EA045943B1 - BISPECIFIC ANTIBODIES ANTI-PSMA X ANTI-CD28 AND THEIR APPLICATIONS - Google Patents

Description

Related applications

This application is related to U.S. Provisional Patent Application No. 62/688227, filed June 21, 2018, U.S. Provisional Patent Application No. 62/781930, filed Dec. 19, 2018, U.S. Provisional Patent Application No. 62/781980, filed 19 December 2018, and U.S. Provisional Patent Application No. 62/815878, filed March 8, 2019, and claims priority thereunder. The entire contents of the foregoing applications are expressly incorporated herein by reference.

List of sequences

This application contains a Sequence Listing which has been submitted electronically in ASCII format and is hereby incorporated by reference in its entirety. The ASCII copy in question, created on June 20, 2019, is named 10367WO01_118003-45220_SL.TXT and is 48690 bytes in size.

Technical field

The present invention relates to bispecific antigen binding molecules that bind CD28 and a target molecule, such as PSMA, and methods of using them.

State of the art

CD28 is a type I transmembrane protein expressed on the surface of T cells that has a single extracellular V-type Ig-like domain assembled as a homodimer. CD28 is a receptor for the CD80 (B7.1) and CD86 (B7.2) proteins and is activated by CD80 or CD86 expressed on APCs. Binding of CD28 to CD80 or CD86 provides costimulatory signals important for T cell activation and survival. Stimulation of T cells by CD28 in addition to the T cell receptor (TCR) provides a potent signal for the production of various interleukins. CD28 enhances cellular signals, such as signaling pathways controlled by the transcription factor NFkB, upon TCR activation. The accompanying CD28 signal is important for efficient T cell activation such as differentiation, proliferation, cytokine release, and cell killing by T cells.

Antibodies to CD28 have been proposed for therapeutic purposes, including T cell activation. One specific CD28 antibody, TGN1412 (CD28 superagonist), was used in a clinical study. TGN1412 caused a cytokine storm that was not predicted by toxicology studies or ex vivo studies with human PBMCs. In 2006, six healthy volunteers were administered intravenously TGN1412 (a CD28 superagonist) at a dose of 0.1 mg/kg. Within 2 hours, all six patients exhibited significant inflammatory reactions (cytokine storm). Within 16 hours, all patients experienced multiple organ failure. Subjects were administered corticosteroids and cytokine levels returned to normal levels after 2-3 days. The starting dose of 0.1 mg/kg in the phase 1 study (related to CRS (cytokine release syndrome)) was based on a 500-fold higher level at which no adverse effects were observed NOAEL of 50 mg/kg in cynomolgus monkeys (Suntharalingam, et al., Cytokine Storm in a Phase 1 Trial of the Anti-CD28 Monoclonal Antibody TGN1412, NEJM 355:1018-1028 (2006)).

A toxicological study in cynomolgus monkeys was unable to predict the cytokine response observed in humans.

PSMA (prostate-specific membrane antigen)/FOLH1 is a well-characterized tumor target. PSMA is a type II transmembrane glycoprotein overexpressed in prostate cancer. It is also known as glutamate carboxypeptidase II (GPC). In the normal human prostate gland, PSMA is associated with the cytoplasm and apical side of the epithelium surrounding the prostatic ducts. Dysplastic and/or neoplastic transformation of prostate tissue leads to the transfer of PSMA from the apical membrane to the luminal surface of the ducts. PSMA undergoes constitutive endocytosis and is not exfoliated. PSMA is a target in various ADC (antibody drug conjugate) clinical studies and imaging modalities. PSMA is highly expressed in human prostate adenocarcinoma and corresponds to metastases (lymph nodes). In prostate tumors, PSMA expression levels increase depending on stage and grade. The transition to androgen-independent prostate cancer ultimately leads to increased expression. Interestingly, PSMA expression has also been detected in neovasculature of several solid tumors (including colon, lung, breast, kidney, and bladder cancer subtypes).

PSMA is also expressed in healthy tissues. The strongest expression is found in epithelial cells of the prostate, duodenum, renal tubular cells, salivary glands and astrocytes. PSMA is expressed at low levels in the fallopian tubes, mammary gland, and is rarely expressed in the cervical endothelium.

Bispecific antigen-binding molecules that bind both CD28 and a target antigen (such as PSMA) would be effective in therapeutic situations where specific targeting and T cell-mediated killing of cells expressing the target antigen is required. There is also a need for an anti-CD28 antibody that is safe for use in

- 1 045943 pharmaceutical composition.

Brief description of the invention

In a first aspect of the present invention, bispecific antigen-binding molecules are provided that bind CD28 and a target antigen. In certain illustrative embodiments, the bispecific antigen binding molecules bind CD28 and PSMA; such bispecific antigen binding molecules are also referred to herein as anti-CD28/anti-PSMA bispecific molecules. The anti-PSMA portion of the αanti-CD28/antiPSMA bispecific molecule is effective in targeting tumor cells expressing PSMA (eg, prostate tumor cells), and the anti-CD28 portion of the bispecific molecule is effective in activating T cells. Simultaneous binding of PSMA on the tumor cell and CD28 on the T cell promotes targeted killing (cell lysis) of the target tumor cell by the activated T cell. Therefore, the anti-CD28/anti-PSMA bispecific molecules of the invention are effective, among other things, for the treatment of diseases and disorders associated with or caused by tumors expressing PSMA (eg, prostate cancer).

Bispecific antigen binding molecules according to this aspect of the present invention comprise a first antigen binding domain that specifically binds human CD28 and a second antigen binding domain that specifically binds PSMA. The present invention includes bispecific αanti-CD28/αanti-PSMA molecules (eg, bispecific antibodies), wherein each antigen binding domain comprises a heavy chain variable region (HCVR) paired with a light chain variable region (LCVR). In certain illustrative embodiments, the anti-CD28 antigen-binding domain and the anti-PSMA antigen-binding domain each comprise a different, separate HCVR paired with a common LCVR.

The present invention relates to bispecific anti-CD28/anti-PSMA molecules, wherein the first antigen binding domain, which specifically binds CD28, contains any of the HCVR amino acid sequences presented in table. 1. The first antigen binding domain, which specifically binds CD28, may also contain any of the LCVR amino acid sequences presented in table. 1. In certain embodiments, the first antigen binding domain that specifically binds CD28 comprises any of the HCVR/LCVR amino acid sequence pairs presented in Table. 1. The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the first antigen binding domain that specifically binds CD28 comprises any of the heavy chain amino acid sequences CDR1CDR2-CDR3 presented in Table 1. 1, and/or any of the light chain amino acid sequences CDR1-CDR2-CDR3 presented in table. 1.

In certain embodiments, the present invention provides bispecific anti-CD28/anti-PSMA molecules, wherein the first antigen binding domain that specifically binds CD28 comprises a heavy chain variable region (HCVR) having an amino acid sequence selected from the group consisting of SEQ ID NO : 10, 26 and 58, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98% or at least 99% sequence identity.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the first antigen binding domain that specifically binds CD28 comprises a light chain variable region (LCVR) having an amino acid sequence selected from the group consisting of SEQ ID NO: 18, 42 and 66, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98%, or at least 99% sequence identity.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the first antigen binding domain that specifically binds CD28 comprises a pair of amino acid sequences HCVR and LCVR (HCVR/LCVR) selected from the group consisting of SEQ ID NO: 10/ 18, 26/42 and 58/66.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the first antigen binding domain that specifically binds CD28 comprises a heavy chain CDR3 domain (HCDR3) having an amino acid sequence selected from the group consisting of SEQ ID NO: 16, 32 and 64, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98%, or at least 99% sequence identity; and a light chain CDR3 domain (LCDR3) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 24, 48 and 72, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity.

In certain embodiments, the first antigen binding domain that specifically binds CD28 comprises the amino acid sequence pair HCDR3/LCDR3 selected from the group consisting of SEQ ID NOs: 16/24, 32/48, and 64/72.

The present invention also relates to bispecific antigen-binding molecules anti

- 2 045943

CD28/aHTu-PSMA, wherein the first antigen binding domain that specifically binds CD28 comprises a heavy chain CDR1 (HCDR1) domain having an amino acid sequence selected from the group consisting of SEQ ID NOs: 12, 28 and 60, or substantially similar to them a sequence having at least 90%, at least 95%, at least 98% or at least 99% sequence identity; heavy chain CDR2 domain (HCDR2) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 14, 30 and 62, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity; light chain CDR1 domain (LCDR1) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 20, 44 and 68, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity; and a light chain CDR2 domain (LCDR2) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 22, 46 and 70, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity.

Some non-limiting exemplary anti-CD28/anti-PSMA bispecific antigen binding molecules of the invention include a first antigen binding domain that specifically binds CD28 comprising the domains HCDR1-HCDR2-HCDR3-LCDR1-LCDR2-LCDR3, respectively, having an amino acid sequence selected from the group consisting from: SEQ ID NO: 12-14-16-20-22-24; 28-30-32-44-46-48; and 60-62-64-68-70-72.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the second antigen binding domain that specifically binds PSMA comprises a heavy chain variable region (HCVR) having an amino acid sequence selected from the group consisting of SEQ ID NO: 2, 34 and 50, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98%, or at least 99% sequence identity.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the second antigen binding domain that specifically binds PSMA comprises a light chain variable region (LCVR) having an amino acid sequence selected from the group consisting of SEQ ID NO: 18, 42 and 66, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98%, or at least 99% sequence identity.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the second antigen binding domain that specifically binds PSMA comprises a pair of amino acid sequences HCVR and LCVR (HCVR/LCVR) selected from the group consisting of SEQ ID NO: 2/ 18, 34/42 and 50/66.

The present invention also provides bispecific anti-CD28/anti-PSMA molecules, wherein the second antigen binding domain that specifically binds PSMA comprises a heavy chain CDR3 domain (HCDR3) having an amino acid sequence selected from the group consisting of SEQ ID NO: 8, 40 and 56, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98%, or at least 99% sequence identity; and a light chain CDR3 domain (LCDR3) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 24, 48 and 72, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity.

In certain embodiments, the second antigen binding domain that specifically binds PSMA comprises the amino acid sequence pair HCDR3/LCDR3 selected from the group consisting of SEQ ID NOs: 8/24, 40/48 and 56/72.

The present invention also provides bispecific antiCD28/anti-PSMA antigen binding molecules, wherein the second antigen binding domain that specifically binds PSMA comprises a heavy chain CDR1 domain (HCDR1) having an amino acid sequence selected from the group consisting of SEQ ID NO: 4, 36 and 52, or a substantially similar sequence thereto having at least 90%, at least 95%, at least 98%, or at least 99% sequence identity; heavy chain CDR2 domain (HCDR2) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 6, 38 and 54, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity; light chain CDR1 domain (LCDR1) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 20, 44 and 68, or a substantially similar sequence thereto, having at least 90%, at least 95%, at least 98% or at least 99% sequence identity; and a light chain CDR2 domain (LCDR2) having an amino acid sequence selected from the group consisting of SEQ ID NOs: 22, 46 and 70, or substantially similar thereto

- 3 045943 identity having at least 90%, at least 95%, at least 98% or at least 99% sequence identity.

Some non-limiting exemplary anti-CD28/anti-PSMA bispecific antigen binding molecules of the invention include a second antigen binding domain that specifically binds PSMA comprising the domains HCDR1-HCDR2-HCDR3-LCDR1-LCDR2-LCDR3, respectively, having amino acid sequences selected from the group consisting from: SEQ ID NO: 4-6-8-20-22-24; 36-38-40-44-46-48; and 52-54-56-68-70-72.

In a related embodiment, the invention includes anti-CD28/anti-PSMA bispecific antigen binding molecules, wherein the second antigen binding domain that specifically binds PSMA comprises heavy and light chain CDR domains contained in heavy and light chain variable region (HCVR/LCVR) sequences, selected from the group consisting of SEQ ID NO: 2/18, 34/42 and 50/66.

In yet another aspect, the present invention provides nucleic acid molecules encoding any of the HCVR, LCVR, or CDR sequences of the anti-CD28/anti-PSMA bispecific antigen binding molecules disclosed herein, including nucleic acid molecules comprising the polynucleotide sequences set forth in Table. 1 herein, as well as nucleic acid molecules containing two or more polynucleotide sequences presented in table. 1, in any functional combination or arrangement. Recombinant expression vectors carrying the nucleic acids of the invention and the host cells into which such vectors have been introduced are also covered by the invention, as are methods for producing antibodies by culturing said host cells under conditions that allow the production of antibodies and the isolation of the resulting antibodies.

The present invention includes bispecific anti-CD28/anti-PSMA antigen-binding molecules, wherein any of the above-mentioned antigen-binding domains that specifically bind CD28 is combined, linked or otherwise associated with any of the above-mentioned antigen-binding domains that specifically bind PSMA to form a bispecific antigen-binding molecule that binds CD28 and PSMA.

The present invention includes anti-CD28/anti-PSMA bispecific antigen-binding molecules having a modified glycosylation profile. For some applications, a modification to remove unwanted glycosylation sites, or an antibody that lacks the fucose moiety present on the oligosaccharide chain, may be useful, for example to enhance the function of antibody-dependent cellular cytotoxicity (ADCC) (see Shield et al. (2002) JBC 277:26733). For other applications, galactosylation modification can be made to modify complement dependent cytotoxicity (CDC).

In yet another aspect of the invention, there is provided a pharmaceutical composition comprising an anti-CD28/anti-PSMA bispecific antigen binding molecule as disclosed herein and a pharmaceutically acceptable carrier. In a related aspect of the invention, a composition is provided that is a combination of an anti-CD28/anti-PSMA bispecific antigen-binding molecule and a second therapeutic agent. In one embodiment, the second therapeutic agent is any agent that can be usefully used in combination with an anti-CD28/anti-PSMA bispecific antigen binding molecule. Examples of agents that may be usefully used in combination with the anti-CD28/anti-PSMA bispecific antigen binding molecule are discussed in detail elsewhere herein.

In yet another aspect of the invention, therapeutic methods are provided for targeting/killing tumor cells expressing PSMA using the anti-CD28/anti-PSMA bispecific antigen binding molecule of the invention, wherein said therapeutic methods comprise administering a therapeutically effective amount of a pharmaceutical composition comprising the anti-CD28/anti-PSMA bispecific antigen binding molecule. -CD28/anti-PSMA according to the present invention to a subject in need thereof.

The present invention also includes the use of the anti-CD28/anti-PSMA bispecific antigen binding molecule of the invention for the manufacture of a medicament for the treatment of a disease or disorder associated with or caused by PSMA expression.

In another aspect of the invention, therapeutic methods are provided for targeting/killing tumor cells expressing PSMA using an anti-CD28/anti-PSMA bispecific antigen binding molecule of the invention, wherein said anti-CD28/anti-PSMA bispecific antigen binding molecule is used in combination with other antitumor bispecific antigen-binding molecules that bind to CD3 (eg anti-CD28/anti-PSMA in combination with anti-CD3/anti-PSMA antibodies).

In yet another aspect of the invention, therapeutic methods are provided for targeting/killing tumor cells expressing PSMA using an anti-CD28/anti-PSMA bispecific antigen binding molecule of the invention, wherein said anti-CD28/anti-PSMA bispecific antigen binding molecule is used in combination with an inhibitor

- 4 045943 checkpoints targeting PD-1 or CTLA-4 (eg aHTu-CD28/aHTu-PSMA in combination with anti-PD-1 antibodies).

In another aspect of the invention, therapeutic methods are provided for targeting/killing tumor cells expressing PSMA using an anti-CD28/anti-PSMA bispecific antigen binding molecule of the invention, wherein said anti-CD28/anti-PSMA bispecific antigen binding molecule is used in combination with other antitumor bispecific antigen-binding molecules that bind to CD3 (eg, anti-CD28/anti-PSMA in combination with anti-CD3/anti-PSMA bispecific antibodies), and a checkpoint inhibitor targeting PD-1 or CTLA-4 (eg, anti-CD28/anti-PSMA in combination with anti-PD-1 antibodies).

Other embodiments will be apparent from a review of the following detailed description of the invention.

Brief description of graphic materials

In fig. 1 schematically shows the structure of an exemplary anti-CD28/PSMA bispecific antibody.

Fig. 2 is a graph showing inhibition of tumor growth in genetically modified cell lines introduced to express a costimulatory ligand. Three tumor cell lines, B16F10.9, EL4, and MS 38, were genetically modified to express a costimulatory ligand or GFP as a control. Fig. 2 is a graph showing the percentage (%) of survival. Mice were sacrificed when tumors grew to a size greater than 2000 mm ³ .

Fig. Figure 3 schematically shows the original and genetically modified bioassay for testing bispecific antibodies to PSMAxCD28.

In fig. 4A and 4B show that activation of CD4+ T cells and genetically modified JRT3.T3/1G4/hCD28 cells was enhanced by anti-hPSMAxhCD28 in primary stimulation conditions (REGN2281) and PSMA expressed on HEK293/hCD20 cells.

In fig. 5A-H show that in genetically modified cell lines with introduced PSMA, bispecific antibodies to PSMAxCD28 enhance T cell activation under conditions of TCR stimulation with bispecific antibodies to CD20xCD3.

In fig. 5A schematically shows the mechanism of action of bispecific antibodies.

Fig. 5B is a graph showing the quantification of the ratio of CD28 fluorescence within and outside the immunological synapse. Genetically modified PSMA target cells (HEK293) and human Jurkat T cells were co-cultured with fluorescently labeled bispecific antibodies (anti-PSMAxCD28, uhtu-CD20xCD3) for 1 h at 37°C, gently fixed and stained with anti-CD28 antibody. The number of analyzed cells in each group is indicated in the histogram.

In fig. 5C and 5D show the proliferation of human T cells cultured with genetically modified PSMA target cells. Data shown are mean ±SEM (standard error of the mean). Data refer to at least two (2) experiments.

In fig. 5C is a graph showing dose titration of CD20xCD3 in the presence of 0.5 nM isotype control hIgG4s or PSMAxCD28.

In fig. 5D is a graph showing dose titration of PSMAxCD28 or the indicated controls in the presence of 5 pM isotype control hIgG4s or CD20xCD3.

Fig. 5E-G are graphs demonstrating the proliferation and cytokine release of human T cells cultured with genetically modified TAA (PSMA or CD20) target cells as indicated at the top of the panel, in the presence of 5 pM isotype control hIgG4s (bottom panel of each figure ) or CD20xCD3 (top panel of each figure). Data represent mean ± SEM. Data refer to at least three (3) experiments.

Fig. 5E is a graph showing proliferation.

Fig. 5F is a graph showing the release of IL-2.

Fig. 5G is a graph showing the release of ZFNy.

Fig. 5H is a graph demonstrating that bispecific antibodies to PSMAxCD28 and PSMAxCD3 can simultaneously bind to cells expressing PSMA. 22RV1 cells were preincubated for 30 min at 4°C in flow cytometry buffer (PBS+1%FBS) with 20 mg/ml PSMAxCD3 or 20 mg/ml anti-PSMA antibody bearing an anti-PSMA arm similar to that contained in the bispecific PSMAxCD28. After incubation, cells were washed with flow cytometry buffer and incubated for 20 min at 4°C with 5 mg/ml PSMAxCD28 directly labeled with Alexa647. After incubation, cells were washed, resuspended in flow cytometry buffer, and analyzed by flow cytometry.

In fig. 6A and 6B show that anti-PSMAxCD28 enhanced the cytotoxic activity of anti

- 5 045943

PSMAxCD3 in the presence of prostate carcinoma cells and human or cynomolgus macaque T cells.

In fig. 7A-H are graphs demonstrating that in cancer cell lines with endogenous PSMA, bispecific antibodies to PSMAxCD28 enhance T cell activation under conditions of TCR stimulation with bispecific antibodies to PSMAxCD3. In fig. 7A-D human T cells were cultured with target cancer cells with endogenous PSMA expression (prostate cancer line C4-2) and the indicated bispecific antibodies for 96 hours.

Fig. 7A is a graph showing the killing of tumor cells. Data shown are the percentage of viable cells.

Fig. 7B is a graph showing the release of ZFNy.

In fig. 7C is a graph showing the number of CD4 T cells and the frequency of CD25' cells as a percentage of CD4 T cells.

In fig. 7D is a graph showing the number of CD8 T cells and the frequency of CD25+ cells as a percentage of CD8 T cells.

In fig. Cynomolgus 7E-G T cells were cultured with endogenously expressing PSMA cancer cells (prostate cancer line C4-2) and the indicated bispecific antibodies for 96 h.

Fig. 7E is a graph showing tumor cell killing. Data shown are the percentage of viable cells.

In fig. 7F is a graph showing the number of CD4 T cells and the frequency of CD25+ cells as a percentage of CD4 T cells.

In fig. 7G shows graphs showing the number of CD8 T cells and the frequency of CD25+ cells as a percentage of CD8 T cells.

In fig. 7H is a graph showing antibody binding to cellular targets measured by flow cytometry.

In fig. 8 shows a comparison of the properties of a TGN superagonist and an anti-CD28 antibody according to the invention.

In fig. 9A and 9B are graphs and scatter plots demonstrating that the PSMAxCD28 bispecific antibody enhances antitumor immunity through PSMAxCD3 and induced T cell activation. MC38/hPSMA tumor cells were subcutaneously implanted into humanized hCD3/hCD28/hPSMA mice. Mice were administered the indicated bispecific antibodies at a dose of 5 mg/kg on days 0, 3 and 7.

In fig. Figure 9A shows the dynamics of tumor volumes. Values are means ± SEM and refer to three (3) experiments with 3-7 mice per group. P values were calculated using two-way analysis of variance. (*, p<0.05; **, p<0.01; ***, p<0.001 and ****, p<0.0001).

In fig. 9B is a graph showing the percentage of cells in each cluster from each treatment group (top panel); overlay of the specified cluster on the viSNE plot (bottom panel).

In fig. 9C and 9D are histograms demonstrating that the PSMAxCD28 bispecific antibody enhances PSMAxCD3-induced cytokine release. Mice were bled for serum cytokine levels 4 hours post-dose on day 0. Statistical significance was calculated using one-way analysis of variance compared with isotype **p<0.01 and ***p<0.0001. n=7 mice per group. Data refer to 3 experiments.

In fig. 9E is a graph and scatter plot demonstrating that PSMAxCD28 bispecific antibody enhances PSMAxCD3-induced T cell activation.

In fig. 10 shows that bispecific anti-PSMAxCD28 antibodies or parental bivalent anti-CD28 antibodies do not induce serum cytokine production in humanized CD3/CD28/PSMA mice.

In fig. Figure 11 shows that treatment with the CD28 superagonist induced a cytokine response after 4 hours in PBMC-engrafted NSG mice, in contrast to the anti-CD28 antibody mAb14226P2.

In fig. 12 shows the potentially safer toxicological profile of the co-stimulatory bispecific antigen binding molecules of the present invention. Ahtu-CD28 x anti-PSMA did not induce a cytokine response, unlike anti-CD3 x anti-PSMA.

In fig. 13A and 13B show that PSMAxCD28 did not exhibit cytokine production or T cell margination compared to PSMAxCD3 and CD28 superagonist. LLOQ: lower limit of quantitation.

In fig. 14 and 17 show that TSAxCD28 and PD-1 blockade activate TCR/CD3 and CD28 on T cells at the tumor site. In fig. 15, 16, and 18 show that bispecific TSAxCD28 and PD-1 blockade synergistically promote T cell activation in vitro. In fig. 14-18 T cell (Jurkat/PD-1) and target cells (Raji WT) are conjugated in the presence of a non-blocking mAb to PD-1 (NB mAb to PD-1) or a blocker (mAb to PD-1) and the CD20xCD3 bispecific.

In fig. 14 shows histograms demonstrating quantification of the localization of PD-1 and CD28 at the immunological synapse. Statistical significance was calculated using

- 6,045,943 paired t-tests (not significant, ns). To quantify the localization of PD-1 and CD28 at the immunological synapse, images of conjugates of T cells (Jurkat/PD-1) and target cells (Raji WT) were obtained in the presence of a non-blocking mAb to PD-1 (NB mAb to PD-1) or blocker (mAb to PD-1) and bispecific antibody to CD20xCD3. The anti-PD-1 mAb was directly labeled with Alexa647, the anti-CD20xCD3 bispecific antibody was directly labeled with Alexa488, the anti-CD28 mAb was directly labeled with PE, and the nuclei were stained with Hoechst 33342 (images not shown).

In fig. 15 shows histograms demonstrating quantification of the localization of PD-1 and CD28 at the immunological synapse. Statistical significance was calculated using an unpaired t test (p<0.0001, ****). To quantify the localization of PD-1 and CD28 at the immunological synapse, images of conjugates of T cells (Jurkat/PD-1) and target cells (Raji/PDL1) were obtained in the presence of a non-blocking mAb to PD-1 (NB mAb to PD-1) or a blocker (mAb to PD-1) and a bispecific antibody to CD20xCD3. The anti-PD-1 mAb was directly labeled with Alexa647 (shown in red), the CD20xCD3 was directly labeled with Alexa488 (shown in green), the anti-CD28 mAb was directly labeled with PE (shown in blue), and the nuclei were stained with Hoechst 33342 (shown in gray). The dotted lines represent cell outlines drawn from the brightfield image.

In fig. 16 shows the percentage (%) of 22RV1-PDL1 cell viability after 96 hours.

Fig. 17 is a graph showing IFNy levels in the supernatant after 96 hours.

Fig. 18 is a graph showing the release of IL-2 after 96 hours.

In fig. 19A-D show that expression of CD28 ligand (CD86) on tumor cells acts synergistically with anti-PD1 molecule treatment in inducing CD8-mediated antitumor immunity. MC38 tumor cells were transduced with CD28 ligand, CD86 (MC38/CD86) or empty vector control (MC38/EV). WT C57BL6 mice were initially implanted with 1x 10 ⁶ tumor cells per mouse and administered anti-PD-1 mAb or rat isotype control at a dose of 5 mg/kg on days 0, 3, 7, 10, and 14 after tumor implantation.

In fig. 19A shows the dynamics of the average tumor volume. Error bars represent ±SEM. Statistical significance was determined using two-way analysis of variance and Tukey's multiple comparison tests.

In fig. 19B shows the dynamics of survival (percentage of mice with tumors <2000 mm ³ ). Statistical significance at 60 days postimplantation was determined using the log-rank test (Cox-Mantel).

In fig. 19C mice were treated with an antibody that depletes the CD8 population (CD8 depletion) or an isotype control (no depletion). Shown are the dynamics of mean tumor volume with CD8 depletion (dashed lines) and without depletion (solid lines) ±SEM. Statistical significance was determined using two-way analysis of variance and Tukey's multiple comparison tests.

In fig. 19D shows re-implantation of tumors (re-challenge with tumor cells) in tumor-free mice implanted with MC38/CD86 and treated with anti-PD1 mAb.

In fig. 19A-D shows data from 1 experiment with 10 mice per group. Data refer to at least 4 separate experiments. Statistical significance is indicated (*p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001).

In fig. 20, 21A, 22, 24, and 25 show that PSMAxCD28 acts synergistically with anti-PD1 mAb treatment to induce antitumor immunity. MC38/hPSMA tumor cells were subcutaneously implanted into hCD3/hCD28/hPSMA mice. Anti-PSMAxCD28 bispecific antibody, anti-PD-1 mAb, or rat isotype control IgG2a were administered as monotherapy or in combination by intraperitoneal injection at 5 mg/kg each. In fig. 21B and 21C show that the combination of PSMAxCD28 and anti-PD-1 mAb increases the frequency of tumor-specific T cells. In fig. 23A and 23B show that PSMAxCD28 acts synergistically with anti-PD1 antibody treatment to induce intratumoral, but not splenic or systemic cytokines.

In fig. 20A-E show that immediate treatment with anti-PSMA x anti-CO28 enhances immunity at the tumor site and acts synergistically with anti-PD-1 antibodies to promote tumor rejection. In fig. 20A shows the dynamics of the average tumor volume. Error bars represent ±SEM. Statistical significance was determined using two-way analysis of variance and Tukey's multiple comparison tests (***, p < 0.001 and ****, p < 0.0001). In fig. 20B shows the dynamics of survival (mice with tumors >2000 mm ³ were sacrificed). Statistical significance was determined using the log-rank test (Cox-Mantel) (**, p < 0.01). In fig. 20C shows the average tumor volume at 21 days after implantation. Error bars represent ±SEM. Statistical significance was determined using one-way analysis of variance and the Holm-Sidak multiple comparison test (****, p < 0.0001). In fig. 20D shows a table with tumor volumes (mm ³ ) on day 21 after implantation. In fig. 20E is a table showing the number of tumor-free mice per group for 6 separate experiments.

In fig. 21A shows repeated challenge with tumor cells in mice without tumors, previously

- 7,045,943 who received a combination of PSMAxCD28 and mAb to PD1. Error bars represent ±SEM. Data refer to 4 experiments.

In fig. 21B shows MC38/PSMA tumor cells implanted into humanized CD3/CD28/PSMA mice that were treated with isotype control, PSMAxCD28, anti-PD1 mAb, or the combination at 5 mg/kg on days 10 and 14 postimplantation. Spleens were collected on day 17. Splenocytes were cultured overnight in T cell medium with 10 mg/ml peptide (p15E or OVA) and 2 mg/ml anti-CD28 antibody. After overnight incubation, intracellular cytokine staining was performed according to standard procedures.

In fig. Figure 21C shows that tumor-free mice treated with α-anti-CD28 x PSMA and anti-PD1, but not anti-CD28 x PSMA and anti-CD3 x PSMA, rejected recurrent tumors.

In fig. 22A-D, 24 and 25 show a delayed/therapeutic treatment regimen (dosing indicated by arrows; FIGS. 22A and 22B on days 9, 16 and 22; FIGS. 24 and 25 on days 7, 11 and 14.

In fig. 22A shows the dynamics of the average tumor volume. Error bars represent ±SEM. Statistical significance was determined using two-way analysis of variance and Tukey's multiple comparison tests. Data refer to 3 experiments.

In fig. 22B shows splenic and intratumoral cytokines ex vivo. Points represent data from individual mice. Bars represent mean ±SEM.

In fig. Figure 22C shows the average tumor size at day 20.

In fig. 22D shows the survival of mice treated with different treatments.

In fig. 23A and 23B show that PSMAxCD28 acts synergistically with anti-PD1 antibody treatment to induce intratumoral, but not splenic or systemic cytokines.

Data in Fig. 23A are similar to FIG. 22V. Splenic and intratumoral cytokines ex vivo. Points represent data from individual mice. Bars represent mean ±SEM.

In fig. 23B, triple humanized CD3/CD28/PSMA mice were implanted with MC38/hPSMA and treated with the antibody at a dose of 5 mg/kg on day 0. Mice were bled and serum collected 4 hours post-dose. In fig. 23A and 23B, statistical significance was calculated using one-way analysis of variance and Tukey's multiple comparison tests. *p<0.05, **p<0.01, ****p<0.0001.

In fig. 24 shows PD1 expression on T cell subsets from the draining lymph node (dLN), spleen (Sp), and tumor (Tu). Points represent data from individual mice. Bars represent mean ±SEM.

In fig. 25 shows the frequency of CD8 T cells in C1 and C2 from the indicated treatment groups.

The data in each of Figs. 22B, 24 and 25 refer to 1 experiment. Fig. 22B and 24: n=4-6 mice per group. Fig. 25: n=10 mice per group.

In fig. 26A-C and 27 show that TAAxCD28 alone or in combination with PD1 therapy does not induce systemic T cell activation compared with a CD28 superagonist in cynomolgus monkeys. Cynomolgus monkeys were administered a single dose of bispecifics at the indicated dose (1 or 10 mg/kg, as indicated). Time refers to the period after dosing (h).

In fig. 26A shows serum cytokines.

In fig. 26B shows the relative numbers of T cells in peripheral blood.

In fig. 26C shows the frequency of Ki67 ⁺ and ICOS+ T cells (% of CD3).

In fig. 26A-C values represent the mean ±SEM. N=3 animals per group.

In fig. 27 shows that triply humanized CD3/CD28/PSMA mice were administered a single dose of antibody (0.25 or 2.5 mg/kg, as indicated). Mice were bled and serum collected 4 hours (day 0) post-dose. Statistical significance was calculated using one-way analysis of variance and the Holm-Sidak multiple comparison test. *p<0.05, **p<0.01, ****p<0.0001.

In fig. 28 showed that PSMAxCD28 ±PD1 did not increase serum cytokine levels in tumor-bearing mice.

In fig. 29A and 29B show that PSMAxCD28 alone or in combination with an anti-PD1 mAb exhibits a safe cytokine profile in vivo in tumor-free mice. The data are similar to those presented in Fig. 27. CD3/CD28/PSMA humanized mice were dosed with a single dose of antibody (0.25 or 2.5 mg/kg, as indicated) in triplicate. Mice were bled and serum collected at 4 hours (day 0) and 72 hours (day 3) post-dose. Statistical significance was calculated using one-way analysis of variance and the Holm-Sidak multiple comparison test. *p<0.05, **p<0.01, ****p<0.0001.

