one comprises an anti CD 16 antibody, so as to engage with CD 16 NK cells. Such bispecific constructs, also called “BiKE” (“Bispecific Killer engager”) have for example been disclosed by McCall et al (1999), Kueahara et al (2020) or Nikkhoi et al (2023). Many examples of anti CD 16 antibodies have been described in literature. Human anti CD 16 antibodies are for example commercially available at Antibody Sy stem under the catalog names NM3E2 and GMA-161.

[0222] According to one embodiment of the bi- or multifunctional molecule, the first binding domain capable of binding to the CD2 antigen is fused to a C-terminus of a CH3 domain of an Fc domain of the second binding domain, in case said second binding domain is an antibody or fragment or derivative that comprises at least one heavy chain and / or at least one Fc domain, directly or via a linker.

[0223] According to one embodiment of the bi- or multifunctional molecule, only one first binding domain capable of binding to the CD2 antigen is fused to the C-terminus of one Fc domain. The inventors have surprisingly shown (see Figure 28) that such format largely avoids NK-cell mediated fratricide. (Nakamura et al, 2013). Without being bound to theory, this format seems to enforce simultaneous engagement of CD 16 and CD2 on the same NK cell (“killer cell”), as shown in Figure 36.

[0224] Due to the absence of a second CD2 binding moiety, this format seems to avoid significant engagement with another NK cell as “target cell”, in such way, interactions between two NK cells brought into direct neighbourhood, leading to fratricide phenomenon, are avoided.

[0225] Without being bound to theory, it appears that both (i) monovalency regarding the first binding domain capable of binding to CD2 (i.e., formats that are not bi- or higher valent for such binding domains) and / or (ii) using 16-92-4 independently reduce fratricide activity.

[0226] In one embodiment, such antibody has an Fc domain which is modified to exhibit increased effector function (like ADCC), as outlined elsewhere herein (e. g. by afucosylation). The inventors have also shown that 16-92-4, in the bispecific and / or ADCC-enhanced monospecific format causes less fratricide than 16-23-7 (see Figure 28).

[0227] According to one embodiment of such bi- or multifunctional molecule, the two Fc domains are heterodimeric relative to one another.

[0228] Technologies to create antibodies with heterodimeric Fc domains are disclosed for example in Brinkmann & Kontermann (2017), the content of which is incorporated herein by reference for enablement purposes.

[0229] According to one embodiment of such bi- or multifunctional molecule, the two Fc domains are provided in knobs-into-holes configuration.

[0230] Knobs-into-holes is a technology in which a CH3 interface is created that favours heterodimeric assembly by replacing small side chains on one CH3 interface with larger ones to generate a knob and replacing large side chains on the other CH3 domain with smaller ones to generate a hole. Examples encompass, but are not limited to, the pairs shown in table 5. In other embodiments, Fc heterodimerization is accomplished by one other means selected from the non-limiting list of examples shown in table 5, taken from Brinkmann & Kontermann (2017), the content of which is incorporated herein by reference for enablement purposes.

[0231] Table 5: Means and technologies to accomplish Fc heterodimerization According to another aspect of the present invention, a combination comprising (a) the bi- or multifunctional molecule according to any one of the aforementioned claims and (b) a therapeutic NK cell product is provided.

[0232] Such therapeutic NK cell product can for example comprise endogenous NK cells or iPSC derived NK cells (=iNK). iNK may have to be modified to express CD2 and to have sufficiently high CD 16 expression. Endogenous NK cells can either be derived from the patient (autologous) or from a healthy donor (allogenic). These cells need to be ex-vivo expanded but need not necessarily be gene modified.

[0233] Technologies and methodologies to make NK cells accessible for therapeutic use are e.g. disclosed in Du et al (2021) and Shin et al (2023), the contents of both of which is incorporated herein by reference for enablement purposes.

[0234] According to one embodiment, the bi- or multifunctional molecule lacks a third domain capable of interacting with, or binding to, CD 16, or comprises a third domain that has reduced interaction or binding capacity to CD 16. This format is also called „Bifunctional T cell costimulator“ (BiTCo).

[0235] As used herein, the term “domain that has reduced interaction or binding capacity to CD 16” relates to a domain that is mutated or otherwise modified relative to a non-mutated or nonmodified domain that has interaction or binding capacity to CD 16.

[0236] In one embodiment, wherein the bi- or multifunctional molecule lacks a third domain capable of interacting with or binding to CD 16, the molecule can adopt different formats, all of which do not comprise an antibody Fc domain.

[0237] In one embodiment, wherein the bi- or multifunctional molecule comprises a third domain that has reduced interaction or binding capacity to CD 16, said molecule may comprise an Fc domain which is Fc-silenced so that it does not interact with CD16, or only to a reduced degree. In this case, the bi- or multifunctional molecule can adopt the form of an IgG-antibody, yet with a silenced Fc domain, or a soluble T cell receptor fused to a silenced Fc domain, or a so-called scFv-Fc construct with a silenced Fc domain,

[0238] In such way, while interaction with T cells is still maintained by means of the CD2 binding domain, effector functions caused by Fc-CD16 interaction, like ADCC are eliminated or at least reduced.

[0239] Fc silencing is a well established technology to avoid ADCC and other effector function. In one embodiment, Fc silencing reduces or eliminates FcyRIIIa binding to the Fc domain. FcyRIIIa (also called CD 16) is a cluster of differentiation molecule found on the surface of inter alia natural killer cells. Hence, antibodies comprising such silenced Fc domains lose the ability to engage NK cells. Principles of Fc silencing are for example disclosed in Wang et al (2019), the content of which is incorporated herein by reference for enablement purposes.

[0240] According to one embodiment, the Fc domain comprises at least one of the following amino acid substitutions: L234A, L235E, A330S, P331S and / or G237A. Table 6 shows a non-limiting list of further examples.

[0241] Table 6: substitutions that can be used for Fc silencing

[0242] In one embodiment, said molecule comprises an Fc domain which has increased affinity to FcRn.

[0243] FcRn is the neonatal Fc receptor and not to be confused with the FcyRIII receptors discussed above. FcRn is a protein that in humans is encoded by the FCGRT gene. It is an IgG Fc receptor which is similar in structure to the MHC class I molecule and also associates with beta-2- microglobulin.

[0244] FcRn extends the half-life of IgG and serum albumin by reducing lysosomal degradation of these proteins in endothelial cells and bone-marrow derived cells. IgG, serum albumin and other serum proteins are continuously internalized into cells through pinocytosis, and transported from early endosomes to lysosomes, where they are degraded. Following entry into cells, the two most abundant serum proteins, IgG and serum albumin, are bound by FcRn at the slightly acidic pH within early endosomes, sorted and recycled to the cell surface where they are released at the neutral pH of the extracellular environment. In this way, IgG and serum albumin are salvaged to avoid lysosomal degradation.

[0245] The identification of FcRn as a central regulator of IgG levels led to the engineering of IgG- FcRn interactions to increase in vivo persistence of IgG (like e.g. YTE applied in the antibody cocktail (Evusheld)), as described e. g. by Dall’Acqua et al (2002), who describe mutations in the Fc domain which increase affinity to FcRn, and hence increase the serum half-life of molecules comprising such mutations. Some selected mutations are shown in table 7.

[0246] Table 7: substitutions that can be used for increasing FcRn affinity

[0247] According to one embodiment of the bi- or multifunctional molecule, the first binding domain capable of binding to the CD2 antigen is fused to the C-terminus of the CL domain, in case said second binding domain is an antibody or fragment or derivative that comprises at least one CL domain, directly or via a linker.

[0248] Without being bound to theory, this format, as shown in Figure 37, seems to present a suitable configuration for binding to the tumor antigen and simultaneously mediating functional engagement of CD2 on effector T cells.

[0249] The inventors found that embodiments in which the first binding domain capable of binding to the CD2 antigen is fused to the C-terminus of the CL domain increases T cell co-stimulation substantially. This effect was particularly evoked by 16-92-4 scFv moieties (while less evoked by 16-23-7, see Figure 17).

[0250] However, other configurations in which CD2 binding or binding of the target antigen is not sterically hindered by other domains of the bi- or multifunctional molecule are likewise possible.

[0251] According to one embodiment of the bi- or multifunctional molecule, the first binding domain capable of binding to the CD2 antigen is fused to the C-terminus of both CL domains.

[0252] In another aspect of the invention, a combination comprising (a) the bi- or multifunctional molecule according to the above description and (b) a CD3 -directed T cell engager, a CAR T cell, an immune checkpoint inhibitor, or a bispecific T cell receptor based molecule is provided.

[0253] CD3-directed T cell engagers are bi- or multispecific binders one domain of which binds to CD3 and at least another domain of which binds to another target - preferably on a cancer cell. Such T cell engagers can be provided in the so-called “BiTE” format (bispecific T cell engagers), or in other formats, which are for example disclosed in Brinkmann & Kontermann (2017), the content of which is incorporated herein by reference for enablement purposes.

[0254] CAR T cell approaches use T cells engineered with a chimeric antigen receptor, which is, essentially, a chimera of the intracellular domain of a T cell receptor (mostly CD3Q, an extracellular domain of a cancer target binding antibody (preferably in the scFv format), and at least one of a hinge domain, a transmembrane domain and a costimulatory domain engineered between the two. In such way, the T cell, which originally would only attack target cells that disclose, on MHC, a non-self peptide antigen, is reprogrammed to attack cells which present on their surface a cancer target. CAR T cell technologies are for example disclosed in Sterner and Sterner (2021), the content of which is incorporated herein by reference for enablement purposes.

[0255] Immune checkpoint inhibitors are molecules that target immune checkpoints, which are key regulators of the immune system that when stimulated can dampen the immune response to an immunologic stimulus. Some cancers can protect themselves from attack by stimulating immune checkpoint targets. Checkpoint therapy can block inhibitory checkpoints, restoring immune system function. Currently approved immune checkpoint inhibitors target CTLA4, PD-1, and PD-L1, yet other experimental immune checkpoint inhibitors target OX-40, Lag3 or Tim3, to name a few.

[0256] Bispecific T cell receptor-based molecules typically comprise a soluble form of a T cell receptor specific for a MHC-presented T cell epitope (i.e. a cancer target), fused to an antibody domain that has a T cell engaging effect, like e.g. anti CD3 or anti TCR. Such molecules are for example developed by Immatics (“TCER”) or Immunocore (ImmTACs)

[0257] In such case, where the combination comprises a CD3 directed T cell engager or a CAR T cell or a bispecific T cell receptor-based molecule, the domain therein which binds to the cancer target is preferably selected to a) bind to another cancer target than the second binding domain of the bi- or multifunctional molecule according to the invention, or b) if binding to the same target the second binding domain of the bi- or multifunctional molecule according to the invention, binds to another epitope thereof.

[0258] According to an embodiment of said combination, the bi- or multifunctional molecule and the CD3-directed T cell engager are provided in separate units.

[0259] In the following, two embodiments of the present invention are again shown in table form: Table 8: Two selected embodiments of the present invention According to another aspect of the invention, said combination is provided for (the manufacture of a medicament for) use in the treatment of a human or animal subject

[0260] • being diagnosed for,

[0261] • suffering from or

[0262] • being at risk of developing cancer.

[0263] This language is deemed to encompass both the swiss type claim language accepted in some countries (in this case, brackets are deemed absent) and EPC2000 language (in this case, brackets and content within the brackets is deemed absent).

[0264] According to another aspect of the invention, a method for treating or preventing a cancer in a human or animal subject is provided, which method comprises administration of a combination according to the above description to a patient, in one or more therapeutically sufficient doses.

[0265] According to one embodiment of said combination or method, the bi- or multifunctional molecule is administered to the subject prior to, simultaneously with, or after, administration of the CD3 -directed T cell engager.

[0266] According to one embodiment of said combination or method, the bi- or multifunctional molecule and CD3 -directed T cell engager are provided in separate units.

[0267] ITEMS OF THE INVENTION

[0268] The present invention also comprises the following items:

[0269] 1. An antibody that binds to CD2, or a target-binding fragment or derivative of such antibody, which a) comprises a set of three heavy chain and three light chain complementarity determining regions (CDR) comprised in the one of the following heavy chain / light variable domain sequence pairs

[0270] SEQ ID NOs 1 and 2;

[0271] SEQ ID NOs 3 and 4; SEQ ID NOs 5 and 6; and / or

[0272] SEQ ID NOs 7 and 8; b) comprises a set of three heavy chain and three light chain complementarity determining regions (CDR) selected from the following sets

[0273] • SEQ ID NOs 9 - 14;

[0274] • SEQ ID NOs 15 - 20;

[0275] • SEQ ID NOs 21 - 26; and / or

[0276] • SEQ ID NOs 27 - 32; c) comprises the set of heavy chain / light chain complementarity determining regions (CDR) of b), with the proviso that at least one of the CDRs has up to 3 amino acid substitutions relative to the respective SEQ ID NOs, and / or d) comprises the set of heavy chain / light chain complementarity determining regions (CDR) of b) or c), with the proviso that at least one of the CDRs has a sequence identity of > 66 % to the respective CDRs comprised in the SEQ ID NOs, wherein the CDRs are embedded in a suitable protein framework, preferably a variable domain framework, so as to be capable to bind to CD2.

[0277] 2. The antibody or fragment of item 1, wherein the CDRs are determined according to the definition of Kabat, Chothia or MacCallum, preferably wherein the CDRs are determined according to the numbering set forth in Table 2.

