The present invention relates to tumor targeted bispecific agonistic antigen binding molecules characterized by monovalent binding to CD28, methods for their production, pharmaceutical compositions containing these antibodies, and methods of using the same.

Description

86 parts

›SEQUENCE LISTING

This application contains a Sequence Listing which has been submitted electronically in ASCII format and is hereby incorporated by reference in its entirety. Said ASCII copy, created on Dec. 16, 2019, is named P35128-US_SeqListing.txt and is 958,930 bytes in size and updated by a file entitled P35128_SeqLisitng_Replacment_2, created on Aug. 8, 2022, which is 961,083 bytes in size.

›FIELD OF THE INVENTION

The present invention relates to tumor-targeted bispecific agonistic CD28 antigen binding molecules characterized by monovalent binding to CD28, methods for their production, pharmaceutical compositions containing these molecules, and their use as immunomodulators in the treatment of cancer.

›BACKGROUND · 1 of 3

Cancer immunotherapy is becoming an increasingly effective therapy option that can result in dramatic and durable responses in cancer types such as melanoma, non-small cell lung cancer and renal cell carcinoma. This is mostly driven by the success of several immune checkpoint blockades including anti-PD-1 (e.g. Keytruda, Merck; Opdivo, BMS), anti-CTLA-4 (e.g. Yervoy, BMS) and anti-PD-L1 (e.g. Tecentriq, Roche). These agents are likely to serve as standard of care for many cancer types, or as the backbone of combination therapies, however, only a fraction of patients (<25%) benefits from such therapies. Furthermore, various cancers (prostate cancer, colorectal cancer, pancreatic cancer, sarcomas, non-triple negative breast cancer etc.) present primary resistance to these immunomodulators. A number of reports indicate that the absence of pre-existing anti-tumor T cells contributes to the absence or poor response of some patients. In summary, despite impressive anti-cancer effects of existing immunotherapies, there is a clear medical need for addressing a large cancer patient population and for developing therapies that aim to induce and enhance novel tumor-specific T cell responses.

CD28 is the founding member of a subfamily of costimulatory molecules characterized by paired V-set immunoglobulin superfamily (IgSF) domains attached to single transmembrane domains and cytoplasmic domains that contain critical signaling motifs (Carreno and Collins, 2002). Other members of the subfamily include ICOS, CTLA-4, PD1, PD1H, TIGIT, and BTLA (Chen and Flies, 2013). CD28 expression is restricted to T cells and prevalent on all naïve and a majority of antigen-experienced subsets, including those that express PD-1 or CTLA-4. CD28 and CTLA-4 are highly homologous and compete for binding to the same B7 molecules CD80 and CD86, which are expressed on dendritic cells, B cells, macrophages, and tumor cells (Linsley et al., 1990). The higher affinity of CTLA-4 for the B7 family of ligands allows CTLA-4 to outcompete CD28 for ligand binding and suppress effector T cells responses (Engelhardt et al., 2006). In contrast, PD-1 was shown to inhibit CD28 signaling by in part dephosphorylating the cytoplasmic domain of CD28 (Hui et al., 2017). Ligation of CD28 by CD80 or CD86 on the surface of professional antigen-presenting cells is strictly required for functional de novo priming of naïve T cells, subsequent clonal expansion, cytokine production, target cell lysis, and formation of long-lived memory. Binding of CD28 ligands also promotes the expression of inducible co-stimulatory receptors such as OX-40, ICOS, and 4-1BB (reviewed in Acuto and Michel, 2003). Upon ligation of CD28, a disulfide-linked homodimer, the membrane proximal YMNM motif and the distal PYAP motif have been shown to complex with several kinases and adaptor proteins (Boomer and Green, 2010). These motifs are important for the induction of IL2 transcription, which is mediated by the CD28-dependent activation of NFAT, AP-1, and NFκB family transcription factors (Fraser et al., 1991) (June et al., 1987) (Thompson et al., 1989). However, additional poorly characterized sites for phosphorylation and ubiquitination are found within the cytoplasmic domain of CD28. As reviewed by (Esensten et al., 2016), CD28-initiated pathways have critical roles in promoting the proliferation and effector function of conventional T cells. CD28 ligation also promotes the anti-inflammatory function of regulatory T cells. CD28 co-stimulates T cells by in part augmenting signals from the T cell receptor, but was also shown to mediate unique signaling events (Acuto and Michel, 2003; Boomer and Green, 2010; June et al., 1987). Signals specifically triggered by CD28 control many important aspects of T cell function, including phosphorylation and other post-translational modifications of downstream proteins (e.g., PI3K mediated phosphorylation), transcriptional changes (eg. Bcl-xL expression), epigenetic changes (e.g. IL-2 promoter), cytoskeletal remodeling (e.g. orientation of the microtubule-organizing center) and changes in the glycolytic rate (e.g. glycolytic flux). CD28-deficient mice have reduced responses to infectious pathogens, allograft antigens, graft-versus-host disease, contact hypersensitivity and asthma (Acuto and Michel, 2003). Lack of CD28-mediated co-stimulation results in reduced T cell proliferation in vitro and in vivo, in severe inhibition of germinal-centre formation and immunoglobulin isotype-class switching, reduced T helper (Th)-cell differentiation and the expression of Th2-type cytokines. CD4-dependent cytotoxic CD8+ T-cell responses are also affected. Importantly, CD28-deficient naïve T cells showed a reduced proliferative response particularly at lower antigen concentrations. A growing body of literature supports the idea that engaging CD28 on T cells has anti-tumor potential. Recent evidence demonstrates that the anti-cancer effects of PD-L1/PD-1 and CTLA-4 checkpoint inhibitors depend on CD28 (Kamphorst et al., 2017; Tai et al., 2007). Clinical studies investigating the therapeutic effects of CTLA-4 and PD-1 blockade have shown exceptionally promising results in patients with advanced melanoma and other cancers. In addition, infusion of genetically engineered T cells expressing artificial chimeric T cell receptors comprising an extracellular antigen recognition domain fused to the intracellular TCR signaling domains (CD3z) and intracellular co-stimulatory domains (CD28 and/or 4-1BB domains) has shown high rates and durability of response in B cell cancers and other cancers.

CD28 agonistic antibodies can be divided into two categories: (i) CD28 superagonistic antibodies and (ii) CD28 conventional agonistic antibodies. Normally, for the activation of naïve T cells both engagement of the T cell antigen receptor (TCR, signal 1) and costimulatory signaling by CD28 (signal 2) is required. CD28 Superagonists (CD28SA) are CD28-specific monoclonal antibodies, which are able to autonomously activate T cells without overt T cell receptor engagement (Hunig, 2012). In rodents, CD28SA activates conventional and regulatory T cells. CD28SA antibodies are therapeutically effective in multiple models of autoimmunity, inflammation and transplantation. However, a phase I study of the human CD28SA antibody TGN1412 resulted in a life-threatening cytokine storm in 2006. Follow-up studies have suggested that the toxicity was caused by dosing errors due to differences in the CD28 responsiveness of human T cells and T cells of preclinical animal models. TGN1412 is currently being re-evaluated in an open-label, multi-center dose escalation study in RA patients and patients with metastatic or unresectable advanced solid malignancies. CD28 conventional agonistic antibodies, such as clone 9.3, mimic CD28 natural ligands and are only able to enhance T cell activation in presence of a T cell receptor signal (signal 1). Published insights indicate that the binding epitope of the antibody has a major impact on whether the agonistic antibody is a superagonist or a conventional agonist (Beyersdorf et al., 2005). The superagonistic TGN1412 binds to a lateral motif of CD28, while the conventional agonistic molecule 9.3 binds close to the ligand binding epitope. As a consequence of the different binding epitopes, superagonistic and conventional agonistic antibodies differ in their ability to form linear complexes of CD28 molecules on the surface of T cells. Precisely, TGN1412 is able to efficiently form linear arrays of CD28, which presumably leads to aggregated signaling components which are sufficient to surpass the threshold for T cell activation. The conventional agonist 9.3, on the other hand, leads to complexes which are not linear in structure. An attempt to convert conventional agonistic binders based on the 9.3 clone has been previously published (Otz et al., 2009) using a recombinant bi-specific single-chain antibody directed to a melanoma-associated proteoglycan and CD28. The reported bispecific single chain antibody was reported to exert “supra-agonistic” activity despite the use of a conventional CD28 agonistic binder 9.3, based in the intrinsic tendency of bispecific single chain antibodies to form multimeric constructs.

›BACKGROUND · 2 of 3

It has been found that a better T cell activation is achieved when limiting amounts of anti-CD3 bispecific antibodies, i.e. T cell bispecific antibodies (TCBs) such as CEA-TCB, are combined with agonistic anti-CD28 molecules. Given, that CD28 is expressed at baseline on T cells in various tumor indications (Lavin et al., 2017; Tirosh et al., 2016, Zheng et al., 2017) and activation of CD28 signaling enhances T cell receptor signals, the combination of a TCB molecule with a tumor-targeted CD28 molecule is expected to act synergistically to induce strong and long-lasting anti-tumor responses. Thus, we herein describe novel tumor-targeted agonistic CD28 molecules which display synergy with TCBs and require CD28 binding monovalency for strict tumor target dependence in the presence of TCB signals.

Immunotherapy of Solid Tumors

The treatment of solid tumors is an ongoing challenge, with little advancements seen over the last years. Typically, the treatment will be a combination of surgery and chemotherapy and/or radiotherapy. While quite a few new treatment modalities have been developed recently, there is still a need for further improvements, to increase survival rates of patients suffering from solid tumors, and improve their quality of life. Solid tumors rarely express one tumor specific antigen. For most solid tumors, it is more common to find a tumor associated antigen (TAA) that is enriched on tumors but also expressed at very low levels on normal tissues. The TAA is preferably presented on the surface of the solid tumor cell or on a cell of the tumor stroma. This is the case for many frequently targeted TAAs for solid tumors, including Fibroblast Activation Protein (FAP), Carcinoembryonic Antigen (CEA), Folate receptor alpha (FolR1), Melanoma-associated Chondroitin Sulfate Proteoglycan (MCSP), Epidermal Growth Factor Receptor (EGFR), human epidermal growth factor receptor 2 (HER2) and p95HER2. Further TAAs include HER3, EpCAM, TPBG (5T4), mesothelin, MUC1, and PSMA. Bispecific agonistic CD28 antigen binding molecule comprising an antigen binding domain that specifically binds to a tumor associated antigen will thus be directed primarily to the tumor surface or tumor microenvironment and will specifically activate T cells in the proximity of the tumor while a systemic activation may be avoided.

Rationale for Targeting CD28 Agonism to B Cell Malignancies

Non-Hodgkin's lymphoma (NHL) is one of the leading causes of cancer death in the United States and in Europe. Follicular lymphoma (FL) is indolent in its course and has a slow rate of evolution, with a median survival of 8 to 10 years; patients in advanced clinical stages usually are not curable. Likewise, in 2% to 3% of patients per year, the phenotype of FL can transform into an aggressive, large cell lymphoma, a critical event in the course of the disease and one that is associated with increased lymphoma-related mortality. Mantle cell lymphoma and diffuse large B cell lymphoma (DLBCL) are more aggressive and, if untreated, have a median survival rate of only 6 months. The lack of curative outcomes for many patients with both indolent and aggressive NHL subtypes remains an unmet medical need, despite significant advances in immune-therapeutics that have extended progression free-survival times. During the past several years, significant prolonged survival in DLBCL has been observed, particularly with the addition of the anti-CD20 monoclonal antibody, rituximab (Rittman®, MabThera®) to intensive cytotoxic chemotherapeutic regimens. However, despite conventional treatment for previously untreated DLBCL is curative in intent, the majority of patients will eventually relapse. Likewise, advanced FL remains largely incurable by current SoC and is characterized by repeated relapses and progressively shorter remissions. Currently, many new generation monoclonal antibodies are in different preclinical and clinical phases of assessment to further improve the outcome of NHL patients and overcome mechanisms of rituximab resistance. High dose chemotherapy with autologous stem cell support or allogeneic stem cell transplantation provides a curative option for only a minority (10%) of patients with relapsed/refractory (r/r) DLBCL and is associated with substantial treatment-related mortality. Other approaches for the treatment of NHL currently in development include molecular targeted compounds like venetoclax and BET-inhibitors. Recently approved novel agents include lenalidomide, idelalisib, and copanlisib. Chimeric antigen receptor (CAR) T-cell therapy has been approved for the treatment of aggressive forms of r/r B-NHL, but this therapy is available only in limited settings and can be associated with fatal neurologic events and cytokine release syndrome (CRS). Bispecific antibody constructs redirecting the lysis of cytotoxic cells to malignant B cells are currently in development and have shown very promising efficacy against NHL. Chemotherapy-free treatments are envisioned for the future of NHL and will likely be based on bispecific antibodies or chimeric antigen receptor T cells (CAR T cells). A CD28 agonist targeted against a B cell surface antigen in combination with immunotherapy shall enhance the survival and/or cure rates for patients with B cell malignancies, without compromising their quality of life.

B Cell Surface Antigens as a Target for B Cell Malignancies

TAAs related to B cell malignancies are B cell surface antigens. The human CD19 antigen is a 95 KDa transmembrane glycoprotein belonging to the immunoglobulin superfamily. CD19 is classified as a type I transmembrane protein, with a single transmembrane domain, a cytoplasmic C-terminus, and extracellular N-terminus. In normal cells, it is the most ubiquitously expressed protein in the B lymphocyte lineage. CD19 expression is maintained in B-lineage cells that have undergone neoplastic transformation, and therefore CD19 is useful in diagnosis of leukemias and lymphomas using monoclonal antibodies (mAbs) and flow cytometry, so does the CD20 antigen. Because B lineage leukemias and lymphomas rarely lose CD19 expression, and because it is not expressed in the pluripotent stem cell, it has become the target for a variety of immunotherapeutic agents, including immunotoxins. CD79 is the signaling component of the B-cell receptor consisting of a covalent heterodimer containing CD79a (Igα, mb-1) and CD79b (Igβ, B29). CD79a and CD79b each contain an extracellular immunoglobulin (Ig) domain, a transmembrane domain, and an intracellular signaling domain, an immunoreceptor tyrosine-based activation motif (ITAM) domain, like other signalling proteins such as CD3 or activatory Fcγ receptor. CD79a and CD79b are thus transmembrane proteins that compose the signalling subunits of the B cell receptor (BCR). CD79b is a 39 KDa protein exclusively expressed on B cells and, in cooperation with CD79a, initiates the signal transduction cascade downstream of the BCR, which leads to internalization of the BCR complex, its translocation to the endosomes, and antigen presentation. In B cells, antigen-induced BCR clustering triggers tyrosine phosphorylation of the ITAM of CD79a and CD79b by Src kinases. This leads to the recruitment and activation of an array of effector molecules belonging to the BCR signalling cascade, including the most notable SYK and BLNK. Further downstream, recruitment of PLCg2, Btk, and ERK facilitate calcium flux and activate the B cells, which are then ready to receive additional co-activating signals that will drive their proliferation and differentiation into memory or effector cells. During this process, B cells become robust APCs and release cytokines that can influence the outcome and quality of the immune response. In addition to their role in BCR signalling, the CD79 subunits are also essential for the transport and display of membrane-bound Ig from the endoplasmic reticulum to the cell surface. The average surface expression of CD79b on NHLs is similar to that on normal B-cells, but with a greater range. Given the expression of CD79b, it is beneficial to produce therapeutic antibodies to the CD79b antigen that create minimal or no antigenicity when administered to patients, especially for chronic treatment.

›BACKGROUND · 3 of 3

It has been found that a better T cell activation is achieved when limiting amounts of anti-CD3 bispecific antibodies, i.e. T cell bispecific antibodies (TCBs) such as for example a CD20/CD3 bispecific antibody, are combined with agonistic anti-CD28 molecules. Given, that CD28 is expressed at baseline on T cells in various tumor indications (Lavin et al., 2017; Tirosh et al., 2016, Zheng et al., 2017) and activation of CD28 signaling enhances T cell receptor signals, the combination of T cell bispecific antibodies with bispecific agonistic CD28 antigen binding molecules targeting a B cell surface antigen is expected to act synergistically to induce strong and long-lasting anti-tumor responses. Thus, we herein describe novel bispecific agonistic CD28 antigen binding molecules targeting a B cell surface antigen which display synergy with TCBs and require CD28 binding monovalency for strict tumor target dependence in the presence of TCB signals.

Immunotherapy in Multiple Myeloma

Affecting ˜75,000 new patients every year in the EU and US, multiple myeloma (MM) is one of the most common hematological malignancies with remaining high unmet medical need. Multiple myeloma is characterized by terminally differentiated plasma cells that secrete non-functional monoclonal immunoglobulins. In the short-term, the immunomodulatory drugs such as lenalidomide and pomalidomide, and proteasome inhibitors such as carfilzomib or bortezomib may remain the backbone of 1st line therapy for multiple myeloma (Moreau et al, 2016). However, these drugs do not target specifically the diseased tumor cells e.g. diseased plasma cells (PC). Efforts have been made towards selectively depleting the plasma cells in multiple myeloma. The lack of surface proteins that specifically mark plasma cells has hampered the development of antibodies or cellular therapies for multiple myeloma. So far, there are few cases of successful biologics, including daratumumab (anti-CD38) and elotuzumab (anti-CD319), with the caveat that both antigens are also expressed on other normal tissues including hematopoietic lineages and immune effector cells, which may limit their long-term clinical use. B cell maturation antigen (BCMA), a transmembrane glycoprotein in the tumor necrosis factor receptor superfamily 17 (TNFRSF17), is expressed at significantly higher levels in all patient MM cells but not on other normal tissues except normal plasma cells. BCMA-chimeric antigen receptor (CAR) T-cells have already shown significant clinical activities in patients with RRMM who have undergone at least three prior treatments, including a proteasome inhibitor and an immunomodulatory agent. Additional modalities, including anti-BCMA antibody-drug conjugate also has achieved significant clinical responses in patients who failed at least three prior lines of therapy, including an anti-CD38 antibody, a proteasome inhibitor, and an immunomodulatory agent (Cho et al, 2018). One challenge of e.g. BCMA- or CD38-targeted therapy lies in the presence of high levels of soluble BCMA or CD38 in the serum of MM patients, which may reduce the amount of active drug in the patient. An alternative might be new targets, such as the G protein-coupled receptor class C group 5 member D (GPRC5D), that is differentially expressed by plasma cells in multiple myeloma versus plasma cells from healthy donors, and has no soluble form. It has been reported that GPRC5D is associated with prognosis and tumor load in multiple myeloma patients (Atamaniuk, J. et al., 2012; and Cohen, Y., et al., 2013). GPRC5D is an orphan receptor with no known ligand(s) and largely unknown biology in men in general and in cancer specifically. The GPRC5D encoding gene, which is mapped on chromosome12p13.3, contains three exons and spans about 9.6 kb (Brauner-Osbome, H. et al. 2001). The large first exon encodes the seven-transmembrane domain. It has been shown that GPRC5D is involved in keratin formation in hair follicles in animals (Gao, Y. et al., 2016, and Inoue, S. et al., 2004). WO 2018/017786 A2 discloses GPRC5D-specific antibodies or antigen-binding fragments.

Rationale for Targeting CD28 Agonism to Diseased Plasma Cells in Multiple Myeloma

CD28 agonism in Multiple Myeloma may exert different biological functions on immune, respective MM plasma cells. While co-activation of T-cells via CD28 is expected to drive anti-tumor responses, CD28 agonism on MM cells mediates pro-survival signaling via regulation of PI3K/Akt, FoxO3a, and Bimm which in turn is described to induce chemotherapeutic resistance in multiple myeloma (Murray M. E. et al, 2014). Over-expression of CD28 on newly diagnosed Multiple Myeloma plasma cells is described to correlate with worse clinical outcome (Bahlis et al., 2007). However, while CD28 activation enhances myeloma cell survival, its activation inhibits myeloma cell proliferation. Agonizing CD28 in presence of a strong immune cell mediated response, such as a T-cell bispecific activation of T-cells, can further boost efficient anti-tumor responses. We herein provide bispecific agonistic CD28 antigen binding molecules that specifically bind a human Multiple Myeloma (MM) cell surface antigen. Particularly, the bispecific agonistic CD28 antigen binding molecules according to the invention targeting the TAAs selected from BCMA, CD38 and GPRC5D and CD28 expressed on T-cells have the potency to treat multiple myeloma as single agent or in combination with other agents such as T cell bispecific antibodies (TCBs) targeting a human MM cell surface antigen.

›SUMMARY · 1 of 7

The present invention describes tumor-targeted bispecific agonistic CD28 antigen binding molecules which achieve a tumor-dependent T cell activation and tumor cell killing without the necessity to form multimers. The bispecific CD28 antigen binding molecules of the present invention are characterized by monovalent binding to CD28 and in that they comprise at least one antigen binding domain capable of specific binding to a tumor-associated antigen (such as Fibroblast Activation Protein (FAP) or Carcinoembryonic Antigen (CEA), CD19 or GPRC5D). Furthermore, they possess an Fc domain composed of a first and a second subunit capable of stable association comprising one or more amino acid substitution that reduces the binding affinity of the antigen binding molecule to an Fc receptor and/or effector function. Fc receptor-mediated cross-linking is thereby abrogated and tumor-specific activation is achieved by cross-linking through binding of the at least one antigen binding domain capable of specific binding to a tumor-associated antigen to its antigen.

Thus, the invention provides a bispecific agonistic CD28 antigen binding molecule characterized by monovalent binding to CD28, comprising

(a) one antigen binding domain capable of specific binding to CD28,

(b) at least one antigen binding domain capable of specific binding to a tumor-associated antigen, and

In one aspect, a bispecific agonistic CD28 antigen binding molecule as defined below is provided, wherein the Fc domain is an IgG, particularly an IgG1 Fc domain or an IgG4 Fc domain. In one particular aspect, the Fc domain composed of a first and a second subunit capable of stable association is an IgG1 Fc domain. In one aspect, the Fc domain comprises the amino acid substitutions L234A and L235A (numbering according to Kabat EU index). In one aspect, the Fc domain is of human IgG1 subclass and comprises the amino acid mutations L234A, L235A and P329G (numbering according to Kabat EU index).