Detailed Description of the Invention

Before reading the description of the present invention, it should be understood that the present invention is not limited to the specific methods and experimental conditions described, since

- 8 045943 such methods and conditions may vary. It should also be understood that the terminology used herein is intended to describe specific embodiments only and is not intended to be limiting, as the scope of the present invention will be limited only by the appended claims.

Unless otherwise specified, all technical and scientific terms used herein have the meaning commonly understood by one skilled in the art to which this invention belongs. As used herein, the term approximately, when applied to a particular numerical value given, means that the stated value may differ from the given value by no more than 1%. For example, as used herein, the expression approximately 100 includes 99 and 101, and all values in between (eg, 99.1, 99.2, 99.3, 99.4, etc.).

Although any methods and materials similar or equivalent to those described herein may be used in the practice or testing of this invention, preferred methods and materials are described below. All patents, applications and non-patent publications mentioned herein are incorporated herein by reference in their entirety.

Definitions

The expression CD28 as used herein refers to an antigen expressed on T cells as a costimulatory receptor. Human CD28 contains the amino acid sequence shown in SEQ ID NO: 74 and/or has the amino acid sequence corresponding to registration number NP_006130.1. All references to proteins, polypeptides and protein fragments herein are intended to refer to the human version of the corresponding protein, polypeptide or protein fragment unless explicitly stated to be from a non-human species. Thus, the expression CD28 means human CD28 unless specified to be of a non-human species, eg, murine CD28, simian CD28, etc.

As used herein, the term CD28-binding antibody or anti-CD28 antibody includes antibodies and antigen-binding fragments thereof that specifically recognize monomeric CD28, as well as antibodies and antigen-binding fragments thereof that specifically recognize dimeric CD28. Antibodies and antigen binding fragments of the present invention can bind soluble CD28 and/or CD28 expressed on the cell surface. Soluble CD28 includes natural CD28 proteins as well as recombinant variants of CD28 proteins, such as, for example, monomeric and dimeric CD28 constructs that lack the transmembrane domain or are not otherwise associated with the cell membrane.

As used herein, the expression CD28 expressed on the cell surface means one or more CD28 proteins that are expressed on the cell surface in vitro or in vivo such that at least a portion of the CD28 protein is external to the extracellular side of the cell membrane and accessible to the antigen binding portion of the antibody. CD28 expressed on the cell surface includes CD28 proteins contained in the context of a functional T cell costimulatory receptor in the cell membrane. CD28 expressed on the cell surface includes the CD28 protein expressed as part of a homodimer on the cell surface. CD28 expressed on the cell surface may contain or consist of CD28 protein expressed on the surface of a cell that normally expresses CD28 protein. Alternatively, cell surface expressed CD28 may contain or consist of CD28 protein expressed on the surface of a cell that does not normally express human CD28 on its surface but has been artificially genetically modified to express CD28 on its surface.

As used herein, the expression anti-CD28 antibody includes both monovalent antibodies with a single specificity and bispecific antibodies comprising a first arm that binds CD28 and a second arm that binds a second antigen (target), wherein the anti-CD28 arm contains any of the sequences HCVR/LCVR or CDR given in table. 1 in this document. Examples of bispecific antibodies to CD28 are described elsewhere herein. The term antigen-binding molecule includes antibodies and antigen-binding fragments of antibodies, including, for example, bispecific antibodies.

The term antibody as used herein means any antigen-binding molecule or molecular complex containing at least one complementarity determining region (CDR) that specifically binds or interacts with a particular antigen (eg, CD28). The term antibody includes immunoglobulin molecules containing four polypeptide chains—two heavy (H) chains and two light (L) chains—linked by disulfide bonds, as well as their multimers (eg, IgM). Each heavy chain contains a heavy chain variable region (abbreviated herein as HCVR or _VH ) and a heavy chain constant region. The heavy chain constant region contains three domains, CH1, CH2 and CH3. Each light chain contains a light chain variable region (abbreviated herein as LCVR or VL) and a light chain constant region. The light chain constant region contains one domain (CL1). The VH and _VL regions can be further subdivided into regions of hypervariability called complementarity determining regions (CDRs), which alternate with

- 9 045943 more conserved regions called framework regions (FR). Each VH and VL consists of three CDRs and four FRs, located from the amino terminus to the carboxy terminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. In various embodiments, FR antibodies to CD28 (or antigen-binding portion thereof) may be identical to human germline sequences, or may be naturally or artificially modified. A consensus amino acid sequence can be determined based on comparative analysis of two or more CDRs.

The term antibody as used herein also includes antigen-binding fragments of full-length antibody molecules. The terms antigen-binding portion of an antibody, antigen-binding fragment of an antibody, etc. as used herein include any naturally occurring, enzymatically produced, synthetic or genetically modified polypeptide or glycoprotein that specifically binds an antigen to form a complex. Antigen-binding antibody fragments can be prepared, for example, from full-length antibody molecules using any suitable standard techniques, such as proteolytic digestion or recombinant genetic engineering techniques involving processing and expression of DNA encoding antibody variable and, optionally, constant domains. Such DNA is known and/or readily available from, for example, commercial sources, DNA libraries (including, for example, phage antibody libraries), or can be synthesized. The DNA can be sequenced and processed chemically or using molecular biology techniques, for example, to place one or more variable and/or constant domains in a suitable configuration or to introduce codons, create cysteine residues, modify, add or remove amino acids, etc.

Non-limiting examples of antigen binding fragments include: (i) Fab fragments; (ii) F(ab')2 fragments; (iii) Fd fragments; (iv) Fv fragments; (v) single chain Fv molecules (scFv); (vi) dAb fragments; and (vii) minimal recognition units consisting of amino acid residues mimicking the hypervariable region of an antibody (eg, a designated complementarity determining region (CDR), e.g., CDR3 peptide), or a conformationally restricted peptide FR3-CDR3-FR4. Other genetically modified molecules, such as domain-specific antibodies, single-domain antibodies, domain-deleted antibodies, chimeric antibodies, CDR-grafted antibodies, diabodies, tri-bodies, tetrabodies, minibodies, nanobodies (e.g., monovalent nanobodies, divalent nanobodies, etc.), small modular protein (SMIP) immunopharmaceuticals and shark IgNAR variable domains are also encompassed by the expression antigen binding fragment as used herein.

An antigen binding fragment of an antibody typically contains at least one variable domain. The variable domain can be of any size or amino acid composition and typically contains at least one CDR that is adjacent to or in reading frame with one or more framework sequences. In antigen binding fragments having a VH domain associated with a VL domain, the VH and VL domains may be arranged relative to each other in any suitable order. For example, the variable region may be dimeric and contain dimers V _H -V _H , V _H -V _L or V _L -V _L . Alternatively, the antigen binding fragment of the antibody may comprise a monomeric V _H or V _L domain.

In certain embodiments, the antigen binding fragment of an antibody may comprise at least one variable domain covalently linked to at least one constant domain. Non-limiting exemplary configurations of variable and constant domains that may be contained in an antigen binding fragment of an antibody according to the present invention^etenia,^include:⁽ⁱ⁾ VH-CH1;⁽ⁱⁱ⁾ VH-CH2;⁽ⁱⁱⁱ⁾ VH-CH3;^(iv) VH-CH1-CH²;^(v)VH^-CH^1-CH^2-CH^{3; (vi)} VH CH^2-CH^{3; (vii)} VH-CL;^(viii) VL-CH1;^(ix) VL-CH²;^(x) VL-CH3; (xi)VL^-CH^1-CH^{2; (xii)} VL^-CH^1-CH^2-CH^{3; (xiii)}VL-CH2-CH3; and (xiv) VL-CL. In any configuration of variable and constant domains, including any of the exemplary configurations listed above, the variable and constant domains may either be directly linked to each other or may be linked by a full or partial hinge or linker region. The hinge region may consist of at least 2 (eg, 5, 10, 15, 20, 40, 60 or more) amino acids, resulting in a flexible or semi-flexible connection between adjacent variable and/or constant domains in a single polypeptide molecule. In addition, an antigen binding antibody fragment may comprise a homodimer or heterodimer (or other multimer) in any of the variable and constant domain configurations listed above, in non-covalent association with each other and/or with one or more monomeric VH or VL domains (e.g. via disulfide bond(s).

As with full-length antibody molecules, antigen-binding fragments can be monospecific or multispecific (eg, bispecific). A multispecific antigen-binding antibody fragment typically contains at least two different variable domains, each capable of specifically binding to its own antigen or to another epitope on the same antigen. Any multispecific antibody format, including exemplary bispecific antibody formats disclosed herein, may be

- 10 045943 is adapted for use in the context of an antigen binding fragment of an antibody according to the present invention using standard methods available in the art.

Antibodies of the present invention may perform their function through complement-dependent cytotoxicity (CDC) or antibody-dependent cell-mediated cytotoxicity (ADCC).

Complement-dependent cytotoxicity (CDC) refers to the lysis of antigen-expressing cells by an antibody of the invention in the presence of complement. Antibody-dependent cell-mediated cytotoxicity (ADCC) refers to a cell-mediated reaction in which nonspecific cytotoxic cells expressing Fc receptors (FcR) (eg, natural killer (NK) cells, neutrophils, and macrophages) recognize bound antibody on a target cell and thereby lead to its lysis. CCZ and ADCC can be measured using assays that are well known and available in the art. (See, for example, US Patent Nos. 5,500,362 and 5,821,337, and Clynes et al. (1998) Proc. Natl. Acad. Sci. (USA) 95:652-656). The constant region of an antibody is important for the antibody's ability to fix complement and mediate cell-dependent cytotoxicity. Thus, the antibody isotype can be selected based on whether the antibody is desired to mediate cytotoxicity.

In certain embodiments, the anti-CD28 antibodies of the invention (monospecific or bispecific) are human antibodies. The term human antibody as used herein is intended to include antibodies having variable and constant regions derived from human germline immunoglobulin sequences. The human antibodies of the invention may include amino acid residues not encoded by human germline immunoglobulin sequences (eg, mutations introduced by random or site-specific mutagenesis in vitro or somatic mutation in vivo), for example, in the CDR and, in particular, in CDR3. However, the term human antibody as used herein is not intended to include antibodies in which CDR sequences derived from the germ line of another mammalian species, such as a mouse, have been grafted onto human framework sequences.

Antibodies of the invention may, in some embodiments, be recombinant human antibodies. The term recombinant human antibody as used herein is intended to include all human antibodies produced, expressed, generated or isolated by recombinant means, such as antibodies expressed using a recombinant expression vector transfected into a host cell (further described below), antibodies, antibodies isolated from a recombinant, combinatorial human antibody library (further described below), antibodies isolated from an animal (e.g., mouse) transgenic for human immunoglobulin genes (see, for example, Taylor et al. (1992) Nucl. Acids Res. 20: 62876295), or antibodies produced, expressed, created or isolated by any other method that involves splicing human immunoglobulin gene sequences to produce other DNA sequences. Such recombinant human antibodies have variable and constant regions derived from human germline immunoglobulin sequences. However, in certain embodiments, such recombinant human antibodies are subjected to in vitro mutagenesis (or, in the case of a human Ig transgenic animal, somatic mutagenesis in vivo) and thus the amino acid sequences of the VH and VL regions of the recombinant antibodies are sequences that, although derived from and related to human germline VH and VL sequences, they cannot naturally exist in the human germline antibody repertoire in vivo.

Human antibodies can exist in two forms, which are related to hinge differences. In one form, the immunoglobulin molecule contains a stable four-chain construct of approximately 150-160 kDa in which the dimers are held together by an interchain disulfide bond between the heavy chains. In the second form, the dimers are not linked by interchain disulfide bonds, and a molecule of approximately 75-80 kDa is formed, consisting of covalently linked light and heavy chains (half-antibody). These forms are extremely difficult to separate, even after affinity purification.

The frequency of occurrence of the second form in various intact IgG isotypes is due to, but not limited to, structural differences associated with the isotype of the antibody hinge region. A single amino acid substitution in the hinge region of the human G4 Ig hinge can significantly reduce the incidence of the second form (Angal et al. (1993) Molecular Immunology 30:105) to levels typically observed with the human G1 Ig hinge. The present invention covers antibodies having one or more mutations in the hinge regions, C _H 2 or C _H 3, which may be desirable, for example, in production, to improve the yield of the desired form of the antibody.

The antibodies of the invention may be isolated antibodies. An isolated antibody as used herein means an antibody that has been identified and separated and/or recovered from at least one component of its natural environment. For example, anti

- 11 045943 a body that has been separated or extracted from at least one component of the body, or from a tissue or cell in which the antibody exists naturally or is produced naturally, is an isolated antibody in the context of the present invention. The isolated antibody also includes the antibody in situ in the recombinant cell. Isolated antibodies are antibodies that have been subjected to at least one purification or isolation step. In certain embodiments, the isolated antibody may be substantially free of other cellular material and/or chemicals.

The present invention also includes single-arm antibodies that bind CD28. As used herein, the term single-arm antibody means an antigen-binding molecule comprising one antibody heavy chain and one antibody light chain. The single arm antibodies of the present invention may contain any of the HCVR/LCVR or CDR amino acid sequences presented in Table 1. 2.

The anti-CD28 antibodies herein or their antigen binding domains may contain one or more amino acid substitutions, insertions and/or deletions in the framework regions and/or CDR regions of the heavy and light chain variable domains as compared to the corresponding germline sequences from which the antigen binding domains were derived. proteins or antigen binding domains. Such mutations can be readily identified by comparing the amino acid sequences disclosed herein with germline sequences available, for example, from publicly available antibody sequence databases. The present invention includes antibodies and antigen binding domains thereof derived from any of the amino acid sequences disclosed herein, in which one or more amino acids in one or more framework regions and/or CDR regions are mutated to the corresponding germline sequence residue(s), from which the antibody was produced, or to the corresponding residue(s) of another human germline sequence, or to a conservative amino acid substitution of the corresponding germline residue(s) (such sequence changes are collectively referred to herein as germline mutations). A person skilled in the art, starting with the heavy and light chain variable region sequences disclosed herein, can readily produce a variety of antibodies and antigen binding fragments containing one or more individual germline mutations or combinations thereof. In certain embodiments, all framework residues and/or CDR residues in the VH and/or _VL domains are mutated back to residues contained in the original germline sequence from which the antibody was derived. In other embodiments, only certain residues are mutated back to the original germline sequence, for example, only the mutated residues contained in the first 8 amino acids of FR1 or the last 8 amino acids of FR4, or only the mutated residues contained in CDR1, CDR2, or CDR3. In other embodiments, one or more framework residues and/or CDR residues are mutated to the corresponding residue(s) of a different germline sequence (ie, a germline sequence that is different from the germline sequence from which the antibody was originally derived). In addition, antibodies or antigen binding domains thereof according to the present invention may contain any combination of two or more germline mutations within framework regions and/or CDR regions, for example, in which some individual residues are mutated to the corresponding residue of a particular germline sequence, while some other residues that differ from the original germline sequence are retained or mutated into the corresponding residue of another germline sequence. Once produced, antibodies or antigen-binding fragments thereof containing one or more germline mutations can be readily tested for one or more desired properties, such as improved binding specificity, increased binding affinity, improved or enhanced antagonistic or agonistic biological properties (as appropriate). ), reduced immunogenicity, etc. Antibodies or antigen binding fragments thereof prepared in this general manner are covered by the present invention.

The present invention also includes anti-CD28 antibodies and antigen-binding molecules containing variants of any of the HCVR, LCVR and/or CDR amino acid sequences disclosed herein. Exemplary variants included in this aspect of the invention include variants of any of the HCVR, LCVR and/or CDR amino acid sequences disclosed herein having one or more conservative substitutions. For example, the present invention includes antibodies and antigen binding molecules to CD28 having amino acid sequences HCVR, LCVR and/or CDRs with, for example, 10 or less, 8 or less, 6 or less, 4 or less, etc. conservative amino acid substitutions relative to any of the amino acid sequences HCVR, LCVR and/or CDR presented in table. 1 in this document.

The term epitope refers to an antigenic determinant that interacts with a specific antigen-binding site in the variable region of an antibody molecule, known as a paratope. One antigen can have more than one epitope. Thus, different antibodies can bind to

- 12 045943 different regions on the antigen and can have different biological effects. Epitopes can be conformational or linear. A conformational epitope is formed by spatially close amino acids from different segments of a linear polypeptide chain. A linear epitope is an epitope formed by adjacent amino acid residues in a polypeptide chain. In certain circumstances, the epitope may include moieties of saccharides, phosphoryl groups, or sulfonyl groups on the antigen.

The term substantially identical or substantially identical, when applied to a nucleic acid or fragment thereof, means that when optimally aligned with corresponding nucleotide insertions or deletions with another nucleic acid (or its complementary strand), there is a nucleotide base identity of the nucleotide sequence of at least approximately 95% and more preferably at least about 96%, 97%, 98% or 99%, as measured by any well known sequence identity algorithm such as FASTA, BLAST or GAP, as discussed below. A nucleic acid molecule substantially identical to a reference nucleic acid molecule may, in certain cases, encode a polypeptide having the same or substantially similar amino acid sequence as the polypeptide encoded by the reference nucleic acid molecule.

When applied to polypeptides, the term substantially similar or substantially similar means that the two peptide sequences, when optimally aligned, for example by the GAP or BESTFIT programs using default gap opening penalties, have at least 95% sequence identity, even more preferably at least 98% or 99% sequence identity. Preferably, positions of residues that are not identical are distinguished by conservative amino acid substitutions. A conservative amino acid substitution is a substitution in which an amino acid residue is replaced by another amino acid residue having a side chain (R group) with similar chemical properties (eg, charge or hydrophobicity). In general, a conservative amino acid substitution does not substantially change the functional properties of the protein. In cases where two or more amino acid sequences differ from each other by conservative substitutions, the percentage identity or degree of sequence similarity can be adjusted upward to correct for the conservative nature of the substitution. Means for making this adjustment are well known to those skilled in the art. See, for example, Pearson (1994) Methods Mol. Biol. 24: 307-331. Examples of groups of amino acids that have side chains with similar chemical properties include: (1) aliphatic side chains: glycine, alanine, valine, leucine and isoleucine; (2) aliphatic hydroxyl side chains: serine and threonine; (3) amide-containing side chains: asparagine and glutamine; (4) aromatic side chains: phenylalanine, tyrosine and tryptophan; (5) main side chains: lysine, arginine and histidine; (6) acidic side chains: aspartate and glutamate, and (7) sulfur-containing side chains: cysteine and methionine. Preferred groups of conservative amino acid substitutions are: valine-leucine-isoleucine, phenylalanine-tyrosine, lysine-arginine, alanine-valine, glutamate-aspartate and asparagine-glutamine. Alternatively, a conservative replacement is any change that has a positive value in the PAM250 log-likelihood matrix disclosed in Gonnet et al. (1992) Science 256: 1443-1445. A moderately conservative replacement is any change that has a non-negative value in the PAM250 log-likelihood matrix.

Sequence similarity for polypeptides, also called sequence identity, is typically measured using sequence analysis software. Protein analysis software finds similar sequences by applying similarity measures assigned to various substitutions, deletions and other modifications, including conserved amino acid substitutions. For example, GCG software contains programs such as Gap and Bestfit that can be used with default parameters to determine homology or sequence identity between closely related polypeptides, such as homologous polypeptides from different species of organisms, or between a wild-type protein and its mutein. See, for example, GCG version 6.1. Polypeptide sequences can also be compared using FASTA with default or recommended parameters - programs in GCG version 6.1. FASTA (eg, FASTA2 and FASTA3) provides alignments and percent sequence identity of the regions of best overlap between a given and a found sequence (see Pearson (2000) above). Another preferred algorithm when comparing a sequence of the invention to a database containing a large number of sequences from different organisms is the BLAST computer program, especially BLASTP or TBLASTN, using default parameters. See, for example, Altschul et al. (1990) J. Mol. Biol. 215:403-410 and Altschul et al. (1997) Nucleic Acids Res. 25:3389-402.

Bispecific antigen-binding molecules.

Antibodies according to the present invention can be monospecific, bispecific

- 13 045943 or multispecific. Multispecific antibodies may be specific for different epitopes of a single target polypeptide or may contain antigen binding domains specific for more than one target polypeptide. See, for example, Tutt et al., 1991, J. Immunol. 147:60-69; Kufer et al., 2004, Trends Biotechnol. 22:238-244. Antibodies to CD28 according to the present invention can be associated with or co-expressed with another functional molecule, for example, another peptide or protein. For example, an antibody or fragment thereof may be operably linked (e.g., by chemical linkage, genetic fusion, non-covalent association, or otherwise) to one or more other molecular entities, such as another antibody or antibody fragment, to produce a bispecific or multispecific antibody with the second is binding specificity.

The use of the expression CD28 antibody herein is intended to include both monospecific CD28 antibodies and bispecific antibodies comprising a CD28 binding arm and a second target antigen binding arm. Thus, the present invention includes bispecific antibodies in which one immunoglobulin arm binds human CD28 and the other immunoglobulin arm is specific for a target antigen. The target antigen to which the other arm of the CD28 bispecific antibody binds can be any antigen expressed on or near the cell, tissue, organ, microorganism or virus against which a directed immune response is desired. The CD28-binding arm may contain any of the amino acid sequences HCVR/LCVR or CDR presented in table. 1 in this document. In certain embodiments, the CD28 binding arm binds human CD28 and induces proliferation of human T cells.

In the context of bispecific antibodies of the present invention, where one arm of the antibody binds CD28 and the other arm binds a target antigen, said target antigen may be a tumor-associated antigen such as PSMA.

In certain illustrative embodiments, the present invention includes bispecific antigen binding molecules that specifically bind CD28 and PSMA. Such molecules may be referred to herein, for example, as anti-CD28/anti-PSMA, or anti-CD28xPSMA, or CD28xPSMA, or anti-PSMA/anti-CD28, or anti-PSMAxCD28, or PSMAxCD28 bispecific molecules, or other similar terminology.

The term PSMA as used herein refers to human PSMA protein unless indicated to be from a non-human species (eg, murine PSMA, simian PSMA, etc.). The human PSMA protein has the amino acid sequence set forth in SEQ ID NO: 73 and/or has the amino acid sequence corresponding to accession number NCBINP_004467.1.

According to certain illustrative embodiments shown in FIGS. 1, bispecific antigen binding molecules (eg, bispecific antibodies) may have an effector arm and a targeting arm. The effector arm may be the first antigen-binding domain (eg, anti-CD28 antibody) that binds to antigens on effector cells (eg, T cells). The targeting arm may be a second antigen-binding domain (eg, an anti-PSMA antibody) that binds to antigens on target cells (eg, tumor cells). In certain illustrative embodiments, the effector arm binds to CD28 and the targeting arm binds to PSMA. The CD28/PSMA bispecific molecule can transmit a co-stimulatory signal to effector cells (eg, T cells). The effector arm is unable to stimulate T cells in the absence of clustering. When clustered, the effector arm itself does not significantly stimulate T cells. In combination with the targeting arm, the effector arm stimulates T cells. The tumor-targeting arm may have imperfect tumor specificity. The antigen that is the target of the targeting arm (eg, PSMA) may be expressed on a portion of the tumor cells. The specificity of the tumor-targeting arm can be increased by bridging in combination with bispecific CD3 antigen-binding molecules (eg, anti-CD3/PSMA bispecific antibody).

As used herein, the expression antigen-binding molecule means a protein, polypeptide, or molecular complex containing or consisting of at least one complementarity determining region (CDR), which alone or in combination with one or more additional CDRs and/or framework regions (FRs) binds specifically to a specific antigen. In certain embodiments, the antigen binding molecule is an antibody or antibody fragment as those terms are defined elsewhere herein.

As used herein, the expression bispecific antigen binding molecule means a protein, polypeptide, or molecular complex comprising at least a first antigen binding domain and a second antigen binding domain. Each antigen binding domain in a bispecific antigen binding molecule contains at least one CDR that, alone or in combination with one or more additional CDRs and/or FRs, specifically binds

- 14 045943 is associated with a specific antigen. In the context of the present invention, the first antigen binding domain specifically binds a first antigen (eg, CD28), and the second antigen binding domain specifically binds a second, different antigen (eg, PSMA).

In certain illustrative embodiments of the present invention, the bispecific antigen binding molecule is a bispecific antibody. Each antigen binding domain of a bispecific antibody contains a heavy chain variable domain (HCVR) and a light chain variable domain (LCVR). In the context of a bispecific antigen binding molecule comprising first and second antigen binding domains (eg, a bispecific antibody), the CDRs of the first antigen binding domain may be prefixed D1 and the CDRs of the second antigen binding domain may be prefixed D2. Thus, the CDRs of the first antigen binding domain may be referred to herein as D1-HCDR1, D1-HCDR2 and D1-HCDR3; a The CDRs of the second antigen binding domain may be referred to herein as D2-HCDR1, D2HCDR2 and D2-HCDR3.

The first antigen binding domain and the second antigen binding domain can be directly or indirectly linked to each other to form a bispecific antigen binding molecule according to the present invention. Alternatively, each of the first antigen binding domain and the second antigen binding domain may be coupled to a separate multimerization domain. The association of one multimerizing domain with another multimerizing domain facilitates the association between the two antigen-binding domains, thereby allowing the formation of a bispecific antigen-binding molecule. As used herein, the term multimerization domain means any macromolecule, protein, polypeptide, peptide or amino acid having the ability to associate with a second multimerization domain of the same or similar structure or composition. For example, the multimerizing domain may be a polypeptide containing an immunoglobulin CH3 domain. A non-limiting example of a multimerizing component is the Fc portion of an immunoglobulin (containing the CH2-CH3 domain), for example, the Fc domain of an IgG selected from the isotypes IgG1, IgG2, IgG3 and IgG4, as well as any allotype within the group of each isotype.

The bispecific antigen binding molecules of the present invention typically comprise two multimerizing domains, for example two Fc domains, each of which is individually part of a distinct antibody heavy chain. The first and second multimerizing domains may be of the same IgG isotype, for example IgG1/IgG1, IgG2/IgG2, IgG4/IgG4. Alternatively, the first and second multimerizing domains may be of different IgG isotypes, for example IgG1/IgG2, IgG1/IgG4, IgG2/IgG4, etc.

In certain embodiments, the multimerization domain is an Fc fragment or amino acid sequence ranging from 1 to about 200 amino acids in length containing at least one cysteine residue. In other embodiments, the multimerization domain is a cysteine residue or a short cysteine-containing peptide. Other multimerizing domains include peptides or polypeptides containing or consisting of a leucine zipper, a helix-loop motif, or a supercoil motif.

Any bispecific antibody format or technology can be used to produce the bispecific antigen binding molecules of the present invention. For example, an antibody or fragment thereof having a first antigen-binding specificity may be operably linked (e.g., by chemical linkage, genetic fusion, non-covalent association, or otherwise) to one or more other molecular entities, such as another antibody or antibody fragment, having a second antigen-binding specificity, in order to obtain a bispecific antigen-binding molecule. Specific exemplary bispecific formats that may be used in the context of the present invention include, but are not limited to, scFv- or diabody-based bispecific formats, IgG-scFv fusions, double variable domain (OVO)-Ig, quadhybridoma, peak-to-trench, common light chain (e.g. common light chain with ridge-to-hole interactions, etc.), CrossMab, CrossFab, (SEEO)body, leucine zipper, Duobody, IgG1/IgG2, dual-acting Fab (OAF)-IgG, and bispecific Mab ² formats (for a review of the above formats, see, for example, Klein et a/. 2012, mAbs 4:6, 111, and references cited therein).

In the context of bispecific antigen binding molecules of the present invention, multimerizing domains, eg Fc domains, may contain one or more amino acid changes (eg, insertions, deletions or substitutions) compared to the naturally occurring wild type version of the Fc domain. For example, the invention includes bispecific antigen binding molecules containing one or more modifications in the Fc domain, resulting in a modified Fc domain having a modified binding interaction (eg, increased or decreased) between Fc and FcRn. In one embodiment, the bispecific antigen binding molecule contains a modification at the CH2 or CH3 region, which modification increases the affinity of the Fc domain for FcRn in an acidic environment (eg, in an endosome where the pH is from about 5.5 to about 6.0). Non-limiting examples of such Fc modifications include, at

- 15 045943 example, modification in position 250 (for example, E or Q); 250 and 428 (for example, L or F); 252 (for example, LN/FIW or T), 254 (for example, S or T) and 256 (for example, S/R/Q/EID or T); or modification at position 428 and/or 433 (eg, UR/S/P/Q or K) and/or 434 (eg, H/F or V); or modification at position 250 and/or 428; or modification at position 307 or 308 (eg, 308F, V308F) and 434. In one embodiment, the modification includes modification 428L (eg, M428L) and 434S (eg, N434S); modification 428L, 2591 (for example, V2591) and 308F (for example, V308F); modification 433K (for example, H433K) and 434 (for example, 434Y); modification 252, 254 and 256 (for example, 252Y, 254T and 256E); modification 250Q and 428L (for example, T250Q and M428L); and modification 307 and/or 308 (for example, 308F or 308P).

The present invention also includes bispecific antigen binding molecules comprising a first CH3 domain and a second CH3 Ig domain, wherein the first and second CH3 Ig domains differ from each other by at least one amino acid, and wherein the difference in at least one amino acid reduces binding of the bispecific antibody to the protein And compared to a bispecific antibody that has no difference in amino acid. In one embodiment, the first CH3 Ig domain binds Protein A, and the second CH3 Ig domain contains a mutation that reduces or eliminates Protein A binding, such as the H95R modification (according to IMGT exon numbering; H435R according to EU numbering). The second CH3 may additionally contain the Y96F modification (according to IMGT; according to Y436F EU). Other modifications that may be contained in the second CH3 include: D16E, L 18M, N44S, K52N, V57M and V821 (according to IMGT; D356E, L358M, N384S, K392N, V397M and V4221 according to EU) in the case of IgG1 antibodies; N44S, K52N and V821 (IMGT; EU N384S, K392N and V4221) for IgG2 antibodies; and Q15R, N44S, K52N, V57M, R69K, E79Q and V821 (according to IMGT; Q355R, N384S, K392N, V397M, R409K, E419Q and V4221 according to EU) for IgG4 antibodies.

In certain embodiments, the Fc domain may be chimeric and comprise a combination of Fc sequences derived from more than one immunoglobulin isotype. For example, a chimeric Fc domain may comprise part or all of a CH2 sequence derived from the CH2 region of a G1 human Ig, G2 human Ig, or G4 human Ig, and part or all of a CH3 sequence derived from a G1 human Ig, G2 human Ig, or G4 human Ig. . The chimeric Fc domain may also comprise a chimeric hinge region. For example, a chimeric hinge may comprise an upper hinge sequence derived from a human G1 Ig, a G2 human Ig, or a G4 human Ig hinge region in combination with a lower hinge sequence derived from a G1 human Ig, a G2 human Ig, or a G4 human Ig. A specific example of a chimeric Fc domain that can be included in any of the antigen binding molecules provided herein contains, from N- to C-terminus: [IgG4 CH1] - [IgG4 upper hinge] - [IgG2 lower hinge] - [ IgG4 CH2] - [IgG4 CH3]. Another example of a chimeric Fc domain that can be included in any of the antigen binding molecules provided herein contains, from N- to C-terminus: [IgG1 CH1] [IgG1 upper hinge] - [IgG2 lower hinge] - [ IgG4 CH2] - [IgG1 CH3]. These and other examples of chimeric Fc domains that can be included in any of the antigen binding molecules of the present invention are described in WO 2014/022540 A1. Chimeric Fc domains having this general structure and variants thereof may exhibit altered binding to the Fc receptor, which in turn affects Fc effector function.

Sequence options.

The antibodies and bispecific antigen binding molecules of the present invention may contain one or more amino acid substitutions, insertions and/or deletions in the framework regions and/or CDR regions of the heavy and light chain variable domains as compared to the corresponding germline sequences from which the individual antigen binding domains were derived. . Such mutations can be readily identified by comparing the amino acid sequences disclosed herein with germline sequences available, for example, from publicly available antibody sequence databases. The antigen binding molecules of the present invention may comprise antigen binding fragments derived from any of the exemplary amino acid sequences disclosed herein, in which one or more amino acids in one or more framework regions and/or CDR regions are mutated to the corresponding germline sequence residue(s) , from which the antibody was derived, or to the corresponding residue(s) of another human germline sequence, or to a conserved amino acid substitution of the corresponding germline residue(s) (such sequence changes are collectively referred to herein as germline mutations). A person skilled in the art, starting with the heavy and light chain variable region sequences disclosed herein, can readily produce a variety of antibodies and antigen binding fragments containing one or more individual germline mutations or combinations thereof. In certain embodiments, all framework residues and/or CDR residues in the VH and/or _VL domains are mutated back to residues contained in the original germline sequence from which the antigen binding domain was originally derived. In other embodiments, only some residues are mutated back

- 16 045943 into the original germline sequence, for example, only the mutated residues contained in the first 8 amino acids of FR1 or the last 8 amino acids of FR4, or only the mutated residues contained in CDR1, CDR2 or CDR3. In other embodiments, one or more framework residues and/or CDR residues are mutated to the corresponding residue(s) of a different germline sequence (ie, a germline sequence that is different from the germline sequence from which the antigen binding domain was originally derived). In addition, antigen binding domains may contain any combination of two or more germline mutations within framework and/or CDR regions, for example, in which some individual residues are mutated to the corresponding residue of a particular germline sequence, while some other residues that differ from original germline sequence, retained or mutated into the corresponding residue of another germline sequence.