[0278] 3. The antibody or fragment according to any one of items 1 - 2, which comprises a) the heavy chain / light chain variable domain (HCVD / LCVD) pairs set forth in the following pairs of SEQ ID NOs:

[0279] • SEQ ID NOs 1 and 2;

[0280] • SEQ ID NOs 3 and 4;

[0281] • SEQ ID NOs 5 and 6; and / or

[0282] • SEQ ID NOs 7 and 8; b) the heavy chain / light chain variable domains (HCVD / LCVD) pairs of a), with the proviso that

[0283] • the HCVD has a sequence identity of > 80 % to the respective SEQ ID NO, and / or

[0284] • the LCVD has a sequence identity of > 80 % to the respective SEQ ID NO, c) the heavy chain / light chain variable domains (VD) pairs of a) or b), with the proviso that at least one of the HCVD or LCVD has up to 10 amino acid substitutions relative to the respective SEQ ID NO, said antibody or fragment still being capable to bind to CD2.

[0285] 4. The antibody or fragment according to any one of items 1 - 3, wherein at least one amino acid substitution is a conservative amino acid substitution.

[0286] 5. An antibody that binds to CD2, or a target-binding fragment or derivative of such antibody, which has a binding affinity of > 50 % to CD2 compared to that of the antibody or fragment according to any one of the aforementioned items.

[0287] 6. An antibody that binds to CD2, or a target-binding fragment or derivative of such antibody, which competes for binding to CD2 with a) an antibody according to any one of items 1 - 5, or b) an antibody selected from clones 16-92-4, 16-92-4_h or 16-92-4_ho.

[0288] 7. An antibody or a target-binding fragment or derivative of such antibody that binds to essentially the same, or the same, region or epitope on CD2 as a) an antibody according to any one of items 1 - 5, or b) an antibody selected from clones 16-92-4, 16-92-4_h or 16-92-4_ho.

[0289] 8. A bi- or multifunctional molecule comprising at least: a) a first binding domain capable of binding to the CD2 antigen, or to a subdomain of CD2, and b) a second binding domain capable of specifically binding to a target antigen other than the CD2 antigen. 9. The bi- or multifunctional molecule according to item 8, wherein the second binding domain binds to a cancer-associated target on a cancer target cell surface or to an immune checkpoint agonist on a cancer cell.

[0290] 10. The bi- or multifunctional molecule according to any one of items 8 - 9 wherein the first binding domain capable of binding the CD2 antigen is an anti-CD2 antibody, or a target binding fragment thereof.

[0291] 11. The bi- or multifunctional molecule according to any one of items 8 - 10, wherein the second binding domain capable of specifically binding to a target antigen other than the CD2 antigen is an antibody, or a target binding fragment thereof.

[0292] 12. The bi- or multifunctional molecule according to item 11, wherein the second binding domain capable of specifically binding to a target antigen other than the CD2 antigen is provided in a molecule that adopts the IgG format.

[0293] 13. The bi- or multifunctional molecule according to any one of items 8 - 10, wherein the second binding domain capable of specifically binding to a target antigen other than the CD2 antigen is a molecule that comprises a soluble T cell receptor, or a target binding fragment thereof.

[0294] 14. The bi- or multifunctional molecule according to any one of items 10 - 13, wherein the anti-CD2 antibody, upon binding to CD2, increases, stabilizes and / or enhances binding between CD2 and CD58.

[0295] 15. The bi- or multifunctional molecule according to any one of items 10 - 13, wherein the anti-CD2 antibody, upon binding to CD2, reduces or inhibits the binding between CD2 and CD58.

[0296] 16. The bi- or multifunctional molecule according to any one of items 10 - 13, wherein the anti-CD2 antibody is an antibody according to any one of items 1 - 7.

[0297] 17. The bi- or multifunctional molecule according to any one of items 8 - 16, which molecule further comprises a third domain capable of interacting with, or binding to, CD 16. 18. The bi- or multifunctional molecule according to item 17, wherein the domain capable of interacting with or binding to CD 16 is at least one selected from the group consisting of

[0298] • an anti-CD16 antibody or a target binding fragment thereof, or

[0299] • an antibody Fc domain

[0300] 19. The bi- or multifunctional molecule according to item 18, wherein the antibody Fc domain is modified to exhibit increased effector function.

[0301] 20. The bi- or multifunctional molecule according to any one of items 8 - 19, wherein a first binding domain capable of binding to the CD2 antigen is fused to a C-terminus of a CH3 domain of an Fc domain of the second binding domain, in case said second binding domain is an antibody or fragment or derivative that comprises at least one heavy chain and / or at least one Fc domain, directly or via a linker.

[0302] 21. The bi- or multifunctional molecule according to item 20, wherein only one first binding domain capable of binding to the CD2 antigen is fused to the C-terminus of one Fc domain only.

[0303] 22. The bi- or multifunctional molecule according to any one of items 20 - 21, wherein the two Fc domains are heterodimeric relative to one another.

[0304] 23. The bi- or multifunctional molecule according to any one of items 20 - 22, wherein the two Fc domains are provided in knobs-into-holes configuration.

[0305] 24. A combination comprising (a) the bi- or multifunctional molecule according to any one of items 8 - 23 and (b) a therapeutic NK cell product.

[0306] 25. The bi- or multifunctional molecule according to any one of items 8 - 23, which molecule lacks a third domain capable of interacting with, or binding to, CD 16, or comprises a third domain that has reduced interaction or binding capacity to CD 16.

[0307] 26. The bi- or multifunctional molecule according to item 25, which molecule comprises an Fc domain which is Fc-silenced. 27. The bi- or multifunctional molecule according to any one of items 8 - 23 or 25 - 26, wherein two first binding domains capable of binding to the CD2 antigen are fused to the C- termini of the CL domain, in case said second binding domain is an antibody or fragment or derivative that comprises at least one CL domain, directly or via a linker.

[0308] 28. The bi- or multifunctional molecule according to item 26, wherein two first binding domains capable of binding to the CD2 antigen are fused to the C-termini of both CL domains.

[0309] 29. A combination comprising (a) the bi- or multifunctional molecule according to any one of items 8 - 23 or 25 - 28 and (b) a CD3 -directed T cell engager, a CAR T cell, an immune checkpoint inhibitor, or a bispecific T cell receptor-based molecule.

[0310] 30. The combination according to item 29, wherein the bi- or multifunctional molecule and the CD3-directed T cell engager are provided in separate units.

[0311] 31. The combination according to any one of items 24 and 29 - 30 for (the manufacture of a medicament for) use in the treatment of a human or animal subject

[0312] • being diagnosed for,

[0313] • suffering from or

[0314] • being at risk of developing cancer.

[0315] 32. A method for treating or preventing a cancer in a human or animal subject, which method comprises administration of a combination according to any one of items 24 and 29 - 31 to a patient, in one or more therapeutically sufficient doses.

[0316] 33. The combination or method according to any one of items 24 and 29 - 32, wherein (a) is administered to the subject prior to, simultaneously with, or after, administration of (b)

[0317] 34. The combination or method according to any one of items 24 and 28 - 33, wherein (a) and (b) are provided in separate units. EXAMPLES

[0318] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

[0319] All amino acid sequences disclosed herein are shown from N-terminus to C-terminus; all nucleic acid sequences disclosed herein are shown 5'->3'.

[0320] Example 1: Generation of anti-CD2 antibodies

[0321] CD2 antibodies were identified by immunizing mice with human CD4+effector memory T cells (TEM) cells and then phenotypically screening for antibodies that preferentially bind TEM versus naive T cells and co-stimulate T cells. TEM cells were chosen for this immunization because they play an important role at the front line of immunity. In cancer the majority of tumor infiltrating T cells present with a TEM phenotype, while in autoimmunity and chronic inflammation, persistence of self-reactive CD4+TEM cells provide a major contribution to the insult.

[0322] Immunization and hybridoma generation:

[0323] NZBWF1 mice were immunized with human CD4+effector memory T cells (TEM) that were FACS-sorted and amassed from peripheral blood mononuclear cells (PBMC) of 13 healthy donors over the course of the immunization regimen.

[0324] Figure 1 shows the paradigm for FACS-based sorting of effector memory T cells (TEM) from MACS-purified CD4+T cells. CD4+TEM cells were identified as CD45RA-negative, CCR7- negative cells. MACS = Magnetic activated cell sorting. Mice were injected with one million CD4+TEM cells in phosphate buffer saline (PBS) per immunization on days 0, 7, 14, 28, and 31 intraperitoneally. Responder animals were determined by testing binding of mouse serum to CD4+T cells before (pre-immune serum) or after (immune serum) immunization by flow cytometry. Mice showing high titers as compared to their respective pre-immune sera were selected for hybridoma generation and screening. To this end, mice were sacrificed on day 35, and single cell suspensions prepared from splenocytes and lymph nodes for generating hybridoma by standard methods. Culture supernatants of hybridoma were screened 10-14 days later for production of antibodies of interest.

[0325] Screening of hybridoma supernatants

[0326] Aliquots of hybridoma supernatants were allowed to bind to single cell suspensions of total primary human CD4+T cells by 30 min incubation on ice. Cell-bound mouse antibodies were detected with a fluorochrome-labelled anti-mouse IgG secondary antibody. Counterstaining for CD45RO, a splice variant of the CD45 receptor whose expression is restricted to memory T cells, allowed to determine the binding ratio of the antibodies to memory T cells versus naive T cells. Figure 2A illustrates the criterium used to define hit binders in the flow cytometry screening of hybridoma. Antibodies were defined as hit binders if they showed at least a 1.8- fold higher binding to CD45RO-positive memory CD4+T cells as compared to CD45RO- negative naive CD4 T cells. Hybridoma supernatants from hit binders were further tested for absence of binding to the human embryonic kidney fibroblast cell line HEK293 to exclude binders that recognize non-T cell selective human antigens. Hybridomas, that showed a binding signal on HEK293 cells that exceeded 50% of the value of buffer control incubations, were excluded from further analysis. The final set of hit hybridomas were subcloned to monocl onals by standard methods and mouse monoclonal IgGs were purified from supernatants of monoclonal hybridoma by Protein A chromatography. The monoclonal mouse hybridoma were also sequenced (VH and VL) as described, for example, in Hanack et al. (AEMB book series, 2016, volume 917) and Meyer et al. (PLoS Online, 2019; 14(6)). Elucidated sequences were cloned as chimeric antibody constructs comprising a mouse variable domain and a human IgG4 S228P constant domain in a modified version of the pcDNA3.1 expression vector. Chimeric antibodies were expressed, and Protein-A purified from HEK293 cultures according to standard methods.

[0327] Screening for T cell receptor (TCR) co-stimulation Briefly, 96-well flat-bottom tissue culture plates were coated with 200 ng / ml anti-CD3 antibody OKT3 overnight at 4°C. Plates were washed and coated overnight at 20 pg / ml with either mouse IgG isotype controls, the CD28 agonist antibody CD28.2 as positive control or test antibodies identified in the above-described screen. Plates were cultivated with CD4+TEM cells and after five days of culture IFN-y released into the supernatants was measured using a commercial kit (Mesoscale Discovery). Figure 2B shows representative examples of results obtained from the plate-bound antibody TCR co-stimulation assay using antibodies in mouse IgG format. Many of the antibodies emerging from this screen enhanced OKT3-mediated IFN-y release revealing co-stimulation. In the chimeric human IgG4 format the co-stimulation mediated by the original murine antibody hits was confirmed. Figure 2C shows representative examples of results obtained from the plate-bound antibody TCR co-stimulation assay using TEM antibodies in chimeric human IgG4 format. The assay was performed as described above.

[0328] Counterscreening against known T cell surface targets

[0329] To rapidly identify binders against known T cell surface targets 18 selected immune checkpoint receptors were purchased (Origene) and transiently expressed in HEK293 cells. All constructs were expressed as C-terminal GFP-fusion proteins to estimate transfection efficiency. Two days after transfection, cells were harvested, and target expression on the cell surface verified by flow cytometry using commercially available antibodies. Figure 2D shows confirmation of the expression of 18 selected immune checkpoint receptors that were transiently expressed in HEK293 cells. The chimeric hit antibodies were then tested for binding to this array of immune checkpoints expressed in HEK293 cells. Figure 2E shows flow cytometry data for three antibodies from the screen in chimeric human IgG4 format that bind to CD2-transfected HEK293 cells (but not to other transfectants, data not shown) as compared to an isotype control antibody. To confirm that the novel CD2 antibodies are able to interact also with endogenously expressed CD2, we additionally tested their binding to wildtype or CD2-knockout primary human CD4+T cells. CD2-knockout T cells were generated by CRISPR-Cas9 technology. Successful knockout of CD2 was demonstrated by flow cytometry using a fluorochrome- labelled, commercially available CD2 antibody. Wildtype (untreated), CD2-knockout or mock- treated primary CD4+T cells were incubated with 10 pg / ml of respective CD2 chimeric antibody or an isotype control on ice. Following a washing step, bound antibodies were visualized by fluorochrome-labeled anti-human IgG secondary antibody and data collected by flow cytometry. Figure 3 is a bar graph representation of flow cytometry data showing antibody binding to CD2 knockout, mock-treated and untreated CD4+primary T cells. The antibodies 16-23-7 and 16-92-4 both bound to untreated and mock-treated, but not to CD2 knockout primary T cells, thus, revealing binding to the endogenous protein and verifying antigen specificity. Antibody 16-38-1 unexpectedly also bound to CD2 knockout primary T cells and is excluded from further analyses.