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as defined herein before, wherein the antigen binding domain capable of specific binding to CD28 comprises

(i) a heavy chain variable region (V H CD28) comprising a heavy chain complementary determining region CDR-H1 of SEQ ID NO: 36, a CDR-H2 of SEQ ID NO: 37, and a CDR-H3 of SEQ ID NO: 38, and a light chain variable region (V L CD28) comprising a light chain complementary determining region CDR-L1 of SEQ ID NO: 39, a CDR-L2 of SEQ ID NO: 40 and a CDR-L3 of SEQ ID NO: 41; or

(ii) a heavy chain variable region (V H CD28) comprising a CDR-H1 of SEQ ID NO: 20, a CDR-H2 of SEQ ID NO: 21, and a CDR-H3 of SEQ ID NO: 22, and a light chain variable region (V L CD28) comprising a CDR-L1 of SEQ ID NO: 23, a CDR-L2 of SEQ ID NO: 24 and a CDR-L3 of SEQ ID NO: 25.

In another aspect, the antigen binding domains capable of specific binding to CD28 of the bispecific agonistic CD28 antigen binding molecule comprises a heavy chain variable region (V H CD28) comprising a CDR-H1 of SEQ ID NO: 20, a CDR-H2 of SEQ ID NO: 21, and a CDR-H3 of SEQ ID NO: 22, and a light chain variable region (V L CD28) comprising a CDR-L1 of SEQ ID NO: 23, a CDR-L2 of SEQ ID NO: 24 and a CDR-L3 of SEQ ID NO: 25.

Furthermore, provided is a bispecific agonistic CD28 antigen binding molecule as defined herein before, wherein the antigen binding domain capable of specific binding to CD28 comprises a heavy chain variable region (V H CD28) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:26, and a light chain variable region (V L CD28) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:27.

In a further aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD28 comprises a heavy chain variable region (V H CD28) comprising an amino acid sequence selected from the group consisting of SEQ ID NO:42, SEQ ID NO:43, SEQ ID NO:44, SEQ ID NO:45, SEQ ID NO:46, SEQ ID NO:47, SEQ ID NO:48, SEQ ID NO:49, SEQ ID NO:50 and SEQ ID NO:51, and a light chain variable region (V L CD28) comprising an amino acid sequence selected from the group consisting of SEQ ID NO:27, SEQ ID NO:52, SEQ ID NO:53, SEQ ID NO:54, SEQ ID NO:55, SEQ ID NO:56, SEQ ID NO:57, SEQ ID NO:58, SEQ ID NO:59, SEQ ID NO:60 and SEQ ID NO:61.

In another aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CD28 comprises

(a) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54, or

(b) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or

(c) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:51 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:61, or

(d) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:53, or

›SUMMARY · 2 of 7

(e) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54, or

(f) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:59, or

(g) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or

(h) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:43 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or

(i) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:53, or

(j) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:59, or

(k) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27.

In one particular aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD28 comprises a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54.

In another particular aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD28 comprises a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:53.

In a further particular aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD28 comprises a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27.

In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to Carcinoembryonic Antigen (CEA).

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to CEA comprises

(i) a heavy chain variable region (V H CEA) comprising a CDR-H1 comprising the amino acid sequence of SEQ ID NO:188, a CDR-H2 comprising the amino acid sequence of SEQ ID NO:189, and a CDR-H3 comprising the amino acid sequence of SEQ ID NO:190, and a light chain variable region (V L CEA) comprising a CDR-L1 comprising the amino acid sequence of SEQ ID NO:191, a CDR-L2 comprising the amino acid sequence of SEQ ID NO:192, and a CDR-L3 comprising the amino acid sequence of SEQ ID NO:193; or

(ii) a heavy chain variable region (V H CEA) comprising a CDR-H1 comprising the amino acid sequence of SEQ ID NO:180, a CDR-H2 comprising the amino acid sequence of SEQ ID NO:181, and a CDR-H3 comprising the amino acid sequence of SEQ ID NO:182, and a light chain variable region (V L CEA) comprising a CDR-L1 comprising the amino acid sequence of SEQ ID NO:183, a CDR-L2 comprising the amino acid sequence of SEQ ID NO:184, and a CDR-L3 comprising the amino acid sequence of SEQ ID NO:185; or

(iii) a heavy chain variable region (V H CEA) comprising a CDR-H1 comprising the amino acid sequence of SEQ ID NO:127, a CDR-H2 comprising the amino acid sequence of SEQ ID NO:128, and a CDR-H3 comprising the amino acid sequence of SEQ ID NO:129, and a light chain variable region (V L CEA) comprising a CDR-L1 comprising the amino acid sequence of SEQ ID NO:130, a CDR-L2 comprising the amino acid sequence of SEQ ID NO:131, and a CDR-L3 comprising the amino acid sequence of SEQ ID NO:132, or

(iv) a heavy chain variable region (V H CEA) comprising a CDR-H1 comprising the amino acid sequence of SEQ ID NO:507, a CDR-H2 comprising the amino acid sequence of SEQ ID NO:508, and a CDR-H3 comprising the amino acid sequence of SEQ ID NO:509, and a light chain variable region (V L CEA) comprising a CDR-L1 comprising the amino acid sequence of SEQ ID NO:510, a CDR-L2 comprising the amino acid sequence of SEQ ID NO:511, and a CDR-L3 comprising the amino acid sequence of SEQ ID NO:512.

In one aspect, the antigen binding domain capable of specific binding to CEA comprises a heavy chain variable region (V H CEA) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:133, and a light chain variable region (V L CEA) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:134. Particularly, the antigen binding domain capable of specific binding to CEA comprises a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:186, and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:187.

In another aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CEA comprises

(a) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:194 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:195, or

(b) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:196 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:197, or

›SUMMARY · 3 of 7

(c) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:198 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:199, or

(d) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:200 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:201, or

(e) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:202 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:203, or

(f) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:204 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:205, or

(g) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:206 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:207, or

(h) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:208 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:209, or

(i) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:210 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:211, or

(j) a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:212 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:213.

Particularly, the antigen binding domain capable of specific binding to CEA comprises a heavy chain variable region (V H CEA) comprising the amino acid sequence of SEQ ID NO:200 and a light chain variable region (V L CEA) comprising the amino acid sequence of SEQ ID NO:201.

In another aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to Fibroblast Activation Protein (FAP). In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to FAP comprises (a) a heavy chain variable region (V H FAP) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:12, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:13, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:14, and a light chain variable region (V L FAP) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:15, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:16, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:17, or (b) a heavy chain variable region (V H FAP) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:4, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:5, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:6, and a light chain variable region (V L FAP) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:7, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:8, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:9. In particular, the antigen binding domain capable of specific binding to FAP comprises a heavy chain variable region (V H FAP) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:12, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:13, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:14, and a light chain variable region (V L FAP) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:15, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:16, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:17. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to FAP comprises (a) a heavy chain variable region (V H FAP) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:18, and a light chain variable region (V L FAP) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:19, or (b) a heavy chain variable region (V H FAP) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:10, and a light chain variable region (V L FAP) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:11. Particularly, the antigen binding domain capable of specific binding to FAP comprises a heavy chain variable region (V H FAP) comprising the amino acid sequence of SEQ ID NO:18 and a light chain variable region (V L FAP) comprising the amino acid sequence of SEQ ID NO:19.

›SUMMARY · 4 of 7

In yet another aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to CD30. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to CD30 comprises a heavy chain variable region (V H CD30) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:531, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:532, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:533, and a light chain variable region (V L CD30) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:534, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:535, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:536. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD30 comprises (a) a heavy chain variable region (V H CD30) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:537, and a light chain variable region (V L CD30) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:538. Particularly, the antigen binding domain capable of specific binding to CD30 comprises a heavy chain variable region (V H CD30) comprising the amino acid sequence of SEQ ID NO:537 and a light chain variable region (V L CD30) comprising the amino acid sequence of SEQ ID NO:538.

Furthermore, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to TBPG. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to TBPG comprises a heavy chain variable region (V H TBPG) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:539, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:540, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:541, and a light chain variable region (V L TBPG) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:542, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:543, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:544. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to TBPG comprises (a) a heavy chain variable region (V H TBPG) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:545, and a light chain variable region (V L TBPG) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:546. Particularly, the antigen binding domain capable of specific binding to TBPG comprises heavy chain variable region (V H TBPG) comprising the amino acid sequence of SEQ ID NO:545 and a light chain variable region (V L TBPG) comprising the amino acid sequence of SEQ ID NO:546.

In a further aspect, the invention provides a bispecific agonistic CD28 antigen binding molecule characterized by monovalent binding to CD28, comprising (a) one antigen binding domain capable of specific binding to CD28, (b) at least one antigen binding domain capable of specific binding to a Multiple Myeloma (MM) cell surface antigen, and (c) an Fc domain composed of a first and a second subunit capable of stable association comprising one or more amino acid substitution that reduces the binding affinity of the antigen binding molecule to an Fc receptor and/or effector function. In one aspect, the Multiple Myeloma (MM) cell surface antigen selected from the group consisting of CD38, BCMA and GPRC5D.

Thus, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to GPRC5D. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to GPRC5D comprises (a) a heavy chain variable region (V H GPRC5D) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:563, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:564, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:565, and a light chain variable region (V L GPRC5D) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:566, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:567, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:568, or (b) a heavy chain variable region (V H GPRC5D) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:579, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:580, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:581, and a light chain variable region (V L GPRC5D) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:582, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:583, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:584. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to GPRC5D comprises a heavy chain variable region (V H GPRC5D) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:569, and a light chain variable region (V L GPRC5D) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:570. Particularly, the antigen binding domain capable of specific binding to GPRC5D comprises a heavy chain variable region (V H GPRC5D) comprising the amino acid sequence of SEQ ID NO:569 and a light chain variable region (V L GPRC5D) comprising the amino acid sequence of SEQ ID NO:570.

›SUMMARY · 5 of 7

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to CD38. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to CD38 comprises a heavy chain variable region (V H CD38) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:547, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:548, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:549, and a light chain variable region (V L CD38) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:550, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:551, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:552. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD38 comprises a heavy chain variable region (V H CD38) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:553, and a light chain variable region (V L CD38) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:554. Particularly, the antigen binding domain capable of specific binding to CD38 comprises a heavy chain variable region (V H CD38) comprising the amino acid sequence of SEQ ID NO:553 and a light chain variable region (V L CD38) comprising the amino acid sequence of SEQ ID NO:554.

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to BCMA. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to BCMA comprises a heavy chain variable region (V H BCMA) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:555, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:556, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:557, and a light chain variable region (V L BCMA) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:558, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:559, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:560. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to BCMA comprises a heavy chain variable region (V H BCMA) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:559, and a light chain variable region (V L BCMA) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:560. Particularly, the antigen binding domain capable of specific binding to BCMA comprises a heavy chain variable region (V H BCMA) comprising the amino acid sequence of SEQ ID NO:561 and a light chain variable region (V L BCMA) comprising the amino acid sequence of SEQ ID NO:562.

In a further aspect, the invention provides a bispecific agonistic CD28 antigen binding molecule characterized by monovalent binding to CD28, comprising (a) one antigen binding domain capable of specific binding to CD28, (b) at least one antigen binding domain capable of specific binding to a B cell surface antigen, and (c) an Fc domain composed of a first and a second subunit capable of stable association comprising one or more amino acid substitution that reduces the binding affinity of the antigen binding molecule to an Fc receptor and/or effector function. In one aspect, the B cell surface antigen selected from the group consisting of CD19, CD79b, CD20, CD22 and CD37.

Thus, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to CD19. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to CD19 comprises (a) a heavy chain variable region (V H CD19) comprising (i) CDR-HI comprising the amino acid sequence of SEQ ID NO:406, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:407, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:408, and a light chain variable region (V L CD19) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:409, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:410, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:411, or (b) a heavy chain variable region (V H CD19) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:414, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:415, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:416, and a light chain variable region (V L CD19) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:417, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:418, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:419. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD19 comprises (a) a heavy chain variable region (V H CD19) comprising an amino acid sequence that is at least about 95%, 98% or 100% identical to the amino acid sequence of SEQ ID NO:412, and a light chain variable region (V L CD19) comprising an amino acid sequence that is at least about 95%, 98% or 100% identical to the amino acid sequence of SEQ ID NO:413, or (b) a heavy chain variable region (V H CD19) comprising an amino acid sequence that is at least about 95%, 98% or 100% identical to the amino acid sequence of SEQ ID NO:420, and a light chain variable region (V L CD19) comprising an amino acid sequence that is at least about 95%, 98% or 100% identical to the amino acid sequence of SEQ ID NO:421. Particularly, the antigen binding domain capable of specific binding to CD19 comprises a heavy chain variable region (V H CD19) comprising an amino acid sequence of SEQ ID NO:412 and a light chain variable region (V L CD19) comprising an amino acid sequence of SEQ ID NO:413.

›SUMMARY · 6 of 7

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to a tumor-associated antigen is an antigen binding domain capable of specific binding to CD79b. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, wherein the antigen binding domain capable of specific binding to CD79b comprises a heavy chain variable region (V H CD79b) comprising (i) CDR-H1 comprising the amino acid sequence of SEQ ID NO:422, (ii) CDR-H2 comprising the amino acid sequence of SEQ ID NO:423, and (iii) CDR-H3 comprising the amino acid sequence of SEQ ID NO:424, and a light chain variable region (V L CD79b) comprising (iv) CDR-L1 comprising the amino acid sequence of SEQ ID NO:425, (v) CDR-L2 comprising the amino acid sequence of SEQ ID NO:426, and (vi) CDR-L3 comprising the amino acid sequence of SEQ ID NO:427. In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding domain capable of specific binding to CD79b comprises a heavy chain variable region (V H CD79b) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:428, and a light chain variable region (V L CD79b) comprising an amino acid sequence that is at least about 95%, 96%, 97%, 98%, 99% or 100% identical to the amino acid sequence of SEQ ID NO:429. Particularly, the antigen binding domain capable of specific binding to CD79b comprises a heavy chain variable region (V H CD79b) comprising the amino acid sequence of SEQ ID NO:428, and a light chain variable region (V L CD79b) comprising the amino acid sequence of SEQ ID NO:429.

In another aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, comprising

(a) one Fab fragment capable of specific binding to CD28,

(b) one crossFab fragment capable of specific binding to a tumor-associated antigen, and

In a further aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, comprising

(a) a first Fab fragment capable of specific binding to CD28, (b) a second Fab fragment capable of specific binding to a tumor-associated antigen, and

In another aspect, a bispecific agonistic CD28 antigen binding molecule as disclosed herein is provided, comprising

(a) a first Fab fragment capable of specific binding to CD28,

(b) a second and a third Fab fragment capable of specific binding to a tumor-associated antigen, and

wherein the first Fab fragment capable of specific binding to CD28 is fused at the C-terminus of the Fab heavy chain to the N-terminus of the Fab heavy chain of the second Fab fragment capable of specific binding to a tumor-associated antigen, which is in turn fused at its C-terminus to the N-terminus of the first Fc domain subunit, and the third Fab fragment capable of specific binding to a tumor-associated antigen is fused at the C-terminus of the Fab heavy chain to the N-terminus of the second Fc domain subunit.

In a further aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein, comprising

(a) a Fab fragment capable of specific binding to CD28,

(b) a VH and VL domain capable of specific binding to a tumor-associated antigen, and

(c) a Fc domain composed of a first and a second subunit capable of stable association comprising one or more amino acid substitution that reduces the binding affinity of the antigen binding molecule to an Fc receptor and/or effector function, wherein the Fab fragment capable of specific binding to CD28 is fused at its C-terminus to the N-terminus of the first Fc domain subunit, and wherein one of the VH and VL domain capable of specific binding to a tumor-associated antigen is fused via a peptide linker to the C-terminus of the first Fc domain subunit and the other one of the VH and VL domain capable of specific binding to a tumor-associated antigen is fused via a peptide linker to the C-terminus of the second Fc domain subunit.

According to another aspect of the invention, there is provided one or more isolated polynucleotide(s) encoding the bispecific agonistic CD28 antigen binding molecule of the invention. The invention further provides one or more vector(s), particularly expression vector(s), comprising the isolated polynucleotide(s) of the invention, and a host cell comprising the isolated polynucleotide(s) or the expression vector(s) of the invention. In some aspects, the host cell is a eukaryotic cell, particularly a mammalian cell. In another aspect, provided is a method of producing a bispecific agonistic CD28 antigen binding molecule as described herein comprising culturing the host cell of the invention under conditions suitable for the expression of the bispecific agonistic CD28 antigen binding molecule. Optionally, the method also comprises recovering the bispecific agonistic CD28 antigen binding molecule. The invention also encompasses a bispecific agonistic CD28 antigen binding molecule produced by the method of the invention.

›SUMMARY · 7 of 7

The invention further provides a pharmaceutical composition comprising a bispecific agonistic CD28 antigen binding molecule of the invention and at least one pharmaceutically acceptable excipient. In one aspect, the pharmaceutical composition is for use in the treatment of cancer.

Also encompassed by the invention are methods of using the bispecific agonistic CD28 antigen binding molecule and the pharmaceutical composition of the invention. In one aspect the invention provides a bispecific agonistic CD28 antigen binding molecule or a pharmaceutical composition according to the invention for use as a medicament. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as described herein for use in (a) enhancing cell activation or (b) enhancing T cell effector functions. In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule or a pharmaceutical composition according to the invention for use in the treatment of a disease. In a specific aspect, the disease is cancer. In another aspect is provided a bispecific agonistic CD28 antigen binding molecule or pharmaceutical composition according to the invention is for use in the treatment of cancer, wherein the bispecific agonistic CD28 antigen binding molecule is administered in combination with a chemotherapeutic agent, radiation therapy and/or other agents for use in cancer immunotherapy. In a further aspect, provided is a bispecific agonistic CD28 antigen binding molecule or a pharmaceutical composition for use in the treatment of cancer, wherein the bispecific agonistic CD28 antigen binding molecule is administered in combination with a T-cell activating anti-CD3 bispecific antibody. In yet another aspect, provided is a bispecific agonistic CD28 antigen binding molecule or a pharmaceutical composition for use in the treatment of cancer, wherein the bispecific agonistic CD28 antigen binding molecule is administered in combination with an anti-PD-L1 antibody or an anti-PD-1 antibody.

Also provided is the use of a bispecific agonistic CD28 antigen binding molecule according to the invention in the manufacture of a medicament for the treatment of a disease; as well as a method of treating a disease in an individual, comprising administering to said individual a therapeutically effective amount of a bispecific agonistic CD28 antigen binding molecule according to the invention or a composition comprising the bispecific agonistic CD28 antigen binding molecule according to the invention in a pharmaceutically acceptable form. In a specific aspect, the disease is cancer. In one aspect, provided is a method (a) enhancing cell activation or (b) enhancing T cell effector functions in an individual, comprising administering a bispecific agonistic CD28 antigen binding molecule according to the invention or a composition comprising the bispecific agonistic CD28 antigen binding molecule according to the invention in a pharmaceutically acceptable form to said individual. In another aspect, provided is the use of a bispecific agonistic CD28 antigen binding molecule according to the invention in the manufacture of a medicament for the treatment of a disease, wherein the treatment comprises co-administration with a chemotherapeutic agent, radiation therapy and/or other agents for use in cancer immunotherapy. In a further aspect, provided is a method of treating a disease in an individual, comprising administering to said individual a therapeutically effective amount of a bispecific agonistic CD28 antigen binding molecule according to the invention or a composition comprising the bispecific agonistic CD28 antigen binding molecule according to the invention in a pharmaceutically acceptable form, wherein the method comprises co-administration with a chemotherapeutic agent, radiation therapy and/or other agents for use in cancer immunotherapy. In a further aspect, provided is a method of treating a disease in an individual, comprising administering to said individual a therapeutically effective amount of a bispecific agonistic CD28 antigen binding molecule according to the invention or a composition comprising the bispecific agonistic CD28 antigen binding molecule according to the invention in a pharmaceutically acceptable form, wherein the method comprises co-administration of a T-cell activating anti-CD3 bispecific antibody. In another aspect, provided is a method of treating a disease in an individual, comprising administering to said individual a therapeutically effective amount of a bispecific agonistic CD28 antigen binding molecule according to the invention or a composition comprising the bispecific agonistic CD28 antigen binding molecule according to the invention in a pharmaceutically acceptable form, wherein the method comprises co-administration of an anti-PD-L1 antibody or an anti-PD-1 antibody. Also provided is a method of inhibiting the growth of tumor cells in an individual comprising administering to the individual an effective amount of the bispecific agonistic CD28 antigen binding molecule according to the invention, or a composition comprising the bispecific agonistic CD28 antigen binding molecule according to the invention in a pharmaceutically acceptable form, to inhibit the growth of the tumor cells. In any of the above aspects, the individual preferably is a mammal, particularly a human.

›BRIEF DESCRIPTION OF THE DRAWINGS · 1 of 6

In FIGS. 1 A to 1 L schematic illustrations of the molecules as described are shown. FIG. 1 A shows the CD28 agonistic antibody CD28(SA) in its huIgG4 isoform (TGN1412). FIG. 1 B illustrates the CD28(SA) agonistic antibody as huIgG1 PGLALA isotype (“Fc silent”).

Bispecific FAP-CD28 antigen binding molecules in 1+1 format, 1+2 format, 2+2 format and 1+4 format are shown in FIGS. 1 C, 1 D, 1 E and 1 F , respectively.

Bispecific CEA-CD28 antigen binding molecules in 1+2 format, 2+2 format and 1+1 format are shown in FIGS. 1 G, 1 H and 1 J , respectively.

FIG. 1 I shows a schematic illustration of the CD28 agonistic antibody variants as monovalent huIgG1 PGLALA isotype (“Fc silent”).

FIG. 1 K shows a bispecific FAP-CD28 antigen binding molecule in 1+1 format, wherein the FAP antigen binding domain is represented as VH and VL domains each fused to one C-terminus of the Fc domain subunits.

FIG. 1 L illustrates a bispecific FAP-CD28 antigen binding molecule in 2+1 format, wherein the CD28 antigen binding domain is represented as crossFab that is fused at its C-terminus to the N-terminus of one of the heavy chains of the “bivalent” FAP antibody.