Once prepared, antigen binding domains containing one or more germline mutations can be readily screened for one or more desired properties, such as improved binding specificity, increased binding affinity, improved or enhanced antagonistic or agonistic biological properties (as appropriate), reduced immunogenicity, etc. Bispecific antigen binding molecules containing one or more antigen binding domains prepared according to this general principle are covered by the present invention.

The present invention also includes antigen binding molecules wherein one or both antigen binding domains comprise variants of any of the HCVR, LCVR and/or CDR amino acid sequences disclosed herein having one or more conservative substitutions. For example, the present invention includes antigen binding molecules comprising an antigen binding domain having the amino acid sequences HCVR, LCVR and/or CDRs with, for example, 10 or less, 8 or less, 6 or less, 4 or less, etc. conservative amino acid substitutions relative to any of the HCVR, LCVR and/or CDR amino acid sequences disclosed herein. A conservative amino acid substitution is a substitution in which an amino acid residue is replaced by another amino acid residue having a side chain (R group) with similar chemical properties (eg, charge or hydrophobicity). In general, a conservative amino acid substitution does not substantially change the functional properties of the protein. Examples of groups of amino acids that have side chains with similar chemical properties include: (1) aliphatic side chains: glycine, alanine, valine, leucine and isoleucine; (2) aliphatic hydroxyl side chains: serine and threonine; (3) amide-containing side chains: asparagine and glutamine; (4) aromatic side chains: phenylalanine, tyrosine and tryptophan; (5) main side chains: lysine, arginine and histidine; (6) acidic side chains: aspartate and glutamate, and (7) sulfur-containing side chains: cysteine and methionine. Preferred groups of conservative amino acid substitutions are: valine-leucine-isoleucine, phenylalanine-tyrosine, lysine-arginine, alanine-valine, glutamate-aspartate and asparagine-glutamine. Alternatively, a conservative replacement is any change that has a positive value in the PAM250 log-likelihood matrix disclosed in Gonnet et al. (1992) Science 256: 1443-1445. A moderately conservative replacement is any change that has a non-negative value in the PAM250 log-likelihood matrix.

The present invention also includes antigen binding molecules comprising an antigen binding domain having an HCVR, LCVR and/or CDR amino acid sequence substantially identical to any of the HCVR, LCVR and/or CDR amino acid sequences disclosed herein. The term substantially identical or substantially identical, when applied to an amino acid sequence, means that two amino acid sequences when optimally aligned, for example using the GAP or BESTFIT programs using default gap opening penalties, have at least 95% sequence identity, even more preferably at least 98% or 99% sequence identity. Preferably, positions of residues that are not identical are distinguished by conservative amino acid substitutions. In cases where two or more amino acid sequences differ from each other by conservative substitutions, the percentage identity or degree of sequence similarity may be adjusted upward to correct for the conservative nature of the substitution. Means for making this adjustment are well known to those skilled in the art. See, for example, Pearson (1994) Methods Mol. Biol. 24: 307-331.

Sequence similarity for polypeptides, also called sequence identity, is typically measured using sequence analysis software. Protein analysis software finds similar sequences by applying similarity measures assigned to various substitutions, deletions and other modifications, including conserved amino acid substitutions. For example, the GCG software contains programs such as Gap and Bestfit, which can be used with default settings to define

- 17 045943 division of homology or sequence identity between closely related polypeptides, such as homologous polypeptides from different species of organisms, or between a wild-type protein and its mutein. See, for example, GCG version 6.1. Polypeptide sequences can also be compared using FASTA with default or recommended parameters - programs in GCG version 6.1. FASTA (eg, FASTA2 and FASTA3) provides alignments and percent sequence identity of the regions of best overlap between a given and a found sequence (see Pearson (2000) above). Another preferred algorithm when comparing a sequence of the invention to a database containing a large number of sequences from different organisms is the BLAST computer program, especially BLASTP or TBLASTN, using default parameters. See, for example, Altschul et al. (1990) J. Mol. Biol. 215:403-410 and Altschul et al. (1997) Nucleic Acids Res. 25:3389-402.

pH-dependent binding.

The present invention includes anti-CD28/anti-PSMA bispecific antigen-binding molecules with pH-dependent binding characteristics. For example, an anti-CD28 antibody of the present invention may exhibit reduced binding to CD28 at acidic pH compared to neutral pH. Alternatively, the anti-PSMA antibodies of the invention may exhibit increased binding to PSMA at acidic pH compared to neutral pH. The expression acidic pH includes pH values less than about 6.2, such as about 6.0, 5.95, 5.9, 5.85, 5.8, 5.75, 5.7, 5.65, 5.6 , 5.55, 5.5, 5.45, 5.4, 5.35, 5.3, 5.25, 5.2, 5.15, 5.1, 5.05, 5.0 or less . As used herein, the expression neutral pH means a pH of from about 7.0 to about 7.4. The expression neutral pH includes pH values of approximately 7.0, 7.05, 7.1, 7.15, 7.2, 7.25, 7.3, 7.35 and 7.4.

In some cases, reduced binding ... at acidic pH compared to neutral pH is expressed by the ratio of the KD value of binding of an antibody to its antigen at acidic pH to the KD value of binding of an antibody to its antigen at neutral pH (or vice versa). For example, an antibody or antigen binding fragment thereof may be considered to exhibit reduced binding to CD28 at acidic pH compared to neutral pH in the context of the present invention if said antibody or antigen binding fragment thereof exhibits an acidic/neutral KD ratio of approximately 3.0 or greater. In certain illustrative embodiments, the acidic/neutral KD ratio for an antibody or antigen binding fragment of the present invention may be about 3.0, 3.5, 4.0, 4.5, 5.0, 5.5, 6.0, 6 ,5, 7.0, 7.5, 8.0, 8.5, 9.0, 9.5, 10.0, 10.5, 11.0, 11.5, 12.0, 12.5 , 13.0, 13.5, 14.0, 14.5, 15.0, 20.0, 25.0, 30.0, 40.0, 50.0, 60.0, 70.0, 100 ,0 or more.

Antibodies with pH-dependent binding characteristics can be obtained, for example, by screening a population of antibodies for decreased (or increased) binding to a particular antigen at an acidic pH compared to a neutral pH. In addition, modifications of the antigen-binding domain at the amino acid level can produce antibodies with pH-dependent characteristics. For example, by replacing one or more amino acids of the antigen binding domain (eg, in the CDR) with a histidine residue, an antibody with reduced antigen binding at acidic pH compared to neutral pH can be generated.

Antibodies containing Fc variants.

Certain embodiments of the present invention provide bispecific anti-CD28/anti-PSMA antigen binding molecules comprising an Fc domain containing one or more mutations that increase or decrease antibody binding to the FcRn receptor, for example, at an acidic pH compared to a neutral pH. For example, the present invention includes antibodies and antigen-binding molecules containing a mutation in the C _H 2 or C _H 3 region of the Fc domain, where the mutation) increases the affinity of the Fc domain for FcRn in an acidic environment (for example, in the endosome, where the pH is from about 5 .5 to approximately 6.0). Such mutations may result in an increase in the serum half-life of the antibody when administered to an animal. Non-limiting examples of such Fc modifications include, for example, modification at position 250 (eg, E or Q); 250 and 428 (for example, L or F); 252 (for example, L/Y/F/W or T), 254 (for example, S or T) and 256 (for example, S/R/Q/E/D or T); or modification at position 428 and/or 433 (eg, H/L/R/S/P/Q or K) and/or 434 (eg, H/F or Y); or modification at position 250 and/or 428; or modification at position 307 or 308 (eg, 308F, V308F) and 434. In one embodiment, the modification includes modification 428L (eg, M428L) and 434S (eg, N434S); modification 428L, 259I (for example, V259I) and 308F (for example, V308F); modification 433K (for example, H433K) and 434 (for example, 434Y); modification 252, 254 and 256 (for example, 252Y, 254T and 256E); modification 250Q and 428L (for example, T250Q and M428L); and modification 307 and/or 308 (for example, 308F or 308P).

For example, the present invention includes anti-CD28/anti-PSMA bispecific antigen-binding molecules comprising an Fc domain containing one or more pairs or groups of mutations selected from the group consisting of: 250Q and 248L (eg, T250Q and M248L); 252Y, 254T and 256E (for example, M252Y, S254T and T256E); 428L and 434S (for example, M428L and N434S); and 433K and 434F (for example, H433K and

- 18 045943

N434F). All possible combinations of the above Fc domain mutations and other mutations in the variable domains of the antibodies disclosed herein are within the scope of the present invention.

Biological characteristics of antibodies and antigen-binding molecules.

The present invention includes antibodies and antigen binding fragments thereof that bind human CD28 and/or PSMA with high affinity. The present invention also includes antibodies and antigen-binding fragments thereof that bind human CD28 and/or PSMA with moderate or low affinity, depending on the therapeutic context and the specific targeting properties desired. For example, in the context of a bispecific antigen-binding molecule in which one arm binds CD28 and the other arm binds a target antigen (e.g., PSMA), it may be desirable that the arm that binds the target antigen binds it with high affinity and the arm binds CD28 bound CD28 with only moderate or low affinity. In this way, selective targeting of the antigen-binding molecule to cells expressing the target antigen can be achieved while preventing indiscriminate/off-target binding of CD28 and associated associated side effects.

In certain embodiments, the present invention includes antibodies and antigen-binding antibody fragments that bind human CD28 (e.g., at 25°C) with a KD of less than about 210 nM as measured by surface plasmon resonance, for example, using the assay format of Example 3 in this document. In certain embodiments, the antibodies or antigen binding fragments of the present invention bind CD28 with a KD of less than about 150 nM, less than about 130 nM, less than about 120 nM, less than about 100 nM, less than about 50 nM, less than about 80 nM , less than about 60 nM, less than about 40 nM, or less than about 30 nM when measured by surface plasmon resonance, for example, using the assay format of Example 3 herein or a substantially similar assay. In certain embodiments, the antibodies or antigen binding fragments of the present invention bind CD28 with a KD of from about 30 nM to about 207 nM.

The present invention also includes antibodies and antigen-binding fragments thereof that bind CD28 with a dissociative half-life (t1/2) greater than approximately 3.5 minutes when measured by surface plasmon resonance at 25 or 37°C, for example, using the assay format according to Example 3 herein or a substantially similar analysis. In certain embodiments, the antibodies or antigen binding fragments of the present invention bind CD28 with a t1/2 of more than about 5 minutes, more than about 5 minutes, more than about 20 minutes, more than about 30 minutes, more than about 5 minutes, more than about 50 minutes , more than about 60 minutes, more than about minutes, more than about 80 minutes, more than about 90 minutes, more than about

100 minutes, more than about 200 minutes, more than about 300 minutes, more than about 400 minutes, more than about 500 minutes, more than about 600 minutes, more than about 700 minutes, more than about 800 minutes, more than about 900 minutes , more than about 1000 minutes, or more than about 1200 minutes when measured by surface plasmon resonance at 25 or 37°C, for example, using the assay format of Example 3 herein or a substantially similar assay.

The present invention includes bispecific antigen binding molecules (eg, bispecific antibodies) capable of simultaneously binding to human CD28 and human PSMA. In certain embodiments, the bispecific antigen binding molecules of the invention specifically interact with cells expressing CD28 and/or PSMA. The extent to which a bispecific antigen-binding molecule binds cells expressing CD28 and/or PSMA can be assessed by fluorescence-activated cell sorting (FACS), as illustrated in Example 4 herein. For example, the present invention includes bispecific antigen binding molecules that specifically bind human cell lines expressing CD28 but not PSMA (e.g., HEK293 genetically modified to express CD28) and human prostate carcinoma cell lines expressing PSMA but not CD28 (e.g. C4-2). The present invention includes bispecific antigen binding molecules that bind any of the above cells and cell lines with an EC ₅₀ value of from about 9.6 x 10 ^-9 to about 3.5 x 10 ^-10 or less as determined using the FACS assay of Example 4 or a substantially similar assay.

The present invention also provides anti-CD28/anti-PSMA bispecific antigen-binding molecules that induce or enhance T cell-mediated killing of tumor cells. For example, the present invention includes anti-CD28xPSMA antibodies that induce or enhance T cell-mediated tumor cell killing with an EC50 of less than about 78 pM as measured in a T cell-mediated tumor killing assay.

- 19 045943 left cells in vitro, for example, using the assay format according to example 6 herein (for example, assessing the degree of killing of human C4-2 tumor cells by PBMCs in the presence of antibodies to CD28xPSMA) or a substantially similar assay. In certain embodiments, the antibodies or antigen-binding fragments of the present invention induce T cell-mediated killing of tumor cells (e.g., PBMC-mediated killing of C4-2 cells) with an EC50 value of less than about 40 pM, less than about 20 pM, less than about 16 pM , less than about 10 pM, less than about 5.0 pM, less than about 4.0 pM, less than about 3.0 pM, less than about 2.5 pM, less than about 2.0 pM, less than about 1.5 pM or less than about 1.45 pM when measured in an in vitro T cell-mediated tumor cell killing assay, for example, using the assay format of Example 6 herein or a substantially similar assay.

The present invention also includes anti-CD28/anti-PSMA bispecific antigen-binding molecules that bind to human CD28-expressing T cells with an EC ₅₀ value of 1.0 pM to 1000 nM. In certain embodiments, the anti-CD28/anti-PSMA bispecific antigen-binding molecules bind to human CD28-expressing T cells with an EC ₅₀ value of 48 nM to 180 nM. For example, the present invention includes anti-CD28/anti-PSMA bispecific antigen binding molecules that bind to human CD28 expressing T cells with an EC ₅₀ value of about 1 pM, about 10 pM, about 100 pM, about 500 pM, about 1 nM, about 2 nM, approximately 5 nM, approximately 10 nM, approximately 20 nM, approximately 30 nM, approximately 40 nM, approximately 50 nM, approximately 60 nM, approximately 70 nM, approximately 80 nM, approximately 90 nM, approximately 100 nM, approximately 200 nM , about 300 nM, about 500 nM, about 800 nM, about 1000 nM or more.

The present invention also includes anti-CD28/anti-PSMA bispecific antigen-binding molecules exhibiting one or more characteristics selected from the group consisting of: (a) inducing T cell proliferation in vitro (see, for example, Example 8 herein); (b) activating T cells, inducing up-regulation of CD25 and PD-1 in human PBMCs (see, for example, Example 8 herein); (c) enhancing human T cell-mediated cytotoxicity against cell lines expressing PSMA (see, for example, Example 8 herein); (d) inducing naïve primate T cell-mediated cytotoxicity against cell lines expressing PSMA (see, for example, Example 8 herein); (e) depleting tumor cells in mice (eg, Example 10 herein); (f) enhancing tumor clearance in mice (eg, Example 10 herein); (g) lack of induction of a cytokine storm (eg, Example 10 herein); (h) lack of induction of systemic T cell effects in cynomolgus monkeys (eg, Example 11 herein); (i) enhancing the effect of PD-1 blockade on T cell activation-induced tumor cell killing (eg, Example 13 herein); (j) enhancing the expansion of memory T cells (eg, Example 13 herein).

The present invention includes anti-CD28/anti-PSMA bispecific antigen-binding molecules capable of reducing the pool of tumor cells in a subject (see, for example, Example 8). For example, certain embodiments provide anti-CD28/anti-PSMA bispecific antigen binding molecules, wherein a single administration of the bispecific antigen binding molecule to a subject (e.g., at a dose of about 0.1 mg/kg, about 0.08 mg/kg, about 0.06 mg/kg about 0.04 mg/kg, about 0.04 mg/kg, about 0.02 mg/kg, about 0.01 mg/kg or less) causes a decrease in the number of tumor cells in a subject.

Epitope mapping and related technologies.

The epitope on CD28 or PSMA to which the antigen-binding molecules of the present invention bind may consist of one contiguous sequence of 3 or more (e.g., 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 or more) amino acids of the CD28 protein or PSMA protein. Alternatively, the epitope may consist of multiple non-contiguous amino acids (or amino acid sequences) of CD28 or PSMA. Antibodies of the invention may react with amino acids contained in a CD28 monomer, or may react with amino acids on two different CD28 chains of a CD28 dimer. The term epitope as used herein refers to an antigenic determinant that interacts with a specific antigen-binding site in the variable region of an antibody molecule, known as a paratope. One antigen can have more than one epitope. Thus, different antibodies may bind to different regions on the antigen and may have different biological effects. Epitopes can be conformational or linear. A conformational epitope is formed by spatially close amino acids from different segments of a linear polypeptide chain. A linear epitope is an epitope formed by adjacent amino acid residues in a polypeptide chain. In certain circumstances, the epitope may include moieties of saccharides, phosphoryl groups or sulfonyl groups

- 20 045943 groups on the antigen.

Various techniques known to those skilled in the art can be used to determine whether the antigen binding domain of an antibody interacts with one or more amino acids in a polypeptide or protein. Exemplary methods that can be used to determine the epitope or binding domain of a particular antibody or antigen binding domain include, for example, a standard cross-competition assay such as that described in Antibodies, Harlow and Lane (Cold Spring Harbor Press, Cold Spring Harb., NY) , spot mutagenesis (e.g., alanine scan mutagenesis, arginine scan mutagenesis, etc.), peptide blot assay (Reineke, 2004, Methods Mol Biol 248:443-463), protease-protected assay, and peptide digestion assay . In addition, techniques such as epitope excision, epitope extraction, and chemical modification of antigens can be used (Tomer, 2000, Protein Science 9:487-496). Another method that can be used to identify the amino acids in the polypeptide with which the antibody interacts is hydrogen-deuterium exchange, determined using mass spectrometry. In general terms, the hydrogen-deuterium exchange method involves labeling a protein of interest with deuterium, followed by binding of an antibody to the deuterium-labeled protein. The protein/antibody complex is then transferred to water to allow hydrogen-deuterium exchange to occur on all residues except the antibody-protected residues (which remain deuterium-labeled). Once the antibody is dissociated, the target protein is subjected to protease digestion and mass spectrometric analysis, thereby identifying deuterium-labeled residues that correspond to the specific amino acids with which the antibody interacts. See, for example, Ehring (1999) Analytical Biochemistry 267(2):252-259; Engen and Smith (2001) Anal. Chem. 73:256A-265A. X-ray diffraction analysis of the crystal structure can also be used to identify the amino acids in the polypeptide with which the antibody reacts.

The present invention further includes anti-CD28 and anti-PSMA antibodies that bind to the same epitope as any of the specific exemplary antibodies described herein (e.g., antibodies containing any of the amino acid sequences presented in Table 1 herein) . Likewise, the present invention also includes anti-CD28 and/or anti-PSMA antibodies that compete for binding to CD28 and/or PSMA with any of the specific exemplary antibodies described herein (e.g., antibodies containing any of the amino acid sequences presented in Table 1 in this document).

The present invention also includes bispecific antigen binding molecules comprising a first antigen binding domain that specifically binds human CD28 and a second antigen binding fragment that specifically binds human PSMA, wherein said first antigen binding domain binds the same epitope on CD28 as any of the specific exemplary CD28s. -specific antigen binding domains described herein, and/or wherein said second antigen binding domain binds to the same epitope on PSMA as any of the specific exemplary PSMA-specific antigen binding domains described herein.

Likewise, the present invention also includes bispecific antigen binding molecules comprising a first antigen binding domain that specifically binds human CD28 and a second antigen binding fragment that specifically binds human PSMA, wherein said first antigen binding domain competes for binding to CD28 with any of the specific exemplary CD28-specific antigen binding domains described herein, and/or wherein said second antigen binding domain competes for binding to PSMA with any of the specific exemplary PSMA-specific antigen binding domains described herein.

It can be readily determined whether a particular antigen binding molecule (e.g., antibody) or antigen binding domain thereof binds to the same epitope as or competes with the reference antigen binding molecule of the present invention using standard methods known in the art. . For example, to determine whether a test antibody binds to the same epitope on CD28 (or PSMA) as a reference bispecific antigen binding molecule of the present invention, the reference bispecific molecule is first allowed to bind to the CD28 protein (or PSMA protein). The ability of the test antibody to bind to the CD28 (or PSMA) molecule is then assessed. If the test antibody is able to bind to CD28 (or PSMA) after saturating binding to the reference bispecific antigen binding molecule, it can be concluded that the test antibody binds to an epitope of CD28 (or PSMA) different from that bound by the reference bispecific antigen binding molecule. On the other hand, if the test antibody is unable to bind to a CD28 (or PSMA) molecule after saturating binding to a reference bispecific antigen binding molecule, then the test antibody may bind to the same CD28 (or PSMA) epitope as the epitope bound by the reference bispecific antigen binding molecule according to invention. Then maybe

- 21 045943 conducted an additional standard experiment (e.g., mutation and peptide binding analysis) to confirm whether the observed lack of binding of the test antibody is actually due to binding to the same epitope as the reference antigen binding molecule, or whether the lack of observed binding is due to steric blocking ( or other phenomenon). Experiments of this type can be performed using ELISA, RIA, Biacore, flow cytometry or any other quantitative or qualitative antibody binding assay available in the art. In certain embodiments of the present invention, two antigen binding proteins bind to the same (or overlapping) epitope if, for example, a 1-, 5-, 10-, 20-, or 100-fold excess of one antigen binding protein inhibits the binding of the other by at least 50%, but preferably 75%, 90% or even 99% when measured in a competitive binding assay (see, for example, Junghans et al., Cancer Res. 1990:50:1495-1502). Alternatively, two antigen binding proteins are considered to bind to the same epitope if substantially all amino acid mutations in the antigen that reduce or eliminate the binding of one antigen binding protein reduce or eliminate the binding of the other. Two antigen binding proteins are considered to have overlapping epitopes if only a subset of amino acid mutations that reduce or eliminate the binding of one antigen binding protein reduce or eliminate the binding of the other.

To determine whether an antibody or its antigen-binding domain competes for binding to a reference antigen-binding molecule, the binding assay described above is used in two orientations: in the first orientation, the reference antigen-binding molecule is allowed to bind to the CD28 protein (or PSMA protein) under saturating conditions, followed by binding assessment test antibody with the CD28 (or PSMA) molecule. In the second orientation, the antibody of interest is allowed to bind to the CD28 (or PSMA) molecule under saturating conditions, followed by assessment of the binding of a reference antigen binding molecule to the CD28 (or PSMA) molecule. If in both orientations only the first (saturating) antigen binding molecule is able to bind to the CD28 (or PSMA) molecule, then it is concluded that the test antibody and the reference antigen binding molecule are competing for binding to CD28 (or PSMA). As will be appreciated by one skilled in the art, an antibody that competes for binding to a reference antigen binding molecule need not bind to the same epitope as the reference antibody, but may sterically block binding of the reference antibody by binding to an overlapping or adjacent epitope.

Preparation of antigen-binding domains and construction of bispecific molecules.

Antigen-binding domains specific for specific antigens can be obtained using any antibody technology known in the art. Once prepared, two different antigen binding domains specific for two different antigens (eg, CD28 and PSMA) can be suitably positioned relative to each other to produce a bispecific antigen binding molecule of the present invention using standard methods. (A discussion of exemplary bispecific antibody formats that can be used to construct bispecific antigen binding molecules of the present invention is provided elsewhere herein.) In certain embodiments, one or more individual components (eg, heavy and light chains) of the multispecific antigen binding molecules of the invention are derived from chimeric, humanized, or fully human antibodies. Methods for producing such antibodies are well known in the art. For example, one or more heavy and/or light chains of the bispecific antigen binding molecules of the present invention can be produced using VELOCIMMUNE™ technology. Using VELOCIMMUNE™ technology (or any other human antibody technology), chimeric antibodies with high affinity for a specific antigen (eg, CD28 or PSMA), having a human variable region and a murine constant region, are initially isolated. Antibodies are characterized and selected for desired characteristics, including affinity, selectivity, epitope, etc. The mouse constant regions are replaced with the desired human constant region to create fully human heavy and/or light chains that can be incorporated into the bispecific antigen binding molecules of the present invention.

Genetically modified animals can be used to produce human bispecific antigen-binding molecules. For example, a genetically modified mouse may be used that is unable to rearrange and express an endogenous murine immunoglobulin light chain variable region sequence, wherein the mouse expresses only one or two human light chain variable domains encoded by human immunoglobulin sequences operably linked to the murine kappa constant region gene at the endogenous mouse kappa locus. Such genetically modified mice can be used to produce fully human bispecific antigen-binding molecules containing two

- 22 045943 different heavy chains associated with an identical light chain containing a variable domain derived from one of two different human light chain variable region gene segments. (For a detailed discussion of such genetically modified mice and their use to produce bispecific antigen-binding molecules, see, for example, US 2011/0195454).

Bioequivalents.

The present invention covers antigen-binding molecules having amino acid sequences that differ from those of the described antibodies, but retain the ability to bind CD28 and/or PSMA. Such variant molecules contain one or more additions, deletions or substitutions of amino acids compared to the original sequence, but exhibit biological activity that is substantially equivalent to that of the described antigen binding molecules. Likewise, DNA sequences of the present invention encoding antigen binding molecules include sequences containing one or more additions, deletions or substitutions of nucleotides compared to the disclosed sequence, but which encode an antigen binding molecule that is substantially bioequivalent to the disclosed antigen binding molecules of the invention. Examples of such variant amino acid and DNA sequences are discussed above.

The present invention includes antigen binding molecules bioequivalent to any of the exemplary antigen binding molecules presented herein. Two antigen-binding proteins or antibodies are considered bioequivalent if, for example, they are pharmaceutical equivalents or pharmaceutical alternatives whose rate and extent of absorption do not show a significant difference when administered at the same molar dose under similar experimental conditions, either in a single dose or in multiple doses. Some antibodies will be considered equivalents or pharmaceutical alternatives if they are equivalent in their extent of absorption, but not in their rate of absorption, and may nevertheless be considered bioequivalent because such differences in rate of absorption are intentional and reflected in the labeling and are not significant. values for achieving effective concentrations of the drug in the body, for example, during chronic use, and are considered medically insignificant for the particular drug being studied.

In one embodiment, two antigen binding proteins are bioequivalent if there are no clinically significant differences in their safety, purity, and potency.

In one embodiment, two antigen binding proteins are bioequivalent if a patient can be switched one or more times from the reference product to the biologic product and vice versa without an expected increase in the risk of adverse effects, including a clinically significant change in immunogenicity or decrease in efficacy, compared to continuous therapy without such translation.

In one embodiment, two antigen binding proteins are bioequivalent if they both act through the same mechanism or mechanisms of action for the condition or conditions of use, to the extent such mechanisms are known.

Bioequivalence can be demonstrated using in vivo and in vitro methods. Options for assessing bioequivalence include, for example, (a) an in vivo test in humans or other mammals that measures the concentration dynamics of the antibody or its metabolites in blood, plasma, serum, or other biological fluid; (b) an in vitro test that has been correlated with and is reasonably predictive of in vivo bioavailability data in humans; (c) an in vivo test in humans or other mammals that measures the time course of the relevant pharmacological action of a single dose of an antibody (or its target); and (d) in a well-controlled clinical study that establishes the safety, efficacy or bioavailability or bioequivalence of the antibody.

Bioequivalent variants of the exemplary antigen-binding molecules provided herein can be constructed, for example, by various substitutions of residues or sequences or the removal of terminal or internal residues or sequences that are not necessary to exhibit biological activity. For example, cysteine residues that are not essential for biological activity can be deleted or replaced with other amino acids to prevent the formation of unnecessary or irregular intramolecular disulfide bridges upon renaturation. In other contexts, bioequivalent antibodies may include exemplary bispecific antigen-binding molecules provided herein containing amino acid substitutions that modify the glycosylation characteristics of the antibodies, for example, mutations that eliminate or eliminate glycosylation.

Species selectivity and cross-reactivity between species.

The present invention, according to certain embodiments, relates to antigen-binding molecules that bind to human CD28, but not to CD28 of other species. The present invention also provides antigen-binding molecules that bind to human PSMA, but not to PSMA of other species. The present invention also includes antigen binding molecules,

- 23 045943 which bind to human CD28 and CD28 of one or more non-human species; and/or antigen-binding molecules that bind to human PSMA and to PSMA of one or more non-human species.

Certain illustrative embodiments of the invention provide antigen binding molecules that bind to human CD28 and/or human PSMA, and may or may not bind, as appropriate, to one or more of mouse, rat, guinea pig CD28 and/or PSMA , hamster, gerbil, pig, cat, dog, rabbit, goat, sheep, cow, horse, camel, cynomolgus monkey, monkey, rhesus monkey or chimpanzee. For example, in a particular illustrative embodiment of the present invention, bispecific antigen binding molecules are provided comprising a first antigen binding domain that binds human CD28 and cynomolgus CD28, and a second antigen binding domain that specifically binds human PSMA.

Immunoconjugates.

The present invention encompasses antigen-binding molecules conjugated to a therapeutic moiety (immunoconjugate), such as a cytotoxin, chemotherapeutic drug, immunosuppressant, or radioisotope. Cytotoxic agents include any agent that destroys cells. Examples of suitable cytotoxic agents and chemotherapeutic agents for the formation of immunoconjugates are known in the art (see, for example, WO 05/103081).

Therapeutic drug and administration.

The present invention relates to pharmaceutical compositions containing antigen binding molecules according to the present invention. The pharmaceutical compositions of the invention are formulated with suitable carriers, excipients and other agents that provide improved transport, delivery, tolerability and the like. Many suitable drugs can be found in a reference book known to all pharmaceutical chemists: Remington's Pharmaceutical Sciences, Mack Publishing Company, Easton, PA. These preparations include, for example, powders, pastes, ointments, jellies, waxes, oils, lipids, lipid-containing (cationic or anionic) vesicles (such as LIPOFECTIN™, Life Technologies, Carlsbad, California), DNA conjugates, anhydrous absorbent pastes, emulsions such as oil in water and water in oil, emulsions based on carbowax (polyethylene glycols with different molecular weights), semi-solid gels and semi-solid mixtures containing carbowax. See also Powell et al. Compendium of excipients for parenteral formulations PDA (1998) J Pharm Sci Technol 52:238–311.

The dose of antigen binding molecule administered to a patient may vary depending on the age and size of the patient, the disease being treated, conditions, route of administration, and the like. The preferred dose is usually calculated based on body weight or body surface area. When using the bispecific antigen binding molecule of the present invention for therapeutic purposes in an adult patient, it may be preferred to administer the bispecific antigen binding molecule of the present invention intravenously, typically in a single dose of from about 0.01 to about 20 mg/kg body weight, more preferably from about 0.02 to about 7, about 0.03 to about 5, or about 0.05 to about 3 mg/kg body weight. Depending on the severity of the condition, the frequency and duration of treatment may be adjusted. Effective dosages and schedules of administration of a bispecific antigen-binding molecule can be determined experimentally; for example, the patient's disease can be monitored by periodic examination and the dose adjusted accordingly. In addition, cross-species dosage scaling can be accomplished using methods well known in the art (eg, Mordenti et al., 1991, Pharmaceut. Res. 8:1351).

Various delivery systems are known and can be used for the administration of the pharmaceutical composition according to the invention, for example, encapsulation in liposomes, microparticles, microcapsules, recombinant cells capable of expressing mutant viruses, receptor-mediated endocytosis (see, for example, Wu et al., 1987, J. Biol Chem 262:4429–4432). Routes of administration include, but are not limited to, intradermal, intramuscular, intraperitoneal, intravenous, subcutaneous, intranasal, epidural, and oral routes.

The composition can be administered by any convenient route, for example, by infusion or bolus injection, by absorption through epithelial or mucocutaneous integuments (for example, oral mucosa, rectal and intestinal mucosa, etc.) and can be administered together with other biologically active agents. Administration may be systemic or local.

The pharmaceutical composition of the present invention can be delivered subcutaneously or intravenously using a standard needle and syringe. Moreover, in the case of subcutaneous delivery, a pen syringe is very suitable for delivering the pharmaceutical composition of the present invention. This syringe pen can be reusable or disposable. A reusable delivery pen typically uses a replaceable cartridge containing a pharmaceutical composition. After introducing all the pharmaceutical composition from the cartridge and emptying it, start

- 24 045943 This cartridge can be easily discarded and replaced with a new cartridge containing the pharmaceutical composition. The pen can then be reused. The disposable pen syringe does not have a replaceable cartridge. In contrast, a disposable pen comes already filled with a pharmaceutical composition contained in a reservoir inside the device. After emptying the reservoir of the pharmaceutical composition, the entire device is discarded.