[0330] Figure 4A is a bar graph representation of ELISA-based assessment of cross-reactivity to human CD2 and CD2 of cynomolgus monkey (Macaca fascicularis, in the following: “cyno CD2”). CD2 antibodies were tested for reactivity to cynomolgus CD2 by ELISA. Recombinant His6x-tagged proteins of the extracellular domain (ECD) of human and cynomolgus CD2 (Biozol, Eching) were diluted to 2 pg / ml in PBS and adsorbed onto 384-well microtiter plates. After blocking with a 1% bovine serum albumin PBS solution, plates were incubated with chimeric CD2 antibodies at 10 pg / ml for one hour at room temperature. Bound CD2 antibodies were detected with a horse-radish peroxidase (HRP)-conjugated goat anti-human IgG secondary antibody. A commercially available positive control mouse anti-human CD2 antibody (clone RPA-2.10) was detected with HRP-conjugated donkey anti-mouse IgG secondary antibody. Bound detector antibodies were visualized by the addition of 3, 3', 5,5'- Tetram ethylbenzidine (TMB) substrate and the ensuing enzymatic reaction was subsequently quenched with addition of one molar sulfuric acid. Absorbances were measured at 450 nm wavelength and data plotted as bar graphs. Proprietary CD2 antibodies 16-23-7 and 16-92-4 both showed comparable binding to plate-coated recombinant human and cyno CD2 ECD.

[0331] Figure 4B is a graphical representation of flow cytometry-based data assessing binding of antibodies to human CD2 and CD2 of mouse (Mus musculus, in the following: “mouse CD2”). Full length constructs of human and mouse CD2 receptors were purchased (Origene) and transiently expressed in HEK293 cells. Both constructs were expressed as C-terminal GFP- fusion proteins to estimate transfection efficiency. Two days after transfection, cells were harvested, and binding by serially titrated mouse variable domain, human IgG4 S228P constant domain chimeric CD2 and isotype control antibodies was tested for. Bound antibodies were detected with a fluorochrome-labeled anti-human IgG secondary antibody. The expression of the mouse CD2 receptor was verified with a rat anti -mouse CD2 antibody that was consequently detected with an anti-rat IgG secondary antibody. Data were collected by flow cytometry and median fluorescence intensities analyzed. Proprietary CD2 antibodies 16-23-7 and 16-92-4 both showed comparable binding to human CD2 but no binding to mouse CD2. To determine the binding kinetic parameters of the antibodies, surface plasmon resonance (SPR) analysis on the Biacore T200 platform was performed. A clinical stage reference CD2 antibody (Siplizumab) and the two proprietary CD2 antibodies were separately captured on an anti -human Fab CM5 chip surface in HBS-EP+ buffer at a flow rate of 10 pl per minute. Seven different concentrations of recombinant CD2 antigen (3 -fold dilutions starting at 200 nM) were tested comprising a cycle of 3 minutes antigen association, 30 minutes dissociation and a regeneration phase of two 60-second runs with glycine buffer, pH 2.1. Experiments were conducted in HBS-EP+ buffer at a flow rate of 30 pl / min at 25°C. Sensorgram data were fitted to global 1 : 1 Langmuir model and dissociation rate constants (kd) and association rate constants (ka) determined. KD values were calculated from the ratio of the kd and ka values (Figure 6). The reference and proprietary CD2 antibodies all showed comparable KD values in the nanomolar range (~4 - 20 nM). Figure 5 shows representative SPR sensorgram overlays of association and dissociation profiles of the two CD2 antibodies 16-23-7 and 16-92-4, and the reference CD2 antibody Siplizumab. Fig 6 shows summary SPR data for Figure 5.

[0332] Figure 7A shows schematic representations of two assay set-ups to study the impact of CD2 antibodies on the CD2-CD58 receptor ligand interaction by flow cytometry. CD58 (LFA-3) is the high affinity ligand for CD2 in humans. To assess whether the binding epitopes of 16-23-7 and 16-92-4 overlap with the ligand binding site of CD2 the binding of recombinant CD58- ECD-Fc protein (containing the extracellular domain of CD58) to T cells was tested in the absence or presence of CD2 antibodies. The impact of a fixed concentration of 10 pg / ml of CD2 antibodies on the CD2-CD58 interaction was studied in two related set-ups: In set-up 1, primary CD8+T cells endogenously expressing CD2 were pre-incubated with biotinylated CD58-Fc prior to addition of anti-CD2 antibodies. In set-up 2, primary CD8+T cells were first pre-incubated with anti-CD2 antibodies prior to addition of biotinylated CD58-Fc. In each case after washing, T cell-bound CD58-Fc was detected with an APC-labelled streptavidin reagent and data collected by flow cytometry.

[0333] Figure 7B shows bar graph representations of flow cytometry data revealing binding intensities of CD58-Fc to CD8+T cells in presence or absence of CD2 antibodies using assay set-ups 1 and 2. For set-up 1, CD8+T cells were pre-incubated with biotinylated recombinant CD58- ECD-Fc protein (R&D Systems) at 4°C for 60 minutes before addition of respective CD2 or control antibodies at a fixed antibody concentration of 10 pg / ml. After an additional incubation at 4°C for 60 minutes and washing, bound biotinylated CD58-ECD-Fc was detected by APC- labeled streptavidin and data collected by flow cytometry. For set-up 2, CD8+T cells were first incubated with respective CD2 or control antibodies at a fixed antibody concentration of 10 pg / ml at 4°C for 60 minutes prior to incubation with biotinylated CD58-ECD-Fc protein also at 4°C for additional 60 minutes. As in the first set-up, after a washing step, bound biotinylated CD58-ECD-Fc was detected by APC-labeled streptavidin and data collected by flow cytometry. Both assay set-ups showed the same opposite effects of the two CD2 antibodies on the CD2- CD58 interaction: while antibody 16-23-7 completely blocked CD58 binding to T cells, antibody 16-92-4 had the opposite effect and rather enhanced binding of recombinant CD58 to T cells by several fold. An isotype control antibody did not affect the CD2-CD58 interaction at all. These data clearly demonstrate that 16-23-7 and 16-92-4 have opposite impacts on the CD2- CD58 interaction.

[0334] Figure 7C is a graphical representation of a concentration response analysis showing the impact of CD2 antibodies on the binding of CD58-Fc to primary CD8+T cells. Concentration response analyses were performed with various antibodies using assay set-up 2 as described in Figure 7A above. Clear dose-dependent blocking and enhancing effects could be demonstrated for antibodies 16-23-7 and 16-92-4 respectively. In the same assay, Siplizumab, a CD2 antibody used in clinical trials in kidney transplant patients, as well as an isotype control antibody showed no impact on CD2-CD58 interaction at any concentration.

[0335] Figure 8 shows bar graph representations of results from a FACS-based epitope binning assay. Given the opposite effects of the proprietary CD2 antibodies 16-23-7 and 16-92-4, and the surprising impact of antibody 16-92-4 on binding of recombinant CD58 to primary T cells, it was investigated whether the antibodies shared CD2 binding epitopes using an epitope binning assay. Commercially available CD2 antibodies (prior art) were also tested in the epitope binning assay. It must be noted that cross-blocking events in epitope binning assays may indicate either shared or overlapping antibody epitopes or steric hindrance without any overlap in epitopes. However, the absence of cross-blocking between antibodies can confidently be interpreted to show that the said antibodies have distinct, non-overlapping binding epitopes. Primary human CD8+T cells endogenously expressing CD2 were incubated for 30 minutes at 4°C with or without 20 pg / ml of unlabeled anti-CD2 antibodies or an isotype control prior to addition of 20-fold lower amount of a biotinylated version of CD2 antibodies: Siplizumab, 16-23-7 or 16- 92-4, to a final concentration of 1 pg / ml of biotinylated antibodies. Following an additional 30- minute, 4°C incubation and subsequent washing steps, T cell-bound biotinylated antibodies were detected with an APC-labelled streptavidin reagent and data collected by flow cytometry. Median fluorescence intensities were normalized to respective signal of binding in the presence of buffer only (no block). Distinct profiles were observed for all tested antibodies. All three biotinylated antibodies were blocked by their unlabeled versions. Binding of Siplizumab was blocked by CD2 antibodies 16-23-7 and MEM-65 but not by RPA-2.10, LT2 and TS1 / 8. Binding of 16-23-7 was blocked by all tested CD2 antibodies except 16-92-4. In contrast to Siplizumab and 16-23-7, binding of 16-92-4 was not significantly impacted by any of the other tested antibodies indicating that 16-92-4 binds to a unique epitope.

[0336] Figure 9 shows the results of domain mapping experiments using HEK293 cells transfected with human CD2 wt and deletion mutants.

[0337] Figure 9A is a schematic representation of C-terminal GFP-fusion constructs of full length (FL- CD2) human CD2 and two domain deletion mutants of human CD2. The CD2delC mutant lacks the membrane proximal extracellular Ig-like C-domain, whereas the CD2delV mutant lacks the membrane distal extracellular Ig-like V-domain.

[0338] Figure 9B is a bar graph representation of flow cytometry data showing binding of 16-23-7, 16- 92-4 and an isotype control antibody to full-length and domain deletion mutants of CD2 expressed in HEK293 cells. All constructs were transfected into HEK293 cells by standard methods. Transiently transfected HEK293 were incubated with 10 pg / ml of respective antibodies for 60 minutes at 4°C and after washing bound antibodies were detected with a DyLight650-labelled goat anti-human IgG secondary antibody and data collected by flow cytometry. To focus analysis on transfected cells, GFP-positive cells were gated and DyLight650 median fluorescence intensity (MFI) of this cell population was quantified. An APC-conjugated commercial anti-CD2 antibody (clone RPA-2.10) was used as a positive control. Flow cytometry data showed that all CD2 antibodies but not the isotype control antibody bound specifically to full length CD2-transfected cells but not to mock-transfected cells (GFP only). Antibody 16-23-7 and the commercial antibody RPA-2, but not antibody 16- 92-4, also bound to the V-domain only construct (i.e. CD2-delC) showing comparable MFI values to their binding to full length CD2 (FL-CD2). These results clearly demonstrate that the binding epitopes for the antibodies 16-23-7 and RPA-2 are contained entirely within the V- domain of CD2, whereas the epitope of antibody 16-92-4 is not. Since none of the antibodies bound the C-domain only construct (i.e. CD2-delV), it is unclear whether this construct gets displayed on the cell surface. Hence, no conclusion can be drawn with regards to whether the epitope of 16-92-4 is fully contained within the C-domain or not. Notwithstanding, the absence of binding by 16-92-4 to the V-domain only construct (CD2-delC) while both 16-23-7 and RPA-2 bound to intensities comparable to their respective binding to full length CD2, suggests that the epitope of 16-92-4 resides outside of the V-domain i.e. likely in the C-domain.

[0339] Mapping the binding epitope of proprietary CD2 antibody 16-92-4

[0340] A strategy was devised to elucidate the epitope of CD2 antibody 16-92-4 based on the hypothesis that the epitope of 16-92-4 resides in the C-domain of human CD2. Taking the absence of cross-reactivity of 16-92-4 with mouse CD2 into account (see Figure 4B), human and mouse CD2 sequences were aligned. Figure 10 is an alignment of the amino acid sequences of full-length human CD2 (Uniprot code: P06729) and mouse CD2 (Uniprot code: P08920). The sequences encompassing the signal peptides (SP), V-domains, C-domains, transmembrane regions (TM) and intracellular domains (ICD) are marked. Additionally, sequences in the C- domain that are non-conserved between human and mouse CD2 were marked and delineated into six regions i.e. regions 1 through 6. It was hypothesized that one or more of regions 1-6 form the epitope of antibody 16-92-4. Hence chimeric human-mouse CD2 full length constructs were designed and generated as expression constructs. Figure 11 A is a schematic representation of full length human CD2 and human / mouse CD2 chimera constructs generated and expressed in HEK293. Chimeric constructs included either N-terminal human V-domain sequence appended to C-terminal mouse C-domain, mouse transmembrane and mouse intracellular sequences or vice-versa i.e., N-terminal mouse V-domain sequence appended to C-terminal human C-domain, human transmembrane and human intracellular sequences. Additionally, six chimeric constructs were generated comprising largely full length human CD2 bearing limited mouse CD2 sequences in parts of the C-domain marked as regions 1 through 6 in Figure 10.

[0341] HEK293 cells were transfected with corresponding constructs. Two days after transfection, cells were harvested for binding studies. Cells were incubated with serially titrated amounts of antibodies 16-23-7, 16-92-4 or an isotype control antibody on ice for 60 minutes. Cells were then washed, and bound primary antibodies detected with a fluorochrome-labeled anti-human IgG secondary antibody and data collected by flow cytometry. Figure 1 IB shows bar graph representations of median fluorescence intensity (MFI) data from flow cytometry analyses investigating binding of antibodies to full length human CD2 and human / mouse CD2 chimera constructs. For clarity, data are shown for the highest antibody concentration in the assay, i.e. 10 pg / ml. Both proprietary CD2 antibodies showed comparable binding to full length CD2 but in line with previous data, only 16-23-7 bound to the CD2-huV-msC_chimera construct. In contrast, 16-92-4 but not 16-23-7 showed binding to the CD2-msV-huC_chimera construct in support of the hypothesis that the epitope of 16-92-4 resides in the membrane-proximal C- domain of human CD2. Data from the chimera constructs comprising full length human CD2 with limited mouse sequences designated huCD2-msRegX_chimera (X = 1, 2, 3, 4, 5, or 6) showed that mutating the sequences in regions 1, 4 or 6 of human CD2 to corresponding mouse CD2 sequences had no impact on binding by 16-92-4 clearly indicating no contribution by these residues to the antibody epitope of 16-92-4. However, mutating the sequences in region 2 or region 5 of human CD2 to corresponding mouse CD2 sequences resulted in marked reduction in binding by antibody 16-92-4 indicating a major contribution of certain residues in regions 2 and 5 to the epitope of 16-92-4. Remarkably, mutating the non-conserved residues in region 3 of human CD2 to the corresponding mouse CD2 sequence resulted in complete loss of binding by 16-92-4 indicating that residues in this region likely form the core of the epitope of CD2 antibody 16-92-4.