FIG. 1 M shows another bispecific FAP-CD28 antigen binding molecule in 1+1 format, wherein the CD28 antigen binding domain is represented as crossFab that is fused at its C-terminus to the N-terminus of the Fab fragment binding to FAP.

FIG. 1 N illustrates a trispecific FAP-CEA-CD28 antigen binding molecule in 1+1+1 format, wherein the CD28 antigen binding domain is represented as Fab that is fused at the C-terminus of both the light and the heavy chain to the N-terminus of both the light and heavy chain of the anti-FAP antigen binding domain on the huIgG1 PG-LALA Fc knob chain and wherein the anti-CEA CrossFab fragment is part of the huIgG1 PG-LALA Fc hole chain.

FIGS. 2 A, 2 B, 2 C, 2 D and 2 E relate to the binding of CD28 agonistic antibodies and FAP-CD28 antigen binding molecules to human CD28 or human FAP on cells. Shown is the binding of CD28(SA) in it IgG4 isoform vs. huIgG1 PGLALA isotype ti human CD28 in FIG. 2 A and the binding of different FAP-CD28 molecules to human CD28 ( FIG. 2 B ) and human FAP ( FIG. 2 C ) on cells. Median fluorescence intensities of binding of different CD28 agonistic antibodies or anti-DP47 targeted molecules to CHO cells expressing human CD28 (parental cell line CHO-kl ATCC #CCL-61, modified to stably overexpress human CD28) or 3T3 cells expressing human FAP (NIH/3T3 cell line (ATCC CRL-1658)) was assessed by flow cytometry. Depicted are technical triplicates with SEM. A comparison of FAP(4B9)-CD28(SA) antigen binding molecules (Molecules D, E and F as described in Example 1) is shown in FIG. 2 D (binding to human CD28) and FIG. 2 E (binding to human FAP).

FIGS. 2 F and 2 G show the binding of different formats of FAP-CD28 antigen binding molecules with monovalent binding to CD28 and FAP, respectively. Shown are the curves for FAP-CD28 CTF 1+1 (P1AE2236, Molecule I), FAP-CD28 1+1 (P1 AD4492, Molecule C), FAP-CD28 H2T 1+1 (P1AE2021, Molecule H), and the two compounds FAP-CD28(SA) 1+2 (P1AD9011, Molecule E) and DP47 as reference. Shown are median fluorescence intensities of binding of the FAP-CD28 antibodies or anti-DP47 antibody (negative control) to CHO cells expressing human CD28 (parental cell line CHO-kl ATCC #CCL-61, modified to stably overexpress human CD28) ( FIG. 2 F ) or 3T3 cells expressing human FAP (NIH/3T3 cell line (ATCC CRL-1658)) ( FIG. 2 G ), assessed by flow cytometry. Shown are technical triplicates with SEM. FIGS. 2 H and 21 show the binding of FAP-CD28 2+1 (P1AE5231, Molecule G) to CD28 and FAP, respectively.

The alignment of the variable domains of CD28(SA) and variants thereof is shown in FIGS. 3 A to 3 D . Alignment of the CD28(SA) VH domain and variants thereof in order to remove cysteine 50 and to reduce the affinity of the resulting anti-CD28 binders to different degrees is shown in FIGS. 3 A and 3 B . Of note, in VH variants i and j, the CDRs of CD28(SA) were grafted from an IGHV1-2 framework into an IGHV3-23 framework ( FIG. 3 B ). In FIGS. 3 C and 3 D , alignment of the CD28(SA) VL domain and variants thereof in order to reduce the affinity of the resulting anti-CD28 binders to different degrees is shown. In variant t, the CDRs were grafted into the framework sequence of the trastuzumab (Herceptin) VL sequence.

In FIGS. 4 A to 4 C the binding of affinity-reduced CD28 agonistic antibody variants in monospecific, monovalent IgG formats from supernatants to human CD28 on cells is shown. Median fluorescence intensities of binding to CHO cells expressing human CD28 (parental cell line CHO-kl ATCC #CCL-61, modified to stably overexpress human CD28) compared to the negative control (anti-DP47) and the original TGN1412, were assessed by flow cytometry. The binding curves of variants 1-10 are shown in FIG. 4 A , those of variants 11 to 22 in FIG. 4 B and those of variants 23 to 31 in FIG. 4 C . Depicted are technical duplicates with SD.

In FIGS. 4 D and 4 E , the binding of FAP-targeted bispecific CD28 agonistic antibody variants in huIgG1 PG-LALA 1+1 format with selected affinity-reduced CD28 agonistic antibody variants to human CD28 on cells is shown. The binding curves of bispecific 1+1 constructs with variants 8, 11, 12, 15, 16 and 17 are shown in FIG. 4 D , whereas the binding curves of bispecific 1+1 constructs with variants 19, 23, 25, 27 and 29 are shown in FIG. 4 E . Selected binders were chosen based on affinities for production in a 1+1, bispecific FAP-targeted format. In FIGS. 4 F and 4 G , the binding of the same FAP-targeted bispecific CD28 agonistic antibody variants in huIgG1 PG-LALA 1+1 format to human FAP is shown. Provided are median fluorescence intensities of binding to CHO cells expressing human CD28 (parental cell line CHO-kl ATCC #CCL-61, modified to stably overexpress human CD28) or to 3T3 cells expressing human FAP (NIH/3T3 cell line (ATCC CRL-1658)) compared to the negative control (anti-DP47) and to TGN1412 (Molecule A), assessed by flow cytometry. Shown are technical triplicates with SEM.

›BRIEF DESCRIPTION OF THE DRAWINGS · 2 of 6

The in vitro potency of selected FAP-targeted bispecific CD28 agonistic antibody variants in huIgG1 PG-LALA 1+1 format is illustrated in FIGS. 4 H, 4 I and 4 J . PBMC T cells were incubated with MCSP- and FAP-expressing MV3 melanoma cells for 5 days in the presence of limiting concentration of MCSP-TCB (5 pM, P1AD2189) and increasing concentration of FAP-CD28 constructs with the indicated CD28 variant binders. In FIG. 4 H is shown the CFSE-dilution as measure for T cell proliferation of CD8 T cells, assessed by flow cytometry. Error bars show SEM, graphs depict technical triplicates of representative results from 2 donors. In FIG. 4 I is shown the correlation of K D (nM) of the CD28 binder variant in relation to potency by area under the curve of (a) as % of the parental TGN1412 clone (CD28(SA)). In FIG. 4 J the target cell killing at 90 h is shown.

FIGS. 5 A to 5 D refer to the establishment of high-density (HD) pre-culture and mode of action of CD28(SA). PBMC T cells were either pre-cultured at high density (HD) for 2 days or used fresh from PBMC isolation and stimulated with increasing concentrations of CD28(SA). Depicted is CFSE-dilution as proxy for T cell proliferation after 5 days of stimulation with CD28(SA) (Molecule A, P1AE1975) ( FIG. 5 A ) and cytokine secretion after 2 days ( FIG. 5 B ) of stimulation. FIG. 5 C shows the percentage of FcγRIIb expression in PBMC monocytes and B cells before and after 2 days HD PBMC pre-culture, assessed by flow cytometry. FIG. 5 D : HD pre-cultured PBMCs were co-cultured with CD28(SA) for 5 days in presence or absence of an FcγRIIb blocking antibody or isotype control and percentage of CFSE-dilution of CD4 T cells was assessed by flow cytometry. Graphs are representative of at least 6 donors ( FIG. 5 A, 5 B ) and 2 donors ( FIG. 5 C, 5 D ), each assessed in independent experiments. The graphs show technical triplicates. Error bars indicate SEM. Statistical analysis was performed by student's t-test. ***: p<0.001. Superagonism of CD28(SA) IgG4 depends on cross-linking to FcγRIth.

In FIGS. 6 A and 6 B the T cell proliferation, i.e. CFSE-dilution of CD4 T cells after 5 days of stimulation with either original Fc wild-type IgG4 CD28(SA) (P1AE1975) or CD28(SA) bearing the P329G-LALA mutation (P1AD9289) is shown. T cells were pre-cultured at high density for 2 days. Graphs are representative of at least 3 independent experiments. Technical triplicates are shown. Fc-silencing abolishes superagonism in TGN1412. Adding a tumor-targeting moiety to Fc-silenced TGN1412 restores superagonism, which is then dependent on the presence of the tumor-target.

In FIGS. 7 A, 7 B, 7 C and 7 D a comparison of FAP-targeted CD28 agonists in different formats (2+2 and 1+2) and with superagonistic (CD28(SA)) binders and conventional agonistic binders (9.3, CD28(CA)) is shown. FAP-targeted CD28 agonists with conventional CD28 agonistic binders do not function as superagonists. PBMC T cells were co-cultured with 3T3-huFAP cells (FAP present) in the presence of increasing concentrations of the FAP-CD28 formats with superagonistic binders (SA, FIG. 7 A ) or conventional agonistic binders (9.3, FIG. 7 B ) for 5 days. T cell proliferation is shown. PBMC T cells were then also co-cultured with 3T3 WT cells (FAP absent), in the presence of increasing concentrations of the FAP-CD28 formats with superagonistic binders (SA, FIG. 7 C ) or conventional agonistic binders (9.3, FIG. 7 D ) for 5 days. Depicted is CFSE-dilution as measure for T cell proliferation of CD8 T cells, assessed by flow cytometry on day 5 post stimulation. Graphs show cumulative data from 3 donors in 3 independent experiments. Error bars show SEM. In the same experimental setup also cytokines were measured from supernatants after 2 days of co-culture. The values are provided in FIG. 7 E .

The ability of FAP-CD28 in various formats with either superagonistic CD28(SA) binders or conventional agonistic binders (CD28(CA)) to induce killing of FAP-expressing RFP-MV3 melanoma cells was assessed over the course of 90 h by live cell imaging using the IncuCyte technology. All molecules including the FAP-TCB (P1AD4645) were used at 10 nM. FIGS. 8 A, 8 B and 8 C show representative results from three donors with technical triplicates, respectively. FIG. 8 D shows the cumulative results expressed as area under the curve (AUC) at t=90 h of 3 donors from 3 independent experiments. Boxes display 25th-75th percentiles, whiskers display min to max. Statistical analysis was performed by paired 1-way ANOVA. ***: p<0.001, ns: not significant.

A comparison of CEA-targeted CD28 agonists in different formats with superagonistic and conventional agonistic binders is shown in FIGS. 9 A and 9 B . The ability of CEA-CD28 in various formats with either superagonistic CD28(SA) binders or conventional agonistic binders (CD28(CA)) to induce killing of CEA-expressing RFP + MKN45 gastric cancer cells was assessed over the course of 90 h by live cell imaging using the IncuCyte technology. All molecules including the CEACAM5-TCB (P1AD5299) were used at 10 nM. FIG. 9 A shows representative results from one donor with technical triplicates. FIG. 9 B shows the statistical analysis of technical triplicates expressed as area under the curve (AUC) at t=90 h of 1 donor in 1 experiment. Boxes display 25th-75th percentiles, whiskers display min to max. Statistical analysis was performed by paired 1-way ANOVA. ***: p<0.001. It is shown that CEA-targeted CD28 agonists with conventional CD28 agonistic binders do not behave superagonistically.

In FIGS. 10 A, 10 B and 10 C it is shown that targeted CD28 agonists with monovalent superagonistic binders are not functionally superagonistic. PBMC T cells were co-cultured for 5 days with 3T3-huFAP cells in presence of increasing concentrations of FAP-CD28 with bivalent CD28 binders (P1AD9011, closed circles) or FAP-CD28 with monovalency for CD28 binding (P1AD4492, open circles). In FIG. 10 A CFSE-dilution of CD8 T cells is shown. Furthermore, activation of T cells was assessed by detection of activation markers CD69 ( FIG. 10 B ) and CD25 ( FIG. 10 C ) by flow cytometry. Mean fluorescent intensity (MFI) of CD69 and CD25 stainings are shown at 5 days post stimulation. Technical triplicates from 1 donor are shown, error bars indicate SEM. It is shown that TGN1412-like superagonism requires multivalent CD28 binding.

›BRIEF DESCRIPTION OF THE DRAWINGS · 3 of 6

FIGS. 11 A and 11 B show that if combined with T cell bispecific antibodies (TCBs) TCB-mediated effector functions are supported by monovalent and bivalent CD28 binding of FAP-targeted agonistic CD28 antigen binding molecules with comparable potency, but CD28 binder monovalency is required to maintain tumor target dependence of CD28 agonists in the presence of TCBs. In FIG. 11 A , for the presence of FAP, PBMC T cells were incubated with MCSP- and FAP-expressing MV3 melanoma cells for 90 h in the presence of a combined limiting concentration of MCSP-TCB (5 pM, P1AD2189) and increasing concentration (range 0-10 nM) of FAP-CD28(SA) with bivalent or monovalent binding to CD28, respectively. Depicted is target cell killing at 90 h as assessed by live cell imaging using the IncuCyte technology. In FIG. 11 B , PBMC T cells were co-cultured for 90 h with FAP-negative CEA-expressing MKN45 gastric cancer cells (FAP absent) in the presence of limiting concentrations of CEACAM5-TCB (10 pM, P1AD5299) in combination and increasing concentration (range 0-10 nM) of FAP-CD28 with bivalent or monovalent CD28 binding, respectively. Depicted is target cell killing at 90 h as assessed by IncuCyte. Data show killing of MKN45 target cells over time, from 1 donor in 1 experiment, technical triplicates, error bars indicate SEM.

In FIGS. 12 A and 12 B it is shown that FAP-CD28(SA) with bivalent binding to CD28 loses FAP-dependence when combined with T cell bispecifics. In FIG. 12 A no TCB is present. PBMC T cells were either co-cultured with CEA-expressing MKN45 and 3T3-huFAP (“FAP present” condition, closed circles) or 3T3-WT (“FAP absent” condition, open circles), respectively, in the presence of increasing concentrations of FAP-CD28(SA) 2+1. The combination with TCB is shown in FIG. 12 B . PBMC T cells were either co-cultured with CEA-expressing MKN45 and 3T3-huFAP (“FAP present” condition, closed circles) or 3T3-WT (“FAP absent” condition, open circles), respectively, in the presence of limiting concentrations of CEACAM5-TCB (10 pM, P1AD5299) and increasing concentrations of FAP-CD28 2+1 SA. Shown is CD8 T cell proliferation after 5 days of stimulation. Data are representative of 2 independent experiments with 2 donors. Results from one donor are shown, data points represent technical triplicates, error bars indicate SEM.

FIGS. 13 A, 13 B and 13 C show the functionality of FAP-CD28(SA) antigen binding molecules with monovalent binding to CD28 in different formats. Molecule C is a FAP-CD28(SA) classical 1+1 format (P1 AD4492), Molecule H is a FAP-CD28(SA) 1+1 “head-to-tail” (H2T) format (P1AE2021), Molecule I is a FAP-CD28(SA) 1+1 format with C-terminal fusion of the FAP binder (P1AE2236) and Molecule G is a FAP-CD28(SA) 2+1 format (P1AD5231). As reference the bivalent CD28 antigen binding molecule (P1AD9011) was used. PBMC T cells were incubated with MCSP- and FAP-expressing MV3 melanoma cells in the presence of limiting concentration of MCSP-TCB (5 pM, P1AD2189) and increasing concentration (range 0-10 nM) of FAP-CD28 in the given formats. Depicted is CFSE-dilution as measure for T cell proliferation of CD8 ( FIG. 13 A ) and CD4 T cells ( FIG. 13 B ) after 5 days, assessed by flow cytometry. FIG. 13 C : Shown is killing of MV3 cells over 84 hours in presence of 5 pM MCSP-TCB alone compared to combination of 5 pM MCSP-TCB and increasing concentrations of FAP-CD28 in the various formats. Killing was assessed by live cell imaging using the IncuCyte system. All molecules were able to support TCB-mediated effector functions. Graphs depict cumulative data from 3 independent experiments and 4 donors 10 pM MCSP-TCB; E:T 20; Statistics: 2-way ANOVA. Stars indicate lowest concentration at which add-on is significant over TCB alone: *p≤0.05, **p≤0.01; ***p≤0.001. Error bars indicate SEM.

Target cell killing of CEA-CD28 1+1 format in combination with TCB is shown in FIG. 14 . PBMC T cells were co-cultured for 90 h with CEA-expressing MKN45 gastric cancer cells in presence of limiting concentrations of CEACAM5-TCB (10 pM, P1AD5299) in combination with 2 nM of CEA-CD28 (P1AE3127) or untargeted CD28 (P1AD8944). Data show killing of MKN45 target cells over time, from 1 donor in 1 experiment. Killing was assessed by live cell imaging using the IncuCyte system. It is shown that only the combination leads to target cell killing, at the given concentration the molecules alone do not induce killing. CEA-CD28 synergizes with CEACAM5-TCB.

In FIG. 15 it is shown that CEA-CD28 enhances CEA-TCB and CEACAM5-TCB and lowers the threshold of CEA-expression for TCBs to induce T cell activation. PBMC T cells were incubated with increasing concentrations of either CEA-TCB (P1AD4646) or CEACAM5-TCB (P1AD5299) and fixed concentrations of CEA-CD28 (P1AE3127) in presence of target cell lines with different CEA expression levels: (i) MKN45 (high expression, approx. 400 000 CEA binding sites/cell), (ii) Lovo (medium expression, approx. 60 000 CEA binding sites/cell), (iii) HT-29 (low expression, approx. 6 000 CEA binding sites/cell). T cell proliferation was assessed as proxy of T cell activation by flow cytometry.

The binding of selected affinity reduced CD28 binder variants in a bispecific CEA-targeted monovalent 1+1 format to CD28 on cells is shown in FIG. 16 . Median fluorescence intensities of binding of the CEA-CD28 antibodies or anti-DP47 antibody (negative control) to CHO cells expressing human CD28 (parental cell line CHO-kl ATCC #CCL-61, modified to stably overexpress human CD28) was assessed by flow cytometry. Shown are technical triplicates with SEM.

FIGS. 17 A, 17 B and 17 C show the functionality of selected affinity reduced CD28 binder variants in a bispecific CEA-targeted monovalent 1+1 format. PBMC T cells were co-cultured with CEA-expressing MKN45 gastric cancer cells in presence of limiting concentrations of CEACAM5-TCB (10 pM, P1AD5299) in combination with 2 nM of the CEA-CD28 1+1 molecules with the CD28 binder variants. CD8 T cell proliferation ( FIG. 17 A ) and CD4 T cell proliferation ( FIG. 17 B ) was assessed by CFSE-dilution by flow cytometry after 5 days of co-culture. Target cell killing was assessed after 90 h of incubation ( FIG. 17 C ). Data show killing of MKN45 target cells over time, from 1 donor in 1 experiment. All molecules are able to support CEACAM5-TCB mediated effector functions.

›BRIEF DESCRIPTION OF THE DRAWINGS · 4 of 6

FIG. 18 shows the binding of humanized CEA(A5B7) huIgG1 P329G LALA variants to MKN-45 as compared to the binding of the parental murine A5B7 antibody. Antibodies were detected with a fluorescently labeled secondary antibody and fluorescence was measured by flow cytometry.

FIGS. 19 A to 19 C are schematic illustrations of the recombinant proteins displaying different domains of the CEACAM5 protein that were used as antigens in the phage display campaign. FIG. 19 A shows construct NABA-avi-His consisting of the 4 Ig-like domains N, Al, B and A2. FIG. 19 B shows the construct N(A2B2)A-avi-His and FIG. 19 C illustrates the construct NA(B2)A-avi-His.

FIGS. 20 A and 20 B show the VH and VL sequences, respectively, of humanized CEA antibody A5H1EL1D wherein the randomized positions are marked with X.

Schematic drawings of the phage vectors of the affinity maturation libraries are shown in FIG. 21 A (CDRH1/H2 affinity maturation library), FIG. 21 B (CDRL1/H2 affinity maturation library) and FIG. 21 C (CDRH3/CDRL3 amplification library).

FIGS. 22 A and 22 B show an alignment of the VH amino acid sequences ( FIG. 22 A ) and VL amino acid sequences ( FIG. 22 B ) of the affinity-matured, humanized CEA(A5H1EL1D) antibody variants.

In FIGS. 23 A to 23 D schematic illustrations of bispecific CEA/CD28 antigen binding molecules as described in Example 11 are shown.

FIG. 23 A shows a bispecific CEA-CD28 antigen binding molecule in 1+1 format, wherein the CEA antigen binding domain is represented as crossFab (VH/VL exchange) and in the Fab fragment bearing the CD28 antigen binding domain there are charged modifications in order to support the correct pairing of the light chains. The Fc domain has knob into hole modifications and the P329G LALA mutations to abrogate the binding to Fcγ receptors. In FIG. 23 B the CD28 antigen binding domain is represented as crossFab (VH/VL exchange) and the Fab fragment bearing the CEA antigen binding domain comprises the charged modifications.

FIG. 23 C illustrates a bispecific CEA-CD28 antigen binding molecule in 2+1 format, wherein the CD28 antigen binding domain is represented as crossFab and two Fab fragments with CEA antigen binding domains are fused to each other via the heavy chain (head-to tail).

FIG. 23 D illustrates a bispecific CEA-CD28 antigen binding molecule in 2+1 format, wherein the CD28 antigen binding domain is represented as crossFab that is fused at its C-terminus to the N-terminus of one of the heavy chains of the “bivalent” CEA antibody (“classical” format).

In FIG. 24 it is shown that affinity-matured anti-CEA clone P002.139 shows improved binding to CEACAM5 on CEA-expressing MV3 cells. Shown is binding of CEA-CD28 bispecific antibodies carrying either the affinity-matured anti-CEA clone P002.139 or the parental A5H1EL1D clone. Median fluorescence intensities of binding of the CEA-CD28 bispecific antibodies or anti-DP47 antibody (negative control) to MV3 cells genetically engineered to express human CEACAM5 was assessed by flow cytometry. Shown are technical duplicates with SEM. Graph is representative of 3 independent experiments.