A variety of reusable pen syringes and syringe tubes are suitable for subcutaneous delivery of the pharmaceutical composition of the present invention. Examples include, but are not limited to, AUTOPEN™ (Owen Mumford, Inc., Woodstock, UK), DISETRONIC™ pen (Disetronic Medical Systems, Burgdorf, Switzerland), HUMALOG MIX 75/25™ pen, HUMALOG pen ™, HUMALIN 70/30™ pen (Eli Lilly and Co., Indianapolis, IN), NOVOPEN™ I, II and III (Novo Nordisk, Copenhagen, Denmark), NOVOPEN JUNIOR™ (Novo Nordisk, Copenhagen, Denmark), BD™ pen (Becton Dickinson, Franklin Lakes, NJ), OPTIPEN™, OPTIPEN PRO™, OPTIPEN STARLET™ and OPTICLIK™ (Sanofi-Aventis, Frankfurt, Germany), and others. Examples of disposable pen syringes suitable for subcutaneous delivery of the pharmaceutical composition of the present invention include, but are not limited to, SOLOSTAR™ pen (Sanofi-Aventis), FLEXPEN™ (Novo Nordisk) and KWIKPEN™ (Eli Lilly) pen tube SURECLICK™ (Amgen, Thousand Oaks, CA), PENLET™ pen (Haselmeier, Stuttgart, Germany), EPIPEN (Dey, L.P.) and HUMIRA™ (Abbott Labs, Abbott Park, IL), and others.

In certain situations, the pharmaceutical composition may be delivered in a controlled release system. In one embodiment, a pump may be used (see Langer, supra; Sefton, 1987, CRC Crit. Ref. Biomed. Eng. 14:201). In yet another embodiment, polymeric materials may be used; see Medical Applications of Controlled Release, Langer and Wise (eds.), 1974, CRC Pres., Boca Raton, Florida. In yet another embodiment, the controlled release system can be placed in close proximity to the target of the composition, resulting in only a fraction of the systemic dose being required (see, e.g., Goodson, 1984, in Medical Applications of Controlled Release, supra, vol. 2, pp. 115-138). Other controlled release systems are discussed in Langer, 1990, Science 249:1527-1533.

Injectable forms may include dosage forms for intravenous, subcutaneous, intradermal and intramuscular injections, drips, etc. These injectable forms can be prepared using generally known methods. For example, injectable forms can be prepared, for example, by dissolving, suspending or emulsifying the antibody or salt thereof described above in a sterile aqueous or oily medium commonly used for injection. As the aqueous injection medium, there are, for example, saline solution, isotonic solution containing glucose and other auxiliary agents, etc., which can be used in combination with a suitable solubilizing agent such as alcohol (eg ethanol), polyhydric alcohol (eg propylene glycol, polyethylene glycol), nonionic surfactant [eg polysorbate 80, HCO-50 (adduct of polyoxyethylene (50 mol) and hydrogenated castor oil)], etc. As the oil medium, for example, sesame oil, soybean oil, etc. are used, which can be used in combination with a solubilizing agent such as benzyl benzoate, benzyl alcohol, etc. The injection thus prepared is preferably filled into a suitable ampoule.

Preferably, the pharmaceutical compositions for oral or parenteral use described above are prepared in dosage forms in a unit dose selected to contain the dose of the active ingredients. Such unit dose dosage forms include, for example, tablets, pills, capsules, injections (ampoules), suppositories, etc. The amount of the above antibody contained is generally from about 5 to about 500 mg per dosage form per unit dose; preferably, especially in injection form, the above antibody is contained in an amount of from about 5 to about 100 mg, and from about 10 to about 250 mg for other dosage forms.

Types of therapeutic applications of antigen-binding molecules.

The present invention includes methods comprising administering to a subject in need thereof a therapeutic composition comprising an anti-CD28 antibody or a bispecific antigen-binding molecule that specifically binds CD28 and a target antigen (eg, PSMA). Said therapeutic composition may comprise any of the antibodies or bispecific antigen binding molecules as disclosed herein and a pharmaceutically acceptable carrier or diluent. As used herein, the expression subject in need thereof means a human or non-human animal who exhibits one or more symptoms or signs of cancer (for example, a subject exhibiting signs of a tumor or suffering from any of the types of cancer referred to below herein), or who would otherwise benefit from inhibition or reduction of PSMA activity or reduction of the PSMA+ cell pool.

Antibodies and bispecific antigen-binding molecules according to the invention (and containing

- 25 045943 their therapeutic compositions) are suitable, inter alia, for the treatment of any disease or disorder that may benefit from stimulation, activation and/or targeting of the immune response. In particular, the anti-CD28/anti-PSMA bispecific antigen-binding molecules of the present invention can be used for the treatment, prevention and/or amelioration of any disease or disorder associated with or mediated by the expression or activity of PSMA or the proliferation of PSMA+ cells. The mechanisms of action by which the therapies of the invention are achieved include killing cells expressing PSMA in the presence of effector cells, such as T cells. Cells expressing PSMA that the bispecific antigen binding molecules of the invention can be used to inhibit or destroy include, for example, tumorigenic prostate cells.

The antigen binding molecules of the present invention can be used to treat, for example, primary and/or metastatic tumors arising from colon, lung, breast, kidney and bladder cancer subtypes. In certain illustrative embodiments, the bispecific antigen binding molecules of the present invention are used to treat prostate cancer.

The present invention also includes methods for treating residual cancer in a subject. As used herein, the term residual cancer means the presence or persistence of one or more cancer cells in a subject following anticancer therapy.

In certain aspects, the present invention provides methods for treating a disease or disorder associated with the expression of PSMA (eg, prostate cancer), comprising administering one or more bispecific antigen binding molecules described elsewhere herein to a subject after it has been determined that the the subject is resistant to other types of anticancer therapy. For example, the present invention includes methods of treating prostate cancer comprising administering an anti-CD28/anti-PSMA bispecific antigen binding molecule to a patient after 1 day, 2 days, 3 days, 4 days, 5 days, 6 days, 1 week, 2 weeks, 3 weeks or 4 weeks, 2 months, 4 months, 6 months, 8 months, 1 year or more after the subject has received standard treatment for patients suffering from cancer, such as prostate cancer. In other aspects, the bispecific antigen binding molecule of the invention (anti-CD28/anti-PSMA bispecific antigen binding molecule) containing the IgG4 Fc domain is initially administered to a subject at one or more time points (e.g., to provide a sustained initial reduction in the prostate cancer cell pool) followed by administration of an equivalent bispecific antigen-binding molecule containing a different IgG domain, such as the IgG1 Fc domain, at subsequent time points. It is contemplated that the anti-CD28/anti-PSMA antibodies of the invention may be used in combination with other bispecific antigen binding molecules, such as an anti-PSMA/anti-CD3 bispecific antibody. It is also contemplated that the bispecific antibodies of the invention will be used in combination with checkpoint inhibitors, such as those targeting PD-1 and CTLA-4, and other targets. Using a combination of two bispecific antibodies targeting the same tumor antigen (eg, PSMA) may provide benefits, but one of the bispecifics must target CD3 on T cells and the other bispecific must target a costimulator molecule such as CD28. This combination can be used alone to enhance tumor cell killing or can be used in combination with a checkpoint inhibitor.

Combination therapies and drugs.

The present invention includes compositions and therapeutic preparations containing any of the illustrative antibodies and bispecific antigen-binding molecules described herein in combination with one or more additional therapeutically active components, and methods of treatment comprising administering such combinations to subjects in need thereof.

Examples of additional therapeutic agents that may be used in combination or that may be administered in combination with an antigen binding molecule of the present invention include, for example, chemotherapy, radiation therapy, checkpoint inhibitors targeting PD-1 (eg, anti-PD-1 antibody , such as pembrolizumab or nivolumab; see also US9987500), CTLA-4, LAG3, TIM3 and others, co-stimulatory agonistic divalent antibodies targeting molecules such as GITR, OX40, 4-1BB and others), bispecific antibodies to CD3 ( see for example WO 2017/053856 A1, WO 2014/047231 A1, WO 2018/067331 A1 and WO 2018/058001 A1), other antibodies targeting PSMA X CD3 (see for example WO 2017/023761 A1) , and other costimulatory bispecific antibodies to CD28.

Other agents that may be beneficial when administered in combination with antibodies of the invention include, for example, tamoxifen, aromatase inhibitors and cytokine inhibitors, including small molecule cytokine inhibitors and antibodies that bind to cytokines such as IL-1, IL-2 , IL-3, IL-4, IL-5, IL-6, IL-8, IL-9, IL-11, IL-12, IL-13, IL-17, IL-18, or their respective receptors . Pharmaceutical compositions according to the present invention (for example, pharmaceutical compositions containing a bispecific antigen-binding molecule

- 26 045943 lu aHTu-CD28/aHTu-PSMA, as disclosed herein) can also be administered as part of a treatment regimen comprising one or more therapeutic combinations selected from ICE: ifosfamide (eg, Ifex®), carboplatin (eg, Paraplatin®), etoposide (eg, Etopophos®, Toposar®, VePesid®, VP-16); DHAP: dexamethasone (eg, Decadron®), cytarabine (eg, CytosarU®, cytosine arabinoside, ara-C), cisplatin (eg, Platinol®-AQ); and ESHAP: etoposide (eg, Etopophos®, Toposar®, VePesid®, VP-16), methylprednisolone (eg, Medrol®), high-dose cytarabine, cisplatin (eg, Platinol®-AQ).

The present invention also includes therapeutic combinations containing any of the antigen binding molecules mentioned herein and an inhibitor of one or more of VEGF, Ang2, DLL4, EGFR, ErbB2, ErbB3, ErbB4, EGFRvllll, cMet, IGF1 R, B-raf, PDGFR -o, PDGFR-I3, FOLH1, PRLR, STEAP1, STEAP2, TMPRSS2, MSLN, CA9, uroplakin, or any of the above cytokines, where the inhibitor is an aptamer, antisense molecule, ribozyme, shRNA, peptibody, nanobody, or antibody fragment (e.g. Fab fragment; F(ab') ₂ fragment; Fd fragment; Fv fragment; scFv; dAb fragment; or other engineered molecules such as diabodies, tribodies, tetrabodies, minibodies and minimal recognition units). The antigen-binding molecules of the invention can also be administered and/or included in a combination formulation with antivirals, antibiotics, analgesics, corticosteroids and/or NSAIDs (non-steroidal anti-inflammatory drugs). The antigen-binding molecules of the invention may also be administered as part of a treatment regimen, which also includes radiation therapy and/or conventional chemotherapy, or biological treatment, including checkpoint inhibitors or other bispecific antibodies.

The present invention includes compositions and therapeutic preparations containing any of the antigen-binding molecules described herein in combination with one or more chemotherapeutic agents. Examples of chemotherapeutic agents include alkylating agents such as thiotepa and cyclophosphamide (Cytoxan™); alkylsulfonates such as busulfan, improsulfan and piposulfan; aziridines such as benzodopa, carboquone, meturedopa and uredopa; ethyleneimines and methylmelamines, including altretamine, triethylenemelamine, triethylenephosphoramide, triethylenethiophosphoramide and trimethylolmelamine; nitrogen mustards such as chlorambucil, chlornaphasine, holophosphamide, estramustine, ifosfamide, mechlorethamine, mechlorethamine oxide hydrochloride, melphalan, novembiquine, phenesterine, prednimustine, trofosfamide, uracil mustard; nitrosoureas such as carmustine, chlorosotocin, fotemustine, lomustine, nimustine, ranimustine; antibiotics such as aclacinomycins, actinomycin, outramycin, azaserin, bleomycins, cactinomycin, calicheamicin, carabicin, carminomycin, carzinophylline, chromomycins, dactinomycin, daunorubicin, detorubicin, 6-diazo-5oxo-E-norleucine, doxorubicin, epirubicin, ezorubicin, idarubicin , marcellomycin, mitomycins, mycophenolic acid, nogalamycin, olivomycins, peplomycin, potfiromycin, puromycin, quelamycin, rhodorubicin, streptonigrin, streptozocin, tubercidin, ubenimex, zinostatin, zorubicin; antimetabolites such as methotrexate and 5-fluorouracil (5-FU); folic acid analogs such as denopterin, methotrexate, pteropterin, trimetrexate; purine analogs such as fludarabine, 6mercaptopurine, thiamiprin, thioguanine; pyrimidine analogues such as antsitabine, azacitidine, 6azauridine, carmofur, cytarabine, dideoxyuridine, doxifluridine, enocytabine, floxuridine; androgens such as calusterone, drostanolone propionate, epithiostanol, mepithiostane, testolactone; drugs that suppress adrenal function, such as aminoglutethimide, mitotane, trilostane; a folic acid replenisher such as frolinic acid; aceglatone; aldophosphamide glycoside; aminolevulinic acid; amsacrine; bestrabucil; bisantrene; edatraxate; defofamine; demecolcine; diazichon; elfornitine; elliptinium acetate; ethoglucide; gallium nitrate; hydroxyurea; lentinan; lonidamine; mitoguazone; mitoxantrone; mopidamole; nitracrine; pentostatin; phenomet; pirarubicin; podophyllic acid; 2-ethylhydrazide; procarbazine; PSK™; razoxane; sisofiran; spirogermanium; tenuazonic acid; triaziquon; 2,2',2-trichlorotriethylamine; urethane; vindesine; dacarbazine; mannomustin; mitobronitol; mitolactol; pipobromance; gacytosine; arabinoside (Ara-C); cyclophosphamide; thiotepa; taxanes, such as paclitaxel (Taxol™; Bristol-Myers Squibb Oncology, Princeton, NJ) and docetaxel (Taxotere™; Aventis, Anthony, France); chlorambucil; gemcitabine; 6-thioguanine; mercaptopurine; methotrexate; platinum analogs such as cisplatin and carboplatin; vinblastine; platinum preparations; etoposide (VP-16); ifosfamide; mitomycin C; mitoxantrone; vincristine; vinorelbine; navelbine; Novantrone; teniposide; daunomycin; aminopterin; xeloda; ibandronate; SRT-11; topoisomerase inhibitor RFS 2000; difluoromethylornithine (DMFO); retinoic acid; esperamycins; capecitabine; and pharmaceutically acceptable salts, acids or derivatives of any of the foregoing. Also included in this definition are antihormonal agents that regulate or inhibit the action of hormones on tumors, such as antiestrogens, including, for example, tamoxifen, raloxifene, aromatase inhibitor 4(5) imidazoles, 4-hydroxytamoxifen, trioxifene, keoxifene, LY 117018, onapristone and toremifene (Fareston); and antiandrogens such as flutamide, nilutamide, bicalutamide, leuprolide and goserelin; and pharmaceutically acceptable salts, acids or derivatives of any of the above.

The additional therapeutically active component(s) may be administered immediately before, simultaneously with, or shortly after administration of the antigen binding molecule of the present invention (in the context of the present invention, such administration schedules are considered to be administration of the antigen binding molecule in combination with the additional therapeutically active component).

The present invention includes pharmaceutical compositions in which the antigen binding molecule of the present invention is contained in combination with one or more additional therapeutically active components, as described elsewhere herein.

Administration schemes.

In certain embodiments of the present invention, multiple doses of an antigen binding molecule (eg, an anti-CD28 antibody or a bispecific antigen binding molecule that specifically binds PSMA and CD28) may be administered to a subject over a period of time. The methods of this aspect of the invention involve sequentially administering multiple doses of an antigen binding molecule of the present invention to a subject. As used herein, sequential administration means that each dose of the antigen binding molecule is administered to the subject at a different point in time, for example, on different days at a predetermined interval (eg, hours, days, weeks or months). The present invention includes methods comprising sequentially administering to a patient a single initial dose of an antigen binding molecule, followed by one or more secondary doses of said antigen binding molecule, and optionally followed by one or more tertiary doses of said antigen binding molecule.

The terms initial dose, secondary doses and tertiary doses refer to the time sequence of administration of the antigen binding molecule according to the invention. Thus, the starting dose is the dose that is administered at the beginning of the treatment regimen (also called the initial dose); secondary doses are doses that are administered after the initial dose; and tertiary doses are doses that are administered after secondary doses. The initial, secondary and tertiary doses may contain the same amount of antigen-binding molecule, but, as a rule, may differ from each other in terms of frequency of administration. In certain embodiments, however, the amount of antigen-binding molecule contained in the initial, secondary, and/or tertiary doses is varied (eg, adjusted up or down as needed) over the course of treatment. In certain embodiments, two or more (eg, 2, 3, 4, or 5) doses are administered at the beginning of the treatment regimen as loading doses, followed by subsequent doses that are administered less frequently (eg, maintenance doses).

In one illustrative embodiment of the present invention, each secondary and/or tertiary dose is administered at 1-26 (e.g., ^1.1 1/2, 2.21/2, 3, 31/2, 4, 4 ^1/2 , 5, 51/2, 6, 6 ^1/2 _, 7, 71/2, 8, 81/2, 9, 91/2, 10, 101/2, 11, 111/2, 12, 121/2, 13 , 131/2, 14, 141/2, 15, 151/2, 16, 161/2, 17, 171/2, 18, 181/2, 19, 191/2, 20, 201/2, 21, 211 /2, 22, 221/2, 23, 231/2, 24, 241/2, 25, 251/2, 26, 261/2 or more) weeks after the immediately preceding dose. The phrase immediately preceding dose as used herein means, in a multiple dosing sequence, the dose of an antigen binding molecule that is administered to a patient prior to administration of the immediately next dose in the sequence, with no intervening doses.

The methods of this aspect of the invention may include administering to the patient any number of secondary and/or tertiary doses of an antigen binding molecule (eg, an anti-CD28 antibody or a bispecific antigen binding molecule that specifically binds PSMA and CD28). For example, in certain embodiments, only one secondary dose is administered to the patient. In other embodiments, the patient is administered two or more (eg, 2, 3, 4, 5, 6, 7, 8 or more) secondary doses. Likewise, in certain embodiments, only one tertiary dose is administered to the patient. In other embodiments, two or more (eg, 2, 3, 4, 5, 6, 7, 8 or more) tertiary doses are administered to the patient.

In embodiments involving the administration of multiple secondary doses, each secondary dose can be administered at the same frequency as the other secondary doses. For example, each secondary dose may be administered to the subject 1-2 weeks after the immediately preceding dose. Likewise, in embodiments involving the administration of multiple tertiary doses, each tertiary dose can be administered at the same frequency as the other tertiary doses. For example, each tertiary dose may be administered to the patient 2-4 weeks after the immediately preceding dose. Alternatively, the frequency with which the subject is administered secondary and/or tertiary doses may vary over the course of treatment. The frequency of administration may also be adjusted by the attending physician during the course of treatment depending on the needs of the individual patient after clinical examination.

Diagnostic applications of antibodies.

The bispecific antibodies of the present invention can also be used to detect and/or measure CD28 or PSMA, or CD28-expressing, or PSMA-expressing

- 28 045943 cells per sample, for example for diagnostic purposes. For example, an anti-CD28 x PSMA antibody or fragment thereof may be used to diagnose a condition or disease characterized by impaired expression (eg, overexpression, underexpression, lack of expression, etc.) of CD28 or PSMA. Exemplary diagnostic assays for CD28 or PSMA may include, for example, contacting a sample obtained from a patient with an antibody of the invention, wherein said antibody is labeled with a detectable label or reporter molecule. Alternatively, the unlabeled antibody may be used for diagnostic purposes in combination with a secondary antibody that is itself labeled in a detectable manner. The detectable label or reporter molecule may be a radioisotope such as ³ H, ¹⁴ C, ³² P, ³⁵ S or ¹²⁵ I; a fluorescent or chemiluminescent moiety such as fluorescein isothiocyanate or rhodamine; or an enzyme such as alkaline phosphatase, beta-galactosidase, horseradish peroxidase or luciferase. Specific examples of assays that can be used to detect or measure CD28 or PSMA in a sample include enzyme-linked immunosorbent assay (ELISA), radioimmunoassay (RIA), and fluorescence-activated cell sorting (FACS). Samples that can be used in diagnostic assays for CD28 or PSMA according to the present invention include any tissue or fluid sample that can be obtained from a patient containing detectable amounts of CD28 or PSMA protein or fragments thereof, either normally or in pathology. Typically, CD28 or PSMA levels will be measured in a specific sample obtained from a healthy patient (eg, a patient not suffering from a disease or condition associated with abnormal CD28 or PSMA levels or activity) to initially establish a baseline or standard CD28 or PSMA level. This baseline level of CD28 or PSMA can then be compared with levels of CD28 or PSMA measured in samples obtained from individuals suspected of having a disease or condition associated with CD28 or PSMA.

Examples

The following examples are provided to provide those skilled in the art with complete disclosure and description of how to make and use the methods and compositions of the invention and are not intended to limit the scope of what the inventors believe to be their invention. Efforts have been made to ensure accuracy in terms of numbers used (e.g. quantities, temperatures, etc.), but some experimental errors and deviations should be taken into account. Unless otherwise indicated, parts are parts by weight, molecular weight is average molecular weight, temperature is in degrees Celsius, and pressure is at or near atmospheric pressure.

Annotation.

Recent clinical success of tumor-specific antigen (TSA) or tumor-associated antigen (TAA)-CD3 bispecific antibodies, such as PSMAxCD3 bispecific antibodies, is changing cancer immunotherapy. Antibodies to TAA-CD3 provide relatively safe and accessible therapeutic solutions that may prove to be an important new class of immunotherapy. However, they may not achieve the efficacy observed with CAR-T cell approaches that involve a second co-stimulatory signal, often mediated by the CD28 intracellular domain. Indeed, CD28-activating antibodies are incredibly effective but also toxic in human studies. The terms TSA and TAA may be used interchangeably in this document.

Herein, we describe a new class of bispecific antibodies, termed TAA-CD28, that appear to be safe, well tolerated alone and in combination with TAA-CD3 to generate significantly enhanced and highly targeted antitumor responses. Using TAAs such as PSMA expressed on prostate tumors, TAAxCD28 was found to increase T cell activation and significantly enhance tumor cell lysis in the presence of tumor target and TCR/CD3 stimulus.

It was demonstrated that the TAAxCD28 bispecific antibodies of the invention significantly enhanced antitumor immunity mediated by TAAxCD3-induced T cell activation in syngeneic tumor models. Pairing of CD3 with a tumor antigen with limited expression in healthy tissue induced little or no peripheral cytokine production when used alone or in combination with TAAxCD28 in non-human primate and humanized mouse models. In addition, TAAxCD28 alone induced little or no serum cytokine secretion and T cell activation in cynomolgus monkeys compared with CD28 superagonist antibody. These results suggest that combining this new class of co-stimulatory bispecific antibodies with the emerging TAA-CD3 class may provide safer, off-the-shelf biological solutions that could markedly improve the efficacy of traditional TAA-CD3 bispecifics.

Introduction.

Although monoclonal antibodies have been recognized as anticancer therapeutic agents over the past two decades, they have limited ability to mobilize

- 29 045943

T cells and effectively apply their cytotoxic activity to the tumor site. The ability of bispecific antibodies to target T cell immunotherapy and subsequently kill tumor cells has been previously demonstrated. Indeed, the bispecific antibody platform aims to recruit immune effector cells by combining an anti-CD3 antibody with a binding domain to a tumor target. Recently, catumaxomab (EpCAMxCD3) and blinatumomab (CD19xCD3) received regulatory approval for the treatment of acute lymphoblastic leukemia, while many other bispecific antibodies are being studied. Thus, bispecific antibodies that recognize both the B-cell marker CD20 and the CD3 component of the T-cell repertoire have been constructed, and they are currently undergoing clinical trials for the treatment of hematological malignancies. Thus, TAA-CD3 provide relatively safe and ready-to-use therapeutic solutions that do not require careful and painstaking adaptation for the individual patient. However, although TAA-CD3 bispecifics may prove to be an important new class of immunotherapy, cross-study comparisons suggest that they may not achieve the efficacy observed for CAR-T approaches.

Two chimeric antigen receptor T cell (CAR-T cell) drugs, Kymriiah and Yescarta, recently received FDA approval for the treatment of B-cell hematologic malignancies, demonstrating the great potential of this approach for personalized cancer immunotherapy. Both products utilize CD19 antigen as a tumor associated antigen (TAA), which is an ideal target for T cell-mediated killing due to its restricted expression on B cells, thereby minimizing off-target toxicity and enhancing antitumor efficacy. However, the high efficacy of CAR-T cells has been associated with side effects such as cytokine release syndrome (CRS) and neurotoxicity. In addition, the treatment does not yet benefit all patients, and the number of variables that can influence the clinical outcome for each patient is relatively large in both autologous and allogeneic approaches. In addition, attempts to target tumor-associated antigens in solid tumors have so far yielded little success, demonstrating either minimal antitumor activity or serious side effects. Inhibitory tumor microenvironment, poor access of CAR-T cells to entire tumor tissue along with labor-intensive production represent some of the current challenges for promising CAR-T cell therapies.

One limitation of current immunotherapy treatments is the optimal induction of the patient's own immune response against tumor cells through specific tumor cell recognition and induction of cytotoxicity. Efficient activation of naïve T cells and induction of antigen-trained memory T cell populations requires co-stimulatory signals (signal 2) in addition to antigen-specific stimulus through the TCR/CD3 complex (signal 1). Agonism of the CD28 and 4-1BB co-stimulatory signaling pathways can provide significant enhancement of target cell lysis, which may have a positive effect on patient resistance to various types of immunotherapy. However, the broader role of costimulation in clinical oncology has not yet been sufficiently appreciated. Many tumors lack co-stimulatory receptors, preventing them from further enhancing CD3xTAA-induced T cell activation.

A blind screen for costimulatory signaling pathways performed here by forcing expression of costimulatory ligands in a panel of syngeneic tumors revealed that CD28 is one of the most potent costimulatory receptors, together with 4-1BB. In addition, herein, co-stimulatory bispecific antibodies were constructed to couple CD28 on the surface of T cells to TAA from prostate tissue to enhance the antitumor response. Herein, we present data demonstrating that combination therapy with TAA-CD3 and TAA-CD28 bispecifics effectively enhanced T cell activation and cytotoxicity under conditions of simultaneous TAA and TCR stimulation, resulting in enhanced antitumor immunity. Indeed, the data disclosed herein suggest that combining this new class of bispecifics (TAA-CD28) with the emerging TAA-CD3 class may provide safer, off-the-shelf biological solutions that may approach the efficacy of personalized CART. cell therapy.

More specifically, exemplary bispecific antibodies to PSMAxCD28 were generated, and the exemplary anti-PSMAxCD28 antibody was demonstrated to enhance PSMAxCD3 or CD20xCD3-induced T cell activation in vitro and safely enhance antitumor efficacy in vivo. In vitro activity was demonstrated by imaging bispecific antibodies localized at the immunological synapse of the T-cell-target cell conjugate, enhancing PSMAxCD3-induced proliferation, cytokine release, and cytotoxicity. Antitumor efficacy in vivo was assessed in mouse tumor models (syngeneic). Tumor volume and serum cytokine levels were monitored to demonstrate response to bispecific antibody treatment. A study was conducted in cynomolgus monkeys to determine the safety and tolerability of the illustrative PSMAxCD28 according to ISO

- 30 045943 shaving in non-human primates. Animals were tested for toxicity through clinical observation and collection of blood samples for analysis of serum cytokine levels and T cell phenotype.

As described in detail below, exemplary bispecific antibodies to PSMAxCD28 were generated to enhance TCR/CD3-dependent T cell activation, thereby mimicking the costimulation (signal 2) provided by professional APCs.

All procedures were performed in accordance with the NTH (National Institutes of Health) Guide for the Care and Use of Laboratory Animals. The protocols were approved by the Regeneron Pharmaceuticals Institutional Animal Care and Use Committee.

Example 1. Construction of antibodies to PSMAxCD28.

Obtaining antibodies to CD28.

Antibodies to CD28 were generated by immunizing a VELOCIMMUNE® mouse (i.e., a genetically modified mouse containing DNA encoding the variable regions of the human immunoglobulin kappa heavy and light chains) with human CD28 protein fused to the Fc portion of mouse IgG2a, or cells expressing CD28, or DNA encoding CD28. The antibody response was monitored using a CD28-specific immunoassay. Once the desired immune response was achieved, splenocytes were collected and fused with murine myeloma cells to maintain their viability and form hybridoma cell lines. Hybridoma cell lines were screened and selected to identify cell lines producing CD28-specific antibodies. Using this method, several chimeric antibodies to CD28 (ie antibodies having human variable domains and mouse constant domains) have been obtained. In addition, several fully human anti-CD28 antibodies have been isolated directly from antigen-positive B cells without fusion with myeloma cells, as described in US 2007/0280945 A1.

Certain biological properties of exemplary anti-CD28 antibodies produced in accordance with the methods used in this example are described in detail in the examples set forth below.

Obtaining antibodies to PSMA.

Antibodies to PSMA were obtained by immunizing a genetically modified mouse with human PSMA antigen or by immunizing a genetically modified mouse containing DNA encoding the variable regions of the human immunoglobulin kappa heavy and light chain with human PSMA antigen. Alternatively, mice were immunized with human prostate cancer cells (LNCaP, ATTC, Manassas, VA, USA) expressing human PSMA (UniProtKB/Swiss-Prot. Q04609). The antibody response was monitored using a PSMA-specific immunoassay. When the desired immune response was achieved, splenocytes were collected from each mouse and either (1) fused with murine myeloma cells to maintain their viability and generate hybridoma cell lines and screened for PSMA specificity, or (2) B cells were sorted ( as described in US application 2007/0280945 A1) using human PSMA with an N-terminal 6-His tag (R&D, cat. no. 4234-ZN) as a sorting reagent that binds and identifies reactive antibodies (antigen-positive B cells ). Initially, chimeric antibodies to PSMA were isolated, having a human variable region and a mouse constant region. Antibodies were characterized and selected for desired characteristics, including affinity, selectivity, etc. When necessary, mouse constant regions were replaced with the desired human constant region, eg, wild-type IgG1 or IgG4 or modified IgG1 or IgG4, to generate a fully human anti-PSMA antibody.

Preparation of bispecific antibodies that bind CD28 and PSMA.

Bispecific antibodies containing a PSMA-specific binding domain and a CD28-specific binding domain were constructed using standard techniques, where the anti-PSMA antigen-binding domain and the anti-CD28 antigen-binding domain each contain different, separate HCVRs paired with a common LCVR. In some cases, bispecific antibodies have been constructed using an anti-CD28 heavy chain, an anti-PSMA heavy chain, and a common light chain (see Table 1). In certain embodiments, the heavy chain amino acid sequence from an anti-CD28 antibody for an exemplary bispecific antibody (bs16429D) is provided in SEQ ID NO: 81. The heavy chain amino acid sequence from an anti-PSMA antibody for an exemplary bispecific antibody (bs16429D) is provided in SEQ ID NO: 82. The amino acid sequence of the common light chain for bs16429D is shown in SEQ ID NO: 83.

Bispecific antibodies generated in accordance with the present example contain two separate antigen-binding domains (ie, binding arms). The first antigen binding domain comprises a heavy chain variable region derived from an anti-CD28 antibody (CD28-VH), and the second antigen binding domain comprises a heavy chain variable region derived from an anti-PSMA antibody (PSMA-VH). Both anti-PSMA and anti-CD28 share a common light chain. CD28VH/PSMA-VH pairing creates antigen-binding domains that specifically recognize CD28 on T cells and PSMA on tumor cells.

Example 2. Amino acid sequences and nucleotide sequences of the variable regions of the heavy and light chains.

- 31 045943

In table 1 shows the amino acid sequence identifiers of the variable regions of the heavy and light chains and CDRs of various bispecific antibodies created in accordance with example 1. The corresponding nucleic acid sequence identifiers are presented in table. 2.

Table 1

Amino acid sequence identifiers

Anti-PSMA x anti-bispecific antibody product identifier CD28 Second antigen binding domain (D2, anti-PSMA) First antigen binding domain (D1, anti-SP28) Light chain variable region D2- HCVR D2- HCDR1 D2- HCDR2 D2- HCDR3 D1- HCVR D1- HCDR1 D1- HCDR2 D1- HCDR3 D3- LCV R D3- LCDR 1 D3- LCDR 2 D3- LCDR 3 bsl6429D shАБ11838Р2 shАБ14226Р2 8567 ^{2 4 6 8 10 12 14 16 18 20 22 24} bsl6430D shАБ11810Р2 shАБ14193Р2 10082 ^{34 36} 38 40 26 28 ^{thirty 32 42 44} 46 48 bs!6431D shАБ11810Р2 shАБ14216Р2 10082 ⁵⁰ 1 ⁵² 1 ⁵⁴ 1 56 ⁵⁸ 1 60 ⁶² 64 66 ⁶⁸ 1 70 | 72

table 2

Nucleic acid sequence identifiers

Anti-PSMA x anti-bispecific antibody product identifier CD28 Second antigen binding domain (D2, anti-PSMA) First antigen binding domain (D1, anti-SP28) Light chain variable region D2- HCVR D2- HCDR1 D2- HCDR2 D2- HCDR3 D1- HCVR D1- HCDR1 D1- HCDR2 D1- HCDR3 D3- LCV R D3- LCDR 1 D3- LCDR 2 D3- LCDR 3 bsl6429D shАБ11838Р2 shАБ14226Р2 8567 ^{1 3 5 7 9 eleven 13 15 17 19 21 23} bsl6430D shАБ11810Р2 shАБ14193Р2 10082 33 1 ³⁵ 1 ³⁷ 1 ³⁹ 25 27 1 ^{29 31 41} I 43 I 45 1 ⁴⁷ bsl6431D shАБ11810Р2 SHAL4216R2 10082 49 1 ⁵¹ 1 ^{53 55} 57 ⁵⁹ 1 ^{61 63 65 67} 69 ⁷¹

Example 3: CD28 and 4-1BB are potent co-stimulatory receptors.