[0342] While additional alanine scanning constructs may help further define the epitope of Ab 16-92- 4 in more detail, the data presented above already delineate the epitope of antibody 16-92-4 to a few clearly defined residues in the membrane proximal C-domain of CD2. Indeed, this epitope as determined to date, is differentiated from the epitopes described for other CD2 antibodies in the prior art. The epitopes of Siplizumab (formerly known as MED 1-507) and other human CD2 antibodies in the prior art were determined to lie within the V-domain of CD2 (Damschroder, et al., 2004). Additionally, two previously described prominent epitopes in the human CD2 (Ti ll and T1 h) also lie within the V-domain (see Damschroder, Melissa M et al.) which is different for 16-92-4 whose epitope clearly lies in the C-domain. To date the only defined epitope encompassing regions in the membrane proximal C-domain (also called D2 domain of CD2) is the so-called CD2-R defined by the T113 antibody. This epitope is understood to be masked in non-activated T cells and exposed upon activation (Li et al 1996). In this respect, antibody 16-92-4 which also binds to non-activated T cells presents with a differentiated binding profile. Li, T et al. defined the epitope of T113 to lie mainly in the linker region between the V-domain and C-domain of human CD2. In particular, the mutations in residues E8, W10, E131 and H140, and K167 each abrogated binding by antibody TI I3 even though expression of the corresponding constructs could be verified with antibodies that bind elsewhere in the V- domain. Of the said residues, H140 and K167 lie in the herein defined regions 3 and 5, respectively. However, mutations at position K139 and K142, both residues in the core epitope of antibody 16-92-4, had no impact on binding by T113. No residues in the herein defined region 2, whose mutation also had a major impact on binding by antibody 16-92-4 was implicated in the epitope of T113. Finally, ability of 16-92-4 to bind to the construct CD2-msV-huC_chimera that lacks the E8 residue, contrasts with the binding properties of TI I3. Taken together, antibody 16-92-4 presents a novel CD2 binding profile.

[0343] Example 2: Bifunctional T cell costimulators

[0344] Figure 12 is a schematic drawing of an exemplary format for a tumor-targeted CD2 bispecific designed to provide a costimulatory signal to effector T cells. Bivalent binding to CD2 is mediated by fusing disulfide-stabilized single chain variable fragment (scFv) moieties directed against CD2 to the C-termini of both light chains of an IgGl molecule directed against a tumor associated antigen (TAA). The asterisks represent mutations that render the Fc moiety silent to prevent CD 16a engagement. Another arrow marks the disulfide stabilization of the anti-CD2 scFv. For all further experiments, Trastuzumab-based CD2 immune cell engager bispecifics as schematically represented in Figure 12 were transiently expressed in CHO cells by standard methods and purified by Protein A and size exclusion chromatography (SEC) to high monomerity. Low endotoxin content was verified with a commercial kit. As controls, bispecific molecules in the same format were also generated by appending scFvs derived from either the CD28-targeting antibody 9-3 (positive control), the mesothelin antibody SSI (negative control) or the RSV antibody palivizumab (negative control).

[0345] Figure 13 is a non-linear regression representation of flow cytometry data showing binding of serially diluted Fc-inactive bispecific antibodies in the IgG-L-scFv format either to primary human CD8+T cells or CD2-transfected HEK293 cells. Primary CD8+T cells or HEK293 cells transiently transfected with a CD2 expression construct were incubated with serially diluted bispecific antibodies on ice for 60 minutes. Following a washing step, bound antibodies were detected by DyLight 650-labelled anti-human IgG secondary antibody and data collected by flow cytometry. Median fluorescence intensity (MFI) values were extracted, and data analyzed by GraphPad Prism. Both CD2 bispecifics Tra-L-16-23-7_AEASS and Tra-L-16-92-4_AEASS as well as the CD28 bispecific Tra-L-9-3_AEASS showed concentration-dependent binding to CD8+T cells with different potencies (EC50) and effective maximum signals (Emax), whereas the negative control bispecific Tra-L-SS1_AEASS showed no binding. The light chain fusion CD2 bispecifics Tra-L-16-23-7_AEASS and Tra-L-16-92-4_AEASS, also showed binding to CD2-transfected HEK293 presenting with comparable concentration-dependent profiles. Since HEK293 cells express neither CD28 nor mesothelin both the CD28 and mesothelin control bispecifics showed only background binding, as expected.

[0346] Figure 14 is a non-linear regression representation of flow cytometry data showing binding of serially diluted Fc-inactive Trastuzumab-based bi specific antibodies in the IgG-L-scFv format to HER2 -positive HCC1954 breast cancer cells. Binding of the bispecifics to the tumor target cell line HCC1954 expressing high levels of HER2 was investigated. HCC1954 cells were incubated with serial dilutions of respective CD2 or control bispecifics on ice for 60 minutes. Following a washing step, bound antibodies were detected by fluorochrome-labelled antihuman IgG secondary antibody and data collected by flow cytometry. MFI values were extracted, and data analyzed by GraphPad Prism. The CD2 bispecifics (Tra-L-16-23 -7_AEASS and Tra-L-16-92-4_AEASS) and the control bispecifics (Tra-L-9-3_AEASS and Tra-L- SS1 AEASS) all showed comparable concentration-dependent binding to the HCC1954 cells via their Trastuzumab-based Fab (fragment antigen-binding) arms.

[0347] Figure 15 shows non-linear regression representations of flow cytometry data showing conjugate formation between CD8+T cells and HCC1954 target cells. The ability of the light chain fusion bispecifics to simultaneously bind two different antigens on T cells and tumor cells was investigated in a conjugate formation assay. To this end, CD8+T cells were labelled with the CellTrace violet dye and HCC1954 tumor cells were separately labelled with the CFSE dye. Pre-labelled CD8+T cells and HCC1954 cells were mixed, treated with serially diluted concentrations of the CD2 bispecifics (Tra-L-16-23 -7_AE ASS and Tra-L-16-92-4_AEASS) or the control bispecifics (Tra-L-9-3_AEASS and Tra-L-SS1_AEASS) and incubated at 37°C for 10 minutes following a short centrifugation step to trigger conjugate formation. Thereafter, cells were fixed with paraformaldehyde to maintain conjugates. After washing, bispecific antibody- induced T cell-tumor cell conjugate formation was determined by flow cytometry as percentage of T cells detected to be positive for both fluorochromes. In contrast to the negative control bispecific, Tra-L-SS1_AEASS, the CD2 bispecifics Tra-L-16-23 -7 AEASS and Tra-L-16-92- 4 AEASS as well as the positive control CD28 bispecific Tra-L-9-3_AEASS showed concentration-dependent conjugate formation between T cells and target HCC1954 cells with comparable potency (ECso). The maximum response (Emax) over the full concentration range showed the following rank order: Tra-L-16-23 -7 AEASS > Tra-L-16-92-4 AEASS > Tra-L- 9-3 AEASS. These results demonstrate that the Trastuzumab-based CD2 bispecifics are able to simultaneously engage with CD2 and HER2 on different cell types.

[0348] Figure 16 is a graphical representation of flow cytometry -based concentration response analysis data showing the impact of Fc-inactive Trastuzumab -based bispecific antibodies in the IgG-L- scFv format on the binding of CD58-Fc to primary CD8+T cells. CD8+T cells were first incubated with serial dilutions of various mono- and bispecific CD2 antibodies or control antibodies at 4°C for 60 minutes prior to incubation with biotinylated CD58-ECD-Fc protein at 4°C for an additional 60 minutes. After a washing step bound biotinylated CD58-ECD-Fc was detected by APC-labelled streptavidin and data were collected by flow cytometry. As expected, Trastuzumab and the negative control bispecific antibody Tra-L-SS1_AEAAS, which both lack a CD2 binding arm, had no effect on CD58-Fc binding to CD8+T cells. For the monospecific parental CD2 antibodies 16-92-4 and 16-23-7 their enhancing and blocking effect, respectively, was confirmed. The bispecific antibody Tra-L-16-23-7_AEAAS showed the same maximal blocking effect as the parental monospecific 16-23-7 antibody but with reduced potency. The bispecific antibody Tra-L-16-92-4_AEAAS did not retain the enhancing effect of the parental monospecific 16-92-4 antibody, but rather also showed a minor inhibitory effect at the two highest concentrations tested. This is probably due to steric hindrance of the CD2-CD58-Fc interaction on the cell surface by the bulky bispecific construct.

[0349] Figure 17A is a non-linear regression representation of anti -tumor cytotoxicity data using primary human T cells. Co-stimulation of anti-tumor cytotoxicity by the Fc-inactive Trastuzumab-based bispecific antibodies in the IgG-L-scFv format was investigated using total human T cells as effector cells. Primary human T cells were purified from PBMCs using Pan T Cell isolation kit (Miltenyi Biotec) following the manufacturer’s protocol. T cells were cocultured with the HCC1954 breast cancer cell line (EpCAM+ and HER2+) at an effector-to- target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of 10 nM fixed concentration of either the CD2 bispecifics (Tra-L-16-23-7_AEASS, Tra-L-16-92-4_AEASS), the positive control CD28 bispecific Tra-L-9-3_AEASS or the mesothelin-binding negative control bispecific Tra-L-SS1_AEASS. After a 2-day co-culture, anti-tumor cytotoxicity was determined by LDH release assay using the CyQUANT™ LDH Cytotoxicity Assay (Thermo Scientific). Data were analyzed using GraphPad Prism. Cytotoxicity was quantified and represented as percentage of maximum non-spontaneous target cell release using the following formula: Cytotoxicity = [(Experimental value - Effector cells spontaneous release control value - Target cells spontaneous release control value) / (Target cells maximum release control value - Target cells spontaneous release control value)] x 100

[0350] The CD2 bispecific Tra-L-16-92-4_AEASS enhanced anti-tumor toxicity comparably to the CD28 control bispecific Tra-L-9-3_AEASS, whereas Tra-L-16-23-7_AEASS showed little effect compared to the negative control Tra-L-SS1_AEASS.

[0351] Fig 17B is a non-linear regression representation of anti -tumor cytotoxicity data using primary human T cells as in Figure 17A albeit using BT474 instead of HCC1954 cells as target cells. At a 2-to-l effector-to-target cell ratio co-cultures of total T cells and BT474 targets cells were treated with serially diluted EpCAM*CD3 BiTE in the presence of 10 nM fixed concentration of either the CD2 bispecific Tra-L-16-92-4_AEASS, the positive control CD28 bispecific Tra- L-9-3 AEASS or the RSV-binding negative control bispecific Tra-L-Pali_AEASS. After 48 hours, anti -tumor cytotoxicity was determined by LDH release assay as in Figure 17A and data were analyzed using GraphPad Prism. The percentage cytotoxicity was normalized to the cytotoxicity caused by 10 nM EpCAM*CD3 BiTE, which represents the concentration that generally mediates highest efficacy of the BiTE. The CD2 bispecific Tra-L-16-92-4_AEASS enhanced anti-tumor toxicity comparably to the CD28 control bispecific Tra-L-9-3_AEASS validating the data achieved using the HCC1954 tumor cells in Figure 17A.

[0352] Figure 18A are non-linear regression representations of CD25 upregulation data. T cell activation was measured by reading out CD25 upregulation on T cells in co-culture assays. Pan- T cells were co-cultured with the HCC1954 breast cancer cell line (EpCAM+ and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of 10 nM fixed concentration of either the CD2 bispecifics (Tra-L-16-23-7_AEASS, Tra-L-16-92-4_AEASS), the positive control CD28 bispecific Tra-L- 9-3 AEASS or the mesothelin-binding negative control bispecific Tra-L-SS1_AEASS. After two days, cell suspensions were stained with fluorochrome-labelled antibodies against CD8, CD4 and CD25 and data collected by flow cytometry and analyzed using GraphPad Prism.

[0353] The CD2 bispecific Tra-L-16-92-4_AEASS enhanced both CD4+ and CD8+ T cell activation as indicated by CD25 upregulation in a signal- 1 dependent manner comparable to the CD28 control bispecific Tra-L-9-3_AEASS. In contrast, Tra-L-16-23-7_AEASS showed little effect on CD4+ T cells and no effect on CD8+ T cells compared to the negative control Tra-L- SSl-AEASS.