In FIGS. 25 A and 25 B it is shown that affinity-matured anti-CEA clone P002.139 shows improved functionality in an IL-2 reporter assay. Shown is luminescence readout after 6 h of co-incubation of MKN45 cells, IL-2 reporter cells with 5 nM CEA-TCB and CEA-CD28 carrying either the affinity-matured clone P002.139 or the parental clone ASH1EL1D. FIG. 25 A shows dose response. Dotted line indicates luminescence achieved by CEA-TCB alone. In FIG. 25 B the Area under the curve values calculated from data depicted in FIG. 25 A are shown. Provided are technical duplicates with SEM. Graph is representative of 3 independent experiments.

FIG. 26 shows the study design of an efficacy study with bispecific CEA-CD28 antibodies (comparison of different CEA clones) in combination with CEA TCB in MKN45 Xenograft in humanized mice. Shown is the design and the different treatment groups.

FIGS. 27 A to 27 E show results of the efficacy study with CEA-CD28 and CEA TCB combination in MKN45 Xenograft in humanized mice. Shown is the average tumor volume ( FIG. 27 A ) or the growth of tumors in individual mice for the four treatment groups as plotted on the y-axis ( FIG. 27 B to 27 E ). FIG. 27 B shows the tumor growth for each individual mouse in the vehicle group, FIG. 27 C of the mice treated with CEA TCB alone, FIG. 27 D of mice treated with CEA TCB and CEA(T84.66)-CD28 (SA_Variant 15) and FIG. 27 E of mice treated with CEA TCB and CEA(A5H1EL1D)-CD28 (SA_Variant 15). It can be seen that there is increased TCB-mediated tumor regression in the presence of both bispecific CEA-CD28 antibodies.

FIG. 28 shows the study design of an efficacy study with bispecific CEA-CD28 antibodies (comparison of different CD28 clones) in combination with CEACAM5 TCB in BXPC3 Xenograft in humanized mice. Shown is the design and the different treatment groups.

FIG. 29 shows the tumor growth kinetics (Mean, +SEM) for all treatment groups, the corresponding TGI values of each treatment arm are shown in Table 33 (Example 13.2).

The ex vivo Immuno-PD data are shown in FIGS. 30 A to 30 D . FIG. 30 A shows representative dot plots (CD3 against CD45 and CD4 against CD8) of the stained tumor single cell suspensions of each treatment arm. The summary of CD3, CD8 and CD4 T cell infiltration is depicted in FIG. 30 B (CD3), FIG. 30 C (CD8) and FIG. 30 D (CD4), respectively.

FIG. 31 shows the study design of an efficacy study with bispecific CEA-CD28 antibody (CEA(A5H1EL1D)-CD28 (SA_Variant 8)) in combination with CEA TCB in MKN45 Xenograft in humanized mice. Shown is the design and the different treatment groups.

FIG. 32 shows the tumor growth kinetics (Mean, +SEM) for all treatment groups, the corresponding TGI values of each treatment arm are shown in Table 35 (Example 13.3).

The ex vivo Immuno-PD data are shown in FIGS. 33 A and 33 B . FIG. 33 A shows representative dot plots of the stained tumor single cell suspensions of each treatment arm. The summary of CD3+ T cell infiltration is depicted in FIG. 33 B .

›BRIEF DESCRIPTION OF THE DRAWINGS · 5 of 6

In FIGS. 34 A to 34 D schematic illustrations of bispecific CD28 antigen binding molecules as described in Example 14 are shown.

FIG. 34 A shows a bispecific EpCAM-CD28 antigen binding molecule in 1+1 format, wherein the CD28 antigen binding domain is represented as crossFab (VH/VL exchange) and in the Fab fragment bearing the EpCAM antigen binding domain there are charged modifications in order to support the correct pairing of the light chains. The Fc domain has knob into hole modifications and the P329G LALA mutations to abrogate the binding to Fey receptors.

In FIG. 34 B the Fab bearing the CD28 antigen binding domain comprises charged modifications and the Fab bearing the HERS antigen binding domain is represented as crossFab (VH/VL exchange).

FIG. 34 C illustrates a bispecific CD30-CD28 antigen binding molecule in 1+1 format, wherein Fab molecule bearing the CD28 antigen binding domain comprises charged modifications and the Fab bearing the CD30 antigen binding domain is represented as crossFab (VH/VL exchange).

FIG. 34 D illustrates a bispecific TPBG-CD28 antigen binding molecule in 1+1 format, wherein Fab molecule bearing the CD28 antigen binding domain comprises charged modifications and the Fab bearing the TPBG (5T4) antigen binding domain is represented as crossFab (VH/VL exchange).

FIGS. 35 A to 35 C relate to the functional characterization of EpCAM-CD28 bispecific antigen binding molecules. In FIG. 35 A it is shown that EpCAM-CD28 (Molecule 14A) binds to human CD28 on CHO-k1 cells expressing CD28, assessed by flow cytometry. The binding to EpCAM on HT29 cells, assessed by flow cytometry, is shown in FIG. 35 B . Anti-DP47 served as negative control for unspecific binding of antibody compounds to cells. Dots represent means of technical duplicates. In FIG. 35 C it is shown that EpCAM-CD28 (P1AE9051) enhances T cell responses to anti-CD3 stimulus in the IL-2 reporter assay. Shown is IL-2 reporter cell activation measured by luminescence readout after 6 hours of co-incubation with HT-29 in presence of suboptimal concentrations of anti-CD3 IgG (10 nM) and increasing concentrations of EpCAM-CD28. Dots represent means of technical duplicates.

FIGS. 36 A to 36 C relate to the functional characterization of HER3-CD28 bispecific antigen binding molecules. In FIG. 36 A it is shown that HER3-CD28 (P1AF0151) binds to human CD28 on CHO-kl cells expressing CD28, assessed by flow cytometry. The binding to of HER3-CD28 to HER3 on T-47D cells, assessed by flow cytometry, is shown in FIG. 36 B . Anti-DP47 served as negative control for unspecific binding of antibody compounds to cells. Dots are means of technical duplicates. In FIG. 36 C it is shown that HER3-CD28 (P1AF0151) enhances T cell responses to anti-CD3 stimulus in the IL-2 reporter assay. Shown is IL-2 reporter cell activation measured by luminescence readout after 6 hours of co-incubation with T-47D cells in presence of suboptimal concentrations of anti-CD3 IgG clone OKT3 (10 nM) and increasing concentrations of HER3-CD28. Dots represent means of technical duplicates.

In FIGS. 37 A to 37 C schematic illustrations of bispecific CD28 antigen binding molecules targeting a Multiple Myeloma (MM) cell surface antigen as described in Example 16 are shown.

FIG. 37 A shows a bispecific GPRC5D-CD28 antigen binding molecule in 1+1 format, wherein the CD28 antigen binding domain is represented as crossFab (VH/VL exchange) and in the Fab fragment bearing the CPRC5D antigen binding domain there are charged modifications in order to support the correct pairing of the light chains. The Fc domain has knob into hole modifications and the P329G LALA mutations to abrogate the binding to Fcγ receptors. In FIG. 37 B the Fab bearing the CD28 antigen binding domain comprises charged modifications and the Fab bearing the CD38 antigen binding domain is represented as crossFab (VH/VL exchange).

FIG. 37 C illustrates a bispecific BCMA-CD28 antigen binding molecule in 1+1 format, wherein Fab molecule bearing the CD28 antigen binding domain is represented as crossFab (VH/VL exchange) and the Fab bearing the BCMA antigen binding domain comprises charged modifications.

FIG. 37 D illustrates the anti-GPRC5D/anti-CD3 bispecific antibody (GPRC5D TCB) in 2+1 format, wherein the Fab molecules bearing the GPCR5D antigen binding domain comprise charged modifications and the Fab bearing the CD3 antigen binding domain is represented as crossFab (VH/VL exchange).

FIGS. 38 A to 38 F relate to the binding of bispecific antigen binding molecules targeting CD28 and a Multiple Myeloma (MM) cell surface antigen to cells (Example 17.1). Shown is the binding of bispecific antigen binding molecules to either human CD28 on CHO huCD28 c145 cells ( FIGS. 38 A and 38 E ), to human CD38 on OCI-Ly18 cells ( FIG. 38 B ), human BCMA (B-cell maturation antigen, FIG. 38 C ) on IM-9 cells and to human GPRC5D on CHO huGPRC5D L2 cells ( FIGS. 38 D and 38 F ) expressed on the indicated cell lines. Depicted are relative median fluorescence valus (MFI) from duplicates with SD. EC 50 values of binding were calculated by GraphPadPrism and are included in Table 38.

The T-cell activation of bispecific antigen binding molecules targeting CD28 and a Multiple Myeloma (MM) cell surface antigen as assessed in the IL-2 reporter assay is shown in FIGS. 39 A to 39 F . Shown is the IL2-reporter cell assay after 5 and 22 hours of incubation, as determined by luminescence. IL2-reporter effector and GPRC5D-expressing target cells were incubated at a effector to target ratio (E:T) of 5:1. GPRC5D-TCB was added at a fixed final assay concentration of 1 nM, the indicated MM-targeted CD28 bispecific antigen binding molecules were titrated as indicated. Representative dose-response curves are depicted for CD38-CD28 in FIG. 39 A (after 5 hours of incubation) and FIG. 39 B (after 22 hours), for BCMA-CD28 in FIG. 39 C (after 5 hours) and FIG. 390 (after 22 hours) and for GPRC5D-CD28 in FIG. 39 E (after 5 hours) and FIG. 39 F (after 22 hours of incubation).

›BRIEF DESCRIPTION OF THE DRAWINGS · 6 of 6

FIGS. 40 A to 40 C show the boosting of T-cell mediated lysis of the GPRC5D-expressing MM cell line NCI-H929 in presence of 0.2 nM of the indicated CD28 bispecific molecules CD38-CD28 ( FIG. 40 A ), BCMA-CD28 ( FIG. 40 B ) and GPRC5D-CD28 ( FIG. 40 C ). Lysis was determined after co-incubation of human pan T-cells and MM tumor target cells at a final E:T ratio of 1:1 for 22 hours. Depicted are technical duplicates with SD. EC50 values and area under the curve values of tumor cell lysis were calculated by GraphPadPrism and are depicted in Table 39.

FIG. 41 A shows a schematic illustration of a bispecific CD19-CD28 antigen binding molecule in 1+1 format as described in Example 18, wherein in the Fab comprising the CD19 antigen binding domain the VH and VL domains are exchanged with each other (VH/VL crossfab) and wherein in the Fab comprising the CD28 antigen binding domain certain amino acids in the CH1 and CL domain are exchanged (charge variants) to allow better pairing with the light chain. FIG. 41 B shows a corresponding molecule wherein the CD19 antigen binding domain has been replaced by a CD79b antigen binding domain (anti-CD79b crossfab).

FIG. 42 relates to the determination of kinetic and thermodynamic parameters of CD79b (polatuzumab) in the construct CD79b (huMA79b.v28)-CD28 (v15) 1+1. Soluble recombinant CD79b-His was captured on CMS chip via an anti-penta-His antibody and bi-specific CD79b (huMA79b.v28)-CD28 (v15) 1+1 was used as an analyte. Smooth lines represent a global fit of the data to a 1:1 interaction mode.

In FIG. 43 A the median fluorescence intensities (MFI) of binding of CD19-CD28 variant 15 (P1AE9040) to four different B cell lines expressing different levels of CD19 are shown. Binding was assessed by flow cytometry. Shown are technical duplicates with SEM. In FIG. 43 B FACS staining for CD19 of the four different B cell lines (MFI) is depicted.

Binding of CD19-CD28 with varying CD28 affinities to human CD19 and CD28 on cells is shown in FIGS. 44 A and 44 B . Median fluorescence intensities (MFI) of binding to CHOk1-CD28 cells ( FIG. 44 A ) and of binding to CD19 on Nalm6 B cells ( FIG. 44 B ) are shown. Dots represent technical duplicates with SEM. Corresponding EC 50 values are shown in Table 42 (CHOk1-CD28) and Table 43 (Nalm6) of Example 20. Binding was assessed by flow cytometry.

In FIGS. 45 A to 45 D it is shown that CD19-CD28v15 enhances CD20-TCB in IL-2 reporter assay in presence of different B cell lines. Shown is 1L-2 reporter cell activation measured by luminescence readout (LUM) after 6 hours of co-incubation with different B cell lines in presence of suboptimal concentrations of CD20-TCB and increasing concentrations of CD19-CD28v15. Dots represent technical duplicates with SEM. Suboptimal CD20-TCB concentration differs with target cell lines: 10 nM for Nalm6, 0.05 nM for RCK8, WSU DLCL2 and Z138.

FIG. 46 illustrates that CD19-CD28 with various CD28 affinities enhances CD20-TCB mediated T cell activation. Shown is IL-2 reporter cell activation measured by luminescence readout (LUM) after 6 hours of co-incubation with Nalm6 B cells in presence of suboptimal concentrations of CD20-TCB (10 nM) and increasing concentrations of CD19-CD28v15. Dots represent technical duplicates with SEM.

The activation status of PBMC-derived T cells after co-culture with CD20-expressing target cells (Nalm6) (E:T ratio 5:1) and CD19-CD28 in absence or presence of CD20-TCB was assessed. The activity of CD19-CD28 in absence or presence of TCR signals is shown in FIG. 47 . Shown is CD69 expression of PBMC-derived CD4 T cells after 48 h of co-incubation with Nalm6 cells, increasing concentrations of CD19-CD28v15 in presence or absence of 10 nM CD20-TCB. Dots represent technical triplicates with SEM.

In FIGS. 48 A to 48 D it is illustrated that CD19-CD28 alone does not induce cytokine secretion in PBMCs. Shown is cytokine release in whole PBMCs after 48 hours of co-culture with CD19-CD28 molecules in presence or absence of CD20-TCB. Bars represent mean+SEM of technical triplicates. Data are representative of 2 donors. Cytokine secretion was assessed by Bio-Plex Pro Human Cytokine 17-plex Assay. Shown are IFNγ ( FIG. 48 A ), IL-2 ( FIG. 48 B ), IL-10 ( FIG. 48 C ), and TNF ( FIG. 48 D ).

In FIGS. 49 A and 49 B , functional data relating to CD79b-CD28 are shown enhances CD20-TCB in IL-2 reporter assay in presence of Z138 B cells. In FIG. 49 A , median fluorescence intensities (MFI) of binding to CD79b on Z138 B cells is shown. In FIG. 49 B it is shown that CD79b-CD28 enhances CD20-TCB in IL-2 reporter assay in presence of Z138 B cells. Shown is IL-2 reporter cell activation measured by luminescence readout (LUM) after 6 hours of co-incubation with different B cell lines in presence of suboptimal concentrations of CD20-TCB and increasing concentrations of CD79b-CD28. Dots represent technical duplicates with SEM.

FIG. 50 shows the study design of an efficacy study with bispecific CD19-CD28 antibodies (comparison of two different CD28 clones) in NALM6 Xenograft in humanized mice. Shown is the design and the different treatment groups.

FIGS. 51 A to 51 D show results of the efficacy study with CD19-CD28 in NALM6 Xenograft in humanized mice. Shown is the average tumor volume ( FIG. 51 A ) or the growth of tumors in individual mice for the three treatment groups as plotted on the y-axis ( FIG. 51 B to 51 D ). FIG. 51 B shows the tumor growth for each individual mouse in the vehicle group, FIG. 51 C of the mice treated with CD19-CD28 (variant 15) and FIG. 51 D of mice treated with CD19-CD28 (variant 8). It can be seen that CD19-CD28 (variant 8) as a single agent induced stronger tumor growth inhibition as compared to CD19-CD28 (variant 15).

DETAILED DESCRIPTION OF THE INVENTION

›Definitions · 1 of 56

Unless defined otherwise, technical and scientific terms used herein have the same meaning as generally used in the art to which this invention belongs. For purposes of interpreting this specification, the following definitions will apply and whenever appropriate, terms used in the singular will also include the plural and vice versa.

As used herein, the term “antigen binding molecule” refers in its broadest sense to a molecule that specifically binds an antigenic determinant. Examples of antigen binding molecules are antibodies, multispecific antibodies (e.g., bispecific antibodies), antibody fragments and scaffold antigen binding proteins.

As used herein, the term “antigen binding domain that binds to a tumor-associated antigen” or “moiety capable of specific binding to a tumor-associated antigen” refers to a polypeptide molecule that specifically binds to an antigenic determinant. In one aspect, the antigen binding domain is able to activate signaling through its target cell antigen. In a particular aspect, the antigen binding domain is able to direct the entity to which it is attached (e.g. the CD28 antibody) to a target site, for example to a specific type of tumor cell or tumor stroma bearing the antigenic determinant. Antigen binding domains capable of specific binding to a target cell antigen include antibodies and fragments thereof as further defined herein. In addition, antigen binding domains capable of specific binding to a target cell antigen include scaffold antigen binding proteins as further defined herein, e.g. binding domains which are based on designed repeat proteins or designed repeat domains (see e.g. WO 2002/020565).

In relation to an antigen binding molecule, i.e. an antibody or fragment thereof, the term “antigen binding domain that binds to a target cell antigen” refers to the part of the molecule that comprises the area which specifically binds to and is complementary to part or all of an antigen. An antigen binding domain capable of specific antigen binding may be provided, for example, by one or more antibody variable domains (also called antibody variable regions). Particularly, an antigen binding domain capable of specific antigen binding comprises an antibody light chain variable region (VL) and an antibody heavy chain variable region (VH). In another aspect, the “antigen binding domain capable of specific binding to a tumor-associated antigen” can also be a Fab fragment or a crossFab fragment. In another aspect, the “antigen binding domain capable of specific binding to a tumor-associated antigen” can also be a Fab fragment or a crossFab fragment. As used herein, the terms “first”, “second” or “third” with respect to antigen binding domains etc., are used for convenience of distinguishing when there is more than one of each type of moiety. Use of these terms is not intended to confer a specific order or orientation of the moiety unless explicitly so stated.

As used herein, the term “antigen binding domain that binds to a B cell surface antigen” or “moiety capable of specific binding to a B cell surface antigen” refers to a polypeptide molecule that specifically binds to an antigenic determinant on the B cell surface. In one aspect, the antigen binding domain is able to activate signaling through its target cell antigen. In a particular aspect, the antigen binding domain is able to direct the entity to which it is attached (e.g. the CD28 agonist) to a target site, e.g. on the B cell. Antigen binding domains capable of specific binding to a B cell surface antigen include antibodies and fragments thereof as further defined herein. In addition, antigen binding domains capable of specific binding to a B cell surface antigen include scaffold antigen binding proteins as further defined herein, e.g. binding domains which are based on designed repeat proteins or designed repeat domains (see e.g. WO 2002/020565).

The term “antigen binding domain that binds to a Multiple Myeloma (MM) cell surface antigen” or “moiety capable of specific binding to a Multiple Myeloma (MM) cell surface antigen” refers to a polypeptide molecule that specifically binds to an antigenic determinant on Multiple Myeloma (MM) cell. In one aspect, the antigen binding domain is able to activate signaling through its target cell antigen. In a particular aspect, the antigen binding domain is able to direct the entity to which it is attached (e.g. the CD28 agonist) to a target site, e.g. on the MM cell. Antigen binding domains capable of specific binding to a Multiple Myeloma (MM) cell surface antigen include antibodies and fragments thereof as further defined herein. In addition, antigen binding domains capable of specific binding to a B cell surface antigen include scaffold antigen binding proteins as further defined herein, e.g. binding domains which are based on designed repeat proteins or designed repeat domains (see e.g. WO 2002/020565).

The term “antibody” herein is used in the broadest sense and encompasses various antibody structures, including but not limited to monoclonal antibodies, polyclonal antibodies, monospecific and multispecific antibodies (e.g., bispecific antibodies), and antibody fragments so long as they exhibit the desired antigen-binding activity.

The term “monoclonal antibody” as used herein refers to an antibody obtained from a population of substantially homogeneous antibodies, i.e., the individual antibodies comprising the population are identical and/or bind the same epitope, except for possible variant antibodies, e.g. containing naturally occurring mutations or arising during production of a monoclonal antibody preparation, such variants generally being present in minor amounts. In contrast to polyclonal antibody preparations, which typically include different antibodies directed against different determinants (epitopes), each monoclonal antibody of a monoclonal antibody preparation is directed against a single determinant on an antigen.

The term “monospecific” antibody as used herein denotes an antibody that has one or more binding sites each of which bind to the same epitope of the same antigen. The term “bispecific” means that the antigen binding molecule is able to specifically bind to at least two distinct antigenic determinants. Typically, a bispecific antigen binding molecule comprises two antigen binding sites, each of which is specific for a different antigenic determinant. However, a bispecific antigen binding molecule may also comprise additional antigen binding sites which bind to further antigenic determinants. In certain aspects, the bispecific antigen binding molecule is capable of simultaneously binding two antigenic determinants, particularly two antigenic determinants expressed on two distinct cells or on the same cell. The term “bispecific” in accordance with the present invention thus may also include a trispecific molecule, e.g. a bispecific molecule comprising a CD28 antibody and two antigen binding domains directed to two different target cell antigens.

›Definitions · 2 of 56

The term “valent” as used within the current application denotes the presence of a specified number of binding sites specific for one distinct antigenic determinant in an antigen binding molecule that are specific for one distinct antigenic determinant. As such, the terms “bivalent”, “tetravalent”, and “hexavalent” denote the presence of two binding sites, four binding sites, and six binding sites specific for a certain antigenic determinant, respectively, in an antigen binding molecule. In particular aspects of the invention, the bispecific antigen binding molecules according to the invention can be monovalent for a certain antigenic determinant, meaning that they have only one binding site for said antigenic determinant or they can be bivalent or tetravalent for a certain antigenic determinant, meaning that they have two binding sites or four binding sites, respectively, for said antigenic determinant.