To determine which costimulatory receptors are effective in providing the costimulatory signal important for T cell activation, a blind screen for costimulatory signaling pathways was performed by forced expression of costimulatory ligands in a panel of syngeneic tumors (Table 3 and Fig. 2), in which again It was found that CD28, along with 4-1BB, is one of the most powerful co-stimulatory receptors. In table Figure 3 summarizes the number of tumor-free mice in the blind screening. Five (5) mice were tested in each group. Assays were performed on three (3) different tumor cell lines that were genetically modified to express seven (7) different costimulatory ligands. Briefly, EL4, MC38, and B16F10.9 tumor cells were genetically modified to express individual co-stimulatory ligands using lentiviral transduction. The cells were implanted subcutaneously into C57BL6 WT mice. Tumor growth was measured on days 18, 24 and 25 after implantation of tumor types EL4, MC38 and B16F10.9, respectively. This was the time point when the volume of tumors from the control groups reached the maximum acceptable size (>2000 mm ³ ). The sizes of the tumors were measured. In accordance with the data in table. 3, CD28 and 4-1BB are also among the most effective in reducing tumor size (data not shown).

- 32 045943

Table 3

Inhibition of tumor growth in genetically modified cell lines with introduced expression of a costimulatory ligand

Costimulatory Ligand Costimulatory Receptor Lymphoma (EL4) Carcinoma (MSZ 8) Melanoma (B16F10.9) 4-1BBL 4-BB 3 4 1 CD80(B7.1) CD28 2 2 2 CD86 (V7.2) CD28 1 0 2 CD70 CD27 5 0 OX40L 0X40 0 0 2 CD40 CD40L 0 1 0 ICOSL ICOS 0 0 0 Blank vector 0 0 0 Parental 0 0

Example 4 Binding affinities and kinetic constants of bispecific antibodies to PSMAxCD28 obtained from surface plasmon resonance analysis.

To determine the binding kinetics of bispecific monoclonal antibodies to PSMAxCD28, the binding affinities and kinetic constants of bispecific antibodies to PSMAxCD28 and the corresponding parental monoclonal antibodies to PSMA and/or CD28 were determined in a surface plasmon resonance assay.

Kinetics of binding of bispecific monoclonal antibodies to PSMAxCD28 with PSMA.

Equilibrium dissociation constants ( _KD values) for the binding of 6h.hPSMA (recombinant human PSMA/FOLH1 protein, R&D, cat. no. 4234-ZN) with purified bispecific anti-PSMAxCD28 monoclonal antibody or divalent parental anti-PSMA monoclonal antibody were determined using a biosensor based on surface plasmon resonance in real time using the Biacore T-200 device. The CM5 Biacore sensor surface was modified by amino linking to a mouse monoclonal anti-human Fc antibody to capture purified anti-PSMAxCD28 bispecific antibodies or parental anti-PSMA and anti-CD28 monoclonal antibodies.

This Biacore binding assay was performed in a buffer consisting of 0.01 MHEPES pH 7.4, 0.15 MNaCl, 0.5 mM MgCl ₂ , 1.0 mM CaCl ₂ , 0.05 vol% surfactant P20 (HBS running buffer -P++). Various concentrations of hPSMA with an N-terminal polyhistidine tag (6h.hPSMA, R&D) were prepared in HBS-P++ running buffer ranging from 10 nM to 0.4 nM with serial 3-fold dilutions for bispecific antibodies to PSMAxCD28 and parental monoclonal antibodies to PSMA or to CD28.

Various concentrations of 6h.hPSMA were passed over the captured monoclonal antibody surface at a flow rate of 50 μL/min. Association of 6h.hPSMA with captured monoclonal antibody was monitored for 3 min, and dissociation of 6h.hPSMA in HBS-P++ running buffer was monitored for 10 min. Kinetic rate constants for association ( _ka ) and dissociation (kd) were determined by fitting real-time sensorgrams to a 1:1 binding model using curve fitting software Scrubber 2.0c (BioLogic Software). Equilibrium binding dissociation constants (KD) and dissociative half-lives (t1/2) were calculated from the kinetic rate constants using the formula: KD (M) = kd/k _a , and t1/2 (min) = 0.693/kd/60.

Binding kinetic parameters for the binding of 6h.hPSMA to purified monoclonal antibodies at 25°C are presented in Table 1 below. 4.

Table 4

Binding affinities of monoclonal antibodies to PSMA at 25°C in Biacore

Antibody ID Common name ka (1/Ms) kd (1/s) _KD (M) T'/ ₂ (min) bsl6429D monoclonal antibody to 1.96E+05 4.92E-05 2.51E- 10 234.6

- 33 045943

PSMAxCD28 bsl6430D monoclonal antibody to PSMAxCD28 TBD TBD TBD TBD bsl6431D monoclonal antibody to PSMAxCD28 2.80E+05 3.85E-05 1.3 7E- 10 300.4 shАБ11810Р2 monoclonal anti-PSMA antibody 4.45E+05 2.99E-05 6.72E- eleven 386.3 mAbll838P2 monoclonal anti-PSMA antibody TBD TBD TBD TBD shАБ14226Р2 monoclonal anti-CD28 antibody N.B. N.B. N.B. N.B. mAbl4193P2 monoclonal anti-CD28 antibody N.B. N.B. N.B. N.B. shАБ14216Р2 monoclonal anti-CD28 antibody N.B. N.B. N.B. N.B.

TBD: not tested

NB: no binding observed

The binding kinetic parameters of 6h.hPSMA to one purified exemplary monoclonal bispecific antibody at 37°C are presented in Table 1 below. 5. One (1) RU (Response Unit) represents 1 pg of protein per mm2 as defined by the manufacturer.

Table 5

Binding affinities of monoclonal antibody to PSMA at 37°C in Biacore

Antibody ID Capture mAb (RU) Binding 10 nM HPSMA.6H (RU) Ka (1/Ms) Kd (1/s) _KD (M) 1½ bsl6429D 256.8±0.9 43.5 2.00E+05 7.93E-05 3.96E-10 145.7 Antibody ID Capture mAb (RU) Binding of 20 nM HPSMA.6H (RU) Ka (1/Ms) Kd (1/s) _KD (M) 1½ bsl6429D 189.7±1.6 73.7 2.93E+05 6.36E+05 2,ΠΕ-ΙΟ 181.6

Kinetics of binding of bispecific monoclonal antibodies to PSMAxCD28 to CD28.

Equilibrium dissociation constants (KD values) for the binding of hCD28.mmh to purified monoclonal antibodies were determined using a real-time surface plasmon resonance biosensor using a Biacore T-200 instrument. The CM5 Biacore sensor surface was modified by amino linking to a mouse anti-human Fc monoclonal antibody to capture purified bispecific anti-PSMAxCD28 monoclonal antibodies and parental anti-PSMA or anti-CD28 monoclonal antibodies.

Various concentrations of hCD28.mmh were passed over the captured monoclonal antibody surface at a flow rate of 50 μL/min. The association of hCD28.mmh with the captured monoclonal antibody was monitored for 5 min, and the dissociation of hCD28.mmh in HBS-P++ running buffer was monitored for 10 min. Kinetic rate constants for association ( _ka ) and dissociation (kd) were determined by fitting real-time sensorgrams to a 1:1 binding model using curve fitting software Scrubber 2.0c. Equilibrium binding dissociation constants (KD) and dissociative half-lives (t1/2) were calculated from the kinetic rate constants using the formula: K _D (M) = kd/k _a , and t1/2 (min) = 0.693/kd/60.

Binding kinetic parameters for binding of hCD28.mmh to purified bispecific anti-PSMAxCD28 monoclonal antibody or divalent parental anti-CD28 monoclonal antibody at 25°C are presented in Table 1 below. 6.

- 34 045943

Table 6

Binding affinities of monoclonal antibodies to CD28 at 25°C in Biacore

Antibody ID Common name ka (1/Ms) kd (1/s) _KD (M) T'/ ₂ (min) monoclonal 3.26E- bsl6429D antibody to PSMAxCD28 2.26E+04 03 1.44E-07 3.5 monoclonal TV bsl6430D antibody to PSMAxCD28 TBD TBD TBD D monoclonal 1.41E- bsl6431D antibody to PSMAxCD28 6.79E+03 03 2.07E-07 8.2 monoclonal mAB11810P2 anti-PSMA antibody N.B. N.B. N.B. N.B. monoclonal TV mAB11838P2 anti-PSMA antibody TBD TBD TBD D mAB14226P2 monoclonal 2.34E+04 3.28E- 1.40E-07 3.5 anti-CD28 antibody 03 monoclonal 2.64E- mAB14193P2 anti-CD28 antibody 8.73E+03 04 3, OZE-08 43.7 monoclonal 1.41E- mAB14216P2 anti-CD28 antibody 1D2E+04 03 1.27E-07 8.2

TBD: not tested

NB: no binding observed

Binding kinetic parameters for the binding of hCD28.mmh to purified bispecific monoclonal antibody to PSMAxCD28 at 37°C are presented in Table 1 below. 7.

Table 7

Binding affinities of monoclonal antibodies to CD28 at 37°C

Antibody ID Capture shab (RU) Binding of 400 nM hCD28.6H (RU) Ka (1/Ms) Kd (1/s) _KD (M) t'/2 bs!6429D 1576.3±5.0 98.9 2.46E+04 b,70E-03 2.72E-07 1.7

As shown in table. 4-7, several anti-CD28 antibodies of the present invention bind CD28 with high affinity. Several anti-PSMA antibodies of the present invention bind PSMA with high affinity. Several PSMAxCD28 bispecific antibodies bind both CD28 and PSMA with high affinity.

Example 5. Efficiency and specificity of binding of bispecific monoclonal antibodies to PSMAxCD28 with PSMA and CD28 on the cell surface.

To evaluate the ability of these antibodies (anti-PSMA, anti-CD28, and anti-PSMAxCD28 monoclonal antibodies) to specifically bind to cell surface proteins, in vitro binding assays were developed using cells expressing human PSMA or CD28 using an electrochemiluminescence (MSD)-based detection platform. . Two studies were conducted. One study assessed the binding efficiency and specificity of monoclonal antibodies to cell surface antigens. These antibodies (anti-PSMA, anti-CD28, and anti-PSMAxCD28) showed specific binding to cell lines expressing human PSMA or human CD28. In an additional experiment, an extension study was performed with the IgG4s antibody isotype.

Methods used to determine the efficiency and specificity of cellular binding of PSMAxCD28 bispecific antibodies to PSMA and CD28.

The human prostate epithelial carcinoma cell line, C4-2 (UroCor), endogenously expresses human PSMA. The cell line expressing HEK293/hCD28 was genetically modified by transducing human embryonic kidney cells from ATCC (CRL-1573) with a neomycin-resistant lentiviral construct encoding human CD28 (hCD28, accession number NP_006130.1). To assess binding specificity, these two cell lines were assessed in parallel with the parental HEK293HZ cell line (which is negative for PSMA and CD28) by fluorescence-activated cell sorting (FACS). Human IgG1 anti-Feld1 antibody was included as a negative control for IgG detection.

The cell lines described above were washed once with 1xPBS without Ca ^{2+ /} Mg ²⁺ buffer (Irvine Scientific, cat. no. 9240) and incubated for 10 min at 37°C with enzyme-free cell dissociation solution (Millipore, cat. no. S-004-C) to separate cells from the flask. An additional wash was performed with lxPBS with Ca ^{2+ /} Mg2 ⁺ (Irvine Scientific, cat. no. 9236). Cells were then counted using a Cellometer™ Auto T4 Cell Counter (Nexcelom Bioscience, Model #Auto T4). Approximately 10,000 cells per well in cell wash buffer were seeded individually into 96-well carbon plate electrode plates (MULTI-ARRAY High Binding Plate, MSD, Meso Scale Discovery, Cat. No. L15XB-3/LX11XB-3) and incubated for 1 h at 37°C to allow cells to attach. Nonspecific binding sites were blocked with 2% BSA (wt/vol) (Sigma, cat. no. A85771L) in PBS for 1 h at room temperature.

Solutions containing anti-PSMA, anti-CD28, anti-PSMAxCD28, or control antibodies in serial dilutions from 1.7 pM to 100 nM, or solutions without antibody, were added to plate-bound cells and incubated for 1 hour at room temperature. Unless otherwise stated, assays were performed in duplicate. The plates were then washed to remove unbound antibodies using an AquaMax2000 plate washer with cell wash head (MDS Analytical Technologies, model #2000). Plate-bound antibodies were detected using SULFO-TAG™-conjugated goat polyclonal anti-human IgG heavy and light chain specific antibody (Jackson ImmunoResearch, cat. no. 109-005-088) for 1 hour at room temperature. After washes, the plates were developed using reading buffer (MSD, Meso Scale Discovery, cat. no. R92TD-2) according to the procedure recommended by the manufacturer, and luminescent signals were recorded using a SECTOR Imager 600 instrument (Meso Scale Discovery, model no. 600) .

Signals from direct binding (in relative light units, RLU) were analyzed as a function of antibody concentration. Data were fitted to a sigmoidal (four-parameter logistic) dose-response model using GraphPad Prism™ software (GraphPad software version no. 6). EC ₅₀ values, defined as the antibody concentration at which 50% of the maximum binding signal is detected, were determined for binding to HEK293/hCD28 and C4-2 cells to indicate the strength of each antibody binding to CD28 or PSMA, respectively. In addition, ratios of antibody binding signals at 11.1 nM on HEK293/hCD28 or C4-2 cells to HEK293HZ cells were calculated. This representative concentration was chosen due to high binding to target cells but low background signal on CD28, PSMA-negative HEK293HZ cells. Antibodies with a binding ratio less than 3 are labeled NB in the table below. 8. NB defined as no specific binding observed under assay conditions.

In an additional study, a separate experiment was performed using the same protocol as described above but including an IgG4 isotype control (see USSN 15/147,791) and the parental anti-CD28 monoclonal antibody. Test solutions contained from 3.4 pM to 200 nM anti-PSMA antibodies, anti-CD28 antibodies, bispecific monoclonal antibodies against PSMAxCD28, or control antibodies in serial dilutions. Binding ratios were calculated at 7.4 nm.

Results, summary and conclusions.

The ability of bispecific antibodies to PSMAxCD28 to specifically bind to cells expressing PSMA or CD28 was assessed in comparison with a cell line negative for PSMA or CD28 expression using an immunobinding assay. Bivalent parental antibodies specific for PSMA or CD28 were included for comparison. Dose-dependent binding of antibody to cells on 96-well high-binding plates (MSD, Meso Scale Discovery, cat. no. L15XB-3/LX11XB-3) with antibody concentrations up to 100 nM was determined using SULFO-TAG™-conjugated anti-human IgG antibody , and electrochemiluminescence binding signals were recorded on a Sector Imager 600 (MSD). RLU values were determined for antibody binding to cells. For cells expressing CD28 or PSMA, EC ₅₀ values were calculated as a measure of efficiency. For samples that resulted in a high background signal, higher concentrations were excluded from the calculation of _EC50 values and the values are marked with an asterisk in the table. 8. Comparison of antibody binding signals at 11.1 nM with HEK293/hCD28 cells or C4-2 negative HEK293HZ cells was used to assess antibody binding specificity. Specific binding is defined as antibodies that bind to CD28- or PSMA-expressing cells 3-fold or more than HEK293HZ cells at that concentration.

The binding results are summarized in Table. 8. At a concentration of 11.1 nM, three illustrative bispecies

- 36 045943 specific antibodies to PSMAxCD28 according to the invention (bs16429D, bs16430D and bs16431D) specifically bound to both HEK293-hCD28 and C4-2 cells, with a 13-31-fold and 4-10-fold ratio relative to HEK293HZ cells, respectively. The effectiveness of bispecific antibodies varied from EC ₅₀ values of 5.31-9.58 nM on HEK293-hCD28 cells to 0.35-5.24 nM on C4-2 cells.

Table 8

Binding of monoclonal antibodies to cell surface antigens

Antibody ID EC ₅₀ (M) HEK293-hCD28 ec ₅₀ (M) S4-2 HEK293hCD28/ HEK293HZ** С4-2/ HEK293HZ*** bsl6429D 9.58E-09 8.53E-10 31 10 wAL14226P2 (CD28) 1.04E-09 N.B. 68 1 SHAL1838R2 (PSMA) N.B. 7.38E-10* <1 4 bsl6430D 8.47E-09 5.24-09 41 10 wAL14193P2 (CD28) 1.02E-08 N.B. 110 1 shаb11810Р2 (PSMA)# N.B. 9.46E-10 1 25 bsl6431D 5.31E-09 3.50E-10* 13 4 wAL14216P2 (CD28) 2.47E-09 N.B. 114 1 CONTROL Antibody to Feld# N.B. N.B. 1 1

NB: non-specific binding as evidenced by a less than 3-fold ratio of cell binding signal to HEK293HZ at 11.1 nM *: concentrations above 11.1 nM are excluded from the EC ₅₀ calculation due to high background #: samples processed as a single dose serial dilution * *: ratio of 11.1 nM antibody-cell binding signal (RLU) for HEK293-hCD28 relative to parental HEK293HZ * **: ratio of 11.1 nM antibody-cell binding signal (RLU) for C4-2 relative to HEK293HZ

Parental antibodies bound specifically to cells corresponding to the antigen used to generate them, as indicated in parentheses in the antibody identifier column. EC ₅₀ values for binding to HEK293-hCD28 and C4-2 cells for parental antibodies ranged from 1.04-10.2 nM on HEK293-hCD28 cells to 0.738-0.946 nM on C4-2 cells. For antibodies, the binding values of mAb11838P2 and mAb11810P2 at higher concentrations on C4-2 cells were excluded from the calculation of EC ₅₀ values to compensate for the high background value on negative cells. As expected, the control IgG antibody did not show specific binding to cell lines expressing CD28 or PSMA.

An additional experiment was performed as described above but included an IgG4 isotype control and a parental anti-CD28 monoclonal antibody. These data were obtained as an extension study to demonstrate negligible background signal using the hIgG4s isotype control. The data are summarized in table. 9. As shown in table. 9, the anti-CD28 antibody mAb14226P2 specifically binds to a cell expressing human CD28,

HEK293-hCD28 P-3. The anti-CD28 antibody mAb14226P2 does not bind to cells that do not express human CD28, whether those cells express PSMA (C4-2) or not (HEK293HZ).

Table 9

Binding of monoclonal antibody isotype to cell surface antigen

Antibody ID EC ₅₀ (M) HEK293hCD28 R-3 EC ₅₀ (M) S4-2 R-7 HEK293hCD28/ HEK293HZ* C42/HEK293HZ yy tAL4226R2 8.41E-10 N.B. 99 1 CONTROL hlgGlKFeldl N.B. N.B. 1 1 hIgG4 _K Feldl N.B. N.B. 1 1

NB: non-specific binding as evidenced by less than 3-fold cell binding signal ratio to HEK293HZ at 11.1 nM*: 7.4 nM antibody cell binding signal ratio (RLU) for HEK293-hCD28 relative to parental HEK293HZ* *: ratio of 7.4 nM antibody-cell binding signal (RLU) for C4-2 relative

HEK293HZ

- 37 045943

Binding of bispecific antibodies to PSMAxCD28 to T cells and target cells.

Experimental technique.

Flow cytometry analysis followed by phycoerythrin (PE)-labeled anti-human IgG antibody was used to determine the binding of bispecific antibodies to PSMAxCD28 to C4-2, 22RV1, RAJI, human and cynomolgus macaque T cells. Briefly, 1x10 ⁵ cells/well were incubated for 30 min at 4°C with serial dilutions of bispecific antibodies to PSMAxCD28 or human IgG4 antibody (see USSN 15/147,791), which binds human antigen without cross-reactivity with human or cynomolgus CD28. ranging from 133 nM to 8.14 pM for human and cynomolgus T cells and ranging from 133 nM to 61 pM for cells expressing PSMA. After incubation, cells were washed twice with cold PBS containing 1% filtered FBS, and PE-conjugated secondary human anti-antibody (Jackson Immunoresearch, cat. no. 709-116-149) was added to the cells and incubated for an additional 30 min. A dye was added to the incubated human and cynomolgus cells to determine cell viability. Wells containing no antibodies or containing only secondary antibodies were used as controls.

After incubation with PSMA-expressing cells, cells were washed, resuspended in 200 μl of cold PBS containing 1% filtered FBS, and analyzed by flow cytometry on a BD FACS Canto II.

After incubation with human or cynomolgus T cells, the cells were washed and stained with a mixture of anti-CD2 (BD, cat. no. 562638), anti-CD16 (BD, cat. no. 562874), anti-CD4 (BD, cat. no. 564305), and CD8 (BD, cat. no. 563795) in Brilliant staining buffer (BD, cat. no. 566349), and incubation continued for another 20 min at 4°C. After washing, cells were resuspended in 200 μl of cold PBS (Gibco, cat. no. 14190-144) containing 1% filtered FBS (TCB, cat. no. 101), gated live/CD2+/CD4+/CD16- or live/CD2+. /CD8+/CD16- and analyzed by flow cytometry on a BD FACS LSR-Fortessa-X20.

Results, summary and conclusions

The binding of bispecific antibodies to PSMAxCD28 to the surface of human T cells was studied by flow cytometry.

bs16429D bound to all T cells with an _EC50 value of 4.80x10'8 M. It bound to both CD4+ and CD8+ T cells, with an _EC50 value of 5.09x10'8 M and 5.89x10'8 M. respectively.

bs16431D bound weakly to all T cells with an EC ₅₀ value of 1.80x10'7. It bound weakly to both CD4+ and CD8+ T cells, with EC ₅₀ values of 1.67E-07 M and 1.80E-07 M, respectively.

The binding of bispecific antibodies to PSMAxCD28 to the surface of cell lines expressing PSMA was studied by flow cytometry.

C4-2 is a CaP (prostate cancer) cell line derived from LNCaP cells (androgen-sensitive human prostate adenocarcinoma cells derived from lymph node metastases; see Wu et al., Int. J. Cancer, 57: 406-412 (1994)). Both bs16429D and bs16431D bound to C4-2 cells (see Liu et al., 2004, Prostate, 60:98-108) with EC ₅₀ values of 3.87 x 10' ⁹ M and 1.50 x 10' ⁹ M, respectively.

22RV1 is an epithelial prostate carcinoma cell line (see In Vitro Cell Dev. Biol. Anim., 1999, 35(7):403-409). Both bs16429D and bs16431D bound to 22RV1 cells with _EC50 values of 3.05x10'9 M and ^6.33x10'09 M, respectively.

These results are summarized in table. 10-12.

Table 10

Binding of bispecific antibodies to PSMAxCD28 to human T cells

Antibody ID Human CD2+ T cells, FACS [M] Human CD4+ T cells, FACS [M] Human CD8+ T cells, FACS [M] bsl6429D 4.80E-08 5.09E-08 5.89E-08 bsl6431D (Weak) 1.80E-07 (Weak) 1.67E-07 (Weak) 1.80E-07 Isotypic Binding Binding Binding control absent absent absent

- 38 045943

Table 11

Binding of bispecific antibodies to PSMAxCD28 to cynomolgus monkey T cells

Antibody ID Cynomolgus CD2+ T cells, FACS [M] Cynomolgus CD4+ T cells, FACS [M] Cynomolgus CD8+ T cells, FACS [M] bsl6429D 1.10E-08 1.93E-08 (weak) 6.84E-08 bsl6431D (weak) 1.34E-07 (weak) 2.81E-07 Very weak binding Isotypic Binding Binding Binding control absent absent absent

Table 12

Binding of bispecific antibodies to PSMAxCD28 to cells expressing PSMA

Identifier antibodies C4-2 cells, FACS [M] 22RV1 cells, FACS [M] Raji cells, FACS [Μ] bsl6429D 3.87E-09 3.05E-09 Binding absent bsl6431D 1.50E-09 6,ZZE-09 Binding absent Isotypic control Binding absent Binding absent Binding absent

Example 6. Original and genetically modified bioassay for bispecific antibodies to PSMAxCD28.

Activation of T cells occurs through stimulation of T cell receptors (TCRs), which recognize specific peptides presented by major histocompatibility complex class I or II proteins (MHC I or MHC II) on antigen presenting cells (APCs) (Goldrath et al., Selecting and maintaining a diverse T-cell repertoire, Nature 402, 255-262 (1999)). The activated TCR in turn initiates a cascade of signaling events that can be monitored by reporter genes driven by various transcription factors, such as activator protein 1 (AP-1), nuclear factor of activated T cells (NFAT), or nuclear factor kappa -bi (NFkB). The T-cell response is then further refined by engaging coreceptors expressed either constitutively or inducibly on T cells, such as CD28, CTLA-4 (cytotoxic T-lymphocyte-associated protein 4), PD-1 (programmed cell death protein 1), LAG-3 (lymphocyte activation gene 3) or other molecules (Sharpe et al., The B7-CD28 Superfamily, Nat. Rev. Immunol., 2(2): 116-26 (2002)). The costimulatory molecule CD28 is activated by its endogenous ligands CD80 or CD86 expressed on APCs. CD28 enhances cellular signals, such as signaling pathways controlled by the transcription factor NFkB, upon TCR activation. The accompanying CD28 signal is important for efficient T cell activation, such as differentiation, proliferation, cytokine release, and cell killing by T cells (Smeets et al., NFkB activation induced by T cell receptor/CD28 costimulation is mediated by protein kinase C-θ, PNAS, 97(7):3394–3399 (2012).

To identify antibodies that enhance T cell activity in the simultaneous presence of both primary stimulation and PSMA target expression, anti-CD28 antibodies and anti-PSMAxCD28 bispecific antibodies were characterized using genetically modified reporter bioassays and cell-based assays using primary human T cells. The assays evaluate the behavior of the PSMA/CD28 bispecific antibody in the presence and absence of primary stimulation, as well as in the presence and absence of target expression. The analysis scheme is presented in Fig. 3. Assays were performed to select bispecific antibodies to PSMAxCD28, which enhance T cell activity under conditions of primary stimulation and target expression. Accordingly, the analyzes assessed the behavior of bispecific antibodies in the presence and absence of primary stimulation and target expression.

1) Luciferase-based reporter assay.

a) Genetic modification of reporter T cells.

The Jurkat-derived T cell clone JRT3.T3.5 (ATCC, no. TIB-153) was transduced with a reporter construct containing the NFkB luciferase gene (NFkB-Luc, SA Biosciences/Qiagen, cat. no. CLS-013L). By

- 39 045943 Following the isolation of a puromycin-resistant clone (JRT3.T3.5/NFkB-Luc clone 1C2), the cells were additionally genetically modified to express the full-length human TCR alpha subunit (1G4A - amino acids M1 to S274) and the TCR beta subunit (1G4B - amino acids M1 to G311) (Robbins et al., Single and Dual Amino Acid Substitutions in TCR CDRs Can Enhance Antigen-Specific T Cell Functions, J. Immunol. 180(9): 6116-31(2008)). After isolation of a single clone (J.RT3-T3.5/NFkBLuc/1G4AB clone 1D2), the cells were further genetically modified to express the full-length human CD8 alpha subunit (hCD8a - amino acids M1 to V235 under registration number NP_001139345) and the human CD8 beta subunit (hCD8b - amino acids M1 to K210 under registration number P10966). A single clone was again obtained (J.RT3-T3.5/NFkBLuc/1G4AB/hCD8ab clone 1D5) and additionally transduced with full-length human CD28 (hCD28 - amino acids M1 to S220, accession no. P10747). Cells were sorted for high CD28 expression and maintained in RPMI + 20% FBS + penicillin/streptomycin/glutamine (P/S/G) + NEAA + NaPyr + 1 μg/ml puromycin + 500 μg/ml G418 + 250 μg/ml hygromycin + 10 µg/ml blasticidin. For faster growth, genetically modified reporter T cells were maintained in cell culture medium without antibiotics and used for luciferase cell experiments as genetically modified reporter T cells. Reagent details are as follows: RPMI 1640, Irvine Scientific, cat. No. 9160; FBS, Seradigm, cat. No. 1500-50; penicillin/streptomycin/glutamine 100x (P/S/G), Thermo Fisher Scientific, cat. No. 10378-016; essential amino acids (NEAA), Irvine Scientific, cat. No. 9304; Sodium pyruvate (NaPyr), Millipore, cat. No. TMS-005-C; puromycin, Sigma, cat. No. Р8833; geneticin (G418), Thermo Fisher Scientific, cat. No. 11811-098; hygromycin; blasticidin.

b) Genetic modification of agro-industrial complex.

A stable cell line HEK293 (ATCC, no. CRL-1573) expressing human CD20 (amino acids M1 to P297 under accession no. NP_068769.2) was transduced with human PSMA (amino acids M1 to A750 under accession no. Q04609). Human PSMA-positive cells were isolated by fluorescence-activated cell sorting (FACS), and the resulting HEK293/CD20/hPSMA high-sorted cell line was maintained in DMEM+10%+P/S/G+NEAA supplemented with 500 μg/ ml G418.

c) T cell/APC stimulation.

In this experiment, genetically modified reporter T cells are stimulated with two bispecific antibodies. The first stimulation is carried out with a T cell activating bispecific antibody, hIgG4 to CD3xCD20 (see WO 14/047231, US 9657102 and USSN 14/661,334), targeting CD3 molecules on genetically modified reporter T cells and CD20 on HEK293 cells. In this case, the first stimulation makes it possible to avoid the activation of TCRs by their natural ligands, which are certain peptides exposed on MHC molecules. The second stimulation is caused by a bispecific antibody to CD28. This antibody mimics the activation of CD28 on T cells by its ligands, CD80/CD86, expressed on APCs. In this case, the antibody interacts with CD28 on T cells and PSMA on genetically modified HEK293 cells and triggers the activation of CD28 on genetically modified reporter T cells. Simultaneous activation of TCR and CD28 leads to increased transcriptional activity of NFkB, which, in turn, induces the production of a reporter gene, luciferase.

d) Luciferase assay design.

RPMI1640 supplemented with 10% FBS and P/S/G was used as assay media to prepare cell suspensions and antibody dilutions to screen for bispecific antibodies to PSMAxCD28.

The day before screening, genetically modified reporter T cells were cultured to a density of 1x10 ⁶ cells/ml in cell culture medium. Threefold (1:3) serially diluted bispecific antibodies to PSMAxCD28 and controls were tested in the presence of a constant concentration of 50 pM anti-CD20xCD3 antibody or isotype control hIgG4. The 10-point dilution range was from 15 pM to 100 nM, with the final dilution containing no anti-PSMAxCD28 antibodies. Reagents were added in the following order: 1) serially diluted antibodies were added to 96-well white flat-bottom plates into the appropriate wells; 2) a fixed concentration of 50 pM antibody to CD20xCD3 or isotype control hIgG4 was added to each well; 3) APCs were added to the plates, re-suspended to a density of ^4x105 cells/ml, with a final concentration of 1x104 cells/well; 4) reporter T cells cultured overnight were resuspended at a density of 2 x 10 ⁶ /ml and added to the plates at a final concentration of 5 x 10 4 cells/well. The plates were incubated for 4-6 hours at 37°C/5% CO ₂ before adding 100 μl of ONE-Glo™ reagent (Promega, cat. no. E6051) to lyse cells and determine luciferase activity. Emitted radiation was recorded in relative light units (RLU) on an Envision multichannel plate analyzer (PerkinElmer, model 2104). All serial dilutions were tested in duplicate.

The _EC50 values of antibodies were determined by approximating the data with a four-parameter log

- 40 045943 static equation on a 10-point dose-response curve using GraphPad Prism™ software. The induction factor was calculated using the following equation:

Multiplicity Average RLU values of antibody G100 nM1 induction = Average RLU values of antibody [0 nM]

2) Functional analysis of IL-2 using primary human CD4+ T cells:

A functional assay in primary CD4 T cells/APCs was developed to evaluate the effect of CD28 activation on IL-2 production when bound to PSMAxCD28 bispecific antibodies.

a) Isolation of primary human CD4+ T cells.