[0354] Fig. 18B shows non-linear regression representations of CD25 upregulation data as in Figure 18A albeit with BT474 target cells. Primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+ and HER2+) at an effector-to-target cell ratio of 2: 1. Cocultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of 10 nM fixed concentration of either the CD2 bispecific Tra-L-16-92-4_AEASS, the positive control CD28 bispecific Tra-L-9-3_AEASS or the RSV-binding negative control bispecific Tra-L- Pali AEASS. After two days, cell suspensions were stained with fluorochrome-labelled antibodies against CD8, CD4 and CD25 and data collected by flow cytometry and analyzed using GraphPad Prism. Both the CD2 bispecific Tra-L-16-92-4_AEASS and the CD28 control bispecific Tra-L-9-3_AEASS comparably enhanced activation of CD4+ and CD8+ T cells as indicated by CD25 upregulation in a signal- 1 dependent manner.

[0355] Figure 19 is a non-linear regression representation of anti -tumor cytotoxicity data using macrophage-depleted human PBMCs. Human PBMCs were depleted of macrophages by cultivating them on tissue culture plates overnight and subsequently harvesting non-adherent cells. HCC1954 breast cancer cells (EpCAM+and HER2+) were stably transfected to express the far-red fluorescent protein mKate2. Macrophage-depleted PBMCs were co-cultured with the mKate2+HCC1954 breast cancer cells at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of 50 nM fixed concentration of either the CD2 bispecifics (Tra-L-16-23-7_AEASS, Tra-L-16-92-4_AEASS), the positive control CD28 bispecific Tra-L-9-3_AEASS or the mesothelin-binding negative control bispecific Tra-L-SS1_AEASS. Cell growth of the mKate2+ HCC1954 cells was tracked using the Incucyte system over a 4-day culture period. Anti-tumor cytotoxicity was determined by calculating area under curve values (AUC) of the mKate2+ HCC1954 cell growth curves and normalizing to HCC1954 cell growth in the untreated control using the following formula:

[0356] % Cytotoxicity = (AUC experimental value / AUC untreated control) * 100

[0357] Data were analyzed using GraphPad Prism. The CD2 bispecific Tra-L-16-92-4_AEASS enhanced anti-tumor toxicity comparably to the positive control CD28 bispecific Tra-L-9-3_AEASS, whereas Tra-L-16-23-7_AEASS showed no increased cytotoxic activity compared to the negative control bispecific Tra-L-SS1_AEASS.

[0358] Figure 20 is a non-linear regression representation of IFN-y release data. T cell activation in the PBMC / HCC1954 co-culture assay was determined by measuring IFN-y levels released into the culture supernatants using a commercial kit from Mesoscale Discovery (MSD). Macrophage- depleted human PBMCs were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of Trastuzumab-based Fc-inactive CD2 bispecifics or control bispecific molecules at a fixed concentration of 50 nM. After a 4-day culture, IFN-y release into the culture supernatant was measured by MSD and data analyzed using GraphPad Prism. The CD2 bispecific antibody Tra-L-16-92-4_AEASS further enhanced IFN-y release mediated by the EpCAM*CD3 BiTE improving both the EC50 as well as the Emax value. In contrast, Tra-L-16-23-7_AEASS showed no improvement over the negative control bispecific Tra-L-SS1_AEASS. The positive control CD28 bispecific Tra-L-9-3_AEASS reached an even higher Emax value than Tra-L-16-92-4_AEASS albeit with comparable potency (EC50).

[0359] Figure 21 is a bar graph representation of anti -tumor cytotoxicity data showing the strict dependence of T cell co-stimulation on signal 1. Macrophage-depleted human PBMCs were co-cultured with the mKate2+HCC1954 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with a 50 nM fixed concentration of either the CD2 bispecifics (Tra-L-16-23-7_AEASS, Tra-L-16-92-4_AEASS), the positive control CD28 bispecific Tra-L-9-3_AEASS or the mesothelin-binding negative control bispecific Tra-L-SS1_AEASS in the absence of providing signal 1 via the EpCAM*CD3 BiTE. Cell growth of the mKate2+HCC1954 cells was tracked in the red fluorescence channel of an Incucyte system over a 4-day culture period. Anti-tumor cytotoxicity was determined by calculating area under curve values (AUC) of the mKate2+HCC1954 cell growth curves and normalizing to HCC954 cell growth in the untreated control. Data were analyzed using GraphPad Prism. None of the bispecifics showed any effect on HCC1954 growth and viability under these conditions, underscoring their strict dependence on signal 1 for mediating costimulatory cytotoxicity. Figure 22A shows non-linear regression representations of anti-tumor T cell cytotoxicity data in a signal 2 concentration response analysis. The dose-dependent effects of co-stimulatory bispecifics on T cell mediated cytotoxicity were investigated using the co-culture assay system. Primary human T cells were co-cultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with a fixed low dose of 2 pM EpCAM*CD3 BiTE in the presence of increasing concentrations of either the CD2 bispecific Tra-L-16-92-4_AEASS, the positive control CD28 bispecific Tra-L-9-

[0360] 3 AEASS or the mesothelin-binding negative control bispecific Tra-L-SS1_AEASS. After 2- days of co-culture, anti-tumor cytotoxicity was determined by LDH release assay using the CyQUANT™ LDH Cytotoxicity Assay (Thermo Scientific) as described above for Figure 17. Data were analyzed using GraphPad Prism. As expected, the negative control bispecific Tra-L- SS1 AEASS showed no co-stimulatory effect at any concentration as compared to the control value for treatment with 2 pM EpCAMxCD3 BiTE. The CD2 bispecific Tra-L-16-92-

[0361] 4 AEASS showed dose-dependent co-stimulation of anti-tumor cytotoxicity as did the CD28 control bispecific Tra-L-9-3_AEASS. Both these bispecifics showed comparable costimulation of anti-tumor cytotoxicity with nanomolar EC50 values.

[0362] Figure 22B shows non-linear regression representations of anti-tumor T cell cytotoxicity data in a signal 2 concentration response analysis with BT474 instead of HCC1954 as target cells. Primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with a fixed low dose of 5 pM EpCAM*CD3 BiTE in the presence of increasing concentrations of either the CD2 bispecific Tra-L-16-92-4_AEASS or the RSV-binding negative control bispecific Tra-L- Pali AEASS. After two days of co-culture, anti -turn or cytotoxicity was determined by LDH release assay using the CyQUANT™ LDH Cytotoxicity Assay (Thermo Scientific) as described above for Figure 17 A. Percentage cytotoxicity was normalized to the cytotoxicity at 10 nM EpCAM*CD3 BiTE, which represents the concentration that generally mediates highest efficacy of the BiTE. Data were analyzed using GraphPad Prism. While the negative control bispecific Tra-L-Pali_AEASS showed no effect, the CD2 bispecific Tra-L-16-92-4_AEASS showed remarkable dose-dependent co-stimulation of anti-tumor cytotoxicity demonstrating sub-nanomolar EC50 value (0.22 nM). Data from both, figures 22A and 22B, validate the dosedependent co-stimulatory impact of the CD2 bispecific Tra-L-16-92-4_AEASS on T cell cytotoxicity. Figure 23 A shows non-linear regression representations of CD25 upregulation data as read-out for T cell activation in a signal 2 concentration response analysis. Dose-dependent effects of various co-stimulatory bispecifics on T cell activation were measured by reading out CD25 upregulation in the pan T cell / HCC1954 cell co-culture assays. Primary human T cells were cocultured with the HCC1954 breast cancer cell line (EpCAM+and HER2+) at an effector-to- target cell ratio of 2: 1. Co-cultures were treated with a fixed low dose of 2 pM EpCAM*CD3 BiTE in the presence of increasing concentrations of either the CD2 bispecific Tra-L- 16-92- 4 AEASS, the positive control CD28 bispecific Tra-L-9-3_AEASS or the mesothelin negative control bispecific Tra-L-SS1_AEASS. After 2-days of co-culture, cell suspensions were stained with a fluorochrome-labelled antibody against CD25 as surface marker for T cell activation and counterstained with fluorochrome-labelled antibodies against CD4 and CD8 to discriminate the two T cell subpopulations. Data were collected by flow cytometry and analyzed using GraphPad Prism. In contrast to the negative control bispecific Tra-L-SSl, the CD2 bispecific Tra-L-16- 92-4 AEASS enhanced, in a dose-dependent manner, T cell activation as indicated by CD25 upregulation in both CD4+and CD8+T cells. Emax values for T cell activation were 6-10 fold above baseline and Tra-L- 16-92-4 AEASS showed sub-nanomolar potency. The CD28 control bispecific Tra-L-9-3_AEASS achieved T cell activation with a ~2-fold higher Emax than Tra- L-16-92-4_AEASS.

[0363] Figure 23B shows non-linear regression representations of CD25 upregulation data as read-out for T cell activation in a signal 2 concentration response analysis with BT474 instead of HCC1954 as target cells. Primary human T cells were co-cultured with the BT474 breast cancer cell line (EpCAM+and HER2+) at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with a fixed low dose of 5 pM EpCAM*CD3 BiTE in the presence of increasing concentrations of either the CD2 bispecific Tra-L- 16-92-4 AEASS or the RSV-binding negative control bispecific Tra-L-Pali_AEASS. After two days of co-culture, cell suspensions were stained with a fluorochrome-labelled antibody against CD25 as surface marker for T cell activation and counterstained with fluorochrome-labelled antibodies against CD4 and CD8 to discriminate the two T cell subpopulations. Data were collected by flow cytometry and analyzed using GraphPad Prism. While the negative control bispecific Tra-L-Pali_AEASS showed no effect, the CD2 bispecific Tra-L- 16-92-4 AEASS mediated a dose-dependent increase in T cell activation in line with co-stimulation of anti-tumor cytotoxicity as shown in Figure 22B. Interestingly, Tra- L-16-92-4 showed a stronger impact on CD8+ T cells (with regards to maximum fold increase percentage CD25 T cells) cells than on CD4+ T cells. Example 3: Trifunctional NK cell costimulators

[0364] Figure 24 is a schematic drawing of an exemplary format for a tumor targeted CD2 bispecific designed to provide a co-stimulatory signal to NK cells. Monovalency for binding to CD2 is achieved by fusing a disulfide-stabilized single chain Fv moiety directed against CD2 to the C- terminus of only one of the two heavy chains of an IgGl molecule whose Fab arms recognize a tumor associated antigen (TAA). To ensure correct heterodimer pairing of the two different heavy chains knob-in-hole mutations were introduced into the Fc moieties.

[0365] Figure 25 shows non-linear regression representations of flow cytometry data demonstrating binding of serially diluted bispecific antibodies or Trastuzumab as control to different cell types expressing CD2. CD8+T cells and NK cells that both endogenously express CD2 or HEK293 cells transfected with a human CD2 expression plasmid were incubated with various Fc- competent Trastuzumab-based CD2 bispecific molecules or with monospecific Trastuzumab IgG at 4 °C for 60 min. After washing away unbound antibodies, DyLight650-labelled antihuman IgG secondary antibody was added to the cell suspension and allowed to bind for 45 min at 4°C before again washing away any excess secondary antibody and determining the median fluorescence intensity of cell-bound antibody signal by flow cytometry. Lacking a CD2 binding moiety, Trastuzumab and Tra-H-SSl used as negative controls showed no binding to T or NK cells and only low background binding to CD2 transfected HEK293 , which was similar in magnitude to non-transfected HEK293 cells (data not shown). The monovalent CD2 bispecific constructs Tra-H-monovalent-16-23-7_KiH and Tra-H-monovalent-16-92-4_KiH showed dose-dependently enhanced binding to CD2-transfected HEK293 cells that was clearly above background. The bivalent CD2 bispecific constructs Tra-H-16-23-7 and Tra-H-16-92-4 showed even further enhanced binding to CD2 transfected HEK293 cells. Similar effects were observed with CD8+T cells and NK cells, but likely because of the lower expression level of endogenous CD2 the difference in avidity between mono- and bivalent CD2 bispecifics resulted in a more pronounced left shift of the binding curves for CD2 bivalent constructs.

[0366] Figure 26 shows non-linear regression representations of flow cytometry data demonstrating binding of serially diluted bispecific antibodies or Trastuzumab control to HCC1954 target cells. HCC1954 tumor cells endogenously expressing high levels of human HER2 were incubated with various Trastuzumab-based bispecific molecules optimized for NK cell engagement or with monospecific Trastuzumab IgG at 4 °C for 60 min. After washing away unbound antibodies, DyLight650-labelled anti-human IgG secondary antibody was added to the cell suspension and allowed to bind for 45 min at 4°C before again washing away any excess antibody and determining the median fluorescence intensity of cell-bound antibody signals by flow cytometry. Both monovalent and bivalent CD2 bispecific constructs (Tra-H-monovalent- 16-23-7_KiH, Tra-H-monovalent-16-92-4_KiH and Tra-H-16-23-7, Tra-H- 16-92-4, respectively) showed similar dose-dependent binding to HCC 1954 tumor cells as did the control antibody Tra H SSl. Maximum binding achieved by Trastuzumab was approximately 2-fold higher than all bispecific constructs.