The terms “full length antibody”, “intact antibody”, and “whole antibody” are used herein interchangeably to refer to an antibody having a structure substantially similar to a native antibody structure. “Native antibodies” refer to naturally occurring immunoglobulin molecules with varying structures. For example, native IgG-class antibodies are heterotetrameric glycoproteins of about 150,000 daltons, composed of two light chains and two heavy chains that are disulfide-bonded. From N- to C-terminus, each heavy chain has a variable region (VH), also called a variable heavy domain or a heavy chain variable domain, followed by three constant domains (CH1, CH2, and CH3), also called a heavy chain constant region. Similarly, from N- to C-terminus, each light chain has a variable region (VL), also called a variable light domain or a light chain variable domain, followed by a light chain constant domain (CL), also called a light chain constant region. The heavy chain of an antibody may be assigned to one of five types, called α (IgA), δ (IgD), ε (IgE), γ (IgG), or μ (IgM), some of which may be further divided into subtypes, e.g. γ1 (IgG1), γ2 (IgG2), γ3 (IgG3), γ4 (IgG4), al (IgA1) and α2 (IgA2). The light chain of an antibody may be assigned to one of two types, called kappa (κ) and lambda (λ), based on the amino acid sequence of its constant domain.

An “antibody fragment” refers to a molecule other than an intact antibody that comprises a portion of an intact antibody that binds the antigen to which the intact antibody binds. Examples of antibody fragments include but are not limited to Fv, Fab, Fab′, Fab′-SH, F(ab′)2; diabodies, triabodies, tetrabodies, crossFab fragments; linear antibodies; single-chain antibody molecules (e.g. scFv); and single domain antibodies. For a review of certain antibody fragments, see Hudson et al., Nat Med 9, 129-134 (2003). For a review of scFv fragments, see e.g. Plückthun, in The Pharmacology of Monoclonal Antibodies, vol. 113, Rosenburg and Moore eds., Springer-Verlag, New York, pp. 269-315 (1994); see also WO 93/16185; and U.S. Pat. Nos. 5,571,894 and 5,587,458. For discussion of Fab and F(ab′)2 fragments comprising salvage receptor binding epitope residues and having increased in vivo half-life, see U.S. Pat. No. 5,869,046. Diabodies are antibody fragments with two antigen-binding sites that may be bivalent or bispecific, see, for example, EP 404,097; WO 1993/01161; Hudson et al., Nat Med 9, 129-134 (2003); and Hollinger et al., Proc Natl Acad Sci USA 90, 6444-6448 (1993). Triabodies and tetrabodies are also described in Hudson et al., Nat Med 9, 129-134 (2003). Single-domain antibodies are antibody fragments comprising all or a portion of the heavy chain variable domain or all or a portion of the light chain variable domain of an antibody. In certain embodiments, a single-domain antibody is a human single-domain antibody (Domantis, Inc., Waltham, Mass.; see e.g. U.S. Pat. No. 6,248,516 B1). Antibody fragments can be made by various techniques, including but not limited to proteolytic digestion of an intact antibody as well as production by recombinant host cells (e.g. E. coli or phage), as described herein.

Papain digestion of intact antibodies produces two identical antigen-binding fragments, called “Fab” fragments containing each the heavy- and light-chain variable domains and also the constant domain of the light chain and the first constant domain (CH1) of the heavy chain. As used herein, Thus, the term “Fab fragment” refers to an antibody fragment comprising a light chain fragment comprising a variable light chain (VL) domain and a constant domain of a light chain (CL), and a variable heavy chain (VH) domain and a first constant domain (CH1) of a heavy chain. Fab′ fragments differ from Fab fragments by the addition of a few residues at the carboxy terminus of the heavy chain CH1 domain including one or more cysteins from the antibody hinge region. Fab′-SH are Fab′ fragments in which the cysteine residue(s) of the constant domains bear a free thiol group. Pepsin treatment yields an F(ab′)2 fragment that has two antigen-combining sites (two Fab fragments) and a part of the Fc region.

The term “crossFab fragment” or “xFab fragment” or “crossover Fab fragment” refers to a Fab fragment, wherein either the variable regions or the constant regions of the heavy and light chain are exchanged. Two different chain compositions of a crossover Fab molecule are possible and comprised in the bispecific antibodies of the invention: On the one hand, the variable regions of the Fab heavy and light chain are exchanged, i.e. the crossover Fab molecule comprises a peptide chain composed of the light chain variable (VL) domain and the heavy chain constant domain (CHI), and a peptide chain composed of the heavy chain variable domain (VH) and the light chain constant domain (CL). This crossover Fab molecule is also referred to as CrossFab (VLVH) . On the other hand, when the constant regions of the Fab heavy and light chain are exchanged, the crossover Fab molecule comprises a peptide chain composed of the heavy chain variable domain (VH) and the light chain constant domain (CL), and a peptide chain composed of the light chain variable domain (VL) and the heavy chain constant domain (CH1). This crossover Fab molecule is also referred to as CrossFab (CLCH1) .

›Definitions · 3 of 56

A “single chain Fab fragment” or “scFab” is a polypeptide consisting of an antibody heavy chain variable domain (VH), an antibody constant domain 1 (CH1), an antibody light chain variable domain (VL), an antibody light chain constant domain (CL) and a linker, wherein said antibody domains and said linker have one of the following orders in N-terminal to C-terminal direction: a) VH-CH1-linker-VL-CL, b) VL-CL-linker-VH-CH1, c) VH-CL-linker-VL-CH1 or d) VL-CH1-linker-VH-CL; and wherein said linker is a polypeptide of at least 30 amino acids, preferably between 32 and 50 amino acids. Said single chain Fab fragments are stabilized via the natural disulfide bond between the CL domain and the CH1 domain. In addition, these single chain Fab molecules might be further stabilized by generation of interchain disulfide bonds via insertion of cysteine residues (e.g. position 44 in the variable heavy chain and position 100 in the variable light chain according to Kabat numbering).

A “crossover single chain Fab fragment” or “x-scFab” is a is a polypeptide consisting of an antibody heavy chain variable domain (VH), an antibody constant domain 1 (CH1), an antibody light chain variable domain (VL), an antibody light chain constant domain (CL) and a linker, wherein said antibody domains and said linker have one of the following orders in N-terminal to C-terminal direction: a) VH-CL-linker-VL-CH1 and b) VL-CH1-linker-VH-CL; wherein VH and VL form together an antigen-binding site which binds specifically to an antigen and wherein said linker is a polypeptide of at least 30 amino acids. In addition, these x-scFab molecules might be further stabilized by generation of interchain disulfide bonds via insertion of cysteine residues (e.g. position 44 in the variable heavy chain and position 100 in the variable light chain according to Kabat numbering).

A “single-chain variable fragment (scFv)” is a fusion protein of the variable regions of the heavy (V H ) and light chains (V L ) of an antibody, connected with a short linker peptide of ten to about 25 amino acids. The linker is usually rich in glycine for flexibility, as well as serine or threonine for solubility, and can either connect the N-terminus of the V H with the C-terminus of the V L , or vice versa. This protein retains the specificity of the original antibody, despite removal of the constant regions and the introduction of the linker. scFv antibodies are, e.g. described in Houston, J. S., Methods in Enzymol. 203 (1991) 46-96). In addition, antibody fragments comprise single chain polypeptides having the characteristics of a VH domain, namely being able to assemble together with a VL domain, or of a VL domain, namely being able to assemble together with a VH domain to a functional antigen binding site and thereby providing the antigen binding property of full length antibodies.

“Scaffold antigen binding proteins” are known in the art, for example, fibronectin and designed ankyrin repeat proteins (DARPins) have been used as alternative scaffolds for antigen-binding domains, see, e.g., Gebauer and Skerra, Engineered protein scaffolds as next-generation antibody therapeutics. Curr Opin Chem Biol 13:245-255 (2009) and Stumpp et al., Darpins: A new generation of protein therapeutics. Drug Discovery Today 13: 695-701 (2008). In one aspect of the invention, a scaffold antigen binding protein is selected from the group consisting of CTLA-4 (Evibody), Lipocalins (Anticalin), a Protein A-derived molecule such as Z-domain of Protein A (Affibody), an A-domain (Avimer/Maxibody), a serum transferrin (trans-body); a designed ankyrin repeat protein (DARPin), a variable domain of antibody light chain or heavy chain (single-domain antibody, sdAb), a variable domain of antibody heavy chain (nanobody, aVH), V NAR fragments, a fibronectin (AdNectin), a C-type lectin domain (Tetranectin); a variable domain of a new antigen receptor beta-lactamase (V NAR fragments), a human gamma-crystallin or ubiquitin (Affilin molecules); a kunitz type domain of human protease inhibitors, microbodies such as the proteins from the knottin family, peptide aptamers and fibronectin (adnectin). CTLA-4 (Cytotoxic T Lymphocyte-associated Antigen 4) is a CD28-family receptor expressed on mainly CD4 + T-cells. Its extracellular domain has a variable domain-like Ig fold. Loops corresponding to CDRs of antibodies can be substituted with heterologous sequence to confer different binding properties. CTLA-4 molecules engineered to have different binding specificities are also known as Evibodies (e.g. U.S. Pat. No. 7,166,697B1). Evibodies are around the same size as the isolated variable region of an antibody (e.g. a domain antibody). For further details, see Journal of Immunological Methods 248 (1-2), 31-45 (2001). Lipocalins are a family of extracellular proteins which transport small hydrophobic molecules such as steroids, bilins, retinoids and lipids. They have a rigid beta-sheet secondary structure with a number of loops at the open end of the conical structure which can be engineered to bind to different target antigens. Anticalins are between 160-180 amino acids in size, and are derived from lipocalins. For further details, see Biochim Biophys Acta 1482: 337-350 (2000), U.S. Pat. No. 7,250,297B1 and US20070224633. An affibody is a scaffold derived from Protein A of Staphylococcus aureus which can be engineered to bind to antigen. The domain consists of a three-helical bundle of approximately 58 amino acids. Libraries have been generated by randomization of surface residues. For further details, see Protein Eng. Des. Sel. 2004, 17, 455-462 and EP 1641818A1. Avimers are multidomain proteins derived from the A-domain scaffold family. The native domains of approximately 35 amino acids adopt a defined disulfide bonded structure. Diversity is generated by shuffling of the natural variation exhibited by the family of A-domains. For further details, see Nature Biotechnology 23(12), 1556-1561 (2005) and Expert Opinion on Investigational Drugs 16(6), 909-917 (June 2007). A transferrin is a monomeric serum transport glycoprotein. Transferrins can be engineered to bind different target antigens by insertion of peptide sequences in a permissive surface loop. Examples of engineered transferrin scaffolds include the Trans-body. For further details, see J. Biol. Chem 274, 24066-24073 (1999). Designed Ankyrin Repeat Proteins (DARPins) are derived from Ankyrin which is a family of proteins that mediate attachment of integral membrane proteins to the cytoskeleton. A single ankyrin repeat is a 33 residue motif consisting of two alpha-helices and a beta-turn. They can be engineered to bind different target antigens by randomizing residues in the first alpha-helix and a beta-turn of each repeat. Their binding interface can be increased by increasing the number of modules (a method of affinity maturation). For further details, see J. Mol. Biol. 332, 489-503 (2003), PNAS 100(4), 1700-1705 (2003) and J. Mol. Biol. 369, 1015-1028 (2007) and US20040132028A1. A single-domain antibody is an antibody fragment consisting of a single monomeric variable antibody domain. The first single domains were derived from the variable domain of the antibody heavy chain from camelids (nanobodies or V H H fragments). Furthermore, the term single-domain antibody includes an autonomous human heavy chain variable domain (aVH) or V NAR fragments derived from sharks. Fibronectin is a scaffold which can be engineered to bind to antigen. Adnectins consists of a backbone of the natural amino acid sequence of the 10th domain of the 15 repeating units of human fibronectin type III (FN3). Three loops at one end of the beta-sandwich can be engineered to enable an Adnectin to specifically recognize a therapeutic target of interest. For further details, see Protein Eng. Des. Sel. 18, 435-444 (2005), US20080139791, WO2005056764 and U.S. Pat. No. 6,818,418B1. Peptide aptamers are combinatorial recognition molecules that consist of a constant scaffold protein, typically thioredoxin (TrxA) which contains a constrained variable peptide loop inserted at the active site. For further details, see Expert Opin. Biol. Ther. 5, 783-797 (2005). Microbodies are derived from naturally occurring microproteins of 25-50 amino acids in length which contain 3-4 cysteine bridges—examples of microproteins include KalataBI and conotoxin and knottins. The microproteins have a loop which can be engineered to include up to 25 amino acids without affecting the overall fold of the microprotein. For further details of engineered knottin domains, see WO2008098796.

›Definitions · 4 of 56

An “antigen binding molecule that binds to the same epitope” as a reference molecule refers to an antigen binding molecule that blocks binding of the reference molecule to its antigen in a competition assay by 50% or more, and conversely, the reference molecule blocks binding of the antigen binding molecule to its antigen in a competition assay by 50% or more.

The term “antigen binding domain” refers to the part of an antigen binding molecule that comprises the area which specifically binds to and is complementary to part or all of an antigen. Where an antigen is large, an antigen binding molecule may only bind to a particular part of the antigen, which part is termed an epitope. An antigen binding domain may be provided by, for example, one or more variable domains (also called variable regions). Preferably, an antigen binding domain comprises an antibody light chain variable domain (VL) and an antibody heavy chain variable domain (VH).

As used herein, the term “antigenic determinant” is synonymous with “antigen” and “epitope,” and refers to a site (e.g. a contiguous stretch of amino acids or a conformational configuration made up of different regions of non-contiguous amino acids) on a polypeptide macromolecule to which an antigen binding moiety binds, forming an antigen binding moiety-antigen complex. Useful antigenic determinants can be found, for example, on the surfaces of tumor cells, on the surfaces of virus-infected cells, on the surfaces of other diseased cells, on the surface of immune cells, free in blood serum, and/or in the extracellular matrix (ECM). The proteins useful as antigens herein can be any native form the proteins from any vertebrate source, including mammals such as primates (e.g. humans) and rodents (e.g. mice and rats), unless otherwise indicated. In a particular embodiment the antigen is a human protein. Where reference is made to a specific protein herein, the term encompasses the “full-length”, unprocessed protein as well as any form of the protein that results from processing in the cell. The term also encompasses naturally occurring variants of the protein, e.g. splice variants or allelic variants.

By “specific binding” is meant that the binding is selective for the antigen and can be discriminated from unwanted or non-specific interactions. The ability of an antigen binding molecule to bind to a specific antigen can be measured either through an enzyme-linked immunosorbent assay (ELISA) or other techniques familiar to one of skill in the art, e.g. Surface Plasmon Resonance (SPR) technique (analyzed on a BIAcore instrument) (Liljeblad et al., Glyco J 17, 323-329 (2000)), and traditional binding assays (Heeley, Endocr Res 28, 217-229 (2002)). In one embodiment, the extent of binding of an antigen binding molecule to an unrelated protein is less than about 10% of the binding of the antigen binding molecule to the antigen as measured, e.g. by SPR. In certain embodiments, an molecule that binds to the antigen has a dissociation constant (Kd) of ≤1 μM, ≤100 nM, ≤10 nM, ≤1 nM, ≤0.1 nM, ≤0.01 nM, or 0.001 nM (e.g. 10 −8 M or less, e.g. from 10 −8 M to 10 −13 M, e.g. from 10 −9 M to 10 −13 M).

“Affinity” or “binding affinity” refers to the strength of the sum total of non-covalent interactions between a single binding site of a molecule (e.g. an antibody) and its binding partner (e.g. an antigen). Unless indicated otherwise, as used herein, “binding affinity” refers to intrinsic binding affinity which reflects a 1:1 interaction between members of a binding pair (e.g. antibody and antigen). The affinity of a molecule X for its partner Y can generally be represented by the dissociation constant (Kd), which is the ratio of dissociation and association rate constants (koff and kon, respectively). Thus, equivalent affinities may comprise different rate constants, as long as the ratio of the rate constants remains the same. Affinity can be measured by common methods known in the art, including those described herein. A particular method for measuring affinity is Surface Plasmon Resonance (SPR).

A “tumor-associated antigen” or TAA as used herein refers to an antigenic determinant presented on the surface of a target cell, for example a cell in a tumor such as a cancer cell, a cell of the tumor stroma, a malignant B lymphocyte or a melanoma cell. In certain aspects, the target cell antigen is an antigen on the surface of a tumor cell. In one aspect, TAA is selected from the group consisting of Fibroblast Activation Protein (FAP), Carcinoembryonic Antigen (CEA), Folate receptor alpha (FolR1), Melanoma-associated Chondroitin Sulfate Proteoglycan (MCSP), Epidermal Growth Factor Receptor (EGFR), human epidermal growth factor receptor 2 (HER2), p95HER2, EpCAM, HER3, CD30 or TPBG (5T4), CD19, CD79b, CD20, CD22, CD37, CD38, BCMA and GPRC5D. In one particular aspect, TAA is selected from the group consisting of Fibroblast Activation Protein (FAP), Carcinoembryonic Antigen (CEA), Folate receptor alpha (FolR1), Melanoma-associated Chondroitin Sulfate Proteoglycan (MCSP), Epidermal Growth Factor Receptor (EGFR), human epidermal growth factor receptor 2 (HER2) and p95HER2. In another particular aspect, TAA is selected from the group consisting of Fibroblast Activation Protein (FAP), Carcinoembryonic Antigen (CEA), EpCAM, HER3, CD30 or TPBG (5T4). In one particular aspect, the tumor-associated antigen is Fibroblast Activation Protein (FAP) or Carcinoembryonic Antigen (CEA). In one aspect, TAA is a B cell surface antigen selected from the group consisting of CD19, CD79b, CD20, CD22 and CD37, in particular CD19 and CD79b. In one aspect, TAA is a Multiple Myeloma (MM) cell surface antigen selected from the group consisting of CD38, BCMA and GPRC5D.

The term “Fibroblast activation protein (FAP)”, also known as Prolyl endopeptidase FAP or Seprase (EC 3.4.21), refers to any native FAP from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The term encompasses “full-length,” unprocessed FAP as well as any form of FAP that results from processing in the cell. The term also encompasses naturally occurring variants of FAP, e.g., splice variants or allelic variants. In one embodiment, the antigen binding molecule of the invention is capable of specific binding to human, mouse and/or cynomolgus FAP. The amino acid sequence of human FAP is shown in UniProt (world wide web.uniprot.org) accession no. Q12884 (version 149, SEQ ID NO:2), or NCBI (world wide web.ncbi.nlm.nih.gov/) RefSeq NP_004451.2. The extracellular domain (ECD) of human FAP extends from amino acid position 26 to 760. The amino acid sequence of a His-tagged human FAP ECD is shown in SEQ ID NO:135. The amino acid sequence of mouse FAP is shown in UniProt accession no. P97321 (version 126, SEQ ID NO:136), or NCBI RefSeq NP_032012.1. The extracellular domain (ECD) of mouse FAP extends from amino acid position 26 to 761. SEQ ID NO:137 shows the amino acid sequence of a His-tagged mouse FAP ECD. SEQ ID NO 138 shows the amino acid sequence, of a His-tagged cynomolgus FAP ECD. Preferably, an anti-FAP binding molecule of the invention binds to the extracellular domain of FAP.

›Definitions · 5 of 56

The term “Carcinoembroynic antigen (CEA)”, also known as Carcinoembryonic antigen-related cell adhesion molecule 5 (CEACAM5), refers to any native CEA from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CEA is shown in UniProt accession no. P06731 (version 151, SEQ ID NO:3). CEA has long been identified as a tumor-associated antigen (Gold and Freedman, J Exp Med., 121:439-462, 1965; Berinstein N. L., J Clin Oncol., 20:2197-2207, 2002). Originally classified as a protein expressed only in fetal tissue, CEA has now been identified in several normal adult tissues. These tissues are primarily epithelial in origin, including cells of the gastrointestinal, respiratory, and urogential tracts, and cells of colon, cervix, sweat glands, and prostate (Nap et al., Tumour Biol., 9(2-3):145-53, 1988; Nap et al., Cancer Res., 52(8):2329-23339, 1992). Tumors of epithelial origin, as well as their metastases, contain CEA as a tumor associated antigen. While the presence of CEA itself does not indicate transformation to a cancerous cell, the distribution of CEA is indicative. In normal tissue, CEA is generally expressed on the apical surface of the cell (Hammarström S., Semin Cancer Biol. 9(2):67-81 (1999)), making it inaccessible to antibody in the blood stream. In contrast to normal tissue, CEA tends to be expressed over the entire surface of cancerous cells (Hammarström S., Semin Cancer Biol. 9(2):67-81 (1999)). This change of expression pattern makes CEA accessible to antibody binding in cancerous cells. In addition, CEA expression increases in cancerous cells. Furthermore, increased CEA expression promotes increased intercellular adhesions, which may lead to metastasis (Marshall J., Semin Oncol., 30(a Suppl. 8):30-6, 2003). The prevalence of CEA expression in various tumor entities is generally very high. In concordance with published data, own analyses performed in tissue samples confirmed its high prevalence, with approximately 95% in colorectal carcinoma (CRC), 90% in pancreatic cancer, 80% in gastric cancer, 60% in non-small cell lung cancer (NSCLC, where it is co-expressed with HER3), and 40% in breast cancer; low expression was found in small cell lung cancer and glioblastoma.

CEA is readily cleaved from the cell surface and shed into the blood stream from tumors, either directly or via the lymphatics. Because of this property, the level of serum CEA has been used as a clinical marker for diagnosis of cancers and screening for recurrence of cancers, particularly colorectal cancer (Goldenberg D M., The International Journal of Biological Markers, 7:183-188, 1992; Chau I., et al., J Clin Oncol., 22:1420-1429, 2004; Flamini et al., Clin Cancer Res; 12(23):6985-6988, 2006).

The term “epithelial cell adhesion molecule (EpCAM)” refers to any native EpCAM from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The term encompasses “full-length,” unprocessed EpCAM as well as any form of EpCAM that results from processing in the cell. The term also encompasses naturally occurring variants of EpCAM, e.g., splice variants or allelic variants. In one embodiment, the antigen binding molecule of the invention is capable of specific binding to human, mouse and/or cynomolgus EpCAM. The amino acid sequence of human EpCAM is shown in UniProt (world wide web.uniprot.org) accession no. P16422 (version 167, SEQ ID NO:68), or NCBI (world wide web.ncbi.nlm.nih.gov/) RefSeq NP_002345.2. The amino acid sequence of mouse EpCAM is shown in UniProt (world wide web.uniprot.org) accession no. Q99JW5 (version 111, SEQ ID NO:75), or NCBI (world wide web.ncbi.nlm.nih.gov/) RefSeq NP_032558.2. Epithelial cell adhesion molecule (EpCAM)—also known as tumor-associated calcium signal transducer 1 (TACSTD1), 17-1A and CD326—is a type I ˜40 kDa transmembrane glycoprotein that is frequently overexpressed in cancers of epithelial origin and by cancer stem cells, and is therefore a molecule of significant interest for therapy and diagnosis. The extracellular domain EpCAM can be cleaved to yield the soluble extracellular domain molecule EpEX, and the intracellular molecule EpICD. EpICD has been shown to associate with other proteins to form a nuclear complex which upregulates the expression of genes promoting cell proliferation. EpCAM may also be involved in the epithelial to mesenchymal cell transition (EMT), and may contribute to the formation of large metastases.