Human peripheral blood mononuclear cells (PBMCs) were isolated from leukocyte concentrate obtained from a healthy donor. PBMCs were isolated by density gradient centrifugation using 50 mL SepMate™ tubes (StemCell Technologies, cat. no. 85450) according to the protocol recommended by the manufacturer. Briefly, 50 mL SepMate tubes were layered with 15 mL of FicollPaque PLUS, followed by the addition of 30 mL of leukocytes diluted 1:2 with D-PBS (Dulbecco's sodium phosphate buffer, Irvine Scientific, cat. no. 9240). Subsequent steps were carried out according to the protocol of the manufacturer SepMate. CD4+ T cells were then isolated from PBMCs using anti-human CD4 microsphere kits from Miltenyi Biotec (cat. no. 130-045-101) following the manufacturer's instructions. Isolated CD4+ T cells were frozen in FBS containing 10% DMSO (Macron Fine Chemicals, cat. no. 4948-02) at a concentration of 5x10 ⁶ cells per vial.

b) Release of IL-2 from primary CD4+ T cells treated with anti-CD28 antibodies.

In this assay, primary CD4 T cells are activated by cross-linking CD3 on their surface using a bispecific antibody to CD20xCD3 in combination with HEK293 cells genetically modified to express human CD20. Binding of the CD20 arm of the CD20xCD3 bispecific antibody to CD20-expressing HEK293 cells triggers clustering of the CD3 receptor, providing the first signal important for T cell stimulation. However, co-stimulation is important to determine the quantitative release of IL-2, which can be achieved by cross-linking CD28 molecules. In this case, bispecific antibodies to PSMAxCD28 interact with CD28 on CD4 T cells and PSMA on genetically modified HEK293/hCD20 cells, and trigger clustering - activation of CD28. The combined engagement of TCR and CD28 leads to increased production of IL-2, which is released into the culture medium. IL-2 is detected and quantified in cell supernatant using a homogeneous, wash-free AlphaLisa kit (PerkinElmer, cat. no. AL221).

Previously isolated and frozen human CD4+ T cells from a donor were thawed on the day of analysis in a stimulating medium (X-VIVO 15 cell culture medium (Lonza, cat. no. 04-418Q) supplemented with 10% FBS, HEPES, NaPyr, NEAA and 0.01 mM BME (P-mercaptoethanol, Sigma-Aldrich, cat. no. M-7522) containing 50 U/ml benzonase nuclease (EMD Millipore, cat. no. 71205-3)). Cells were centrifuged at 1200 rpm for 10 min, resuspended in stimulation medium and seeded into 96-well round-bottom plates at a concentration of 1x10 ⁵ cells/well. HEK293 cells genetically modified to express human CD20, alone or in combination with human PSMA, were treated with 15 μg/ml mitomycin C (Sigma-Aldrich, cat. no. M4287) in primary plating stimulation medium at a concentration of 10 x 10 ⁶ cells/ml. After incubation for 1 hour at 37°C, 5% CO ₂ , HEK293 cells were washed 3 times with D-PBS containing 2% FBS and added to wells containing CD4+ T cells at a final concentration of 1 x 10 ⁴ cells per well.

Subsequently, serially diluted 1:3 bispecific anti-PSMAxCD28 antibodies or control antibodies ranging from 15 pM to 100 nM were added to the wells in the presence of 50 pM anti-CD20xCD3 or hIgG4 isotype control. The 10-point dilution endpoint did not contain anti-PSMAxCD28 or anti-CD28 antibodies. After incubating the plates for 72 h at 37°C, 5% CO ₂ , they were centrifuged to pellet the cells and 40 μl of supernatant medium was collected. From this volume, 5 μl was tested in the human IL-2 assay by AlphaLISA according to the manufacturer's protocol. Measurements were made on an Envision multichannel plate analyzer (PerkinElmer, model 2104). A calibration curve of known IL-2 concentrations was used to determine the IL-2 concentrations generated in the assay wells. All serial dilutions were tested in duplicate.

Antibody EC50 values were determined by fitting the data to a four-parameter logistic equation on a 10-point dose-response curve using GraphPad Prism™ software. The induction factor was calculated using the following equation:

Multiplicity Average values of IL-2 antibodies G100 nM1 induction = Average values of IL-2 antibodies [0 nM]

Results, summary and conclusions.

As shown in FIG. 4A and 4B, activation of CD4 T cells (measured by IL-2 release) and genetically modified JRT3.T3/1G4/hCD28 cells (measured by luciferase activity) was enhanced by antibodies to hPSMA x hCD28 under primary stimulation conditions (REGN2281 to CD20xCD3)

- 41 045943 and PSMA expressed on HEK293/hCD20 cells. The divalent anti-CD28 antibody, mAb14193P2, modestly enhances T cell activity in the presence of priming and not significantly in a genetically modified bioassay in the absence of priming. The CD28 superagonist, TGN1412, enhances T cell activation in both prime and genetically engineered assays under CD20xCD3 stimulation conditions, although to a lesser extent than the PSMAxCD28 bispecifics under PSMA target and prime stimulation conditions.

1) Luciferase-based reporter assay.

The values of EC50 and induction factor are summarized in table. 13 and 14 for genetically modified reporter T cells co-incubated with HEK293/hCD20 or HEK293/hCD20/hPSMA cells in addition to a constant concentration of 50 pM isotype control hIgG4 or anti-CD3xCD20 bispecific antibody (T cell stimulating bispecific antibody).

Table 13

Luciferase activity in genetically modified reporter T cells in the absence of TCR stimulation

Antibodies HEK293/hCD20 HEK293/hCD20/hPSMA ec ₅₀ [M] Multiplicity ec ₅₀ [M] Multiplicity induction induction bsl6429D - 0.81 - 1.04 bsl6430D - 0.87 - 0.89 bsl6431D - 0.94 - 0.98 shАБ14226Р2 5.11 E-09 4.76 4.90 E-09 3.59 (CD28) shАБ14193Р2 - 0.86 - 0.89 (CD28) shАБ14216Р2 η/s (not 2.80 p/s 3.95 (CD28) calculated ) One shoulder - 0.83 - 0.88 tAL14226P2 One shoulder - 0.88 - 0.86 shАБ14193Р2 One shoulder - 0.89 - 0.91 shАБ14216Р2

In table 14 summarizes the results of EC ₅₀ values and fold induction for luciferase activity in genetically modified T cells co-incubated with HEK293/hCD20 or HEK293/hCD20/hPSMA cells and a constant concentration of 50 pM isotype control hIgG4.

- 42 045943

Table 14

Luciferase activity in genetically modified reporter T cells under TCR stimulation

Antibodies HEK293/hCD20 HEK293/hCD20/hPSMA ec ₅₀ [M] Induction ratio ec ₅₀ [M] Induction ratio bsl6429D - 1.00 1.72 E-10 16.15 bsl6430D - 0.92 4.57 E-10 13.61 bsl6431D - 0.95 2.78 E-10 24.47 wAL14226P2 (CD28) 3.59 E-09 3.16 3.27 E-09 3.03 wAL14193P2 (CD28) 2.23 E-08 1.27 - 1.37 wAL14216P2 (CD28) p/s 3.26 p/s 3.15 One shoulder shАБ14226Р2 - 0.99 - 0.95 One shoulder shАБ14193Р2 2.24 E-08 1,Yu - 1.15 One shoulder shАБ14216Р2 - 0.96 - 0.94

In table 14 summarizes the results of EC50 values and fold induction for luciferase activity in genetically modified T cells co-incubated with HEK293/hCD20 or HEK293/hCD20/hPSMA cells and a constant concentration of 50 pM anti-CD3xCD20.

When T cells and APCs were treated with 50 pM isotype control hIgG4, none of the anti-CD28 bispecific antibodies showed an increase in luciferase activity in the absence of TCR stimulation, regardless of the APC line used in the assay. Little luciferase activation was observed with one of the parental anti-CD28 antibodies (mAb14226P2) on HEK293/hCD20 cells (4.76x) and HEK293/hCD20/hSPMA cells (3.59x), as shown in Table 1. 14.

In contrast, when cells were treated with 50 pM of the bispecific antibody to CD3xCD20, all three bispecific antibodies to PSMAxCD28, bs16429D, bs16430D, and bs16431D, highly induced luciferase activity when co-incubated with APCs expressing hPSMA on their surface. Regardless of the APC line, virtually no activation was observed in single-arm controls (mAb14226P2 single arm, mAb14193P2, and mAb14216P2). Mild luciferase activation was observed for all three parental anti-CD28 antibodies (mAb14226P2, mAb14193P2, and mAb14216P2), as shown in Table 1. 14.

2) Functional analysis of IL-2 using primary human CD4+ T cells.

The ability of PSMAxCD28 bispecific antibodies to provide co-stimulation through CD28 on T cells in the absence or presence of PSMA target expression was assessed in a functional assay of primary CD4 T cells, measuring IL-2 cytokine production.

The values of EC ₅₀ and induction ratio are summarized in table. 15 for CD4 T cells co-incubated with HEK293/hCD20 or HEK293/hCD20/hPSMA cells in addition to a constant concentration of 50 pM isotype control hIgG4 or anti-CD3xCD20 bispecific antibody (T cell stimulating bispecific antibody).

As expected, no measurable IL-2 release was observed in wells containing constant amounts of the hIgG4 isotype control, since there was no primary stimulation of T cells.

In contrast, measurable levels of IL-2 were detected in samples treated with the CD3xCD20 bispecific antibody. Under these conditions, when human CD4 T cells were co-incubated with HEK293/hCD20 cells, all anti-CD28 monoclonal antibodies tested, including anti-CD28 and anti-PSMA x CD28 bispecific antibodies, with the exception of bs16430D and bs16431D, demonstrated increased levels of IL-2 (Table 1). 15). The parent antibody mAb14226P2 showed the highest fold induction and _EC50 of approximately 6 nM. IL-2 release was detected for all three anti-PSMAxCD28 bispecific antibodies (bs16429D, bs16430D, and bs16431D) when co-administered

- 43 045943 culturing CD4+ T cells with APCs expressing hPSMA and a bispecific antibody to CD3xCD20. Lower levels of IL-2 were measured for their single arm controls and parental antibodies under the same conditions as in Table 1. 15.

Table 15

IL-2 production by primary CD4+ T cells under TCR stimulation

Antibodies HEK293/hCD20 HEK293/hCD20/hPSMA ec ₅₀ [M] Multiplicity ec ₅₀ [M] Multiplicity induction induction bsl6429D p/s 22 0.27 E-09 606 bsl6430D - 1 0.15 E-09 165 bsl6431D - 1 0.22 E-09 463 SHAL4226R2 (CD28) 6.05 E-09 110 8.69 E-09 46 SHAL4193R2 (CD28) p/s 27 p/s 13 SHAL4216R2 (CD28) p/s 13 p/s AND One shoulder SHAL4226R2 p/s 21 p/s 14 One shoulder SHAL4193R2 13.78 E-09 21 14.52 E-09 13 One shoulder SHAL4216R2 p/s 9 p/s 6

In table 15 summarizes the results of EC ₅₀ values and fold induction for IL-2 production by primary CD4+ T cells co-incubated with HEK293/hCD20 or HEK293/hCD20/hPSMA cells and a constant concentration of 50 pM anti-CD3xCD20 antibody.

Example 7 Bispecific antibodies to PSMAxCD28 enhance T cell activation in the simultaneous presence of both PSMA and TCR stimulation with anti-CD20xCD3.

Exemplary bispecific antibodies to PSMAxCD28 were generated and tested by the Biacore assay (see, for example, Examples 1 and 4). A pair of bispecific antibodies (CD20xCD3 and exemplary PSMAxCD28 bispecific antibodies of the invention) were used to induce clustering of T cells and co-stimulatory receptors on tumor cells by binding to PSMA (CD20 and prostate-specific membrane tumor antigen - PSMA) (Fig. 5A). To demonstrate that PSMAxCD28 binds and activates CD28 in the presence of PSMA expressed on target cells and TCR activation, a series of in vitro cell-based assays were performed.

Localization of CD28.

Cell lines.

The HEK293 stable cell line (ATCC, no. CRL-1573) was used to generate the HEK293/hCD20, HEK293/hPSMA, and HEK293/hCD20/hPSMA cell lines. To generate the HEK293/hPSMA cell line, stable transfection was performed using a mammalian expression vector encoding ubiquitin-C promoter-driven PSMA (amino acids M1 to A750 under accession number Q04609) and a neomycin resistance gene. Similarly, the HEK293/hCD20 cell line was generated using a mammalian cell expression vector encoding the ubiquitin promoter-driven hCD20 (amino acids M1 to P297 under accession number NP_068769.2). Transfected cells were cultured in 500 μg/ml geneticin A to select stably expressing cell lines.

To generate HEK293/hCD20/hPSMA cells, a lentiviral plasmid encoding human PSMA (amino acids M1 to A750 under accession number Q04609) and a neomycin resistance gene was used to transfect HEK293T cells, producing viral particles that were subsequently used to infect HEK293 cells /hCD20. Human PSMA-positive cells were isolated using fluorescence-activated cell sorting (FACS). All resulting cell lines were maintained in DMEM+10% FBS+P/S/G+NEAA supplemented with 500 μg/ml G418. Clone Jurkat E6-1 (ATCC, no. TIB-152) was cultivated in accordance with the recommended

- 44 045943 with the new ATCC protocol.

To generate MC38/hPSMA cells, a lentiviral plasmid encoding human PSMA (amino acids M1 to A750 under accession number Q04609) and a neomycin resistance gene was used to transfect HEK293T cells, producing viral particles that were subsequently used to infect parental MC38 cells. Human PSMA-positive cells were isolated using FACS. MC38/hPMA was maintained in DMEM+10% FBS+P/S/G+NEAA supplemented with 500 μg/ml G418.

Amnis Image stream.

Jurkat T cells and target cells (HEK293/hPSMA or HEK293/hPSMA/hCD20) were incubated with CD20xCD3-Alexa488 (REGN2280, 0.5 μg/ml), alone or together with PSMAxCD28-Alexa647 (bs16429D, 1 μg/ml) , for 1 hour at 37°C. Cells were gently washed twice with FACS buffer (3% FBS, 2 mM EDTA in PBS) and stained with anti-CD28 antibody with biotin (REGN1412, 2 μg/ml) for 15 min at 4°C, followed by streptavidin-PE-CF594 ( BD 562284, 1 μg/ml) and Hoechst 33342 (Thermo Fisher H3570, 1 μM) for 15 min at 4°C. Cells were washed with PBS and stored in BD stabilizing fixative (BD 338036). Cell images were acquired on an Amnis® imaging flow cytometer and analyzed using IDEAS® software. Cells were gated for brightfield doublet, doublet nucleus, nuclear focus, single spot count, CD28 singlet. The area of synapses was determined by applying a Valley mask based on nuclear staining. Cells with the incorrect Valley mask were removed by gating across the overlap region between the Valley mask and CD28. The ratio of CD28 in synapse/out of synapse was calculated using the following formula:

CD28 in/out of synapse = CD28 intensity in synapse/(total CD28 intensity - CD28 intensity in synapse) x 100%.

Results, summary and conclusions.

T cells were co-cultured with genetically modified target cells (HEK293/hPSMA or HEK293/hCD20/hPSMA) overexpressing CD20 and PSMA, and fluorescently labeled bispecifics (CD20xCD3 in green, PSMAxCD28 in red). To determine the localization of CD28, cells were fixed and stained with anti-CD28 antibody after incubation for 1 h at 37°C. Images of T cell and target cell conjugates were acquired using an Amnis ImageStream imaging flow cytometer. In the absence of PSMA expression on target cells, the CD20xCD3 bispecific alone did not induce CD28 clustering on T cells. When PSMA was expressed on target cells, CD20xCD3 localized at the interface between T cell and target cell conjugates and formed an immunological synapse (IS) where CD28 was localized. PSMAxCD28 together with CD20xCD3 further enhanced the accumulation of CD28 in the IS. CD28 distribution was quantified by calculating the ratio of CD28 staining within the IS to outside the IS ( Fig. 5B ).

It was concluded that PSMAxCD28, in the presence of PSMAxCD3 and target cells expressing PSMA, causes a sustained accumulation of CD28 in the IS, the site where signaling for T cell activation occurs.

Release of cytokines.

To further examine the effect of TAAxCD28 on T cell activation, T cell proliferation and release of the cytokines IL-2 and IFNy were measured after incubation with titrated doses of the bispecific CD20xCD3 or PSMAxCD28 in co-culture with primary human T cells and genetically modified target cells overexpressing various TAAs ( HEK293/hPSMA, HEK293/hCD20, or HEK293/hCD20/hPSMA) (Figures 5C-5G). It was confirmed that the exemplary anti-PSMAxCD28 antibodies of the invention and the bispecific anti-PSMAxCD3 antibodies do not compete for binding to cells expressing PSMA and, therefore, bind to different epitopes (Fig. 5H). In addition, it was confirmed that PSMAxCD28 induced T cell activation under conditions of simultaneous CD3 stimulation and PSMA expression on target cells (HEK293/hCD20/hPSMA plus 5 pM CD20xCD3). T cell activation by exemplary PSMAxCD28 bispecific antibodies was not observed in the absence of PSMA expression on target cells (coculture with HEK293/hCD20) or in the absence of CD3 stimulation (coculture with HEK293/PSMA plus 5 pM CD20xCD3 or HEK293/hCD20/hPSMA plus 5 pM isotopic control) (Fig. 5E-5G). In general, exemplary anti-PSMAxCD28 antibodies of the invention have been demonstrated to stimulate T cell activation in the presence of PSMAxCD3 and PSMA-expressing target cells, resulting in increased proliferation and cytokine secretion.

Example 8 Destruction of cells expressing PSMA.

Two FACS-based cytotoxicity studies were conducted. The first study examined FACS-based cytotoxicity on C4-2 cells in the presence of human peripheral blood mononuclear cells (PBMCs) with or without anti-PSMAxCD28 stimulation. Stimulation was performed in the presence of a fixed concentration of bispecific antibody to PSMAxCD28 and serial dilutions of bispecific antibody to PSMAxCD3. WTO

- 45 045943 This study was identical to the first, except that cynomolgus monkey PBMCs were used instead of human PBMCs. It was concluded that bispecific antibodies to PSMAxCD28 enhance T cell activation and cytotoxicity against prostate tumor cells when TCR is stimulated by PSMAxCD3.

Experimental techniques.

Isolation of primary human CD4+ T cells.

Human peripheral blood mononuclear cells (PBMCs) were isolated from leukocyte concentrate obtained from a healthy donor. PBMCs were isolated by density gradient centrifugation using 50 mL SepMate™ tubes according to the manufacturer's protocol. CD4+ T cells were then isolated from PBMCs using the EasySep™ Human CD4+ T Cell Isolation Kit from StemCell Technologies following the instructions recommended by the manufacturer. Isolated CD4' T cells were frozen in FBS containing 10% DMSO at a concentration of 50 x 10 ⁶ cells per vial.

Primary human T cell activation assay.

Previously isolated and frozen human CD4+ T cells were thawed on the day of analysis in stimulating medium (X-VIVO 15 cell culture medium supplemented with 10% FBS, HEPES, NaPyr, NEAA and 0.01 μM BME) containing 50 U/ml benzonase nuclease. Cells were centrifuged at 1200 rpm for 10 min, resuspended in stimulation medium and seeded into 96-well round-bottom plates at a concentration of 1 x 10 ⁵ cells per well. HEK293 cells (HEK293/hPSMA, HEK293/hCD20, or HEK293/hPSMA/hCD20) were treated with 15 μg/ml mitomycin C in priming stimulation medium at a concentration of 10 x 10 ⁶ cells/ml. After incubation for 1 hour at 37°C, 5% CO ₂ , HEK293 cells were washed 3 times with D-PBS containing 2% FBS and added to wells containing CD4+ T cells at a final concentration of 2 x 10 ⁴ cells per well.

To determine the suboptimal concentration of CD20xCD3 bispecific antibody for T cell activation, CD20xCD3 was serially diluted 1:3 ranging from 1.5 pM to 10 nM in the presence of 500 pM PSMAxCD28 bispecific or hIgG4s isotype control. A constant concentration of 5 μM bispecific CD20xCD3 or control hIgG4 was chosen, and the following antibodies: 1) PSMAxCD28; 2) control non-targeting x CD28; 3) parental CD28; 4) CD28SA (superagonist); 5) hIgG4 isotype control and 6) hIgG4s isotype control, titrated from 15 pM to 100 nM with a 1:3 dilution. The 10-point dilution endpoint contained no titratable antibodies, only 5 pM CD20xCD3 bispecific or control hIgG4.

After incubating the plates for 48 h at 37°C, 5% CO ₂ , they were centrifuged to pellet the cells and 50 μl of supernatant medium was collected. From this volume, 5 μl was tested in the human IL-2 and human IFNy assay by AlphaLISA according to the manufacturer's protocol. Measurements were made on a Perkin Elmer Envision multichannel tablet analyzer. A calibration curve of known concentrations of IL-2 or IFNy was constructed to extrapolate the pg/ml IL-2 or IFNy generated in the assay wells. All serial dilutions were tested in duplicate. Pelleted cells were incubated with [methyl-3H]-thymidine, 0.25 µCi/well for 16 h at 37°C, 5% CO ₂ . Cells were harvested onto Perkin Elmer Unifilter plates using the Unifilter 96 Cell Harvester from Perkin Elmer. After adding 30 μl of scintillation fluid, the plates were sealed and the counts per minute were recorded for each well using TopCount NXT from Perkin Elmer.

EC ₅₀ antibody values were determined from a four-parameter logistic equation on a 10-point dose-response curve using GraphPad Prism™ software.

FACS-based cytotoxicity assay.

To monitor the killing of PSMA+ cells in the presence of a combination of anti-PSMAxCD3 and anti-PSMAxCD28 antibodies, C4-2 cells were labeled with 1 μM fluorescent tracking dye Violet Cell Tracker (Invitrogen, cat. no. 34557). After labeling, cells were plated overnight at 37°C. Separately, human PBMCs (New York Blood Center) or cynomolgus PBMCs (Covance, Cranford, NJ) were plated in supplemented RPMI medium at a concentration of 1 x ¹⁰⁶ cells/ml and incubated overnight at 37°C to enrich in lymphocytes at the expense of adhesive macrophages, dendritic cells and some monocytes. The next day, target cells were co-incubated with adherent cell-depleted naive PBMCs (effector/target cell ratio 4:1) and serial dilutions of PSMAxCD3 bispecific antibody or IgG4 control (concentration range: 0.42 nM - 0.1 pM), according to alone or in combination with a fixed concentration of the costimulatory molecules to PSMAxCD28, bs16429D or bs16431D, at 2.5 μg/ml (16.7 nM) for 96 h at 37°C.

After incubation, cells were removed from cell culture plates using trypsin-EDTA dissociation buffer (Millipore, cat. no. SM-2004-C) and analyzed by FACS on a FACS BD LSRFortessa-X20 (BD).

For FACS analysis, cells were stained using cell tracker far-red dye.

- 46 045943 for viability determination (Invitrogen) and directly conjugated antibodies to CD2, CD4, CD8 and CD25 (BD). Samples were analyzed with calibration microspheres for cell counting. To assess specificity of killing, target cells were gated as populations positive for the Violet cell tracker. The percentage of living target cells was calculated as follows: percentage (%) of viable cells = (R1/R2)x100, where RI^po^ht (%) of living target cells in the presence of antibody, and K2 = percentage (%) of living cells - targets in the absence of the test antibody. T cell activation was measured as the percentage of activated (CD25+) T cells from CD2+/CD4+ or CD2+/CD8+ T cells. T cell numbers were measured by calculating the number of live CD4+ or CD8+ cells per calibration microsphere.

The levels of cytokines accumulated in the medium were analyzed using the human Th1/Th2/Th17 Cytokine Assay Flow Cytometry Bead Kit, BD cytometric Bead Array (CBA) human Th1/Th2/Th17 Cytokine kit, according to the manufacturer's protocol.

For FACS analysis, cells were stained with near-infrared viability assay dye (Invitrogen, cat. no. L34976). Five hundred thousand (5x105) calibration microspheres were added to each well immediately prior to FACS analysis. One hundred thousand (1x105) microspheres were collected for each sample. To assess killing specificity, cells were gated on Violet-labeled living populations. The percentage of the population alive was recorded and used to calculate survival.

T cell activation and PD-1 up-regulation were assessed by incubating cells with directly conjugated antibodies to CD2, CD4, CD8, CD25 and PD-1, and by recording the percentage of late activated (CD25+/CD8+) T cells and PD- 1+/CD4+ T cells out of total T cells (CD2+). Information about directly conjugated antibodies is as follows: CD2, PE:CD2 (C1:RPA-2.1), BD, cat. No. 555327; CD4, PerCP-Cy5.5:CD4 (Cl:OKT-4), Biolegend, cat. No. 317428; CD8, APC:CD8 (C1:RPA-T8), Biolegend, cat. No. 301049; CD25, BV510:CD25 (Cl:M-A251), BD, cat. No. 563352; and PD-1, PE-Cy7:PDl (Cl:EH12.2H7), Biolegend, cat. No. 329918.

The supernatant of the assay wells from the human PBMC assay was assessed for Th1/Th2 cytokine release using the BD cytometric bead array human kit (BD, cat. no. 560484) and following the manufacturer's protocol.

Results, summary and conclusions.

In fig. 6A and 6B show that anti-PSMAxCD28 enhanced the cytotoxic activity of bs13644D (anti-PSMAxCD3) in the presence of human or cynomolgus T cells. The anti-PSMAxCD3 bispecific antibody was tested for its ability to induce killing of human PSMA-expressing target cells by naïve human T cells, either as a single agent or in the presence of a co-stimulatory anti-PSMAxCD28 bispecific antibody. In the study shown in FIG. 6A and 6B, the concentration of anti-PSMAxCD28 antibody was fixed at 2.5 μg/ml. Antu-PSMAxCD3 was serially diluted 1:4. Antibodies were incubated with cells for 96 hours. Table. 16 summarizes the components of the antibodies used in this study.

Table 16

Components of bispecific antibodies in cytotoxic efficacy studies

Antibody Product ID Antibody format Shoulder PSMA [VH] Shoulder CD28 [VH] Light chain bsl6429D PSMAxCD28 mAbll838P2 mAbl4226P2 3-20 ULC bsl6431D PSMAxCD28 mAb!1810P2 mAb!4216P2 1-39 ULC bsl3644D PSMA (mABl 1835P2)xCD3 (7221G) 1-39 ULC bs!7664D Control EGFRvIIIxCD3(7221G) 1-39 ULC

An exemplary anti-PSMAxCD28 bispecific antibody was tested to enhance T cell activation and cytotoxicity against prostate cancer cells by targeting PSMA. An exemplary bispecific antibody to PSMAxCD28 in combination with PSMAxCD3 was tested using FACS-based cytotoxicity and T cell activation assays (Figures 6, 7A-7D). Human PBMCs containing T cells were cocultured with C4-2 prostate cancer cells expressing endogenously high levels of PSMA (data not shown). PSMAxCD28 significantly increased the efficiency of cytotoxicity induced by PSMAxCD3 alone, shifting the EC50 from 4.3x10'11 to 1.5x10-1 (a shift in efficiency of more than a logarithm) (Figs. 6 and 7A). Consistent with the induction of T cell cytotoxicity, PSMAxCD28 increased the levels of PSMAxCD3-induced IFNy release by 4-fold (Fig. 7B). Similarly, the combination of PSMAxCD28 and PSMAxCD3 increased the number of CD4 and CD8 T cells and the expression of the activation marker CD25 (Figures 7C and 7D). No effect on cytotoxicity or T cell activation was observed when exemplary PSMAxCD28 bispecific antibodies were used in combination with nontargeting

- 47 045943 bispecific antibody to CD3. Additionally, to demonstrate that an exemplary anti-PSMAxCD28 bispecific antibody can enhance PSMAxCD3 cytotoxicity and T cell activation in non-human primates, the same assays were performed using PBMCs from cynomolgus monkeys with similar results (Figures 6, 7E-7G). These results demonstrated that the exemplary PSMAxCD28 bispecific antibody can effectively enhance PSMAxCD3-mediated T cell activation not only through proliferation and cytokine release, but also through cytotoxicity. In fig. 7H shows that an exemplary anti-PSMAxCD28 bispecific antibody binds to cellular targets as determined by flow cytometry analysis.

The anti-PSMA x CD3 bispecific antibody activated and directed human T cells to reduce the pool of C4-2 cells (a human prostate cancer cell line). Target cell killing was monitored in the presence of a bispecific antibody to PSMAxCD3, and C4-2 cells were killed in a dose-dependent manner with EC50 values in the picomolar range (Table 17). The observed target cell lysis was associated with up-regulation of CD25+ and PD-1+ cells on CD2+ T cells, again with EC ₅₀ values in the picomolar range (Table 17). Bispecific antibody to PSMAxCD3 induced the release of human cytokines. The cytotoxic activity observed for the PSMAxCD3 bispecific antibody when used as a single agent was enhanced in the presence of a fixed concentration of PSMAxCD28 co-stimulatory molecules (bispecific antibodies) (Table 17).

In summary, costimulation with PSMAxCD28 bispecific antibody increased T cell activation, PD-1 upregulation, and cytokine release compared with observations made with PSMAxCD3 bispecific antibody as a single agent. In table 16 summarizes the results of experiments using human PBMCs.

Table 17

Effect of anti-PSMAxCD28 on the cytotoxicity of anti-PSMAxCD3 against C4-2 cells in the presence of human PBMCs

Antibody EC50 destruction of C4-2 [Μ] EC50 activation of T cells [Μ] (CD8+/CD25+) EC ₅₀ up-regulation of PD-1 [M] (CD4+/PD1+) Anti-R SM AxCD3 4.35Ε-11 7.53Ε-11 3, OZE-11 Anti-R SM AxCD3 +bsl6429D 1.45Ε-12 4.95Ε-12 3.71E-12 Anti-R SM AxCD3 +bsl6431D 4.88Ε-12 6.44Ε-12 4.38E-12

The anti-PSMAxCD3 bispecific antibody was also tested for its ability to induce killing of human PSMA-expressing target cells by naïve cynomolgus T cells, either as a single agent or in the presence of a co-stimulatory anti-PSMAxCD28 bispecific antibody. At the selected antibody titer, the PSMAxCD3 bispecific antibody activated human T cells but did not direct T cells to reduce the C4-2 cell pool (Table 18). Costimulation with anti-PSMAxCD28 antibody resulted in increased T cell activation, increased cytotoxic activity, and up-regulation of the PD-1 marker on T cells (Table 18).

Table 18

Effect of anti-PSMAxCD28 on the cytotoxicity of anti-PSMAxCD3 against C4-2 cells in the presence of cynomolgus monkey PBMCs

Antibody EU50 destruction of S4-2 [M] EC50 T cell activation [M] (CD8+/CD25+) EC50 up-regulation of PD-1 [M]

- 48 045943

(CD4+/PD1+) Anti-R SM AxCD3 n/a l,71E-10 2.29E-10 Anti-R SM AxCD3 +bsl6429D 1.55E-11 1.64E-11 5.64E-12 Anti-R SM AxCD3 +bs!6431D 7.72E-11 2.76E-11 2.66E-11

Example 9 Release of cytokines from cells.

As noted elsewhere herein, ex vivo cytokine release assays using human PBMCs in a soluble format failed to predict cytokine release. Covered analysis formats have been developed for this purpose.

The anti-PSMA/CD28 bispecific antibodies and anti-CD28 antibodies of the present invention were evaluated for their ability to induce human PBMC proliferation and cytokine release from cells using a coated assay format. To analyze PBMC proliferation, freshly isolated PBMCs were used in an amount of ^1x105 /well. For co-culture, C4-2 cells treated with mitomycin C were added in an amount of 1 x 10 ⁴ well. Antibody dilutions were applied to plates either wet or dry overnight. The plates were washed before adding PBMCs. Supernatants were collected at approximately 54 h for cytokine analysis using Meso Scale Diagnostics (MSD, Rockville, MD). ³ N thymidine was added for 18 h and proliferation was measured.

The results showed that a soluble bispecific antibody to PSMAxCD28 induced proliferation of human PBMCs in the presence of CD3 stimulus. In contrast, a CD28 superagonist induced human PBMC proliferation in the absence of CD3 stimulus (data not shown).

In addition, wet-applied bs16429D (PSMAxCD28 A) and mAb14226P2 (CD28 A parent) induced less cytokine release than wet-applied CD28 superagonist, whereas bs16431D (PSMAxCD28 B) and mAb14216P2 (CD28 B parent) did not induce significantly greater cytokine release compared with controls (data not shown).

The anti-CD28 antibody of the present invention exhibits several properties that differ from the superagonist TGN1412. Conventional agonistic CD28 monoclonal antibodies (non-superagonists) are thought to bind to membrane-distal epitopes, providing only monovalent antibody binding (Dennehy et al., Cutting Edge: monovalency of CD28 maintains the antigen dependence of T cell costimulatory responses, J. of Immunol 176(10): 5725-29 (2006)). Antibodies to CD28 drive optimal T cell activation in the presence of antigen-specific peptide-MHC recognition by the TCR. CD28 signaling regulates the TCR activation threshold and significantly reduces the number of TCR interactions required for efficient T cell activation. In contrast, the anti-CD28 TGN antibody had some superagonist properties (Luhder et al., Topological requirements and signaling properties of T cell-activating, anti-CD28 antibody superagonists, J. of Exp. Med. 197(8): 955-966 (2003 )); Riley et al., the CD28 family: a T-cell rheostat for therapeutic control of Tcell activation, Blood, 105(1): 13-21 (2005). For example, anti-CD28 TGN induced potent T cell proliferation and IL2 production in vitro and in vivo even in the absence of TCR signaling, enhanced NF-κB activity more effectively than conventional anti-CD28/CD3 signaling, and induced AP activation. 1/SRE, and bound to CD28 epitopes near the cell surface, allowing for divalent binding.