[0367] Figure 27 shows non-linear regression representations of flow cytometry data demonstrating conjugate formation by Trastuzumab or Trastuzumab-based CD2 bispecifics between CD8 T cells and HCC 1954 target cells or NK cells and HCC 1954 target cells. CD8+T cells and NK cells were separately prelabelled with the CellTrace violet dye and HCC 1954 tumor cells were also separately prelabelled with the CFSE dye. Pre-labelled CD8+T cells and HCC 1954 or NK cells and HCC 1954 cells were mixed, treated with serially diluted concentrations of the CD2 bispecifics (Tra-H-monovalent-16-23-7_KiH, Tra-H-monovalent-16-92-4_KiH, Tra-H-16-23- 7, and Tra-H- 16-92-4), the negative control bispecific (Tra-H-SSl) or Trastuzumab IgG and incubated at 37°C for 10 minutes following a short centrifugation step to trigger conjugate formation. Thereafter, cells were fixed with paraformaldehyde to maintain conjugates. After washing, bispecific antibody -induced T cell-tumor cell conjugate formation was determined by flow cytometry by measuring the percentage of T cells orNK cells detected as positive for both fluorochromes. Compared to the negative control bispecific and Trastuzumab IgG, all CD2 bispecifics showed concentration-dependent conjugate formation between either CD8+T cells and the target HCC 1954 cells or NK cells and the target HCC 1954 cells with bivalent CD2 bispecifics showing better potency than monovalent constructs. These results clearly show that the heavy chain fused, Trastuzumab-based CD2 bispecifics can simultaneously engage with CD2 and HER2 on separate cells.

[0368] For assessing the functional effects of the Fc-competent, heavy chain fused CD2 bispecifics on NK cell activation and cytotoxicity we employed the NK-92 cell line as effector cells. The NK- 92 cell line is widely used for ADCC assays and possesses characteristics of highly active blood NK cells but with even broader and greater cytotoxicity. Since the parental NK-92 line lacks endogenous expression of CD 16a, we transduced NK-92 cells with a cDNA construct to over- express the high-affinity 158V variant of CD16. The resulting human CD16 expressing NK-92 line was cultured under antibiotic selection conditions and CD 16 expression was assessed in parallel with each functional experiment to guarantee -80% of CD16+NK-92 cells.

[0369] Figure 28 is a non-linear regression representation of FACS-based concentration response analysis data showing the impact of Fc-competent CD2 bispecifics on NK cell fratricide. NK cell fratricide is a term used for NK cells killing other NK cells, a phenomenon that has been described for Fc-competent antibodies with binding specificity for another NK cell surface marker, like for example the monoclonal anti-CD2 antibody Siplizumab, which we used as positive control in fratricide assays. To evaluate the potential of Fc-competent, heavy chain fused CD2 bispecifics for such unwanted fratricide activity, NK cell degranulation was investigated in the absence of tumor target cells. CD16+NK-92 cells were cultured for 4 hours in the presence of serially titrated concentrations of CD2 bispecifics and control antibodies. NK cell degranulation was monitored as appearance of LAMP1 (CD107a) on the cell surface by adding a Brilliant Violet-labelled anti-CD107A antibody and quantifying the proportion of CD107A bright CD16+NK-92 cells. As expected, the negative control bispecific Tra-H-SSl caused no NK cell degranulation as it does not bind CD2 (or any other NK cell surface antigen for that matter), whereas Siplizumab, used as positive control, even at very low concentrations potently induced NK cell degranulation due to fratricide showing a maximal effect (Emax) of approximately 30%. The bivalent CD2 bispecific Tra-H-16-23-7 also mediated strong NK cell degranulation in the absence of HER2+ tumor target cells showing Emax values comparable to Siplizumab albeit with lower overall potency. In contrast to Siplizumab and Tra-H-16-23-7, the bivalent CD2 bispecific Tra-H-16-92-4 caused little NK cell degranulation due to fratricide with Emax values plateauing below 10%. The monovalent CD2 bispecifics Tra-H-monovalent- 16-92-4_KiH and Tra-H-monovalent-16-23-7_KiH induced even less NK cell degranulation mediating small increases in CD 107a exposure only at the two highest doses used. These data demonstrate that fratricide can be largely avoided by monovalency for CD2. Moreover, the scFv moiety derived from 16-92-4 has the unique property of mediating only very limited fratricide even when used in a bivalent format as C-terminal fusion to the heavy chain of a tumor targeting antibody.

[0370] Figure 29 is a non-linear regression representation of tumor target cell-mediated NK cell degranulation data. To address direct tumor cell killing mediated by CD2 bispecifics, NK cell degranulation was monitored in a short-term co-culture of NK-92 cells with HER2+tumor cells in the presence of serial dilutions of the three CD2 bispecifics that showed little fratricide (Tra- H-16-92-4, Tra-H-monovalent-16-92-4_KiH, and Tra-H-monovalent-16-23-7_KiH). Briefly, CD16+NK-92 cells were co-cultured with CTFR (Cell Trace Far Red)-prelabelled HER2+HCC1954 tumor cells at an effector-to-target ratio of 2: 1 in the presence of increasing concentrations of the various heavy chain fused Trastuzumab-based CD2 bispecifics or control molecules. Brilliant-Violet-421 -conjugated anti-CD107A (LAMP1) antibody was added during co-culture to allow monitoring NK cell degranulation as appearance of LAMP 1 on the cell surface over time. After 4 hours, NK cell degranulation was assessed by FACS quantification of the proportion of CD107A bright, CTFR-negative NK cells. A dose-dependent increase in NK cell degranulation was observed for Trastuzumab and all Trastuzumab-based bispecifics. This is expected as the Trastuzumab binding moiety in conjunction with a non-silenced Fc moiety can mediate antibody-dependent cellular cytotoxicity (ADCC). But the presence of a CD2 binding moiety in the CD2xHER2 bispecifics clearly further enhanced NK cell degranulation compared to the parental Trastuzumab molecule or a Tra-H-SSl negative control bispecific. The extent of degranulation was similar for bispecifics with 16-92-4-based scFv fusions and the one with a 16-23-7-based scFv fusion. Unexpectedly, there was no further enhancement of anti-tumor NK cell degranulation observed with bivalency as compared to monovalency for CD2 (Tra-H- 16-92-4 versus Tra-H-monovalent-16-92-4_KiH) indicating that bivalency for CD2 does not result in a strong avidity effect in this assay.

[0371] Figure 30 is a non-linear regression representation of NK cell-mediated anti-tumor cytotoxicity data. To assess NK cell-mediated cytotoxicity, tumor cell killing after short term co-culture was determined. Briefly, CD16+NK-92 cells were co-cultured for 4 hours with CTFR-prelabeled HER2+HCC1954 tumor cells at a 2: 1 effector-to-target ratio in the presence of increasing concentrations of Trastuzumab, Trastuzumab-based CD2 bispecifics or a negative control CD2 bispecific. At the end of the co-culture period, the percentage of dead CTFR+HCC1954 cells was determined by FACS quantification. In agreement with the NK cell degranulation data, a corresponding dose-dependent increase in NK cell-mediated cytotoxicity was observed for Trastuzumab and all Trastuzumab-based bispecifics. The presence of a CD2 binding moiety in CD2xHER2 bispecifics further enhanced both the potency and efficacy of tumor cell killing as compared to the parental Trastuzumab molecule or a Tra H SSl negative control bispecific. The two CD2 monovalent bispecifics (Tra-H-monovalent-16-23-7_KiH and Tra-H- monovalent-16-92-4_KiH) and the bivalent Tra-H-16-92-4 showed comparable dose response curves. Figure 31 is a non-linear regression representation of NK cell secretion activity (IFN-y release) following co-culture with tumor cells. The impact of CD2 bispecifics on NK cell anti-tumor activity mediated by cytokine release was assessed by quantifying IFN-y release of NK cells by MesoScaleDiscovery (MSD) electrochemiluminescence measurements. Briefly, after 4 hours co-culture of CD16+NK-92 cells with HER2+HCC1954 tumor cells in the presence of increasing concentrations of Trastuzumab, Trastuzumab-based CD2 bispecifics or a negative control CD2 bispecific, co-culture supernatants were analyzed with the U-PLEX Human IFN- y assay kit from MSD according to the manufacturer’s instructions. While Trastuzumab and the negative control construct Tra-H-SSl triggered only minimal IFN-y release at the highest concentrations used, clear dose-dependent increases in IFN-y release were observed for the Trastuzumab-based CD2 bispecifics. The monovalent CD2 bispecific Tra-H-monovalent-16- 92-4_KiH and the bivalent CD2 counterpart Tra-H- 16-92-4 both enhanced IFN-y release with similar potencies and maximal effect levels. The Tra-H-monovalent-16-92-4_KiH was more potent and reached a higher maximal effect level than the Tra-H-monoval ent- 16-23 -7_KiH construct.

[0372] Figure 32A is a bar graph representation of NK cell degranulation data demonstrating dependence of NK cell activation on CD16. To demonstrate the importance of CD16 engagement for the co-stimulatory activity of Fc-competent CD2 bispecifics, we tested in parallel their effect on NK cell degranulation of CD16+NK-92 cells and the parental NK-92 cell line lacking CD 16 expression. Briefly, the different NK-92 cell populations were cocultured for 4 hours with CTFR-prelabeled HER2+HCC1954 tumor cells at a 2: 1 effector-to- target ratio and treated with 50 nM of Trastuzumab, Tra-H-monoval ent- 16-92-4_KiH or a negative control bispecific molecule. The appearance of LAMP 1 on the NK cell surface was monitored with a Brilliant Violet 421-labelled anti-CD107A antibody in FACS. In contrast to clear co-stimulatory activity observed on CD16+NK92 cells, on the CD 16 negative parental NK-92 cells, Tra-H-monoval ent- 16-92-4_KiH did not increase degranulation above a low baseline level of around 5% that was also observed with Trastuzumab or the bispecific negative control (Tra-H-SSl).

[0373] Figure 32B is a bar graph representation of NK cell-mediated anti-tumor cytotoxicity data. CD16+NK-92 and CD16negparental NK-92 cells were separately co-cultured with CTFR- prelabelled HER2+HCC1954 tumor cells at a 2: 1 effector-to-target ratio and treated with 50 nM of Trastuzumab, Tra-H-monovalent-16-92-4_KiH or a negative control bispecific Tra-H- SS1. After 4 hours, anti -turn or cytotoxicity was measured by determining the percentage of dead CTFR+HCC1954 cells by flow cytometry and data were analyzed using GraphPad Prism. In co-cultures with CD16+NK-92 cells addition of Tra-H-monovalent-16-92-4_KiH substantially increased HCC1954 cell killing as compared to addition of Trastuzumab or Tra- H-SS1. However, in co-cultures with NK-92 cells lacking CD16 expression, addition of none of the biologies led to an increase in the percentage of dead HCC1954 cells that exceeded what was seen as background level in the absence of NK cells. This level is indicated in Figure 32B as a dotted line labelled “w / o NK”.

[0374] Figure 32C is a bar graph representation of IFN-y release showing the impact of Tra-H- monovalent-16-92-4_KiH and control molecules on cytokine secretion by CD16+or CD16negNK-92 cells. NK cells were cultured in the presence of 50 nM of the respective antibodies with CTFR-prelabeled HER2+HCC1954 tumor cells at a 2: 1 effector-to-target cell ratio for 4 hours. Then IFN-y levels in the culture supernatant were determined by MSD measurements. In agreement with the NK cell degranulation and NK cell-mediated cytotoxicity assessments, a significant increase of IFN-y release was triggered by Tra-H-monovalent-16-92-4_KiH only from CD16+NK-92 cells but not from parental NK-92 cells lacking CD 16 expression.

[0375] Example 4: Enhancement of ADCC through afucosylation

[0376] Figure 33 shows non-linear regression representations of concentration response analysis data showing the impact of fucosylated and afucosylated Fc-competent CD2 bi specifics on CD16+NK-92 cells over a 4h culture period in the absence of tumor targets.

[0377] Antibodies that are afucosylated show enhanced binding to the Fey receptor IIIA (CD 16a) with accompanying enhanced ADCC activity. Among others, afucosylation of antibodies can be achieved by producing the antibodies in media supplemented with analogs of L-fucose that inhibit antibody fucosylation. Afucosylated versions of Trastuzumab and the Trastuzumabbased CD2 bispecific Tra-H-monovalent-16-92-4_KiH were produced in this way and characterized for enhanced functional activity. To evaluate the potential of the afucosylated Fc- competent, heavy chain fused CD2 bispecific to mediate unwanted fratricide activity (see Figure 28), NK cell degranulation and IFN-y production in the absence of target tumor cells was investigated. CD16+NK-92 cells were cultured for 4 hours in the presence of serially titrated concentrations of CD2 bispecifics and control antibodies. NK cell degranulation was monitored as appearance of LAMP 1 (CD 107a) on the cell surface by flow cytometry and the IFN-y secreted into the culture was determined by MSD using the manufacturer’s protocol. In line with previous experimental data, the negative control bispecific Tra-H-SSl caused no NK cell degranulation and mediated no IFN-y release by the NK cells. Siplizumab, used as positive control for fratricide activity induced potent dose-dependent NK cell degranulation (Emax = 35%) and IFN-y release due to fratricide also in line with previous data. In contrast to Siplizumab, the afucosylated version of the CD2 bispecific Tra-H-monovalent-16-92-4_KiH mediated little NK cell degranulation due to fratricide with Emax values plateauing at 10% accompanied by very little IFN-y release. The fucosylated monovalent CD2 bispecific Tra-H- monovalent-16-92-4_KiH induced even less NK cell degranulation mediating small increases in CD107a exposure (~5%) only at the highest dose used and causing negligible IFN-y release. These data demonstrate that even the afucosylated version of the monovalent CD2 bispecific Tra-H-monovalent-16-92-4_KiH largely avoids fratricide. This may be due to the inherent property of the scFv moiety derived from 16-92-4 to mediate only very limited fratricide even when used in a bivalent format as C-terminal fusion to the heavy chain of a tumor targeting antibody.