“CD30” or “TNFRSF8” is a member of the tumor necrosis factor receptor superfamily. It is characteristically expressed in certain hematopoietic malignancies, including anaplastic large cell lymphoma and Hodgkin lymphoma, among others. The variable expression of CD30 on both normal and malignant lymphoid cells has focused research efforts on understanding the pathogenesis of CD30 upregulation, its contribution to lymphomagenesis through anti-apoptotic mechanisms, and its effect on cell survival. Given the restriction of CD30 to certain tumor types, the logical extension of this has been to attempt to exploit it as a therapeutic target. The CD30 is a 120 kD transmembrane glycoprotein receptor belonging to the tumor necrosis factor receptor (TNFR) superfamily, with intracellular, trans-membrane and extracellular domains and the amino acid sequence of human CD30 is shown in UniProt accession no. P28908 (SEQ ID NO:472).

The term “TPBG” refers to Trophoblast glycoprotein, also referred to as “5T4”. TBPG is a leucine-rich transmembrane glycoprotein involved in cell adhesion. In adults this protein is highly expressed in many tumor cells and is associated with poor clinical outcome in numerous cancers. It refers to any native TPBG from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human TPBG is shown in UniProt accession no. Q13641 (SEQ ID NO: 473).

›Definitions · 6 of 56

The term “FolR1” refers to Folate receptor alpha and has been identified as a potential prognostic and therapeutic target in a number of cancers. It refers to any native FolR1 from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human FolR1 is shown in UniProt accession no. P15328 (SEQ ID NO: 139), murine FolR1 has the amino acid sequence of UniProt accession no. P35846 (SEQ ID NO:140) and cynomolgus FolR1 has the amino acid sequence as shown in UniProt accession no. G7PR14 (SEQ ID NO:141). FolR1 is an N-glycosylated protein expressed on plasma membrane of cells. Foal has a high affinity for folic acid and for several reduced folic acid derivatives and mediates delivery of the physiological folate, 5-methyltetrahydrofolate, to the interior of cells. FOLR1 is a desirable target for FOLR1-directed cancer therapy as it is overexpressed in vast majority of ovarian cancers, as well as in many uterine, endometrial, pancreatic, renal, lung, and breast cancers, while the expression of FOLR1 on normal tissues is restricted to the apical membrane of epithelial cells in the kidney proximal tubules, alveolar pneumocytes of the lung, bladder, testes, choroid plexus, and thyroid. Recent studies have identified that FolR1 expression is particularly high in triple negative breast cancers (Necela et al. PloS One 2015, 10(3), e0127133).

The term “Melanoma-associated Chondroitin Sulfate Proteoglycan (MCSP)”, also known as Chondroitin Sulfate Proteoglycan 4 (CSPG4) refers to any native MCSP from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human MCSP is shown in UniProt accession no. Q6UVK1 (version 103, SEQ ID NO:142). MCSP is a highly glycosylated integral membrane chondroitin sulfate proteoglycan consisting of an N-linked 280 kDa glycoprotein component and a 450-kDa chondroitin sulfate proteoglycan component expressed on the cell membrane (Ross et al., Arch. Biochem. Biophys. 1983, 225:370-38). MCSP is more broadly distributed in a number of normal and transformed cells. In particular, MCSP is found in almost all basal cells of the epidermis. MCSP is differentially expressed in melanoma cells, and was found to be expressed in more than 90% of benign nevi and melanoma lesions analyzed. MCSP has also been found to be expressed in tumors of nonmelanocytic origin, including basal cell carcinoma, various tumors of neural crest origin, and in breast carcinomas.

The term “Epidermal Growth Factor Receptor (EGFR)”, also named Proto-oncogene c-ErbB-1 or Receptor tyrosine-protein kinase erbB-1, refers to any native EGFR from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human EGFR is shown in UniProt accession no. P00533 (version 211, SEQ ID NO:143). The proto-oncogene “HER2”, (human epidermal growth factor receptor 2.) encodes a protein tyrosine kinase (p185HER2) that is related to and somewhat homologous to the human epidermal growth factor receptor. HER2 is also known in the field as c-erbB-2, and sometimes by the name of the rat homolog, neu. Amplification and/or overexpression of HER2 is associated with multiple human malignancies and appears to be integrally involved in progression of 25-30% of human breast and ovarian cancers. Furthermore, the extent of amplification is inversely correlated with the observed median patient survival time (Slamon, D. J. et al., Science 244:707-712 (1989)). The amino acid sequence of human HER2 is shown in UniProt accession no. P04626 (version 230, SEQ ID NO:144). The term “p95HER2” as used herein refers to a carboxy terminal fragment (CTF) of the HER2 receptor protein, which is also known as “611-CTF” or “100-115 kDa p95HER2”. The p95HER2 fragment is generated in the cell through initiation of translation of the HER2 mRNA at codon position 611 of the full-length HER2 molecule (Anido et al, EMBO J 25; 3234-44 (2006)). It has a molecular weight of 100 to 115 kDa and is expressed at the cell membrane, where it can form homodimers maintained by intermolecular disulfide bonds (Pedersen et al., Mol Cell Biol 29, 3319-31 (2009)). An exemplary sequence of human p95HER2 is given in SEQ ID NO: 145.

“HER3” or “ErbB3” (human epidermal growth factor receptor 3), like the other members of the ErbB receptor tyrosine kinase family, consists of an extracellular domain, a transmembrane domain, and an intracellular domain. The extracellular domain contains four subdomains (I-IV). Subdomains I and III are leucine-rich and are primarily involved in ligand binding. Subdomains II and IV are cysteine-rich and most likely contribute to protein conformation and stability through the formation of disulfide bonds. Subdomain II also contains the dimerization loop required for dimer formation. The cytoplasmic domain contains a juxtamembrane segment, a kinase domain, and a C-terminal domain. While no evidence has been found that ErbB3 overexpression, constitutive activation, or mutation alone is oncogenic, en.wikipedia.org/wiki/ERBB3—cite_note-pmid8632008-18 the protein as a heterodimerization partner, most critically with ErbB2, is implicated in growth, proliferation, chemotherapeutic resistance, and the promotion of invasion and metastasis. ErbB3 is associated with targeted therapeutic resistance in numerous cancers. The amino acid sequence of human HER3 is shown in UniProt accession no. P21860 (version 224, SEQ ID NO:471)

A “B cell surface antigen” as used herein refers to an antigenic determinant presented on the surface of a B lymphocyte, particularly a malignant B lymphocyte (in that case the antigen also being referred to as “malignant B-cell surface antigen”). Several B-cell surface antigens are interesting in terms of immunotherapy of hematologic malignant neoplasms. In one aspect, the B cell surface antigen is selected from the group consisting of CD19, CD79b, CD20, CD22 and CD37.

›Definitions · 7 of 56

The term “CD19” refers to B-lymphocyte antigen CD19, also known as B-lymphocyte surface antigen B4 or T-cell surface antigen Leu-12 and includes any native CD19 from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CD19 is shown in Uniprot accession no. P15391 (version 160, SEQ ID NO:434). The term encompasses “full-length” unprocessed human CD19 as well as any form of human CD19 that results from processing in the cell as long as the antibody as reported herein binds thereto. CD19 is a structurally distinct cell surface receptor expressed on the surface of human B cells, including, but not limited to, pre-B cells, B cells in early development {i.e., immature B cells), mature B cells through terminal differentiation into plasma cells, and malignant B cells. CD19 is expressed by most pre-B acute lymphoblastic leukemias (ALL), non-Hodgkin's lymphomas, B cell chronic lymphocytic leukemias (CLL), pro-lymphocytic leukemias, hairy cell leukemias, common acute lymphocytic leukemias, and some Null-acute lymphoblastic leukemias. The expression of CD19 on plasma cells further suggests it may be expressed on differentiated B cell tumors such as multiple myeloma. Therefore, the CD19 antigen is a target for immunotherapy in the treatment of non-Hodgkin's lymphoma, chronic lymphocytic leukemia and/or acute lymphoblastic leukemia.

“CD79b” refers to B-cell antigen receptor complex-associated protein beta chain, also known as Ig-beta or B cell specific glycoprotein B29, and includes any native CD79b from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CD79b is shown in Uniprot accession no. P40259 (version 180, SEQ ID NO:435). CD79b is a 39 KDa protein exclusively expressed on B cells and, in cooperation with CD79a, initiates the signal transduction cascade downstream of the BCR, which leads to internalization of the BCR complex, its translocation to the endosomes, and antigen presentation. CD79 (composed of subunits CD79a and CD79b) is a heterodimeric signal-transduction component of the B-cell receptor, ubiquitously expressed in mature B-cell lymphomas and placed on the cell surface by the earliest committed B-cell progenitors before expression of immunoglobulin μ. The term “CD79b” encompasses “full-length,” unprocessed CD79b as well as any form of CD79b that results from processing in the cell. The term also encompasses naturally occurring variants of CD79b, e.g., splice variants or allelic variants.

“CD20” refers to B-lymphocyte antigen CD20, also known as B-lymphocyte surface antigen B1 or Leukocyte surface antigen Leu-16, and includes any native CD20 from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CD20 is shown in Uniprot accession no. P11836 (version 149, SEQ ID NO:436). CD20 is a hydrophobic transmembrane protein with a molecular weight of approximately 35 kD expressed on pre-B and mature B lymphocytes. The corresponding human gene is membrane-spanning 4-domains, subfamily A, member 1, also known as MS4A1. This gene encodes a member of the membrane-spanning 4A gene family. Members of this nascent protein family are characterized by common structural features and similar intron/exon splice boundaries and display unique expression patterns among hematopoietic cells and nonlymphoid tissues. This gene encodes the B-lymphocyte surface molecule which plays a role in the development and differentiation of B-cells into plasma cells. This family member is localized to 11q12, among a cluster of family members. Alternative splicing of this gene results in two transcript variants which encode the same protein. The term “CD20” encompasses “full-length,” unprocessed CD20 as well as any form of CD20 that results from processing in the cell. The term also encompasses naturally occurring variants of CD20, e.g., splice variants or allelic variants.

“CD22” refers to B-cell receptor CD22, also known as B-lymphocyte cell adhesion molecule or SIGLEC2, and includes any native CD22 from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CD22 is shown in Uniprot accession no. P20273 (version 209, SEQ ID NO:437). CD22 is a molecule belonging to the SIGLEC family of lectins and is found on the surface of mature B cells and to lesser extent on some immature B cells. CD22 is thus a B cell restricted cell surface phosphoglycoprotein of 130-150 kDa and is capable of modulating B lymphocyte antigen receptor (BCR)-mediated signals, as well as the generation of BCR-independent signals. The term “CD22” encompasses “full-length,” unprocessed CD22 as well as any form of CD22 that results from processing in the cell. The term also encompasses naturally occurring variants of CD22, e.g., splice variants or allelic variants.

“CD37” refers to Leukocyte antigen CD37, also known as Tetraspanin-26 (Tspan-26), and includes any native CD37 from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CD37 is shown in Uniprot accession no. P11049 (version 162, SEQ ID NO:438). CD37 expression is restricted to cells of the immune system, with highest abundance on mature B cells, and lower expression is found on T cells and myeloid cells. The glycoprotein CD37 is a member of the transmembrane 4 superfamily and controls both humoral and cellular immune responses. The term “CD37” encompasses “full-length,” unprocessed CD37 as well as any form of CD37 that results from processing in the cell. The term also encompasses naturally occurring variants of CD37, e.g., splice variants or allelic variants.

›Definitions · 8 of 56

A “Multiple Myeloma (MM) cell surface antigen” as used herein refers to an antigenic determinant presented on the surface of a Multiple Myeloma (MM) cell. Several MM cell surface antigens are interesting in terms of immunotherapy of Multiple Myeloma. In one aspect, the MM cell surface antigen is selected from the group consisting of CD38, BCMA and GPRC5D.

The term “CD38”, also known as cluster of differentiation 38 or cyclic ADP ribose hydrolase, is a glycoprotein found on the surface of many immune cells (white blood cells), including CD4+, CD8+, B lymphocytes and natural killer cells. CD38 also functions in cell adhesion, signal transduction and calcium signaling. Under normal conditions, CD38 is expressed at relatively low levels on myeloid and lymphoid cells and in some non-hematopoietic tissues. In contrast, normal plasma cells and multiple myeloma (MM) cells have high levels of CD38 expression, which makes CD38 an interesting target for targeting cell surface molecules in MM. CD38 as used herein refers to any CD38 protein from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human CD38 is shown in UniProt (world wide web.uniprot.org) accession no. P28907 (SEQ ID NO:474).

The term “BCMA” refers to B cell maturation antigen, also termed tumor necrosis factor receptor superfamily member 17 (TNFRS17), and is a type III transmembrane protein without a signal-peptide and containing cysteine-rich extracellular domains. BCMA is expressed at significantly higher levels in all patient MM cells but not on other normal tissues except normal plasma cells. BCMA, along with two related TNFR superfamily B-cell activation factor receptor (BAFF-R) and transmembrane activator and calcium modulator and cyclophilin ligand interactor (TACI), critically regulate B cell proliferation and survival, as well as maturation and differentiation into plasma cells. These three functionally related receptors support long-term survival of B cells at different stages of development by binding to BAFF and/or APRIL, their cognate ligands. BCMA as used herein refers to any BCMA protein from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human BCMA is shown in UniProt (world wide web.uniprot.org) accession no. Q02223 (SEQ ID NO:475).

The term “GPRC5D” refers to G protein-coupled receptor class C group 5 member D, a target identified from plasma cells in multiple myeloma using RNA-sequencing. It has been reported that GPRC5D is associated with poor prognosis and tumour load in multiple myeloma patients. GPRC5D refers to any GPRC5D protein from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. The amino acid sequence of human GPRC5D is shown in UniProt (world wide web.uniprot.org) accession no. Q9NZD1 (SEQ ID NO:476).

The term “CD28” (Cluster of differentiation 28, Tp44) refers to any CD28 protein from any vertebrate source, including mammals such as primates (e.g. humans) non-human primates (e.g. cynomolgus monkeys) and rodents (e.g. mice and rats), unless otherwise indicated. CD28 is expressed on T cells and provides co-stimulatory signals required for T cell activation and survival. T cell stimulation through CD28 in addition to the T-cell receptor (TCR) can provide a potent signal for the production of various interleukins. CD28 is the receptor for CD80 (B7.1) and CD86 (B7.2) proteins and is the only B7 receptor constitutively expressed on naïve T cells. The amino acid sequence of human CD28 is shown in UniProt (world wide web.uniprot.org) accession no. P10747 (SEQ ID NO:1).

An “agonistic antibody” refers to an antibody that comprises an agonistic function against a given receptor. In general, when an agonist ligand (factor) hinds to a receptor, the tertiary structure of the receptor protein changes, and the receptor is activated (when the receptor is a membrane protein, a cell growth signal or such is usually transducted). If the receptor is a dimer-forming type, an agonistic antibody can dimerize the receptor at an appropriate distance and angle, thus acting similarly to a ligand. An appropriate anti-receptor antibody can mimic dimerization of receptors performed by ligands, and thus can become an agonistic antibody.

A “CD28 agonistic antigen binding molecule” or “CD28 conventional agonistic antigen binding molecule” is an antigen binding molecule that mimicks CD28 natural ligands (CD80 or CD86) in their role to enhance T cell activation in presence of a T cell receptor signal (“signal 2”). A T cell needs two signals to become fully activated. Under physiological conditions “signal 1” arises form the interaction of T cell receptor (TCR) molecules with peptide/major histocompatibility complex (MHC) complexes on antigen presenting cells (APCs) and “signal 2” is provided by engagement of a costimulatory receptor, e.g. CD28. A CD28 agonistic antigen binding molecule is able to costimulate T cells (signal 2). It is also able to induce T cell proliferation and cytokine secretion in combination with a molecule with specificity for the TCR complex, however the CD28 agonistic antigen binding molecule is not capable of fully activating T cells without additional stimulation of the TCR. There is however a subclass of CD28 specific antigen binding molecules, the so-called CD28 superagonistic antigen binding molecules. A “CD28 superagonistic antigen binding molecule” is a CD28 antigen binding molecule which is capable of fully activating T cells without additional stimulation of the TCR. A CD28 superagonistic antigen binding molecule is capable to induce T cell proliferation and cytokine secretion without prior T cell activation (signal

›Definitions · 9 of 56

The term “variable domain” or “variable region” refers to the domain of an antibody heavy or light chain that is involved in binding the antigen binding molecule to antigen. The variable domains of the heavy chain and light chain (VH and VL, respectively) of a native antibody generally have similar structures, with each domain comprising four conserved framework regions (FRs) and three hypervariable regions (HVRs). See, e.g., Kindt et al., Kuby Immunology, 6th ed. W.H. Freeman and Co., page 91 (2007). A single VH or VL domain may be sufficient to confer antigen-binding specificity.

The term “hypervariable region” or “HVR” as used herein refers to each of the regions of an antigen binding variable domain which are hypervariable in sequence and which determine antigen binding specificity, for example “complementarity determining regions” (“CDRs”). Generally, antigen binding domains comprise six CDRs: three in the VH (CDR-H1, CDR-H2, CDR-H3), and three in the VL (CDR-L1, CDR-L2, CDR-L3). Exemplary CDRs herein include:

(a) hypervariable loops occurring at amino acid residues 26-32 (L1), 50-52 (L2), 91-96 (L3), 26-32 (H1), 53-55 (H2), and 96-101 (H3) (Chothia and Lesk, J. Mol. Biol. 196:901-917 (1987));

(b) CDRs occurring at amino acid residues 24-34 (L1), 50-56 (L2), 89-97 (L3), 31-35b (H1), 50-65 (H2), and 95-102 (H3) (Kabat et al., Sequences of Proteins of Immunological Interest, 5th Ed. Public Health Service, National Institutes of Health, Bethesda, Md. (1991)); and

(c) antigen contacts occurring at amino acid residues 27c-36 (L1), 46-55 (L2), 89-96 (L3), 30-35b (H1), 47-58 (H2), and 93-101 (H3) (MacCallum et al. J. Mol. Biol. 262: 732-745 (1996)).

Unless otherwise indicated, the CDRs are determined according to Kabat et al., supra. One of skill in the art will understand that the CDR designations can also be determined according to Chothia, supra, McCallum, supra, or any other scientifically accepted nomenclature. Kabat et al. also defined a numbering system for variable region sequences that is applicable to any antibody. One of ordinary skill in the art can unambiguously assign this system of “Kabat numbering” to any variable region sequence, without reliance on any experimental data beyond the sequence itself. As used herein, “Kabat numbering” refers to the numbering system set forth by Kabat et al., U.S. Dept. of Health and Human Services, “Sequence of Proteins of Immunological Interest” (1983). Unless otherwise specified, references to the numbering of specific amino acid residue positions in an antibody variable region are according to the Kabat numbering system.

As used herein, the term “affinity matured” in the context of antigen binding molecules (e.g., antibodies) refers to an antigen binding molecule that is derived from a reference antigen binding molecule, e.g., by mutation, binds to the same antigen, preferably binds to the same epitope, as the reference antibody; and has a higher affinity for the antigen than that of the reference antigen binding molecule. Affinity maturation generally involves modification of one or more amino acid residues in one or more CDRs of the antigen binding molecule. Typically, the affinity matured antigen binding molecule binds to the same epitope as the initial reference antigen binding molecule.

“Framework” or “FR” refers to variable domain residues other than hypervariable region (HVR) residues. The FR of a variable domain generally consists of four FR domains: FR1, FR2, FR3, and FR4. Accordingly, the HVR and FR sequences generally appear in the following sequence in VH (or VL): FR1-H1(L1)-FR2-H2(L2)-FR3-H3(L3)-FR4.

An “acceptor human framework” for the purposes herein is a framework comprising the amino acid sequence of a light chain variable domain (VL) framework or a heavy chain variable domain (VH) framework derived from a human immunoglobulin framework or a human consensus framework, as defined below. An acceptor human framework “derived from” a human immunoglobulin framework or a human consensus framework may comprise the same amino acid sequence thereof, or it may contain amino acid sequence changes. In some embodiments, the number of amino acid changes are 10 or less, 9 or less, 8 or less, 7 or less, 6 or less, 5 or less, 4 or less, 3 or less, or 2 or less. In some embodiments, the VL acceptor human framework is identical in sequence to the VL human immunoglobulin framework sequence or human consensus framework sequence.

The term “chimeric” antibody refers to an antibody in which a portion of the heavy and/or light chain is derived from a particular source or species, while the remainder of the heavy and/or light chain is derived from a different source or species.

The “class” of an antibody refers to the type of constant domain or constant region possessed by its heavy chain. There are five major classes of antibodies: IgA, IgD, IgE, IgG, and IgM, and several of these may be further divided into subclasses (isotypes), e.g. IgG 1 , IgG 2 , IgG 3 , IgG 4 , IgA 1 , and IgA 2 . The heavy chain constant domains that correspond to the different classes of immunoglobulins are called α, δ, ε, γ, and μ respectively.

A “humanized” antibody refers to a chimeric antibody comprising amino acid residues from non-human HVRs and amino acid residues from human FRs. In certain embodiments, a humanized antibody will comprise substantially all of at least one, and typically two, variable domains, in which all or substantially all of the HVRs (e.g., CDRs) correspond to those of a non-human antibody, and all or substantially all of the FRs correspond to those of a human antibody. A humanized antibody optionally may comprise at least a portion of an antibody constant region derived from a human antibody. A “humanized form” of an antibody, e.g., a non-human antibody, refers to an antibody that has undergone humanization. Other forms of “humanized antibodies” encompassed by the present invention are those in which the constant region has been additionally modified or changed from that of the original antibody to generate the properties according to the invention, especially in regard to Clq binding and/or Fc receptor (FcR) binding.