Accordingly, as shown in FIG. 8, the CD28 superagonist TGN 1412 activated the AP1 reporter in the absence of primary CD3 stimulation. In contrast, the anti-CD28 antibody of the present invention (the parent antibody of PSMAxCD28) minimally activated AP-1 in the AP-1 luciferase reporter bioassay.

Example 10. In vivo study of anti-PSMAxCD28 antibody.

Syngeneic tumor model.

Consistent with the in vitro study in Example 8, the combination of tumor antigen-targeted bispecific antibodies to CD3xPSMA and CD28xPSMA enhanced tumor clearance in a mouse model. To test the effectiveness of the anti-PSMAxCD28 bispecific antibody as monotherapy or in combination with the anti-PSMAxCD3 bispecific antibody, experiments were performed with syngeneic tumors in mice expressing human CD28, human CD3, and human PSMA instead of the corresponding mouse genes generated using Velocigene's proprietary technology (hCD3 mice /hCD28/hPSMA). It was concluded that PSMAxCD28 bispecific enhances antitumor immunity in vivo through anti-PSMAxCD3-induced T cell activation.

T cell activation is initiated upon binding of the T cell receptor (TCR)/CD3 complex to peptide-MHC complexes (signal 1); then activation is enhanced by the involvement of the second

- 49 045943 costimulatory receptor, such as the CD28 receptor on T cells, binding to its cognate ligand(s) on the target cell (signal 2). Recently described CD3-based bispecific antibodies act by replacing conventional signal 1, linking T cells to tumor cells by binding tumor-specific antigen (TSA) to one arm of the bispecific antibody and coupling via the other arm to TCR/CD3. Although some of these TSAxCD3 bispecifics have demonstrated promising antitumor efficacy in cancer patients, their activity remains to be optimized. As described elsewhere herein, the present invention introduces a new class of bispecific antibodies that mimic signal 2 by coupling a second TSA to the costimulatory receptor CD28 on T cells. These bispecific antibodies are called TSAxCD28 bispecifics or TSA/CD28 bispecifics. As described herein, one exemplary antibody of the present invention is specific for prostate cancer antigens (eg, PSMA). Unlike CD28 superagonists, which activate a broad range of T cells and in some cases resulted in significant toxicity in early clinical trials, these TSAxCD28 bispecifics exhibit limited activity and are not toxic when used alone in genetically humanized immunocompetent mouse or primate models . However, when combined with anti-TSAxCD3 bispecific antibodies, an exemplary antibody of the invention increased artificial synapse between a T cell and its target cell, enhanced T cell activation, and markedly improved the antitumor activity of CD3 bispecifics in syngeneic tumor models. Combining this new class of CD28 co-stimulatory bispecific antibodies with the emerging TSAxCD3 class of bispecific antibodies may provide well-tolerated, off-the-shelf antibody therapy options with potentially enhanced antitumor efficacy.

The ability of T cells to recognize and destroy their cellular targets, such as virus-infected cells or tumor cells, depends on a coordinated set of interactions. Primarily, they involve recognition and binding of the target cell by the TCR complex (which includes the associated γ-, δ-, ε-, ζ-chains of CD3); this interaction has been termed signal 1 for T cell activation. The TCR can recognize viral or tumor peptide presented in the groove of MHC proteins expressed on the surface of target cells. This binding is usually of low affinity; Therefore, for signal 1 to be successfully triggered, it is important to have many TCR complexes clustered along the interface between a T cell and its target cell, and this interface has been called the immune synapse (J. B. Huppa, M. M. Davis, T-cell-antigen recognition and the immunological synapse. Nat Rev Immunol 3, 973-983 (2003). T cell activation and proliferation are then further stimulated by additional interactions with costimulatory receptors such as CD28 (signal 2) (J.H. Esensten, Y.A. Helou, G. Chopra, A. Weiss, J.A. Bluestone, CD28 Costimulation: From Mechanism to Therapy. Immunity 44 , 973-988 (2016)). When a T cell recognizes a target cell through the TCR complex and engages signal 2 through the binding of CD28 to its cognate ligand(s) (CD80/B7.1 and/or CD86/B7.2) on a professional antigen presenting cell or target cell, activation T cells are increased. As with signal 1, CD28-mediated signal 2 likely arises from co-clustering at the immune synapse.

Conventional monoclonal antibodies targeting tumor-specific antigens (TSAs) have been used as anticancer therapeutics over the past two decades (G. Salles et al., Rituximab in B-Cell Hematologic Malignancies: A Review of 20 Years of Clinical Experience. Adv Ther 34, 2232-2273 (2017); M. V. Mateos et al., Daratumumab plus Bortezomib, Melphalan, and Prednisone for Untreated Myeloma. N Engl J Med 378, 518-528 (2018): W. Eiermann, G. International Herceptin Study, Trastuzumab combined with chemotherapy for the treatment of HER2-positive metastatic breast cancer: pivotal trial data. Ann Oncol 12 Suppl 1, S57-62 (2001); J. M. Connors et al., Brentuximab Vedotin with Chemotherapy for Stage III or IV Hodgkin's Lymphoma. N Engl J Med 378, 331-344 (2018); V. Dieras et al., Trastuzumab emtansine versus capecitabine plus lapatinib in patients with previously treated HER2positive advanced breast cancer (EMILIA): a descriptive analysis of final overall survival results from a randomized , open-label, phase 3 trial. Lancet Oncol 18, 732-742 (2017)). However, this class of antibodies had a limited ability to induce T cell-mediated cytotoxicity, and instead acted by stimulating antibody-dependent cellular cytotoxicity (ADCC) and/or complement-dependent cytotoxicity (CDC), or by delivering the toxin to tumor cells. Recently, a new class of bispecific antibodies (TSAxCD3) has emerged that can effectively trigger T cell-mediated killing of tumor cells by binding the T cell to the tumor cell and activating the CD3/TCR complex (usually via the CD3 epsilon chain) through a replacement mechanism, thereby mimicking signal 1. An early version of this bispecific (one arm binds to CD19 on leukemia cells and the other binds to CD3) recently received regulatory approval for the treatment of B-cell acute lymphoblastic leukemia (R. Bargou et al., Tumor regression in cancer patients by very low doses of a T cell attracting antibody. Science 321, 974-977 (2008); H. Kantarjian et al., Blinatumomab versus Chemotherapy for Advanced Acute Lymphoblastic Leukemia. N Engl J Med 376, 836-847 (2017)). Recently, improved versions of bispecifics have been shown to have good activity against

- 50 045943 Zhkin lymphomas, targeting CD20 to these lymphomas (E. J. Smith et al., A novel, native-format bispecific antibody triggering T-cell killing of Bcells is robustly active in mouse tumor models and cynomolgus monkeys. Sci Rep 5 , 17943 (2015); L. L. Sun et al., Anti-CD20/CD3 T cell-dependent bispecific antibody for the treatment of B cell malignancies. Sci TranslMed 7, 287ra270 (2015); M. Bacac et al., CD20-TCB with Obinutuzumab Pretreatment as Next-Generation Treatment of Hematologic Malignancies. Clin Cancer Res 24, 4785-4797 (2018); R. Bannerji et al., Emerging Clinical Activity of REGN1979, an Anti-CD20 x Anti-CD3 Bispecific Antibody, in Patients with Relapsed/Refractory Follicular Lymphoma (FL), Diffuse Large B-Cell Lymphoma (DLBCL), and Other B-Cell Non-Hodgkin Lymphoma (B-NHL) Subtypes American Society of Hematology, (2018); L. Budde et al ., Mosunetuzumab, a Full-Length Bispecific CD20/CD3 Antibody, Displays Clinical Activity in Relapsed/Refractory B-Cell Non-Hodgkin Lymphoma (NHL): Interim Safety and Efficacy Results from a Phase 1 Study. American Society of Hematology, (2018)). However, although TSAxCD3 bispecifics have emerged as an important new class of immunotherapies for hematological malignancies, cross-study comparisons (E. A. Zhukovsky, R. J. Morse, M. V. Maus, Bispecific antibodies and CARs: generalized immunotherapeutics harnessing T cell redirection. Curr Opin Immunol 40, 24 -35 (2016)) suggest that in some cases they may not achieve the level of efficacy observed for personalized chimeric antigen receptor T-cell (CAR-T) therapy.

One reason for the high efficacy of CAR-T cell therapy is that the chimeric antigen receptor (CAR) is genetically modified to provide both signal 1 (through part of the CD3z cytodomain) and signal 2 (for example, through part of the CD28 cytodomain) upon binding with its target on the tumor cell. Two CAR-T cell therapies have recently received FDA approval for the treatment of B-cell malignancies, both of which act by binding to and targeting the CD19 antigen (S. S. Neelapu et al ., Axicabtagene Ciloleucel CAR T-Cell Therapy in Refractory Large B Cell Lymphoma. N Engl J Med 377, 2531-2544 (2017); S. J. Schuster et al., Chimeric Antigen Receptor T Cells in Refractory B-Cell Lymphoma. N Engl J Med 377, 2545-2554 (2017)). CAR-T cell approaches may be associated with serious side effects such as cytokine release syndrome (CRS) and neurotoxicity (S. S. Neelapu et al., Chimeric antigen receptor T-cell therapy - assessment and management of toxicities. Nat Rev Clin Oncol 15, 47-62 (2018); J. Gust et al., Endothelial Activation and Blood-Brain Barrier Disruption in Neurotoxicity after Adoptive Immunotherapy with CD19 CAR-T Cells. Cancer Discov 7, 1404-1419 (2017); A. Shimabukuro-Vornhagen et al., Cytokine release syndrome. J Immunother Cancer 6, 56 (2018)); and due to highly personalized manufacturing processes and the need for chemotherapy pretreatment (S. S. Neelapu et al., Axicabtagene Ciloleucel CAR T-Cell Therapy in Refractory Large B Cell Lymphoma. N Engl J Med 377, 25312544 (2017); S. J. Schuster et al ., Chimeric Antigen Receptor T Cells in Refractory B-Cell Lymphomas. N Engl J Med 377, 2545-2554 (2017); P. Salmikangas, N. Kinsella, P. Chamberlain, Chimeric Antigen Receptor T-Cells (CART-Cells) for Cancer Immunotherapy - Moving Target for Industry? Pharm Res 35, 152 (2018)) many patients are not considered suitable candidates.

The benefits of TSAxCD3 bispecifics as relatively well-tolerated and ready-to-use therapeutic solutions for broader patient populations would be enhanced if their antitumor activity could be further optimized, especially if this could be done without compromising tolerability or perhaps even increasing specificity to tumor cells relative to healthy cells. To this end, it was hypothesized that using TSAxCD3 bispecifics in tandem with a new class of bispecifics that independently activate signal 2 could provide potentially increased efficacy as well as the potential for increased specificity. Therefore, a second class of bispecifics was developed. These bispecifics may engage either a second epitope on the same tumor-specific antigen or a second, separate tumor antigen to interact with the co-stimulatory receptor CD28 (TSAxCD28 bispecifics) expressed on T cells. It has been hypothesized that combining TSAlxCD3 with TSA2xCD28 should provide targeted and enhanced mimic activation of T cells by triggering both signal 1 and signal 2, with a specificity targeting only tumor cells expressing both epitopes or both antigens, thereby providing greater antitumor activity. activity together with the potential to increase specificity.

This document describes the generation and testing of costimulatory bispecific anti-TSAxCD28 antibodies targeting prostate cancer (PSMAxCD28, which binds prostate-specific membrane tumor antigen). Toxicological studies in genetically humanized immunocompetent mice as well as cynomolgus monkeys demonstrate that these bispecifics have limited activity and are not toxic when used as monotherapy. However, these new costimulatory bispecifics can be effectively combined with the emerging TSAxCD3 class of bispecifics to enhance antitumor responses in syngeneic tumor models. Taken together, these data suggest that combining this new class of CD28-based bispecifics (TSAxCD28) with CD3-based bispecifics (TSAxCD3) may provide well-tolerated, ready-to-use biological solutions with

- 51 045943 noticeably enhanced and synergistic antitumor activity.

Materials and methods.

Examples 10 and 13 used the following materials and methods.

Studies on syngeneic tumors.

Mice expressing human CD28, human CD3 and human PSMA instead of the corresponding mouse genes (referred to as hCD3/hCD28/hPSMA humanized mice) were generated using Velocigene® technology as described previously (Valenzuela (2003), Nat Biotechnol, Jun;21( 6):652-9; Crawford et al. 2018, manuscript in preparation). For each humanized mouse, correct gene targeting in the F1H4 embryonic stem (ES) cell clones (C57BL/6 x 129 hybrid) was identified using allelic loss analysis as described previously (Poueymirou et al. (2007), Nat Biotechnol, Jan; 25(1):91-9). Targeted ES cells were injected into Swiss Webster 8-cell stage embryos to generate fully heterozygous F0 generation mice for crossing with C57BL/6N mice (Taconic, Rensselaer, NY) to homozygosity. hCD3/hCD28/hPSMA mice (4-8 mice/group, 8-16 weeks old) were injected subcutaneously with 1x10 ⁶ MC38/hPSMA tumor cells. Bispecific antibody to PSMAxCD28, bispecific antibody to PSMAxCD3 or human IgG4 isotype control was administered as monotherapy or in combination by intraperitoneal injection on days 0, 3 and 7 at a dose of 5 mg/kg.

Tumor growth dynamics were monitored using calipers to measure X and Y diameters. Tumor volume (X*Y*(X/2)) was calculated. Mice were sacrificed when tumors grew to a size greater than 2000 mm ³ .

Measurement of cytokine levels in mouse serum.

At the indicated time points, blood was collected by submandibular vein puncture into microtainer serum tubes (BD 365967). Cytokine levels were analyzed using the V-plex Human Prolnflammatory-10 Plex kit according to the manufacturer's instructions (Meso Scale Diagnostics, Rockville, MD).

Results, summary and conclusions.

Expression of human CD28 on T and NK cells was confirmed by FACS (data not shown). The functional characteristics of the triply humanized mice were confirmed using an in vitro T cell proliferation assay (data not shown). PSMA expression was confirmed by qPCR (quantitative PCR) (data not shown). In triply humanized hCD3/hCD28/hPSMA mice (described above), MC38/hPSMA tumor growth dynamics were monitored (Fig. 9A). Compared with isotype control, both PSMAxCD3 and PSMAxCD28 monotherapy significantly inhibited tumor growth (p<0.001 and p<0.0001, respectively). Tumor growth was further inhibited by combination therapy (p<0.00001). Compared with the isotype control, the combination of bispecifics PSMAxCD3 and PSMAxCD28 provided the greatest cytokine production (Fig. 9C and 9D). To better understand the role of combination therapy on the activation state of intratumoral CD8' T cells, viSNE analysis was performed (Fig. 9B). viSNE is a tool for visualizing high-dimensional cytometric data in 2D while preserving high-dimensional structure. Each treatment resulted in unique clusters of CD8+ T cells in the spleen and tumor. Combination therapy stimulated expansion of the activated/memory T cell phenotype (expressing TCF1, CD1-2, CD127, PD-1, ICOS, KLRG1, and CD38) as shown in cluster 4 in the tumor.

Studies were conducted to determine serum levels of cytokines (IFNy, IL-2, IL-6, IL-10, TNFa, IL-4 and IL-5) in hCD3 ^+/+ /hPSMA ^+/+ /hCD28 ^+/+ mice , and in hCD3 ^+/- /hPSMA ^+/ /hCD28 ^+/- mice. Blood was taken at 0 (4 hours), 3, 7 and 11 days after treatment. With the exception of IL-10 in hCD3 ^+/+ /hPSMA ^+/+ /hCD28 ^+/+ mice, at day 0, cytokine levels were significantly increased by treatment that included anti-PSMAxCD3 antibodies, and combination therapy of anti-PSMAxCD3 and anti-PSMAxCD28 resulted in the greatest release of cytokines. In contrast, control IgG and antiPSMAxCD28 monotherapy did not increase cytokine production. With the exception of IL-5 and IL-10, cytokine production decreased after three (3) days to similar levels in all treatment groups. IL-5 production was significantly higher on day 3, although lower than on day 0, in hCD3 ^+/+ /hPSMA ^+/+ /hCD28 ^+/+ mice, but not in hCD3 ^+/- /hPSMA ^+/- mice /hCD28 ^+/- who received combination treatment. In hCD3 ^+/+ /hPSMA ^+/+ /hCD28 ^+/+ mice, IL-10 production was similar for all treatment groups on day 0 and day 3, but combination therapy produced significantly higher levels in combination therapy on days 7 and 11 day. In hCD3 ^+/- /hPSMA ^+/- /hCD28 ^+/- mice, combination therapy induced significantly more IL-10 on day 0, day 3, and day 11, but not on day 7. Anti-PSMAxCD3 monotherapy induced significantly more IL-10 on day 0 and day 3, but not on day 7 and day 11. AntiPSMAxCD28 monotherapy induced significantly more IL-10 only on day 11 (data not shown).

As shown in FIG. 9A, in contrast to previous in vitro assays in which CD28 bispecifics had very limited activity when used as monotherapy (see Example 8 above), CD28 bispecifics in these syngeneic MC38/hPSMA models had greater activity when

- 52 045943 use as monotherapy. This suggests that signal 1 was already activated to some extent in this MC38 model. Consistent with these data, it was previously shown that MC38 tumor cells express high levels of reactivated endogenous retroviral proteins, such as p15E, and that C57BL6 mice can generate endogenous T cells that recognize and respond to this neoepitope (J.S. Yang, D. Perry-Lalley, The envelope protein of an endogenous murine retrovirus is a tumor-associated T-cell antigen for multiple murine tumors. J Immunother 23, 177-183 (2000); H. J. Zeh, 3rd, D. Perry- Lalley, M. E. Dudley, S. A. Rosenberg, J. C. Yang, High avidity CTLs for two selfantigens demonstrate superior in vitro and in vivo antitumor efficacy. J Immunol 162, 989-994 (1999). Indeed, it has been confirmed that intratumoral T cells responding to this p15E neoantigen can be readily detected in MS38 models. Thus, CD28 bispecifics in these syngeneic tumor models may enhance endogenous TCR/CD3-dependent T cell responses, which can then be further enhanced by providing additional signal 1 activation via the CD3 bispecific antibody.

To determine the cellular mechanism underlying the combination therapy, tumor-infiltrating and splenic CD8+ T cells were profiled from these experiments using high-dimensional flow cytometry and using unsupervised clustering approaches. Each treatment was found to result in unique clusters of CD8+ T cells in the spleen and tumor. Single-dose treatment regimens reduced levels of intratumoral CD8+ T cells with a less activated phenotype (with lower levels of ICOS, KLRG1, Ki67, PD1, CD38, and LAG3; Fig. 9E), as shown in cluster C35. However, combination therapy significantly promoted the expansion of a more activated/memory T cell phenotype (expressing Tcf1, CD122, CD127, PD1, ICOS, KLRG1, and CD38; Fig. 9B), as shown in cluster C4.

Ahtu-CD28 x anti-PSMA does not induce increases in serum cytokine levels in humanized CD28/CD3/PSMA mice in the absence or presence of tumor.

As discussed elsewhere herein, TGN1412, a CD28 superagonist antibody, caused a cytokine storm that was detrimental to the patient. The anti-CD28 antibody and the bispecific anti-CD28xPSMA antibody of the invention did not cause cytokine storm. In this study, humanized hCD3/hCD28/hPSMA mice were administered several antibodies (summarized in Table 19) at a dosage of 2.5 mg/kg or 0.25 mg/kg. Blood was collected from mice 4 hours and 3 days after the administration of antibodies. As shown in FIG. 10, bispecific anti-PSMAxCD28 antibodies or parental bivalent anti-CD28 antibodies did not induce serum cytokine production in humanized CD3/CD28/PSMA mice in the absence of tumor.

Table 19

Antibodies to study cytokine storm in humanized hCD3/hCD28/hPSMA mice

Antibody ID Batch number Name REGN1945 REGN1945-L49 h!gG4 tAL14226P2 REGN5705-L1 mAb to CD28 REGN2329 (TeGenero) REGN2329-L3 Superagonist (SA) to CD28 bsl6429D bsl6429D-L2 PSMAxCD28 BE0146 614216J2B PD1, RMP1-14 bs!5780D bs!5780D-L2 PSMAxCD3

At 2.5 mg/kg, TGN1412 can cause significant increases in blood levels of IL-2, IL-4, IL-5, and TNFa (data not shown).

Moreover, as shown in FIG. 28, treatment with anti-PSMAxCD28 alone or combination treatment with anti-PD1 antibody did not increase serum cytokine levels in tumor-bearing mice. In contrast, treatment with anti-PSMAxCD3, alone or in combination with anti-PD1 antibody, increased serum levels of cytokines such as TNFa, IL-5, IL-10, IL-2, and IL-4 in tumor-bearing mice. Anti-PSMAxCD3 treatment induced cytokine expression 4 h post-dose. The increase in cytokine levels was not maintained after 7 days. This study was performed on humanized hCD3/hCD28/hPSMA mice bearing MC38/hPSMA tumors. Serum was collected from humanized hCD3/hCD28/hPSMA mice bearing MC38/hPSMA tumors 4 hours post-dose on day 0 and day 7, 4 days later with 5 mg/kg of the indicated antibodies or bispecifics. Serum cytokine levels were measured using a 10-plex mouse cytokine MSD kit according to the manufacturer's protocol.

Moreover, as shown in FIG. 11, anti-CD28 superagonist treatment induced a cytokine response after 4 hours in PBMC-engrafted NSG mice, unlike the anti-CD28 antibody of the invention. In the study shown in FIG. 11, immunodeficient NSG mice were inoculated with 5x106 PBMCs from burrows

- 53 045943 small healthy donor on day -10. On day -1, systemic T cell engraftment was confirmed by staining peripheral blood for human T cell markers. On day 0, mice were injected intraperitoneally with 50 μg of isotype control antibody IgG4, 5 μg or 50 μg of anti-CD28 superagonist, or 5 μg or 50 μg of anti-CD28 antibody mAb14226P2. Four hours after antibody injection, blood was collected from the animals and serum was obtained.

Serum cytokine concentrations were analyzed using a multiplex assay (Meso Scale Discovery V-PLEX kit). While the anti-CD28 superagonist induced increased serum levels of interferon-gamma (IFN-γ), IL-2, IL-6, and TNFa compared to animals treated with the isotype control antibody, no increases were observed in the serum of animals treated with the anti-CD28 anti-CD28 superagonist. increasing the cytokine response. Accordingly, the CD28xPSMA bispecific has a potentially safer toxicological profile among co-stimulatory bispecific antigen-binding molecules. Bispecific to CD28xPSMA did not induce a cytokine response, unlike bispecific to CD3xPSMA. As shown in FIG. 12, anti-CD3x PSMA caused an increase in IFN-γ in humanized mice, in contrast to anti-CD28xPSMA. In this study, sera were collected from humanized CD3/CD28/PSMA or CD3/CD28 mice bearing MC38/hPSMA tumors 4 hours after dosing with 5 mg/kg of the indicated bispecifics. Serum cytokine levels were measured using a 10-plex mouse cytokine MSD kit according to the manufacturer's protocol. The points on the graph represent individual cytokine levels per mouse. The bars in the graph represent the mean for each treatment group. Error bars represent ±SEM. Statistical significance was determined by one-way analysis of variance and Holm-Sidak multiple comparison test using isotype-treated mice as controls (**, p < 0.01).

Consistent with data obtained in humanized mice, a bispecific antibody to CD3xPSMA caused increased CRP (C-reactive protein) levels and systemic cytokine release in the non-human primate, cynomolgus macaques. In a monkey study, CD3xPSMA bispecific antibody was administered to cynomolgus monkeys at dosages of 0.01, 0.1, and 0.5 mg/kg. The monkey exhibited the following symptoms: clinical signs on day 1 such as vomiting, arching of the back, red/discolored skin; increased CRP levels (approximately 10 vs. 1 mg/dL in control mice); increased levels of plasma cytokines (IL-6, TNFa, IFN-γ, IL-2 and MCP); and a fall in the absolute number of T cells in the blood (data not shown).

Moreover, mice treated with anti-CD28xPSMA did not exhibit cytokine production or T cell margination compared to anti-PSMAxCD3 or anti-CD28 superagonist treatment (FIGS. 13A and 13B).

Summary.

It has long been known that T cell activation through the TCR complex (signal 1) can be markedly enhanced by co-stimulatory signals, such as those mediated by the interaction of the CD28 receptor on T cells with its ligands (CD80/B7.1 and CD86/B7. 2) on target cells (signal 2) (J. H. Esensten, Y. A. Helou, G. Chopra, A. Weiss, J. A. Bluestone, CD28 Costimulation: From Mechanism to Therapy. Immunity 44, 973-988 (2016)). Consistent with the data disclosed herein, the potential of CD28 costimulation to enhance T cell antitumor activity was first demonstrated in studies in which B7 ligands were overexpressed on tumor cells (R. H. Schwartz, Costimulation of T lymphocytes: the role of CD28, CTLA -4, and B7/BB1 in interleukin-2 production and immunotherapy. Cell 71, 1065-1068 (1992); L. Chen et al., Costimulation of antitumor immunity by the B7 counterreceptor for the T lymphocyte molecules CD28 and CTLA-4 Cell 71, 1093-1102 (1992)), which showed improved T cell rejection of such B7-expressing tumors. This potential has inspired efforts to evaluate CD28-activating antibodies in human trials. Unfortunately, a trial of such an antibody (TGN1412) in 2006 resulted in life-threatening complications in all six volunteers (G. Suntharalingam et al., Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. N Engl J Med 355 , 1018-1028 (2006)), due to multiple organ failure resulting from massive cytokine release syndrome (CRS). This disaster led to the cessation of any further testing of CD28-activating antibodies in humans.

We describe herein a new class of CD28 co-stimulatory bispecific antibodies that can markedly and safely stimulate antitumor activity by providing co-stimulatory signal 2. These CD28 bispecifics themselves have limited activity (in the absence of signal 1) but can markedly enhance antitumor activity in signal 1 conditions, which can be achieved by using these OE28 bispecifics in tandem with an emerging class of OE3 bispecifics (or if these OE28 bispecifics are used in settings in which endogenous populations of tumor-specific T cells are already present). This document describes the generation and testing of costimulatory anti-TSAxCD28 bispecific antibodies targeting prostate cancer (PSMAxCD28). In the absence of signal 1, these CD28 bispecifics have been shown to have minimal activity in vitro or in vivo. However, these OO28-bispecifics

- 54 045943 can be used in conjunction with CD3 bispecifics to form artificial immune synapses containing tumor antigens, as well as TCR and CD28 complexes. Moreover, when paired with appropriate CD3 bispecifics in vitro, these CD28 bispecifics can effectively and specifically promote T cell activation and tumor cell killing in an antigen-dependent manner. In addition, these CD28 bispecifics also effectively enhance the antitumor activity of CD3 bispecifics in vivo in a tumor antigen-dependent manner in syngeneic tumor models; in such models, CD28 bispecifics have minimal activity when used alone unless tumor-specific T cells are already present, and in such conditions they appear to enhance this specific activity in a tumor antigen-dependent manner. In addition, combination therapy of TSAxCD28 and TSAxCD3 significantly stimulates the expansion of the intratumoral phenotype of activated/memory T cells in vivo. Finally, toxicology studies in genetically humanized immunocompetent mice as well as cynomolgus monkeys demonstrate that these bispecifics have limited activity and are not toxic when used alone compared to conventional CD28-activating antibodies.

Often, characterization of human-specific clinical drug candidates in the field of immuno-oncology is limited to testing in xenogeneic tumor models engrafted with human immune cells. Although these xenogeneic models can be very useful, they have limitations. Mice used in such xenogeneic models do not express the human tumor target in their healthy tissues, thereby preventing the evaluation of the test agent under conditions of normal tissue expression of the target. Indeed, if the target is typically also expressed at high levels in healthy tissues, this may limit antitumor efficacy by diverting the test agent from the tumor and may lead to toxicity to these healthy tissues, none of which can be assessed in a syngeneic xenogeneic model. An additional limitation may be the activity of engrafted human peripheral blood mononuclear cells (PBMCs) transferred into an immunodeficient mouse, which may differ from the activity of normal host T cells present in an immunocompetent system. To overcome these limitations and improve models for testing human-specific clinical drug candidates, the present invention generated doubly and triply genetically humanized mice. In these models, tumor antigens were genetically humanized to ensure their normal expression in appropriate host tissues, and CD3 and/or CD28 components were genetically humanized to allow immunocompetent host cells to respond to human-specific clinical drug candidates. In these genetically humanized immunocompetent syngeneic animal models, it was found that CD28 bispecifics to the tumor target PSMA enhanced the antitumor activity of the corresponding CD3 bispecifics. This increase in antitumor efficacy by various TSAxCD28 bispecifics in several preclinical models suggests that this therapeutic modality is robust and not limited to a specific tumor model, but may have broader application as a new class of targeted immunotherapy combination. Overall, the results highlight that TSAxCD28 bispecifics may act synergistically with TSAxCD3 bispecifics and may provide a biological solution that can markedly enhance the efficacy of the well-studied TSAxCD3 bispecifics in a manner sufficiently safe and well tolerated to justify testing in human studies.

TSAxCD3 bispecifics represent a promising emerging class of immunotherapy, but further optimization of antitumor activity will undoubtedly be important in many settings. Similar to how CAR-T cell approaches used chimeric receptors that artificially activate both signal 1 and signal 2 to improve their antitumor activity (E. A. Zhukovsky, R. J. Morse, M. V. Maus, Bispecific antibodies and CARs: generalized immunotherapeutics harnessing T cell redirection. Curr Opin Immunol 40, 24-35 (2016); S. L. Maude et al., Tisagenlecleucel in Children and Young Adults with B-Cell Lymphoblastic Leukemia. N Engl J Med 378, 439-448 (2018) ), this invention demonstrates the potential benefit of combining CD3-specific drugs (which provide signal 1) with CD28-bispecific drugs (which provide signal 2) to enhance antitumor activity. In addition to the practical advantages that this approach may have over CAR-T cell therapy, it does not require the time-consuming preparation of cell therapy, which must be individually tailored to each patient, and does not require patients to first undergo lymphodepletion by toxic chemotherapy so that they can receive such cell therapy, often associated with side effects (A. Shimabukuro-Vornhagen et al., Cytokine release syndrome. J Immunother Cancer 6, 56 (2018); C. H. June, R. S. O'Connor , O. U. Kawalekar, S. Ghassemi, M. S. Milone, CAR T cell immunotherapy for human cancer. Science 359, 1361-1365 (2018)), - the bispecific approach of the invention provides the potential for increased efficacy, as well as increased safety and specificity actions. Thus, it is possible to take advantage of combination targeting by combining a CD3 bispecific for one antigen with a CD28 bispecific that is specific for one antigen.

- 55 045943 to the second antigen - increased efficacy will only occur on tumor cells expressing both antigens - thus focusing T cell killing only on tumor cells expressing both antigens, while limiting off-target toxicity in healthy tissues expressing only one of the antigens . Taken together, the data disclosed herein suggest that combining CD28-based bispecifics with CD3-based bispecifics may provide well-tolerated, off-the-shelf biological solutions with markedly enhanced and synergistic antitumor activity. Initial testing of this possibility in human studies will take place this year. Example 11: Toxicological studies on cynomolgus monkeys. These studies demonstrated that PSMAxCD28 alone or in combination therapy did not induce systemic T cell activity compared with a CD28 superagonist in cynomolgus monkeys. Exemplary PSMAxCD28 bispecific antibodies of the invention enhanced activation of cynomolgus T cells by TAAxCD3 (Example 8, Fig. 7E-G). To determine the safety and tolerability of the exemplary anti-PSMAxCD28 bispecific antibodies of the present invention, alone or in combination with anti-PSMAxCD3, a single dose toxicity study was conducted in cynomolgus monkeys. Female or male cynomolgus macaques were assigned to treatment groups as indicated in Table 1 . 20.