[0378] Figure 34 shows non-linear regression representations of concentration response analysis data showing the impact of Fc-competent CD2 bispecifics, fucosylated and afucosylated, on a coculture of CD16+NK-92 cells with HCC1954 tumor cells over a 4h culture period. NK cell degranulation, monitored as CD 107a accumulation at the cell surface, anti -tumor cytotoxicity and IFN-y release were measured. Impact of afucosylated and normal fucosylated antibodies were compared. CD16+NK-92 cells were co-cultured with HER2+tumor cells in the presence of serial dilutions of the fucosylated or afucosylated versions of the CD2 bispecific Tra-H- monovalent-16-92-4_KiH, fucosylated or afucosylated versions of Trastuzumab or a fucosylated negative control bispecific Tra-H-SSl. Briefly, CD16+NK-92 cells were cocultured with CTFR (Cell Trace Far Red)-prelabeled HER2+HCC1954 tumor cells at an effector-to-target ratio of 2: 1 in the presence of increasing concentrations of the various heavy chain fused Trastuzumab-based bispecifics. After 4 hours, three separate readouts were collected from the experiment. Exposure of CD 107a by the NK cells was monitored by flow cytometry as measurement of NK cell degranulation. The percentage of dead CTFR-positive HCC1954 cells was determined by flow cytometry as readout for anti -tumor cytotoxicity. IFN- y released into the culture media was quantified by MSD using the manufacturer’s protocol as measure of cytokine release by activated NK cells. Expectedly, a dose-dependent increase in NK cell degranulation, anti-tumor cytotoxicity (ADCC) and IFN-y release were observed for Trastuzumab and all Trastuzumab-based bispecifics. However, the afucosylated version of Trastuzumab showed marked enhancement in NK cell anti-tumor response in potency and efficacy in all three readouts. Impressively, normal fucosylated Tra-H-monovalent- 16-92- 4_KiH matched the impact of afucosylated Trastuzumab in ADCC and was more potent in both NK degranulation and IFN-y release readouts. Tra-H-monovalent- 16-92-4_KiH_Afucosylated was the most potent molecule in both NK degranulation and IFN-y release readouts and showed comparable anti -turn or cytotoxicity to the afucosylated version of Trastuzumab. These data show that appending the monovalent CD2 binding moiety of antibody 16-92-4 to Trastuzumab, further enhances the anti-tumor response even beyond the effect achieved by afucosylation. Tra-H-monovalent- 16-92-4_KiH_Afucosylated has best-in-class ADCC properties.

[0379] Figure 35 shows non-linear regression representation of anti -turn or cytotoxicity mediated by Fc-competent CD2 bispecifics, fucosylated and afucosylated, in a co-culture of primary human PBMCs with CTFR-prelabeled HER2+HCC1954 tumor cells.

[0380] To validate the results obtained with CD16+NK92 cells, comparable assays as in Figure 34 were performed using PBMCs as source of effector NK cells. Human PBMCs were depleted of macrophages by cultivating them on tissue culture plates overnight and subsequently harvesting non-adherent cells. HCC1954 breast cancer cells that were stably transfected to express the far- red fluorescent protein mKate2 were co-cultured with the PBMCs at an effector-to-target cell ratio of 2: 1. Co-cultures were treated with serial dilutions of the fucosylated or afucosylated versions of the CD2 bispecific Tra-H-monovalent- 16-92-4_KiH, fucosylated or afucosylated versions of Trastuzumab or a fucosylated negative control bispecific Tra-H-SSl. After a 4-day incubation, anti-tumor cytotoxicity was determined by calculating area under curve values (AUC) of the mKate2+HCC1954 cell growth curves and normalizing to HCC1954 cell growth in the untreated control using the following formula:

[0381] % Cytotoxicity = (AUC experimental value / AUC untreated control) * 100

[0382] A dose-dependent anti-tumor cytotoxicity (ADCC) was observed for Trastuzumab and all Trastuzumab-based bispecifics. However, the afucosylated version of Trastuzumab showed marked enhancement in the efficacy of ADCC but with no meaningful improvement in potency. Remarkably, normal fucosylated Tra-H-monovalent-16-92-4_KiH matched the efficacy of fucosylated Trastuzumab but showed enhanced potency by ~8-fold. Tra-H-monovalent-16-92- 4_KiH_Afucosylated was again the most potent molecule in this assay exceeding the potency of the normal fucosylated Tra-H-monovalent-16-92-4_KiH by 4-fold. Taken together, these data validate the data generated using CD 16+ NK92 as effector cells and confirm that Tra-H- monovalent-16-92-4_KiH_Afucosylated is a best-in-class ADCC molecule.

[0383] Example 5: Impact of endogenous CD2-CD58 interaction on activity of Bifunctional T cell co-stimulators (BiTco) using CD58-knockout HCC1954 tumor cells.

[0384] Figure 39 shows dot-plot representations of flow cytometry-based analysis of antigen expression by wildtype and CD58-knockout HCC1954 cells. A CD58-knockout pool of HCC1954 cells was generated by standard CRISPR-Cas9 methodology. Successful knockout of CD58 was evaluated by flow cytometry using a commercially available fluorochrome- labelled CD58 antibody. Similarly, expressions of HER2 and EpCAM were evaluated using commercial antibodies. Absence of CD58 expression in the CD58-knockout pool was confirmed by flow cytometry. The expression of HER2 and EpCAM was identical in both wildtype and CD58-knockout HCC1954 cells.

[0385] Figure 40 is a non-linear regression representation of data comparing cytotoxicity against wildtype and CD58-knockout target cells using macrophage-depleted human PBMCs as effector cells. Human PBMCs were depleted of macrophages by cultivating them on tissue culture plates overnight and subsequently harvesting non-adherent cells. Macrophage-depleted PBMCs were co-cultured at an effector-to-target cell ratio of 2: 1 either with wildtype or CD58- knockout HCC1954 breast cancer cells (both EpCAM+and HER2+and also expressing the far- red fluorescent protein mKate2). Co-cultures were treated with serially diluted EpCAM*CD3 BiTE. Cell growth of the mKate2+HCC1954 cells was tracked using the Incucyte system over a 4-day culture period. Anti-tumor cytotoxicity was determined by calculating area under curve values of the mKate2+HCC1954 cell growth curves and normalizing to HCC1954 cell growth in the untreated control as described in Figure 19. The anti -tumor cytotoxicity mediated by the EpCAM*CD3 BiTE showed a 6-fold lower potency in the CD58-knockout HCC1954 cells as compared to the wildtype HCC1954 cells. This indicates that the endogenous CD58-CD2 axis contributes to the overall cytotoxicity of the EpCAM*CD3 BiTE. Figure 41 shows non-linear regression representations of anti -tumor cytotoxicity data using macrophage-depleted human PBMCs as effector cells and CD58-knockout HCC1954 as target cells. Human PBMCs were depleted of macrophages by cultivating them on tissue culture plates overnight and subsequently harvesting non-adherent cells. Macrophage-depleted PBMCs were co-cultured at an effector-to-target cell ratio of 2: 1 with CD58-knockout mKate2+HCC1954 breast cancer cells (EpCAM+and HER2+). Co-cultures were treated with serially diluted EpCAM*CD3 BiTE in the presence of 50 nM fixed concentration of either the CD2 bispecifics Tra-L-16-92-4_AEASS, the positive control CD28 bispecific Tra-L-9-3_AEASS or the mesothelin-binding negative control bispecific Tra-L-SS1_AEASS. Cell growth of the mKate2+HCC1954 cells was tracked using the Incucyte system over a 4-day culture period. Anti-tumor cytotoxicity was determined by calculating area under curve values of the mKate2+HCC1954 cell growth curves and normalizing to HCC1954 cell growth in the untreated control as described in Figure 19.

[0386] The CD2 bispecific Tra-L-16-92-4_AEASS enhanced anti-tumor toxicity comparably to the positive control CD28 bispecific. These data demonstrates that the co-stimulatory activity of the CD2 bispecific Tra-L-16-92-4_AEASS is independent of the endogenous CD58-CD2 interaction. Further, in the absence of the endogenous CD2-CD58 interaction (e. g. due to downregulation of CD58 on target tumor cells) treatment with CD2 bispecific Tra-L-16-92-4 can overcome this defect.

[0387] Example 6: Bifunctional T cell costimulators (BiTco) against targets for B cell malignancies

[0388] Figure 42 shows non-linear regression representations of anti-tumor cytotoxicity data using macrophage-depleted human PBMCs as effector cells and Ramos B cell line as target cells. Human PBMCs were depleted of macrophages by cultivating them on tissue culture plates overnight and subsequently harvesting non-adherent cells. Macrophage-depleted PBMCs were co-cultured at an effector-to-target cell ratio of 2: 1 with mKate2+Ramos B cells (CD19+, CD20+and CD22+). Co-cultures were treated with serially diluted CD20*CD3 T cell engager mosunetuzumab, in the presence of 50 nM fixed concentration of either the CD2 bispecifics (Epra-L-16-92-4_AEASS and FMC63-L-16-92-4 AEASS), corresponding positive control CD28 bispecifics (Epra-L-9-3_AEASS and FMC63-L-9-3 AEASS) or corresponding mesothelin-binding negative control bispecifics (Epra-L-SS1_AEASS and FMC63-L- SS1 AEASS). Cell growth of the mKate2+Ramos cells was tracked using the Incyte system over a 4-day culture period. Anti-tumor cytotoxicity was determined by calculating area under curve values (AUC) of the mKate2+Ramos cell growth curves and normalizing to Ramos cell growth in the untreated control as described in Figure 19A above. Both CD2 bispecifics Epra- L-16-92-4 AEASS and FMC63-L-16-92-4 AEASS markedly enhanced anti -turn or toxicity comparably to their corresponding positive control CD28 bispecific mediating ~80-fold and >300-fold increase in potency, respectively. These data demonstrate a wider application of CD2 bifunctional T cell costimulators across different tumor-associated antigens, tumor cell types and cancer indications.

[0389] Example 7: Exemplary alternative format for Bifunctional T cell costimulators (BiTco)

[0390] Figure 43 is a schematic drawing of an exemplary alternative format for a tumor-targeted CD2 bispecific designed to provide a co-stimulatory signal to effector T cells and distinct from the format presented in Figure 12 above. Here, bivalent binding to CD2 is mediated by the Fab domain of the CD2-targeted IgG. Fusing disulfide-stabilized single chain variable fragment (scFv) moieties directed against the tumor-associated antigen (TAA) at the C-termini of both heavy chains of the IgG provides tumor specificity. The star symbols represent mutations that render the Fc moiety silent to prevent CD 16a engagement. Another arrow points to the disulfide stabilization of the anti-TAA scFv. CD2-targeted bifunctional T cell co-stimulator bispecifics as schematically represented in Figure 43 were transiently expressed in CHO cells by standard methods and purified by Protein A and size exclusion chromatography (SEC) to high monomerity. Low endotoxin content was verified with a commercial kit.

[0391] Figure 44 shows non-linear regression representations of anti-tumor cytotoxicity data using BiTco molecules in two different formats. Co-cultures of primary human T cells and BT474 target cells at an effector-to-target cell ratio of 2: 1 were treated with serially diluted EpCAM*CD3 BiTE in the presence of 10 nM fixed concentration of CD2 bispecifics in different formats (Tra-L-16-92-4_AEASS and 16-92-4-H-Tra_AEASS), the positive control CD28 bispecific Tra-L-9-3_AEASS or the RSV-binding negative control bispecific Tra-L- Pali AEASS. After a 2-day co-culture, anti -turn or cytotoxicity was determined by LDH release assay as in Figure 17A and data were analyzed using GraphPad Prism. Percentage cytotoxicity was normalized to the cytotoxicity by 10 nM EpCAM*CD3 BiTE, which represents the concentration that generally mediates highest efficacy of the BiTE.CD2 bispecifics in both formats Tra-L-16-92-4_AEASS and 16-92-4-H-Tra_AEASS enhanced anti-tumor toxicity comparably to the CD28 control bispecific Tra-L-9-3_AEASS, in a signal- 1 dependent manner. These data demonstrate that co-stimulation of cytotoxicity can be achieved for different formats of CD2 BiTco molecules designed based on the CD2 antibody 16-92-4. Herein, co-stimulation was demonstrated for CD2 BiTco molecules with the CD2 binding moiety either as scFv or IgG.

[0392] Figure 45 shows non-linear regression representations of T cell activation data as visualized by CD25 upregulation using BiTco molecules in two different formats. Co-cultures of primary human T cells and BT474 target cells at an effector-to-target cell ratio of 2: 1 were treated with serially diluted EpCAM*CD3 BiTE in the presence of 10 nM fixed concentration of CD2 bispecifics in different formats (Tra-L-16-92-4_AEASS and 16-92-4-H-Tra_AEASS), the positive control CD28 bispecific Tra-L-9-3_AEASS or the RSV-binding negative control bispecific Tra-L-Pali_AEASS. After 2-days of co-culture, T cell activation was determined by staining for CD25 and counterstaining for CD4 and CD8 T cell subsets. Data were collected by flow cytometry and analyzed using GraphPad Prism. CD2 bispecifics in both formats Tra-L- 16-92-4 AEASS and 16-92-4-H-Tra_AEASS enhanced CD25 upregulation in T cells comparably to the CD28 control bispecific Tra-L-9-3_AEASS and in a signal-1 dependent manner. These data demonstrate that co-stimulation of T cell activation can be achieved for different formats of CD2 BiTco molecules based on the CD2 antibody 16-92-4. Herein, costimulation of T cell activation was demonstrated for CD2 BiTco molecules with the CD2 binding moiety either as scFv or IgG.