›Definitions · 10 of 56

A “human” antibody is one which possesses an amino acid sequence which corresponds to that of an antibody produced by a human or a human cell or derived from a non-human source that utilizes human antibody repertoires or other human antibody-encoding sequences. This definition of a human antibody specifically excludes a humanized antibody comprising non-human antigen-binding residues.

The term “CI-11 domain” denotes the part of an antibody heavy chain polypeptide that extends approximately from EU position 118 to EU position 215 (EU numbering system according to Kabat). In one aspect, a CH1 domain has the amino acid sequence of ASTKGPSVFP LAPSSKSTSG GTAALGCLVK DYFPEPVTVS WNSGALTSGV HTFPAVLQSS GLYSLSSVVT VPSSSLGTQT YICNVNHKPS NTKVDKKV (SEQ ID NO: 477). Usually, a segment having the amino acid sequence of EPKSC (SEQ ID NO:480) is following to link the CH1 domain to the hinge region,

The term “hinge region” denotes the part of an antibody heavy chain polypeptide that joins in a wild-type antibody heavy chain the CH1 domain and the CH2 domain, e. g. from about position 216 to about position 230 according to the EU number system of Kabat, or from about position 226 to about position 230 according to the EU number system of Kabat. The hinge regions of other IgG subclasses can be determined by aligning with the hinge-region cysteine residues of the IgG1 subclass sequence. The hinge region is normally a dimeric molecule consisting of two polypeptides with identical amino acid sequence. The hinge region generally comprises up to 25 amino acid residues and is flexible allowing the associated target binding sites to move independently. The hinge region can be subdivided into three domains: the upper, the middle, and the lower hinge domain (see e.g. Roux, et al., J. Immunol. 161 (1998) 4083).

In one aspect, the hinge region has the amino acid sequence DKTHTCPXCP (SEQ ID NO: 481), wherein X is either S or P. In one aspect, the hinge region has the amino acid sequence HTCPXCP (SEQ ID NO: 482), wherein X is either S or P. In one aspect, the hinge region has the amino acid sequence CPXCP (SEQ ID NO:483), wherein X is either S or P.

The term “Fc domain” or “Fc region” herein is used to define a C-terminal region of an antibody heavy chain that contains at least a portion of the constant region. The term includes native sequence Fc regions and variant Fc regions. An IgG Fc region comprises an IgG CH2 and an IgG CH3 domain.

The “CH2 domain” of a human IgG Fc region usually extends from an amino acid residue at about EU position 231 to an amino acid residue at about EU position 340 (EU numbering system according to Kabat). In one aspect, a CH2 domain has the amino acid sequence of APELLGGPSV FLFPPKPKDT LMISRTPEVT CVWDVSHEDP EVKFNWYVDG VEVHNAKTKP REEQESTYRW SVLTVLHQDW LNGKEYKCKV SNKALPAPIE KTISKAK (SEQ ID NO: 478). The C112 domain is unique in that it is not closely paired with another domain. Rather, two N-linked branched carbohydrate chains are interposed between the two CH2 domains of an intact native Fc-region. It has been speculated that the carbohydrate may provide a substitute for the domain-domain pairing and help stabilize the CH2 domain. Burton, Mol. Immunol. 22 (1985) 161-206. In one embodiment, a carbohydrate chain is attached to the CH2 domain. The CH2 domain herein may be a native sequence CH2 domain or variant CH2 domain.

The “CH3 domain” comprises the stretch of residues C-terminal to a CH2 domain in an Fc region denotes the part of an antibody heavy chain polypeptide that extends approximately from EU position 341 to EU position 446 (EU numbering system according to Kabat). In one aspect, the CH3 domain has the amino acid sequence of GQPREPQVYT LPPSRDELTK NQVSLTCLVK GFYPSDIAVE WESNGQPENN YKTTPPVLDS DGSFFLYSKL TVDKSRWQQG NVFSCSVMHE ALHNHYTQKS LSLSPG (SEQ ID NO: 479). The CH3 region herein may be a native sequence CH3 domain or a variant CH3 domain (e.g. a C113 domain with an introduced “protuberance” (“knob”) in one chain thereof and a corresponding introduced “cavity” (“hole”) in the other chain thereof; see U.S. Pat. No. 5,821,333, expressly incorporated herein by reference). Such variant CH3 domains may be used to promote heterodimerization of two non-identical antibody heavy chains as herein described. In one embodiment, a human IgG heavy chain Fc region extends from Cys226, or from Pro230, to the carboxyl-terminus of the heavy chain. However, the C-terminal lysine (Lys447) of the Fc region may or may not be present. Unless otherwise specified herein, numbering of amino acid residues in the Fc region or constant region is according to the EU numbering system, also called the EU index, as described in Kabat et al., Sequences of Proteins of Immunological Interest, 5th Ed. Public Health Service, National Institutes of Health, Bethesda, Md., 1991.

The “knob-into-hole” technology is described e.g. in U.S. Pat. Nos. 5,731,168; 7,695,936; Ridgway et al., Prot Eng 9, 617-621 (1996) and Carter, J Immunol Meth 248, 7-15 (2001). Generally, the method involves introducing a protuberance (“knob”) at the interface of a first polypeptide and a corresponding cavity (“hole”) in the interface of a second polypeptide, such that the protuberance can be positioned in the cavity so as to promote heterodimer formation and hinder homodimer formation. Protuberances are constructed by replacing small amino acid side chains from the interface of the first polypeptide with larger side chains (e.g. tyrosine or tryptophan). Compensatory cavities of identical or similar size to the protuberances are created in the interface of the second polypeptide by replacing large amino acid side chains with smaller ones (e.g. alanine or threonine). The protuberance and cavity can be made by altering the nucleic acid encoding the polypeptides, e.g. by site-specific mutagenesis, or by peptide synthesis. In a specific embodiment a knob modification comprises the amino acid substitution T366W in one of the two subunits of the Fc domain, and the hole modification comprises the amino acid substitutions T366S, L368A and Y407V in the other one of the two subunits of the Fc domain. In a further specific embodiment, the subunit of the Fc domain comprising the knob modification additionally comprises the amino acid substitution S354C, and the subunit of the Fc domain comprising the hole modification additionally comprises the amino acid substitution Y349C. Introduction of these two cysteine residues results in the formation of a disulfide bridge between the two subunits of the Fc region, thus further stabilizing the dimer (Carter, J Immunol Methods 248, 7-15 (2001)).

›Definitions · 11 of 56

A “region equivalent to the Fc region of an immunoglobulin” is intended to include naturally occurring allelic variants of the Fc region of an immunoglobulin as well as variants having alterations which produce substitutions, additions, or deletions but which do not decrease substantially the ability of the immunoglobulin to mediate effector functions (such as antibody-dependent cellular cytotoxicity). For example, one or more amino acids can be deleted from the N-terminus or C-terminus of the Fc region of an immunoglobulin without substantial loss of biological function. Such variants can be selected according to general rules known in the art so as to have minimal effect on activity (see, e.g., Bowie, J. U. et al., Science 247:1306-10 (1990)).

The term “wild-type Fc domain” denotes an amino acid sequence identical to the amino acid sequence of an Fc domain found in nature. Wild-type human Fc domains include a native human IgG1 Fc-region (non-A and A allotypes), native human IgG2 Fc-region, native human IgG3 Fc-region, and native human IgG4 Fc-region as well as naturally occurring variants thereof. Wild-type Fc-regions are denoted in SEQ ID NO: 484 (IgG1, caucasian allotype), SEQ ID NO: 485 (IgG1, afroamerican allotype), SEQ ID NO: 486 (IgG2), SEQ ID NO: 487 (IgG3) and SEQ ID NO:488 (IgG4).

The term “variant (human) Fc domain” denotes an amino acid sequence which differs from that of a “wild-type” (human) Fc domain amino acid sequence by virtue of at least one “amino acid mutation”. In one aspect, the variant Fc-region has at least one amino acid mutation compared to a native Fc-region, e.g. from about one to about ten amino acid mutations, and in one aspect from about one to about five amino acid mutations in a native Fc-region. In one aspect, the (variant) Fc-region has at least about 95% homology with a wild-type Fc-region.

The term “effector functions” refers to those biological activities attributable to the Fc region of an antibody, which vary with the antibody isotype. Examples of antibody effector functions include: Clq binding and complement dependent cytotoxicity (CDC), Fc receptor binding, antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP), cytokine secretion, immune complex-mediated antigen uptake by antigen presenting cells, down regulation of cell surface receptors (e.g. B cell receptor), and B cell activation.

Fc receptor binding dependent effector functions can be mediated by the interaction of the Fc-region of an antibody with Fc receptors (FcRs), which are specialized cell surface receptors on hematopoietic cells. Fc receptors belong to the immunoglobulin superfamily, and have been shown to mediate both the removal of antibody-coated pathogens by phagocytosis of immune complexes, and the lysis of erythrocytes and various other cellular targets (e.g. tumor cells) coated with the corresponding antibody, via antibody dependent cell mediated cytotoxicity (ADCC) (see e.g. Van de Winkel, J. G. and Anderson, C. L., J. Leukoc. Biol. 49 (1991) 511-524). FcRs are defined by their specificity for immunoglobulin isotypes: Fc receptors for IgG antibodies are referred to as FcγR. Fc receptor binding is described e.g. in Ravetch, J. V. and Kinet, J. P., Annu. Rev. Immunol. 9 (1991) 457-492; Capel, P. J., et al., Immunomethods 4 (1994) 25-34; de Haas, M., et al., J. Lab. Clin. Med. 126 (1995) 330-341; and Gessner, J. E., et al., Ann. Hematol. 76 (1998) 231-248.

Cross-linking of receptors for the Fc-region of IgG antibodies (FcγR) triggers a wide variety of effector functions including phagocytosis, antibody-dependent cellular cytotoxicity, and release of inflammatory mediators, as well as immune complex clearance and regulation of antibody production. In humans, three classes of FcγR have been characterized, which are:

FcγRI (CD64) binds monomeric IgG with high affinity and is expressed on macrophages, monocytes, neutrophils and eosinophils. Modification in the Fc-region IgG at least at one of the amino acid residues E233-G236, P238, D265, N297, A327 and P329 (numbering according to EU index of Kabat) reduce binding to FcγRI. IgG2 residues at positions 233-236, substituted into IgG1 and IgG4, reduced binding to FcγRI by 10 3 -fold and eliminated the human monocyte response to antibody-sensitized red blood cells (Armour, K. L., et al., Eur. J. Immunol. 29 (1999) 2613-2624). FcγRII (CD32) binds complexed IgG with medium to low affinity and is widely expressed. This receptor can be divided into two sub-types, FcγRIIA and FcγRIIB. FcγRIIA is found on many cells involved in killing (e.g. macrophages, monocytes, neutrophils) and seems able to activate the killing process. FcγRIIB seems to play a role in inhibitory processes and is found on B cells, macrophages and on mast cells and eosinophils. On B-cells it seems to function to suppress further immunoglobulin production and isotype switching to, for example, the IgE class. On macrophages, FcγRIIB acts to inhibit phagocytosis as mediated through FcγRIIA. On eosinophils and mast cells the B-form may help to suppress activation of these cells through IgE binding to its separate receptor. Reduced binding for FcγRIIA is found e.g. for antibodies comprising an IgG Fc-region with mutations at least at one of the amino acid residues E233-G236, P238, D265, N297, A327, P329, D270, Q295, A327, R292, and K414 (numbering according to EU index of Kabat). FcγRIII (CD16) binds IgG with medium to low affinity and exists as two types. FcγRIIIA is found on NK cells, macrophages, eosinophils and some monocytes and T cells and mediates ADCC. FcγRIIIB is highly expressed on neutrophils. Reduced binding to FcγRIIIA is found e.g. for antibodies comprising an IgG Fc-region with mutation at least at one of the amino acid residues E233-G236, P238, D265, N297, A327, P329, D270, Q295, A327, 5239, E269, E293, Y296, V303, A327, K338 and D376 (numbering according to EU index of Kabat).

Mapping of the binding sites on human IgG1 for Fc receptors, the above mentioned mutation sites and methods for measuring binding to FcγRI and FcγRIIA are described in Shields, R. L., et al. J. Biol. Chem. 276 (2001) 6591-6604.

›Definitions · 12 of 56

The term “ADCC” or “antibody-dependent cellular cytotoxicity” is an immune mechanism leading to lysis of antibody-coated target cells by immune effector cells. The target cells are cells to which antibodies or derivatives thereof comprising an Fc region specifically bind, generally via the protein part that is N-terminal to the Fc region. As used herein, the term “reduced ADCC” is defined as either a reduction in the number of target cells that are lysed in a given time, at a given concentration of antibody in the medium surrounding the target cells, by the mechanism of ADCC defined above, and/or an increase in the concentration of antibody in the medium surrounding the target cells, required to achieve the lysis of a given number of target cells in a given time, by the mechanism of ADCC. The reduction in ADCC is relative to the ADCC mediated by the same antibody produced by the same type of host cells, using the same standard production, purification, formulation and storage methods (which are known to those skilled in the art), but that has not been engineered. For example, the reduction in ADCC mediated by an antibody comprising in its Fc domain an amino acid substitution that reduces ADCC, is relative to the ADCC mediated by the same antibody without this amino acid substitution in the Fc domain. Suitable assays to measure ADCC are well known in the art (see e.g. PCT publication no. WO 2006/082515 or PCT publication no. WO 2012/130831). For example, the capacity of the antibody to induce the initial steps mediating ADCC is investigated by measuring their binding to Fcγ receptors expressing cells, such as cells, recombinantly expressing FcγRI and/or FcγRIIA or NK cells (expressing essentially FcγRIIIA). In particular, binding to FcγR on NK cells is measured.

An “activating Fc receptor” is an Fc receptor that following engagement by an Fc region of an antibody elicits signaling events that stimulate the receptor-bearing cell to perform effector functions. Activating Fc receptors include FcγRIIIa (CD16a), FcγRI (CD64), FcγRIIa (CD32), and FcαRI (CD89). A particular activating Fc receptor is human FcγRIIIa (see UniProt accession no. P08637, version 141).

An “ectodomain” is the domain of a membrane protein that extends into the extracellular space (i.e. the space outside the target cell). Ectodomains are usually the parts of proteins that initiate contact with surfaces, which leads to signal transduction.

The term “peptide linker” refers to a peptide comprising one or more amino acids, typically about 2 to 20 amino acids. Peptide linkers are known in the art or are described herein. Suitable, non-immunogenic linker peptides are, for example, (G 4 S) n , (SG 4 ) n or G 4 (SG 4 ) n peptide linkers, wherein “n” is generally a number between 1 and 5, typically between 2 and 4, in particular 2, i.e. the peptides selected from the group consisting of GGGGS (SEQ ID NO:146) GGGGSGGGGS (SEQ ID NO:147), SGGGGSGGGG (SEQ ID NO:148) and GGGGSGGGGSGGGG (SEQ ID NO:149), but also include the sequences GSPGSSSSGS (SEQ ID NO:150), (G4S) 3 (SEQ ID NO:151), (G4S) 4 (SEQ ID NO:152), GSGSGSGS (SEQ ID NO:153), GSGSGNGS (SEQ ID NO:154), GGSGSGSG (SEQ ID NO:155), GGSGSG (SEQ ID NO:156), GGSG (SEQ ID NO:157), GGSGNGSG (SEQ ID NO:158), GGNGSGSG (SEQ ID NO:159) and GGNGSG (SEQ ID NO:160). Peptide linkers of particular interest are (G4S) (SEQ ID NO:146), (G 4 S) 2 or GGGGSGGGGS (SEQ ID NO:147), (G4S) 3 (SEQ ID NO:151) and (G4S) 4 (SEQ ID NO:152).

The term “amino acid” as used within this application denotes the group of naturally occurring carboxy α-amino acids comprising alanine (three letter code: ala, one letter code: A), arginine (arg, R), asparagine (asn, N), aspartic acid (asp, D), cysteine (cys, C), glutamine (gln, Q), glutamic acid (glu, E), glycine (gly, G), histidine (his, H), isoleucine (ile, I), leucine (leu, L), lysine (lys, K), methionine (met, M), phenylalanine (phe, F), proline (pro, P), serine (ser, S), threonine (thr, T), tryptophan (trp, W), tyrosine (tyr, Y), and valine (val, V).

By “fused” or “connected” is meant that the components (e.g. a polypeptide and an ectodomain of said TNF ligand family member) are linked by peptide bonds, either directly or via one or more peptide linkers.

“Percent (%) amino acid sequence identity” with respect to a reference polypeptide (protein) sequence is defined as the percentage of amino acid residues in a candidate sequence that are identical with the amino acid residues in the reference polypeptide sequence, after aligning the sequences and introducing gaps, if necessary, to achieve the maximum percent sequence identity, and not considering any conservative substitutions as part of the sequence identity. Alignment for purposes of determining percent amino acid sequence identity can be achieved in various ways that are within the skill in the art, for instance, using publicly available computer software such as BLAST, BLAST-2, ALIGN. SAWI or Megalign (DNASTAR) software. Those skilled in the art can determine appropriate parameters for aligning sequences, including any algorithms needed to achieve maximal alignment over the full length of the sequences being compared. For purposes herein, however, % amino acid sequence identity values are generated using the sequence comparison computer program ALIGN-2. The ALIGN-2 sequence comparison computer program was authored by Genentech, Inc., and the source code has been filed with user documentation in the U.S. Copyright Office, Washington D.C., 20559, where it is registered under U.S. Copyright Registration No. TXU510087. The ALIGN-2 program is publicly available from Genentech, Inc., South San Francisco, Calif., or may be compiled from the source code. The ALIGN-2 program should be compiled for use on a UNIX operating system, including digital UNIX V4.0D. All sequence comparison parameters are set by the ALIGN-2 program and do not vary. In situations where ALIGN-2 is employed for amino acid sequence comparisons, the % amino acid sequence identity of a given amino acid sequence A to, with, or against a given amino acid sequence B (which can alternatively be phrased as a given amino acid sequence A that has or comprises a certain % amino acid sequence identity to, with, or against a given amino acid sequence B) is calculated as follows: 100 times the fraction X/Y, where X is the number of amino acid residues scored as identical matches by the sequence alignment program ALIGN-2 in that program's alignment of A and B, and where Y is the total number of amino acid residues in B. It will be appreciated that where the length of amino acid sequence A is not equal to the length of amino acid sequence B, the % amino acid sequence identity of A to B will not equal the % amino acid sequence identity of B to A. Unless specifically stated otherwise, all % amino acid sequence identity values used herein are obtained as described in the immediately preceding paragraph using the ALIGN-2 computer program.

›Definitions · 13 of 56

In certain embodiments, amino acid sequence variants of the CD28 antigen binding molecules provided herein are contemplated. For example, it may be desirable to improve the binding affinity and/or other biological properties of the CD28 antigen binding molecules. Amino acid sequence variants of the CD28 antigen binding molecules may be prepared by introducing appropriate modifications into the nucleotide sequence encoding the molecules, or by peptide synthesis. Such modifications include, for example, deletions from, and/or insertions into and/or substitutions of residues within the amino acid sequences of the antibody. Any combination of deletion, insertion, and substitution can be made to arrive at the final construct, provided that the final construct possesses the desired characteristics, e.g., antigen-binding. Sites of interest for substitutional mutagenesis include the HVRs and Framework (FRs). Conservative substitutions are provided in Table B under the heading “Preferred Substitutions” and further described below in reference to amino acid side chain classes (1) to (6). Amino acid substitutions may be introduced into the molecule of interest and the products screened for a desired activity, e.g., retained/improved antigen binding, decreased immunogenicity, or improved ADCC or CDC.

Amino acids may be grouped according to common side-chain properties:

(1) hydrophobic: Norleucine, Met, Ala, Val, Leu, Ile; (2) neutral hydrophilic: Cys, Ser, Thr, Asn, Gln; (3) acidic: Asp, Glu; (4) basic: His, Lys, Arg; (5) residues that influence chain orientation: Gly, Pro; (6) aromatic: Trp, Tyr, Phe.

Non-conservative substitutions will entail exchanging a member of one of these classes for another class.

The term “amino acid sequence variants” includes substantial variants wherein there are amino acid substitutions in one or more hypervariable region residues of a parent antigen binding molecule (e.g. a humanized or human antibody). Generally, the resulting variant(s) selected for further study will have modifications (e.g., improvements) in certain biological properties (e.g., increased affinity, reduced immunogenicity) relative to the parent antigen binding molecule and/or will have substantially retained certain biological properties of the parent antigen binding molecule. An exemplary substitutional variant is an affinity matured antibody, which may be conveniently generated, e.g., using phage display-based affinity maturation techniques such as those described herein. Briefly, one or more HVR residues are mutated and the variant antigen binding molecules displayed on phage and screened for a particular biological activity (e.g. binding affinity). In certain embodiments, substitutions, insertions, or deletions may occur within one or more HVRs so long as such alterations do not substantially reduce the ability of the antigen binding molecule to bind antigen. For example, conservative alterations (e.g., conservative substitutions as provided herein) that do not substantially reduce binding affinity may be made in HVRs. A useful method for identification of residues or regions of an antibody that may be targeted for mutagenesis is called “alanine scanning mutagenesis” as described by Cunningham and Wells (1989) Science, 244:1081-1085. In this method, a residue or group of target residues (e.g., charged residues such as Arg, Asp, His, Lys, and Glu) are identified and replaced by a neutral or negatively charged amino acid (e.g., alanine or polyalanine) to determine whether the interaction of the antibody with antigen is affected. Further substitutions may be introduced at the amino acid locations demonstrating functional sensitivity to the initial substitutions. Alternatively, or additionally, a crystal structure of an antigen-antigen binding molecule complex to identify contact points between the antibody and antigen. Such contact residues and neighboring residues may be targeted or eliminated as candidates for substitution. Variants may be screened to determine whether they contain the desired properties.