[0393] References

[0394] • Kohler, G. & Milstein, C. (1975): Continuous cultures of fused cells secreting antibody of predefined specificity. In: Nature. Bd. 256, S. 495-497.

[0395] • Jonsson and Malmquist, Advances in Biosensors, 2:291-336 (1992)

[0396] • Wu et al. Proc. Natl. Acad. Sci. USA, 95:6037-6042 (1998)

[0397] • Banik, SSR; Doranz, BJ (2010). "Mapping complex antibody epitopes". Genetic Engineering & Biotechnology News. 3 (2): 25-8

[0398] • DeLisser, HM (1999). Epitope mapping. Methods Mol Biol. 96. pp. 11-20 • Finco et al, Comparison of competitive ligand-binding assay and bioassay formats for the measurement of neutralizing antibodies to protein therapeutics. J Pharm Biomed Anal. 2011 Jan 25;54(2):351-8.

[0399] • Deng et al., Enhancing antibody patent protection using epitope mapping information MAbs. 2018 Feb-Mar; 10(2): 204-209

[0400] • Huston et al., Cell Biophysics, 22: 189-224 (1993);

[0401] • Phickthun and Skerra, Meth. Enzymol., 178:497-515 (1989) and in Day, E. D., Advanced Immunochemistry, Second Ed., Wiley-Liss, Inc., New York, N.Y. (1990)

[0402] • Harding, The immunogenicity of humanized and fully human antibodies. MAbs. 2010 May-Jun; 2(3): 256-265.

[0403] • Eylenstein, et al, Molecular basis of in vitro affinity maturation and functional evolution of a neutralizing anti -human GM-CSF antibody, mAbs, 8: 1, 176-186 (2016)

[0404] • Kabat et al., J. Biol. Chem. 252, 6609-6616 (1977)

[0405] • Chothia et al., J. Mol. Biol. 196:901-917 (1987)

[0406] • MacCallum et al., J. Mol. Biol. 262:732-745 (1996)

[0407] • van der Horst HJ, Nijhof IS, Mutis T, Chamuleau MED. Fc-Engineered Antibodies with Enhanced Fc-Effector Function for the Treatment of B-Cell Malignancies. Cancers (Basel). 2020 Oct 19; 12(10):3041.

[0408] • McCall AM, Adams GP, Amoroso AR, Nielsen UB, Zhang L, Horak E, Simmons H, Schier R, Marks JD, Weiner LM. Isolation and characterization of an anti-CD16 single-chain Fv fragment and construction of an anti-HER2 / neu / anti-CD16 bispecific scFv that triggers CD16-dependent tumor cytolysis. Mol Immunol. 1999 May;36(7):433-45.

[0409] • Kuwahara A, Nagai K, Nakanishi T, Kumagai I, Asano R. Functional Domain Order of an Anti-EGFR x Anti-CD16 Bispecific Diabody Involving NK Cell Activation. Int J Mol Sci. 2020 Nov 24;21(23):8914

[0410] • Nikkhoi SK, Li G, Eleya S, Yang G, Vandavasi VG, Hatefi A. Bispecific killer cell engager with high affinity and specificity toward CD 16a on NK cells for cancer immunotherapy. Front Immunol. 2023 Jan 6; 13: 1039969.

[0411] • Leitner J, Herndler-Brandstetter D, Zlabinger GJ, Grubeck-Loebenstein B, Steinberger P. CD58 / CD2 Is the Primary Costimulatory Pathway in Human CD28-CD8+ T Cells. J Immunol. 2015 Jul 15;195(2):477-87. • Duan Z, Ho M. T-Cell Receptor Mimic Antibodies for Cancer Immunotherapy. Mol Cancer Then 2021 Sep;20(9):1533-1541.

[0412] • Sterner, R.C., Sterner, R.M. CAR-T cell therapy: current limitations and potential strategies. Blood Cancer J. 11, 69 (2021).

[0413] • Menon, A.P.; Moreno, B.; Meraviglia-Crivelli, D.; Nonatelli, F.; Villanueva, H.; Barainka, M.; Zheleva, A.; van Santen, H.M.; Pastor, F. Modulating T Cell Responses by Targeting CD3. Cancers 2023, 15, 1189.

[0414] • Brinkmann U, Kontermann RE. The making of bispecific antibodies. MAbs. 2017 Feb / Mar;9(2): 182-212

[0415] • Nakamura K, Nakayama M, Kawano M, Ishii T, Harigae H, Ogasawara K. NK-cell fratricide: Dynamic crosstalk between NK and cancer cells. Oncoimmunology. 2013 Nov 1;2(11)

[0416] • Du N, Guo F, Wang Y, Cui J. NK Cell Therapy: A Rising Star in Cancer Treatment. Cancers (Basel). 2021 Aug 17; 13(16):4129.

[0417] • Shin E, Bak SH, Park T, Kim JW, Yoon S-R, Jung H and Noh J-Y (2023) Understanding NK cell biology for harnessing NK cell therapies: targeting cancer and beyond. Front. Immunol. 14: 1192907.

[0418] • Dall'Acqua WF, Woods RM, Ward ES, Palaszynski SR, Patel NK, Brewah YA, et al. (November 2002). "Increasing the affinity of a human IgGl for the neonatal Fc receptor: biological consequences". Journal of Immunology. 169 (9): 5171-5180.

[0419] • Baeuerle PA, Reinhardt C: Bispecific T-cell engaging antibodies for cancer therapy. In: Cancer Res. 69. Jahrgang, Nr. 12, Juni 2009, S. 4941-4

[0420] • Harmsen MM, De Haard HJ (November 2007). "Properties, production, and applications of camelid single-domain antibody fragments". Applied Microbiology and Biotechnology. 77 (1): 13-22.

[0421] • Eck SC, Chang D, Wells AD, Turka LA. Differential down-regulation of CD28 by B7-1 and B 7-2 engagement. Transplantation. 1997 Nov 27;64(10): 1497-9

[0422] • Okeley, Nicole M et al. “Development of orally active inhibitors of protein and cellular fucosylation.” Proceedings of the National Academy of Sciences of the United States of America vol. 110,14 (2013): 5404-9.

[0423] • Li, J et al. “Ligand-induced conformational change within the CD2 ectodomain accompanies receptor clustering: implication for molecular lattice formation.” Journal of molecular biology vol. 263,2 (1996): 209-26. • Damschroder, Melissa M et al. “Analysis of human and primate CD2 molecules by protein sequence and epitope mapping with anti-human CD2 antibodies.” Molecular immunology vol. 41,10 (2004): 985-1000.

[0424] Sequences

[0425] The following sequences form part of the disclosure of the present application. A WIPO ST 26 compatible electronic sequence listing is provided with this application, too. For the avoidance of doubt, if discrepancies exist between the sequences in the following table and the electronic sequence listing, the sequences in this table shall be deemed to be the correct ones.

[0426] In some cases, signal peptides may be encompassed in the reproduced sequences. In such case, the sequences shall be deemed disclosed with and without signal peptides. A readily available tool to identify signal peptides in a given protein sequence is SignalP - 6.0 provided by Dansk Technical University under http s : / / servi ces . healthtech , dtu . dk / servi ce . php ? Si gnalP . The same applies to His tags or C-Myc tags or NLS sequences, if existing.

[0427] Table 9: Sequence listing

[0428]

Claims

What is claimed:

1. An antibody that binds to CD2, or a target-binding fragment or derivative of such antibody, which a) comprises a set of three heavy chain and three light chain complementarity determining regions (CDR) comprised in the one of the following heavy chain / light variable domain sequence pairs• SEQ ID NOs 1 and 2;• SEQ ID NOs 3 and 4;• SEQ ID NOs 5 and 6; and / or• SEQ ID NOs 7 and 8; b) comprises a set of three heavy chain and three light chain complementarity determining regions (CDR) selected from the following sets• SEQ ID NOs 9 - 14;• SEQ ID NOs 15 - 20;• SEQ ID NOs 21 - 26; and / or• SEQ ID NOs 27 - 32; c) comprises the set of heavy chain / light chain complementarity determining regions (CDR) of b), with the proviso that at least one of the CDRs has up to 3 amino acid substitutions relative to the respective SEQ ID NOs, and / or d) comprises the set of heavy chain / light chain complementarity determining regions (CDR) of b) or c), with the proviso that at least one of the CDRs has a sequence identity of > 66 % to the respective CDRs comprised in the SEQ ID NOs, wherein the CDRs are embedded in a suitable protein framework, preferably a variable domain framework, so as to be capable to bind to CD2.

2. The antibody or fragment according to claim 1, which comprises a) the heavy chain / light chain variable domain (HCVD / LCVD) pairs set forth in the following pairs of SEQ ID NOs:• SEQ ID NOs 1 and 2;• SEQ ID NOs 3 and 4;• SEQ ID NOs 5 and 6; and / or• SEQ ID NOs 7 and 8; b) the heavy chain / light chain variable domains (HCVD / LCVD) pairs of a), with the proviso that• the HCVD has a sequence identity of > 80 % to the respective SEQ ID NO, and / or• the LCVD has a sequence identity of > 80 % to the respective SEQ ID NO, c) the heavy chain / light chain variable domains (VD) pairs of a) or b), with the proviso that at least one of the HCVD or LCVD has up to 10 amino acid substitutions relative to the respective SEQ ID NO, said antibody or fragment still being capable to bind to CD2.

3. An antibody that binds to CD2, or a target-binding fragment or derivative of such antibody, which(i) competes for binding to CD2 with a) an antibody according to any one of claims 1 - 2, or b) an antibody selected from clones 16-92-4, 16-92-4_h or 16-92-4_ho, or(ii) binds to essentially the same, or the same, region or epitope on CD2 as a) an antibody according to any one of claims 1 - 2, or b) an antibody selected from clones 16-92-4, 16-92-4_h or 16-92-4_ho, or4. A bi- or multifunctional molecule comprising at least: a) a first binding domain capable of binding to the CD2 antigen, or to a subdomain of CD2, and b) a second binding domain capable of specifically binding to a target antigen other than the CD2 antigen, optionally wherein the second binding domain binds to a cancer-associated target on a cancer target cell surface or to an immune checkpoint agonist on a cancer cell.

5. The bi- or multifunctional molecule according to claim 4wherein the first binding domain capable of binding the CD2 antigen is an anti-CD2 antibody, or a target binding fragment thereof, and / or wherein the second binding domain capable of specifically binding to a target antigen other than the CD2 antigen is an antibody, or a target binding fragment thereof.

6. The bi- or multifunctional molecule according to any one of claims 4 - 5, wherein the anti-CD2 antibody, upon binding to CD2, a) increases, stabilizes and / or enhances binding between CD2 and CD58, or b) reduces or inhibits the binding between CD2 and CD58.

7. The bi- or multifunctional molecule according to any one of claims 4 - 6, wherein the anti-CD2 antibody is an antibody according to any one of claims 1 - 3.

8. The bi- or multifunctional molecule according to any one of claims 4 - 7, which molecule further comprises a third domain capable of interacting with, or binding to, CD 16, which third domain is optionally at least one selected from the group consisting of• an anti-CD16 antibody or a target binding fragment thereof, or• an antibody Fc domain.

9. The bi- or multifunctional molecule according to any one of claims, 4 - 8 wherein a first binding domain capable of binding to the CD2 antigen is fused to a C-terminus of a CH3 domain of an Fc domain of the second binding domain, in case said second binding domain is an antibody or fragment or derivative that comprises at least one heavy chain and / or at least one Fc domain, directly or via a linker. wherein optionally only one first binding domain capable of binding to the CD2 antigen is fused to the C-terminus of one Fc domain only.

10. The bi- or multifunctional molecule according to any one of claims 4 - 9, wherein the two Fc domains are heterodimeric relative to one another, optionally wherein the two Fc domains are provided in knobs-into-holes configuration.

11. A combination comprising (a) the bi- or multifunctional molecule according to any one of claims 4 - 10 and (b) a therapeutic NK cell product.

12. The bi- or multifunctional molecule according to any one of claims 4 - 10, a) which molecule lacks a third domain capable of interacting with, or binding to, CD16, or comprises a third domain that has reduced interaction or binding capacity to CD1, optionally which molecule comprises an Fc domain which is Fc-silenced, and / or b) wherein two first binding domains capable of binding to the CD2 antigen are fused to the C-termini of the CL domain, in case said second binding domain is an antibody or fragment or derivative that comprises at least one CL domain, directly or via a linker13. A combination comprising (a) the bi- or multifunctional molecule according to any one of claims 4 - 10 or 12 and (b) a CD3 -directed T cell engager, a CAR T cell, an immune checkpoint inhibitor, or a bispecific T cell receptor based molecule.

14. The combination according to any one of claims 11 or 13 for (the manufacture of a medicament for) use in the treatment of a human or animal subject• being diagnosed for,• suffering from or• being at risk of developing cancer.

15. The combination or method according to any one of claims 11 or 13, wherein (a) is administered to the subject prior to, simultaneously with, or after, administration of (b), and / or wherein (a) and (b) are provided in separate units.

Images