Amino acid sequence insertions include amino- and/or carboxyl-terminal fusions ranging in length from one residue to polypeptides containing a hundred or more residues, as well as intrasequence insertions of single or multiple amino acid residues. Examples of insertions include CD28 antigen binding molecules with a fusion to the N- or C-terminus to a polypeptide which increases the serum half-life of the CD28 antigen binding molecules.

In certain embodiments, the CD28 antigen binding molecules provided herein are altered to increase or decrease the extent to which the antibody is glycosylated. Glycosylation variants of the molecules may be conveniently obtained by altering the amino acid sequence such that one or more glycosylation sites is created or removed. Where the agonistic ICOS-binding molecule comprises an Fc domain, the carbohydrate attached thereto may be altered. Native antibodies produced by mammalian cells typically comprise a branched, biantennary oligosaccharide that is generally attached by an N-linkage to Asn297 of the CH2 domain of the Fc region. See, e.g., Wright et al. TIBTECH 15:26-32 (1997). The oligosaccharide may include various carbohydrates, e.g., mannose, N-acetyl glucosamine (GlcNAc), galactose, and sialic acid, as well as a fucose attached to a GlcNAc in the “stem” of the biantennary oligosaccharide structure. In some embodiments, modifications of the oligosaccharide in agonistic ICOS-binding molecules may be made in order to create variants with certain improved properties. In one aspect, variants of agonistic ICOS-binding molecules are provided having a carbohydrate structure that lacks fucose attached (directly or indirectly) to an Fc region. Such fucosylation variants may have improved ADCC function, see e.g. US Patent Publication Nos. US 2003/0157108 (Presta, L.) or US 2004/0093621 (Kyowa Hakko Kogyo Co., Ltd). Further variants of the CD28 antigen binding molecules of the invention include those with bisected oligosaccharides, e.g., in which a biantennary oligosaccharide attached to the Fc region is bisected by GlcNAc. Such variants may have reduced fucosylation and/or improved ADCC function., see for example WO 2003/011878 (Jean-Mairet et al.); U.S. Pat. No. 6,602,684 (Umana et al.); and US 2005/0123546 (Umana et al.). Variants with at least one galactose residue in the oligosaccharide attached to the Fc region are also provided. Such antibody variants may have improved CDC function and are described, e.g., in WO 1997/30087 (Patel et al.); WO 1998/58964 (Raju, S.); and WO 1999/22764 (Raju, S.).

›Definitions · 14 of 56

In certain embodiments, it may be desirable to create cysteine engineered variants of the CD28 antigen binding molecules of the invention, e.g., “thioMAbs,” in which one or more residues of the molecule are substituted with cysteine residues. In particular embodiments, the substituted residues occur at accessible sites of the molecule. By substituting those residues with cysteine, reactive thiol groups are thereby positioned at accessible sites of the antibody and may be used to conjugate the antibody to other moieties, such as drug moieties or linker-drug moieties, to create an immunoconjugate. In certain embodiments, any one or more of the following residues may be substituted with cysteine: V205 (Kabat numbering) of the light chain; A118 (EU numbering) of the heavy chain; and 5400 (EU numbering) of the heavy chain Fc region. Cysteine engineered antigen binding molecules may be generated as described, e.g., in U.S. Pat. No. 7,521,541.

In certain aspects, the CD28 antigen binding molecules provided herein may be further modified to contain additional non-proteinaceous moieties that are known in the art and readily available. The moieties suitable for derivatization of the antibody include but are not limited to water soluble polymers. Non-limiting examples of water soluble polymers include, but are not limited to, polyethylene glycol (PEG), copolymers of ethylene glycol/propylene glycol, carboxymethylcellulose, dextran, polyvinyl alcohol, polyvinyl pyrrolidone, poly-1, 3-dioxolane, poly-1,3,6-trioxane, ethylene/maleic anhydride copolymer, polyaminoacids (either homopolymers or random copolymers), and dextran or poly(n-vinyl pyrrolidone)polyethylene glycol, propropylene glycol homopolymers, prolypropylene oxide/ethylene oxide co-polymers, polyoxyethylated polyols (e.g., glycerol), polyvinyl alcohol, and mixtures thereof. Polyethylene glycol propionaldehyde may have advantages in manufacturing due to its stability in water. The polymer may be of any molecular weight, and may be branched or unbranched. The number of polymers attached to the antibody may vary, and if more than one polymer is attached, they can be the same or different molecules. In general, the number and/or type of polymers used for derivatization can be determined based on considerations including, but not limited to, the particular properties or functions of the antibody to be improved, whether the bispecific antibody derivative will be used in a therapy under defined conditions, etc. In another aspect, conjugates of an antibody and non-proteinaceous moiety that may be selectively heated by exposure to radiation are provided. In one embodiment, the non-proteinaceous moiety is a carbon nanotube (Kam, N. W. et al., Proc. Natl. Acad. Sci. USA 102 (2005) 11600-11605). The radiation may be of any wavelength, and includes, but is not limited to, wavelengths that do not harm ordinary cells, but which heat the non-proteinaceous moiety to a temperature at which cells proximal to the antibody-non-proteinaceous moiety are killed. In another aspect, immunoconjugates of the CD28 antigen binding molecules provided herein maybe obtained. An “immunoconjugate” is an antibody conjugated to one or more heterologous molecule(s), including but not limited to a cytotoxic agent.

The term “polynucleotide” refers to an isolated nucleic acid molecule or construct, e.g. messenger RNA (mRNA), virally-derived RNA, or plasmid DNA (pDNA). A polynucleotide may comprise a conventional phosphodiester bond or a non-conventional bond (e.g. an amide bond, such as found in peptide nucleic acids (PNA). The term “nucleic acid molecule” refers to any one or more nucleic acid segments, e.g. DNA or RNA fragments, present in a polynucleotide. Each nucleotide is composed of a base, specifically a purine- or pyrimidine base (i.e. cytosine (C), guanine (G), adenine (A), thymine (T) or uracil (U)), a sugar (i.e. deoxyribose or ribose), and a phosphate group. Often, the nucleic acid molecule is described by the sequence of bases, whereby said bases represent the primary structure (linear structure) of a nucleic acid molecule. The sequence of bases is typically represented from 5′ to 3′. Herein, the term nucleic acid molecule encompasses deoxyribonucleic acid (DNA) including e.g., complementary DNA (cDNA) and genomic DNA, ribonucleic acid (RNA), in particular messenger RNA (mRNA), synthetic forms of DNA or RNA, and mixed polymers comprising two or more of these molecules. The nucleic acid molecule may be linear or circular. In addition, the term nucleic acid molecule includes both, sense and antisense strands, as well as single stranded and double stranded forms. Moreover, the herein described nucleic acid molecule can contain naturally occurring or non-naturally occurring nucleotides. Examples of non-naturally occurring nucleotides include modified nucleotide bases with derivatized sugars or phosphate backbone linkages or chemically modified residues. Nucleic acid molecules also encompass DNA and RNA molecules which are suitable as a vector for direct expression of an antibody of the invention in vitro and/or in vivo, e.g., in a host or patient. Such DNA (e.g., cDNA) or RNA (e.g., mRNA) vectors, can be unmodified or modified. For example, mRNA can be chemically modified to enhance the stability of the RNA vector and/or expression of the encoded molecule so that mRNA can be injected into a subject to generate the antibody in vivo (see e.g., Stadler et al. (2017) Nature Medicine 23:815-817, or EP 2 101 823 B1).

By “isolated” nucleic acid molecule or polynucleotide is intended a nucleic acid molecule, DNA or RNA, which has been removed from its native environment. For example, a recombinant polynucleotide encoding a polypeptide contained in a vector is considered isolated for the purposes of the present invention. Further examples of an isolated polynucleotide include recombinant polynucleotides maintained in heterologous host cells or purified (partially or substantially) polynucleotides in solution. An isolated polynucleotide includes a polynucleotide molecule contained in cells that ordinarily contain the polynucleotide molecule, but the polynucleotide molecule is present extrachromosomally or at a chromosomal location that is different from its natural chromosomal location. Isolated RNA molecules include in vivo or in vitro RNA transcripts of the present invention, as well as positive and negative strand forms, and double-stranded forms. Isolated polynucleotides or nucleic acids according to the present invention further include such molecules produced synthetically. In addition, a polynucleotide or a nucleic acid may be or may include a regulatory element such as a promoter, ribosome binding site, or a transcription terminator.

›Definitions · 15 of 56

By a nucleic acid or polynucleotide having a nucleotide sequence at least, for example, 95% “identical” to a reference nucleotide sequence of the present invention, it is intended that the nucleotide sequence of the polynucleotide is identical to the reference sequence except that the polynucleotide sequence may include up to five point mutations per each 100 nucleotides of the reference nucleotide sequence. In other words, to obtain a polynucleotide having a nucleotide sequence at least 95% identical to a reference nucleotide sequence, up to 5% of the nucleotides in the reference sequence may be deleted or substituted with another nucleotide, or a number of nucleotides up to 5% of the total nucleotides in the reference sequence may be inserted into the reference sequence. These alterations of the reference sequence may occur at the 5′ or 3′ terminal positions of the reference nucleotide sequence or anywhere between those terminal positions, interspersed either individually among residues in the reference sequence or in one or more contiguous groups within the reference sequence. As a practical matter, whether any particular polynucleotide sequence is at least 80%, 85%, 90%, 95%, 96%, 97%, 98% or 99% identical to a nucleotide sequence of the present invention can be determined conventionally using known computer programs, such as the ones discussed above for polypeptides (e.g. ALIGN-2).

The term “expression cassette” refers to a polynucleotide generated recombinantly or synthetically, with a series of specified nucleic acid elements that permit transcription of a particular nucleic acid in a target cell. The recombinant expression cassette can be incorporated into a plasmid, chromosome, mitochondrial DNA, plastid DNA, virus, or nucleic acid fragment. Typically, the recombinant expression cassette portion of an expression vector includes, among other sequences, a nucleic acid sequence to be transcribed and a promoter. In certain embodiments, the expression cassette of the invention comprises polynucleotide sequences that encode bispecific antigen binding molecules of the invention or fragments thereof.

The term “vector” or “expression vector” is synonymous with “expression construct” and refers to a DNA molecule that is used to introduce and direct the expression of a specific gene to which it is operably associated in a target cell. The term includes the vector as a self-replicating nucleic acid structure as well as the vector incorporated into the genome of a host cell into which it has been introduced. The expression vector of the present invention comprises an expression cassette. Expression vectors allow transcription of large amounts of stable mRNA. Once the expression vector is inside the target cell, the ribonucleic acid molecule or protein that is encoded by the gene is produced by the cellular transcription and/or translation machinery. In one embodiment, the expression vector of the invention comprises an expression cassette that comprises polynucleotide sequences that encode bispecific antigen binding molecules of the invention or fragments thereof.

The terms “host cell”, “host cell line,” and “host cell culture” are used interchangeably and refer to cells into which exogenous nucleic acid has been introduced, including the progeny of such cells. Host cells include “transformants” and “transformed cells,” which include the primary transformed cell and progeny derived therefrom without regard to the number of passages. Progeny may not be completely identical in nucleic acid content to a parent cell, but may contain mutations. Mutant progeny that have the same function or biological activity as screened or selected for in the originally transformed cell are included herein. A host cell is any type of cellular system that can be used to generate the bispecific antigen binding molecules of the present invention. Host cells include cultured cells, e.g. mammalian cultured cells, such as CHO cells, BHK cells, NS0 cells, SP2/0 cells, Y0 myeloma cells, P3X63 mouse myeloma cells, PER cells, PER.C6 cells or hybridoma cells, yeast cells, insect cells, and plant cells, to name only a few, but also cells comprised within a transgenic animal, transgenic plant or cultured plant or animal tissue.

An “effective amount” of an agent refers to the amount that is necessary to result in a physiological change in the cell or tissue to which it is administered.

A “therapeutically effective amount” of an agent, e.g. a pharmaceutical composition, refers to an amount effective, at dosages and for periods of time necessary, to achieve the desired therapeutic or prophylactic result. A therapeutically effective amount of an agent for example eliminates, decreases, delays, minimizes or prevents adverse effects of a disease.

An “individual” or “subject” is a mammal. Mammals include, but are not limited to, domesticated animals (e.g. cows, sheep, cats, dogs, and horses), primates (e.g. humans and non-human primates such as monkeys), rabbits, and rodents (e.g. mice and rats). Particularly, the individual or subject is a human.

The term “pharmaceutical composition” refers to a preparation which is in such form as to permit the biological activity of an active ingredient contained therein to be effective, and which contains no additional components which are unacceptably toxic to a subject to which the formulation would be administered.

A “pharmaceutically acceptable excipient” refers to an ingredient in a pharmaceutical composition, other than an active ingredient, which is nontoxic to a subject. A pharmaceutically acceptable excipient includes, but is not limited to, a buffer, a stabilizer, or a preservative.

The term “package insert” is used to refer to instructions customarily included in commercial packages of therapeutic products, that contain information about the indications, usage, dosage, administration, combination therapy, contraindications and/or warnings concerning the use of such therapeutic products.

As used herein, “treatment” (and grammatical variations thereof such as “treat” or “treating”) refers to clinical intervention in an attempt to alter the natural course of the individual being treated, and can be performed either for prophylaxis or during the course of clinical pathology. Desirable effects of treatment include, but are not limited to, preventing occurrence or recurrence of disease, alleviation of symptoms, diminishment of any direct or indirect pathological consequences of the disease, preventing metastasis, decreasing the rate of disease progression, amelioration or palliation of the disease state, and remission or improved prognosis. In some embodiments, the molecules of the invention are used to delay development of a disease or to slow the progression of a disease.

›Definitions · 16 of 56

The term “combination treatment” or “co-administration” as noted herein encompasses combined administration (where two or more therapeutic agents are included in the same or separate formulations), and separate administration, in which case, administration of an antibody as reported herein can occur prior to, simultaneously, and/or following, administration of the additional therapeutic agent or agents, preferably an antibody or antibodies.

By “B cell proliferative disorder” is meant a disease wherein the number of B cells in a patient is increased as compared to the number of B cells in a healthy subject, and particularly wherein the increase in the number of B cells is the cause or hallmark of the disease.

The term “hematological cancer” refers to or describes the physiological condition in mammals that is typically characterized by unregulated cell growth/proliferation. Thus, the term cancer as used herein refers to proliferative diseases, such as carcinoma, lymphomas (e.g., Hodgkin's and non-Hodgkin's lymphoma), blastoma, sarcoma, and leukemia. In particular, the term cancer refers to a B-cell proliferative disorder. In one aspect, the cancer is selected from the group consisting of Non-Hodgkin lymphoma (NHL), acute lymphocytic leukemia (ALL), chronic lymphocytic leukemia (CLL), diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), mantle-cell lymphoma (MCL), marginal zone lymphoma (MZL), Multiple myeloma (MM), and Hodgkin lymphoma (HL).

The term “cancer” refers to or describes the physiological condition in mammals that is typically characterized by unregulated cell growth/proliferation. Thus, the term cancer as used herein refers to proliferative diseases, such as carcinoma, lymphomas (e.g., Hodgkin's and non-Hodgkin's lymphoma), blastoma, sarcoma, and leukemia. In particular, the term cancer includes lymphocytic leukemias, lung cancer, non-small cell lung (NSCL) cancer, bronchioloalviolar cell lung cancer, bone cancer, pancreatic cancer, skin cancer, cancer of the head or neck, cutaneous or intraocular melanoma, uterine cancer, ovarian cancer, rectal cancer, cancer of the anal region, stomach cancer, gastric cancer, colon cancer, breast cancer, uterine cancer, carcinoma of the fallopian tubes, carcinoma of the endometrium, carcinoma of the cervix, carcinoma of the vagina, carcinoma of the vulva, Hodgkin's Disease, cancer of the esophagus, cancer of the small intestine, cancer of the endocrine system, cancer of the thyroid gland, cancer of the parathyroid gland, cancer of the adrenal gland, sarcoma of soft tissue, cancer of the urethra, cancer of the penis, prostate cancer, cancer of the bladder, cancer of the kidney or ureter, renal cell carcinoma, carcinoma of the renal pelvis, mesothelioma, hepatocellular cancer, biliary cancer, neoplasms of the central nervous system (CNS), spinal axis tumors, brain stem glioma, glioblastoma multiforme, astrocytomas, schwanomas, ependymonas, medulloblastomas, meningiomas, squamous cell carcinomas, pituitary adenoma and Ewings sarcoma, including refractory versions of any of the above cancers, or a combination of one or more of the above cancers. In one aspect, the cancer is a solid tumor. In another aspect, the cancer is a haematological cancer, particularly leukemia, most particularly acute lymphoblastic leukemia (ALL) or acute myelogenous leukemia (AML).

Bispecific Agonistic CD28 Antigen Binding Molecules of the Invention

The invention provides novel bispecific agonistic CD28 antigen binding molecules with particularly advantageous properties such as producibility, stability, binding affinity, biological activity, targeting efficiency, reduced toxicity, an extended dosage range that can be given to a patient and thereby a possibly enhanced efficacy. The novel bispecific agonistic CD28 antigen binding molecules comprise an Fc domain composed of a first and a second subunit capable of stable association comprising one or more amino acid substitution that reduces the binding affinity of the antigen binding molecule to an Fc receptor and/or effector function (Fc silent) and thus unspecific cross-linking via Fc receptors is avoided. Instead, they comprise at least one antigen binding domain capable of specific binding to a tumor-associated antigen such as Fibroblast Activation Protein (FAP) or Carcinoembryonic Antigen (CEA) which causes cross-linking at the tumor site. Thus, tumor-specific T cell activation is achieved.

Herein provided is a bispecific agonistic CD28 antigen binding molecule with monovalent binding to CD28, comprising

(a) one antigen binding domains capable of specific binding to CD28,

(b) at least one antigen binding domain capable of specific binding to a tumor-associated antigen, and

In one aspect, a bispecific agonistic CD28 antigen binding molecule as defined herein before is provided, wherein the Fc domain is an IgG, particularly an IgG1 Fc domain or an IgG4 Fc domain. In one particular aspect, the Fc domain composed of a first and a second subunit capable of stable association is an IgG1 Fc domain. The Fc domain comprises one or more amino acid substitution that reduces the binding affinity of the antigen binding molecule to an Fc receptor and/or reduces or abolishes effector function. In one aspect, the Fc domain comprises the amino acid substitutions L234A and L235A (numbering according to Kabat EU index). In one aspect, the Fc domain is of human IgG1 subclass and comprises the amino acid mutations L234A, L235A and P329G (numbering according to Kabat EU index). In one aspect, a bispecific agonistic CD28 antigen binding molecule is provided, wherein the antigen binding molecule comprises an Fc domain composed of a first and a second subunit capable of stable association, wherein the first subunit comprises the amino acid sequence of SEQ ID NO:176 and the second subunit comprise the amino acid sequence of SEQ ID NO:177.

›Definitions · 17 of 56

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule as defined herein before, wherein the antigen binding domain capable of specific binding to CD28 comprises

(i) a heavy chain variable region (V H CD28) comprising a heavy chain complementary determining region CDR-H1 of SEQ ID NO: 20, a CDR-H2 of SEQ ID NO: 21, and a CDR-H3 of SEQ ID NO: 22, and a light chain variable region (V L CD28) comprising a light chain complementary determining region CDR-L1 of SEQ ID NO: 23, a CDR-L2 of SEQ ID NO: 24 and a CDR-L3 of SEQ ID NO: 25; or

(ii) a heavy chain variable region (V H CD28) comprising a CDR-H1 of SEQ ID NO: 36, a CDR-H2 of SEQ ID NO: 37, and a CDR-H3 of SEQ ID NO: 38, and a light chain variable region (V L CD28) comprising a CDR-L1 of SEQ ID NO: 39, a CDR-L2 of SEQ ID NO: 40 and a CDR-L3 of SEQ ID NO: 41.

In another aspect, provided is bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CD28 comprises

(a) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54, or

(b) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or

(c) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:51 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:61, or

(d) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:53, or

(e) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54, or

(f) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:59, or

(g) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or

(h) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:43 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or

(i) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:53, or

(j) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:59, or

(k) the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27.

In one aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CD28 comprises the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54. In another aspect, the antigen binding domain capable of specific binding to CD28 of the bispecific agonistic CD28 antigen binding molecule comprises a heavy chain variable region (V H CD28) comprising a CDR-H1 of SEQ ID NO: 489, a CDR-H2 of SEQ ID NO: 490, and a CDR-H3 of SEQ TD NO: 491, and a light chain variable region (V L CD28) comprising a CDR-L1 of SEQ ID NO: 492, a CDR-L2 of SEQ ID NO: 493 and a CDR-L3 of SEQ ID NO: 494.

In another aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CD28 comprises the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:46 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:53. In another aspect, the antigen binding domain capable of specific binding to CD28 of the bispecific agonistic CD28 antigen binding molecule comprises a heavy chain variable region (V H CD28) comprising a CDR-H1 of SEQ ID NO: 495, a CDR-H2 of SEQ ID NO: 496, and a CDR-H3 of SEQ ID NO: 497, and a light chain variable region (V L CD28) comprising a CDR-L1 of SEQ ID NO: 498, a CDR-L2 of SEQ ID NO: 499 and a CDR-L3 of SEQ ID NO: 500.

›Definitions · 18 of 56

In another aspect, provided is a bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CD28 comprises the CDRs of the heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:42 and the CDRs of the light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27. In another aspect, the antigen binding domain capable of specific binding to CD28 of the bispecific agonistic CD28 antigen binding molecule comprises a heavy chain variable region (V H CD28) comprising a CDR-H1 of SEQ ID NO: 501, a CDR-H2 of SEQ ID NO: 502, and a CDR-H3 of SEQ ID NO: 503, and a light chain variable region (V L CD28) comprising a CDR-L1 of SEQ ID NO: 504, a CDR-L2 of SEQ ID NO: 505 and a CDR-L3 of SEQ ID NO: 506.

In another aspect, provided is bispecific agonistic CD28 antigen binding molecule, wherein the antigen binding domain capable of specific binding to CD28 comprises

(a) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ ID NO:47 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:54, or

(b) a heavy chain variable region (V H CD28) comprising the amino acid sequence of SEQ II) NO:47 and a light chain variable region (V L CD28) comprising the amino acid sequence of SEQ ID NO:27, or