Table 20

Exploratory single dose toxicity study of anti-PSMAxCD28 in monkeys

Group Treatment (single dose) Total number of animals Dose levels/escalation (30-minute IV infusion) Males 1 PSMAxCD3 3 0.1 mg/kg 2 bsl6429D 3 1 mg/kg 3 bsl6429D 3 10 mg/kg 4 Super agonist "SA" to CD28 3 10 mg/kg 5 L816429O+antibody to PD-1 3 10 mg/kg+10 mg/kg 6 bsl643 lD+anti-PD-1 antibody 3 10 mg/kg+10 mg/kg 7 Ab isotype control 3 10 mg/kg

The cynomolgus macaque study was conducted in accordance with IACUC (Institutional Animal Care and Use Committee) guidelines. Male cynomolgus monkeys (3 animals/group) received a single dose of each study drug via intravenous infusion over approximately 30 min (the combination treatment was administered as a separate infusion over a total of 1 h). Toxicity assessment was based on clinical observations, qualitative indicators of feed intake, body weight, neurological examinations, vital signs (body temperature, heart rate, pulse oximetry and respiratory rate), as well as laboratory diagnostics and pathological anatomy. Blood and tissue samples were collected for cytokine analysis, FACS immunophenotyping, histopathology, and toxicokinetic evaluation. CRP levels were analyzed on a Roche Modular P 800 system. Cytokines were measured using Meso Scale Diagnostics (MSD, Rockville, MD). For peripheral blood flow cytometry, blood was collected into potassium-EDTA tubes, lysed, stained with indicated antibodies such as anti-CD3, anti-Ki67, and anti-ICOS (BD Biosciences), and analyzed using FACS Canto II. Following administration of a single dose of PSMAxCD28 at 1 or 10 mg/kg, no significant cytokine release, T cell margination, or up-regulation of a T cell activation marker was observed (Table 21).

- 56 045943

Table 21

Summary of Toxicology Study on Cynomolgus Macaques

Molecules a bsl6429D bsl6429D REGN232 9 (TeGenero) Description antiPSMAxCD2 8 antiPSMAxCD2 8 superagonist to CD28 Dose (mg/kg) 1 10 10 Liu 1 day - nab wedge l-ichesk ----- Not obs. 2-4 I days Absolute number of T cells (U/μl) Pre- th 5 hours tes T 7.2 6.58 2 5.0 4.91 9 4.0 4.50 4 5.7 6.03 3 4.5 4.64 3 10.0 7.1 6 6 0.2 7.02 6 0.1 7.06 9 11.8 0.6 7 6 Proliferating KΚΪ67+ T cells (U3/μl) Predv168 h test T 0.62 0.59 0.42 0.32 0.63 0.48 0.33 0.35 0.38 0.64 0.88 0.39 0.60 5.66 0.79 7.29 10.9 1.13 6 CRP (mg/dl) 24 hours 0.5 0.5 1.1 0.2 0.2 1.3 15.4 15 15.3 Plasma cytokines 5 hours after dosing (pg/ml) IL-IL-IF TN IL-IL-ILIL-6 8 10 Ν-γ F-α 2 4 5 BLQ BL BL BL BL BL BL BLQ * Q Q Q Q Q Q BL BL BL BL BL BL BLQ BLQ Q Q Q Q Q Q BL BL BL BL BL BL BLQ BLQ Q Q Q Q Q Q BL BL BL BL BL BL BLQ BLQ Q Q Q Q Q Q BL BL BL BL BL BL BLQ BLQ Q Q Q Q Q Q BL BL BL BL BL BL BLQ BLQ Q Q Q Q Q Q BL BL BL BL BL 116 764 3364 Q Q Q Q Q 9 109 1715 BL BL 2907 333 238 615 5 5 Q Q 532 150 5226 7918 163 522 342 198 4 9 mAblO154 RZ EGFRVIIIAb 5.6 6.39 2 7.7 7.43 2 4.6 3.66 1 0.62 0.39 0.86 0.52 0.52 0.69 0.8 AND 0.2 BL BL BL BL BLQ BLQ 151 335 Q Q Q Q BL BL BL BL BLQ BLQ 267 378 Q Q Q Q BL BL BL BL BL BL BLQ BLQ Q Q Q Q Q Q

BLQ: Below the lower limit of quantitation

In contrast, significant cytokine release, lymphocyte margination, and T cell activation were observed in monkeys administered a CD28 superagonist. Additionally, these results were confirmed using a human T cell proliferation assay with dry and wet application (Example 9). Indeed, induction of human T cell proliferation by the exemplary bispecific anti-PSMAxCD28 antibodies as well as the parental divalent anti-CD28 antibodies was not observed compared to the anti-CD28 superagonist antibody. Overall, the exploratory single dose toxicity study in monkeys and in vitro human T cell assays suggest that the exemplary PSMAxCD28 bispecific antibodies of the invention are safe and well tolerated.

As shown in table. 21, TGN1412, a CD28 superagonist, caused moderate cytokine release and a transient increase in T cell levels (measured at day 15). The study was conducted on male cynomolgus monkeys with weekly doses of 5 to 50 mg/kg for 4 weeks. FACS analysis of leukocyte subpopulations was performed on a male cynomolgus monkey after intravenous injection of weekly (on days 1, 8, 15, 22) increasing doses (5, 10, 25, 50 mg/kg) of TGN1412. There was a moderate increase in the levels of IL-2, IL-5, IL-6, IFN-γ in the first 2-24 hours (2-20 times). Significant changes in IL-4 or UTA were observed. An exploratory single dose toxicity study of anti-CD28xPSMA in monkeys was conducted to establish the safety and pharmacokinetic profile of the anti-CD28xPSMA bispecific antibody.

Thus, anti-CD28xPSMA antibodies were shown to enhance immunity at the tumor site. Ahtu-CD28xPSMA converted tumor cells into antigen presenting cells (APCs). Antibodies to CD28xPSMA do not induce T cell proliferation or stimulate the release of cytokines by

- 57 045943 compared with a CD28 superagonist.

Example 12 Epitope mapping of mAb14226P2 binding to CD28 via hydrogen-deuterium exchange.

Epitope mapping by H/D exchange and mass spectrometry (HDX-MS) was performed to identify amino acid residues in CD28 (recombinant human CD28, represented as hCD28 ecto (N19-P152).mmh; SEQ ID NO: 75), interacting with the monoclonal antibody to hCD28. A general description of the H/D exchange method is given, for example, in Ehring (1999) Analytical Biochemistry 267(2):252-259; and Engen and Smith (2001) Anal. Chem. 73:256A-265A.

HDX-MS experiments were performed on an integrated HDX/MS platform consisting of a Leaptec HDX PAL system for deuterium labeling and reaction stop, Waters Acquity M-Class (solvent assist system) for digestion and sample loading, Waters Acquity M- Class (μ-dual-channel solvent system) for analytical gradient and Thermo Q Exactive HF mass spectrometer for peptide mass measurement.

The labeling solution was prepared as PBS buffer in D ₂ O at pD 7.0 (10 mM phosphate buffer, 140 mM NaCl and 3 mM KCl, equivalent to pH 7.4 at 25°C). For deuterium labeling, 11 μl of CD28.mmH (REGN2011 protein, self-produced by Regeneron, 127 μM) or CD28.mmH, premixed with REGN5705 in a molar ratio of 1:0.6 (antigen-antibody complex), was incubated at 20°C with 44 µl of D ₂ O labeling solution for different time points, each in duplicate (eg, non-deuterated control=0 seconds; deuterated labeled for 5 min and 10 min). The deuteration reaction was stopped by adding 55 μl of pre-cooled stop buffer (0.5 M TCEP-HC1, 8 M urea and 1% formic acid) to each sample and continued incubation for 5 min at 20°C. The stalled sample was then introduced into Waters HDX Manager for rapid pepsin/protease XIII digestion. Digested peptides were separated on a C8 column (1.0x50 mm, NovaBioassays) with a 13-min gradient of 10–32% B (mobile phase A: 0.5% formic acid in water, mobile phase B: 0.1% formic acid in acetonitrile ). The eluted peptides were analyzed by mass spectrometry on Q Exactive HF in LC-MS/MS or LCMS mode.

LC-MS/MS data from a nondeuterated CD28 sample were compared to a database including CD28 and its randomized sequence using the Byonic search engine (Protein Metrics). Search parameters (in the ELN (electronic laboratory journal)) were set to default using nonspecific enzymatic digestion and glycosylation of human proteins as a common variable modification. The list of identified peptides was then imported into HDX Workbench software (version 3.3) to calculate the deuterium uptake of each LC-MS-determined peptide from all deuterated samples. For an existing peptide, the centroid mass (intensity-weighted average mass) at each time point was used to calculate deuterium uptake (D) and percent deuterium uptake (%D).

Deuterium absorption (D absorption) =. Average weight (deuterated) - average weight (non-deuterated)

Uptake of D by peptide in each

Percentage of deuterium uptake point in time X 100% (%D) = Maximum uptake of D by peptide (defined in ELN)

A total of 73 peptides from hCD28.mmH (SEQ ID NO: 75) were identified from both hCD28.mmH alone and hCD28.mmH complexed with anti-CD28 antibody samples, representing 85.8% sequence coverage of hCD28. Any peptide that exhibited a D absorbance percentage greater than 5% was characterized as significantly protected. For hCD28.mmH (SEQ ID NO: 75) regions corresponding to amino acids 5-20 (VKQSPMLVAYDNAVNL; SEQ ID NO: 77), 29-38 (FSREFRASLH; SEQ ID NO: 78), 80-84 (YLQNL; SEQ ID NO : 79) and 91-108 (IYFCKIEVMYPPPYLDNE; SEQ ID NO: 80), were significantly protected by the anti-CD28 antibody, with amino acids 91108 (IYFCKIEVMYPPPYLDNE; SEQ ID NO: 80) identified as the major epitope on CD28. Protection of these residues by anti-CD28 antibody was confirmed using hCD28.mFc (SEQ ID NO: 76). See also table. 22 below, which summarizes the results of this study.

- 58 045943

Table 22

Selected CD28.mmH peptides with significant protection when bound to anti-CD28 antibody

Remainder and Charge (+) 5 minutes 10 min REGN2011 REGN2011 + SHAL4226R2 REGN2011 REGN2011 + shAL4226Р2 CD28 Centroi Centroid Centroid Centroid Δ% ΔΌ bottom ΔΌ naya MN ⁺ naya MN ⁺ naya MN ⁺ D MN ⁺ 5-11 1 805.74 805.53 -0.21 805.83 805.58 -0.25 -7.1 5-11 2 806.69 806.48 -0.21 806.76 806.51 -0.24 -7.1 8-11 1 449.28 449.11 -0.16 449.33 449.16 -0.17 -10.5 12-19 1 869.61 869.30 -0.31 869.72 869.33 -0.39 -7.2 12-19 2 869.95 869.72 -0.23 870.08 869.69 -0.39 -6.4 12-20 1 982.96 982.54 -0.42 983.05 982.63 -0.41 -7.4 15-20 1 648.02 647.61 -0.40 648.06 647.68 -0.38 -12.3 29-32 1 540.44 540.06 -0.38 540.51 540.08 -0.44 -25.4 29-37 2 1116.66 1116.15 -0.51 1116.79 1116.20 -0.60 -9.9 29-37 3 1117.61 1117.74 -0.50 1117.11 1117.17 -0.57 -9.6 30-37 2 969.13 968.85 -0.28 969.23 968.90 -0.33 -6.4 30-38 3 1107.35 1106.93 -0.42 1107.48 1107.02 -0.46 -7.8 33-37 1 595.17 595.23 -0.12 595.05 595.08 -0.15 -5.8 33-38 1 732.40 732.24 -0.16 732.45 732.25 -0.19 -5.5 33-38 2 733.43 733.25 -0.18 733.48 733.29 -0.19 -5.7 80-84 1 652.81 652.51 -0.30 652.91 652.60 -0.31 -12.6 81-84 1 489.26 488.99 -0.28 489.34 489.04 -0.30 -17.9 91-93 1 443.49 443.50 -0.07 443.42 443.41 -0.09 -10.1 93-97 1 641.40 641.48 -0.58 640.83 640.89 -0.59 -24.4 94-105 2 1458.33 1456.70 -1.63 1458.63 1456.83 -1.80 -30.7 94-107 2 1688.09 1686.36 -1.73 1688.33 1686.43 -1.90 -25.2 94-108 2 1817.69 1816.15 -1.54 1817.88 1816.19 -1.69 -20.2 96-107 2 1455.75 1455.07 -0.68 1455.91 1455.09 -0.82 -13.4 98-107 2 1212.67 1212.07 -0.60 1212.78 1212.07 -0.70 -16.3 98-108 2 1342.17 1341.57 -0.60 1342.10 1341.61 -0.48 -5.0

Example 13 Bispecific antibodies to PSMAxCD28 effectively enhance the antitumor efficacy of anti-PD-1 immunotherapy.

Annotation.

T cell activation is enhanced by interaction with a second costimulatory receptor (signal 2), in addition to antigen-specific TCR/CD3 activation (signal 1). The goal of cancer immunotherapy is to optimally activate and mobilize T cells to detect and kill tumor cells. However, current treatments generally fail to effectively and selectively activate T cells at the tumor site, often failing to achieve durable responses and/or leading to unwanted toxicities. Herein, a new tumor-targeted immunotherapy approach combining PD-1 inhibition together with bispecific antibodies was presented. Bispecific antibodies bind a tumor-specific antigen (TSA) (eg, PSMA) with one arm and the costimulatory receptor CD28 on T cells with the other arm. Indeed, inhibition of PD-1-PD-L1 signaling significantly increases the ratio of CD28 accumulated at the immunological synapse, allowing the TSAxCD28 bispecific to exert its effects.

- 59 045943

This combination immunotherapy has been tested using bispecific antibodies specific for prostate antigen (eg, PSMA). In contrast to nonspecific CD28 superagonists, which activate a broad spectrum of T cells, TSAxCD28 bispecifics were well tolerated when used alone or in combination with a PD-1 blocker in genetically humanized immunocompetent mouse or primate models. Importantly, in the presence of an endogenous TCR/CD3 trigger, TSAxCD28 dramatically improved the anti-PD-1 anti-tumor activity associated with durable anti-tumor responses. Combination therapy specifically enhanced intratumoral T cell activation, promoting a memory T cell-like effector T cell phenotype without systemic cytokine secretion in various syngeneic and xenograft human tumor models. Combining this class of CD28 costimulatory bispecific antibodies with a clinically successful anti-PD-1 antibody treatment may provide well-tolerated, off-the-shelf antibody therapy options with measurably increased antitumor efficacy.

Introduction.

A variety of monoclonal antibodies (mAbs) designed to enhance T cell activation are in clinical trials as anticancer therapeutics (M. K. Callahan, M. A. Postow, J. D. Wolchok, Targeting T Cell Co-receptors for Cancer Therapy. Immunity 44, 10691078 (2016) ). However, most current treatments face the challenge of overcoming the inhibitory nature of the tumor microenvironment, which results in an inability to generate effective tumor-specific T cell activation and subsequent tumor cell killing (K. G. Anderson, I. M. Stromnes, P. D. Greenberg, Obstacles Posed by the Tumor Microenvironment to T cell Activity : A Case for Synergistic Therapies. Cancer Cell 31, 311-325 (2017). For several blocking monoclonal antibodies directed against checkpoint inhibitors such as cytotoxic T-lymphocyte-associated protein (CTLA-4) and programmed cell death protein 1 (PD-1)/programmed cell death protein ligand 1 (PD-L1), approved clinical trials for the treatment of melanoma, renal cell carcinoma, non-small cell lung cancer, and advanced metastatic squamous cell carcinoma of the skin (J. S. Weber et al., Nivolumab versus chemotherapy in patients with advanced melanoma who progressed after anti-CTLA-4 treatment (CheckMate 037): a randomized, controlled, open-label, phase 3 trial. Lancet Oncol 16, 375-384 (2015); S. L. Topalian et al., Survival, durable tumor remission, and long-term safety in patients with advanced melanoma receiving nivolumab. J Clin Oncol 32, 1020-1030 (2014); M. A. Postow, M. K. Callahan, J. D. Wolchok, Immune Checkpoint Blockade in Cancer Therapy. J Clin Oncol 33, 1974-1982 (2015); M. R. Migden et al., PD-1 Blockade with Cemiplimab in Advanced Cutaneous Squamous-Cell Carcinoma. N Engl J Med 379, 341-351 (2018)). Blocking PD-1 relieves restrictions on T cell activation, but its effectiveness as monotherapy is often insufficient to achieve tumor clearance and durable antitumor responses. The combination of αnτu-PD-1/PD-L1 and anti-CTLA-4 does produce high response rates in some tumor types, but significant toxic effects are often observed (J. Larkin et al., Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma N Engl J Med 373, 23-34 (2015); D. B. Johnson et al., Fulminant Myocarditis with Combination Immune Checkpoint Blockade. N Engl J Med 375, 1749-1755 (2016); M. H. Pollack et al., Safety of resuming anti-PD-1 in patients with immune-related adverse events (irAEs) during combined anti-CTLA-4 and anti-PD1 in metastatic melanoma. Ann Oncol 29, 250-255 (2018); J. D. Wolchok et al., Nivolumab plus ipilimumab in advanced melanoma. N Engl J Med 369, 122-133 (2013). Consequently, significant efforts continue to be made to identify patients who are more likely to respond to checkpoint inhibition by identifying biomarkers predictive of response to therapy (R. Cristescu et al., Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy Science 362, (2018)). In addition, combination therapies are being evaluated preclinically and clinically to improve the efficacy of PD-1 blockade and the duration of antitumor response by combining with agonistic costimulatory receptor-triggering antibodies or other modalities such as chemotherapy or radiation therapy (S. Hu-Lieskovan , A. Ribas, New Combination Strategies Using Programmed Cell Death 1/Programmed Cell Death Ligand 1 Checkpoint Inhibitors as a Backbone. Cancer J 23, 10-22 (2017); Y. K. Chae et al., Current landscape and future of dual anti- CTLA4 and PD-1/PD-L1 blockade immunotherapy in cancer; lessons learned from clinical trials with melanoma and non-small cell lung cancer (NSCLC). J Immunother Cancer 6, 39 (2018); P. S. Chowdhury, K. Chamoto, T B. Wang et al., Combination cancer immunotherapy targeting PD-1 and GITR can rescue CD8(+) T cell dysfunction and maintain memory phenotype. Sci Immunol 3, (2018); S. Chen et al., Combination of 4-1BB agonist and PD-1 antagonist promotes antitumor effector/memory CD8 T cells in a poorly immunogenic tumor model. Cancer Immunol Res 3, 149-160 (2015)). Microsatellite instability and high mutational load generate potential endogenous antigens that are expressed in certain cancers (K. W. Mouw, M. S. Goldberg, P. A. Konstantinopoulos, A. D. D'Andrea, DNA Damage and Repair Biomarkers of Immunotherapy Response. Cancer Discov 7, 675-693 (2017 )). T cells recognize these turbidities

- 60 045943 peptides as neoantigens (M. Efremova, F. Finotello, D. Rieder, Z. Trajanoski, Neoantigens Generated by Individual Mutations and Their Role in Cancer Immunity and Immunotherapy. Front Immunol 8, 1679 (2017)). However, in isolated conditions, the presentation of these antigens is not sufficient to promote sustained activation of T cells to generate antitumor activity (S. Spranger, R. Wao, T. F. Gajewski, Melanoma-intrinsic beta-catenin signaling prevents anti-tumour immunity. Nature 523 , 231-235 (2015)). This is likely due to the immunoinhibitory tumor microenvironment.

Herein, we describe a novel anti-TSAxCD28 bispecific immunotherapy targeting the prostate cancer TSA PSMAxCD28, which in combination with a PD-1 blocking antibody induced long-lasting antitumor immunity and promoted sustained intratumoral T cell activation and T cell memory without evidence of systemic cytokine release in animal tumor models. Toxicology studies in genetically humanized immunocompetent mice and cynomolgus monkeys have demonstrated that these bispecifics do not exhibit toxicity alone or in combination with anti-PD-1 antibody. Taken together, these data suggest that combining this class of CD28-based bispecific agents (TSAxCD28) with PD-1 inhibition may provide well-tolerated, off-the-shelf biologic solutions with markedly enhanced, specific, and synergistic antitumor activity.

Material and methods.

In Example 13 the following materials and methods were used.

Research plan.

An example of one of the objectives of the present invention is to develop a bispecific antibody to TSAxCD28 and demonstrate that TSAxCD28 enhances PD-1-induced T cell activation in vitro and safely enhances antitumor efficacy in vivo. In vitro activity was demonstrated by imaging the bispecific antibodies, PD-1 and CD28, localized at the immunological synapse of T cell-target cell conjugates, enhancing T cell cytokine release by PD-1. Antitumor efficacy in vivo was assessed using a syngeneic tumor model in mice. The dynamics of tumor volumes and serum cytokine levels were monitored to demonstrate response to bispecific antibody treatment. One of the objectives of the cynomolgus macaque studies was to determine the safety and tolerability (pharmacological and toxicological profile) of TSAxCD28 as monotherapy or in combination with PD-1 in non-human primates. Animals were examined for toxicity through clinical observations and collection of blood samples for analysis of serum cytokine levels and T-cell phenotype.

Animal studies.

All procedures were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals. The protocols were approved by the Regeneron Pharmaceuticals Institutional Animal Care and Use Committee.

Cell lines.

Clone Jurkat E6-1 (ATCC, no. TIB-152), Raji (ATCC, no. CCL-86™), cell line HEK293 (ATCC, no. CRL-1573) and A-431 (ATCC, CRL-1555™) were cultured in in accordance with the recommended ATCC protocol. The HEK293/hCD20 cell line was generated using a mammalian expression vector encoding the ubiquitin promoter-driven hCD20 (amino acids M1 to P297 under accession number NP_068769.2). Transfected cells were cultured in 500 μg/ml Geneticin A (G418) to select stably expressing cell lines. To obtain a cell expressing hCD80 or hPDL1, we used a lentiviral plasmid encoding human CD80 (length 288 aa; registration no. NM_005191.4) and the neomycin resistance gene, or human PDL1 (length 290 aa; registration no. NM_14143.4) and the resistance gene to puromycin to transfect HEK293T cells, promoting the production of viral particles that were subsequently used to infect HEK293/hCD20 or Raji cells. Human CD80- or PDL1-positive cells were isolated by FACS. Jurkat cells were transduced with NFkB-Luc using lentivirus from Qiagen (cat. no. CLS-013L) and a lentiviral plasmid encoding human PD-1 and the puromycin resistance gene. All resulting cell lines were maintained in DMEM+10% FBS+P/S/G+NEAA supplemented with 500 μg/ml G418 and/or 0.5 μg/ml puromycin.

The DU145/hPSMA cell line was generated by transducing DU145 cells (ATCC, NTB-81) with viral particles that were produced by HEK293T cells transfected with a lentiviral plasmid encoding human PSMA (amino acids M1 to A750 under accession number Q04609) and the neomycin resistance gene. After infection, cells were cultured in 500 μg/ml Geneticin A (G418) to select for cells stably expressing PSMA. The established cell line, DU145/PSMA, was maintained in MEM+10% FBS+P/S/G with 500 μg/ml G418.

To create tumor cell lines genetically modified to express co-stimulatory ligands, we used the lentiviral plasmid pLVX with the EF1a promoter, encoding

- 61 045943 mouse CD86 or empty vector and puromycin resistance gene (pLVX.EF1a.CD86-puro and pLVX.EF1a.EV-puro, respectively), to transfect HEK293T cells, promoting the production of viral particles that were subsequently used to infect MC38 (US National Cancer Institute, Laboratory of Tumor Immunology and Biology). Genetically modified cell lines expressing CD86 were isolated using fluorescence-activated cell sorting (FACS). Cells were maintained under conditions recommended by ATCC in the presence of 0.5 μg/ml puromycin. The resulting cell lines were named MC38/CD86 and MC38/EV.

To generate MC38/hPSMA cells, a lentiviral plasmid encoding human PSMA (amino acids M1 to A750 under accession number Q04609) and a neomycin resistance gene was used to transfect HEK293T cells, producing viral particles that were subsequently used to infect parental MC38 cells. Human PSMA-positive cells were isolated using FACS. MC38/hPMA was maintained in DMEM+10% FBS+P/S/G+NEAA supplemented with 500 μg/ml G418.

Amnis Image Stream.

Amnis Image Stream was performed as described in example 7.

Isolation of primary human CD3+ T cells.

Isolation of T cells was carried out as described in example 8.

Release of IL-2 from primary CD3+ T cells in MLR reaction with DU145/PSMA cells.

Previously isolated and frozen human CD3+ T cells were thawed on the day of analysis in stimulating medium (X-VIVO 15 cell culture medium supplemented with 10% FBS, HEPES, NaPyr, NEAA and 0.01 mM BME) containing 50 U/ml benzonase nuclease. Cells were centrifuged at 1200 rpm for 10 min, resuspended in stimulation medium and seeded into 96-well round-bottom plates at a concentration of 1 x 10 ⁵ cells/well. Parental DU145 cells or DU145 cells genetically modified to express human PSMA were treated with 25 μg/ml mitomycin C in primary plating stimulation medium at a concentration of 10 x 10 ⁶ cells/ml. After incubation for 1 hour at 37°C, 5% CO2, mitomycin C-treated cells were washed 3 times with D-PBS containing 2% FBS and added to wells containing CD3+ T cells at a final concentration of 5 x 10 ⁴ cells per well. To prevent potential CD28 agonist activity from occurring through anchoring of the anti-CD28 Fc domain to Fc receptors, a saturating amount of nonspecific human IgG antibody (100 nM each: hIgG1, hIgG4, and hIgG4s) was included in each assay well. Antibodies PSMAxCD28, control nontargeting xCD28, or isotype control hIgG4s were then titrated from 30 pM to 200 nM at a 1:3 dilution and added to the wells. The 10-point dilution endpoint contained no titratable antibodies. Because DU145 cells endogenously express PD-L1, the effect of T cell suppression by PD-1 was assessed by adding a constant concentration of 20 nM of the PD-1 antagonist REGN2810 to the wells. A condition in the absence of PD-1 inhibition was also included, where 20 nM of the isotype-matched hIgG4 control was used instead. The plates were incubated for 72 hours at 37°C, 5% CO2, and then centrifuged to pellet the cells. 50 μl of the supernatant medium was collected, and from this volume, 5 μl was tested in the human IL-2 assay by AlphaLISA according to the manufacturer's protocol. Measurements were made on a Perkin Elmer Envision multichannel tablet analyzer. A calibration curve of known IL-2 concentrations was constructed to extrapolate the pg/mL IL-2 or IFNy generated in the assay wells. All serial dilutions were tested in duplicate. EC ₅₀ antibody values were determined from a four-parameter logistic equation on a 10-point dose-response curve using GraphPad Prism™ software.

FACS-based cytotoxicity assay.

FACS-based cytotoxicity assay was performed as previously described (Example 8).

Studies on syngeneic tumors.

MC38/EV and MC38/CD86 were cultured according to ATCC guidelines. 1x10 ⁶ MC38/EV or MC38/CD86 were implanted subcutaneously into C57BL/6 mice. Mice were treated with anti-PD-1 antibody (RPM114, BioXcell) or rat isotype control IgG2a (BioXcell) at a dose of 5 mg/kg by intraperitoneal injection on days 0, 3, 7, 10, and 14 after tumor implantation. Tumor sizes were measured twice weekly using a caliper (Roboz RS-6466). Tumor volume was calculated using the formula X*Y*(X/2), where Y is the largest dimension and X is the dimension perpendicular to it. Mice with tumors larger than 2000 mm ³ or with ulcerated tumors were killed by CO2 asphyxiation.

Humanized hCD3/hCD28/hPSMA mice were prepared as described in Example 10. Antibody treatment was also similar, except that the antibodies or bispecific antibodies were administered as monotherapy or in combination by intraperitoneal injection on days 0, 7, and 14 (prophylactic treatment) or on days 9, 13 and 22 (delayed treatment) at a dose of 5 mg/kg.

Ex vivo tissue cytokine analysis.

On day 29 postimplantation, CD3/CD28/PSMA mice bearing MC38/hPSMA tumors were sacrificed by carbon dioxide asphyxiation. Spleens and tumors were collected and stored in medium on ice. All subsequent steps were performed on ice or at 4°C unless otherwise noted. Tumors were cut into small pieces and the fragments were converted into single cell suspensions using the Miltenyi Mouse Tumor Dissociation Kit according to the manufacturing protocol (Miltenyi 130-096-730). Spleens were reduced to a single cell suspension using MACS mild mechanical dissociation (Spleen Program 4) and rubbing through a 70 μm filter using the rubber end of a 3 ml syringe. Cells were pelleted by centrifugation at 1200 rpm for 5 min. Erythrocytes were lysed by resuspending the cell pellet in 1 ml of ACK (ammonium chloride-potassium) lysis buffer and incubating on ice for 5 min. Lysis buffer ACK was quenched with FACS buffer. Cells were pelleted by centrifugation at 1200 rpm for 5 min. The cell suspension was resuspended in 1 ml of medium and seeded in 0.2 ml into 96-well plates (20-400 thousand tumor cells or 50-70 thousand spleen cells). Cells were incubated overnight at 37°C and the culture supernatant was collected. Cytokine levels in tissue culture supernatant were measured using the V-Plex Proinflammatory MSD kit according to the manufacturer's protocol (Meso Scale Diagnostics K15048D-4). The number of cells seeded per well was determined using FACS analysis. Cytokine levels were normalized to the number of cells plated. Calibration microspheres were analyzed along with cells to accurately measure cell numbers using the following calculation:

Number of cells = (number of microspheres injected x number of cells counted by FACS)/number of microspheres counted by FACS.

Measurement of cytokine levels in mouse serum.

Measurements of cytokine levels in mouse serum were performed as described previously (Example 10).

Analysis by flow cytometry.

For flow cytometry analysis of in vivo experiments, tumors were harvested, single cell suspensions were prepared, and red blood cells were lysed using ACK lysis buffer (ThermoFisher Scientific). Discrimination between live and dead cells was performed using the Live/dead fixable blue cell staining kit (ThermoFisher Scientific). Samples were acquired on Symphony (BD Bioscience) and analyzed using Cytobank software (Cytobank, Santa Clara, CA). The analysis was performed with an equal number of events per sample. Event rates were determined by the sample with the fewest recorded events. The CITRUS assay from Cytobank was used to automatically cluster T cells based on specific markers.

Toxicological studies on cynomolgus macaques.

The cynomolgus monkey study was performed as previously described (Example 11).

Results.

PD-1 checkpoint inhibition increases the relative abundance of CD28 at the immunological synapse, allowing TSAxCD28 bispecifics to markedly enhance the ability of anti-PD-1 antibody to stimulate T cell activation in vitro.

To test whether co-stimulatory bispecific agonists can complement checkpoint inhibition, an exemplary PSMAxCD28 bispecific of the present invention (bs16429D) was tested for its ability to enhance the efficacy of PD-1 blockade on TCR/CD3-dependent T cell activation. Indeed, efficient T cell activation depends on the co-clustering of TCR/CD3 and CD28 complexes at the immune synapse (IS). However, activation signals from both TCR/CD3 and CD28 are directly inhibited by phosphorylation of PD-1-Shp2 with subsequent clustering of PD-1/PD-L1 at the synapse (E. Hui et al., T cell costimulatory receptor CD28 is a primary target for PD-1-mediated inhibition. Science 355, 1428-1433 (2017); J. M. Chemnitz, R. V. Parry, K. E. Nichols, S. H. June, J. L. Riley, SHP-1 and SHP-2 associate with immunoreceptor tyrosine-based switch motif of programmed death 1 upon primary human T cell stimulation, but only receptor ligation prevents T cell activation. J Immunol 173, 945-954 (2004)). To determine the relative localization of CD28 and PD-1 in the IS, an in vitro Amnis Image Stream assay was developed as described in Example 7 using Jurkat T cells overexpressing PD-1 and Raji target tumor cells genetically modified to overexpress PD-1. L1. Fluorescently labeled bispecific antibody CD20xCD3 (E. J. Smith et al., A novel, native-format bispecific antibody triggering T-cell killing of B-cells is robustly active in mouse tumor models and cynomolgus monkeys. Sci Rep 5, 17943 (2015)) was used to recapitulate MHC peptide binding to the TCR and to visualize T cell interactions with IS-forming target cells. Two different fluorescently labeled monoclonal antibodies to PD-1 (PD-1 mAb), blocker (PD-1 mAb, REGN2810, (E. Burova et al., Characterization of the Anti-PD-1 Antibody REGN2810 and Its Antitumor Activity in Human PD-1 Knock-In Mice. Mol Cancer Ther 16, 861-870 (2017)) and a non-blocking antibody (NB anti-PD-1 mAb) were used to simultaneously block interaction with PD-L1 and visualize PD-1 localization. found that in the absence of PD-L1 expression on target cells and in the presence of either PD-1 mAb, there was no change in the relative abundance of PD-1 or CD28 in the IS (Fig. 14).However, PD-L1 expression on cells- targets in the presence of a non-blocking PD-1 mAb decreased CD28 content at the synapse and promoted high accumulation of PD-1 in the IS.In contrast, in the presence of a blocking PD-1 mAb, PD-1 localization was significantly reduced and the relative levels of CD28 in the IS were maintained. Distribution of PD-1 and CD28 quantitatively

- 63 045943 but was assessed by calculating the ratio of antibody staining within the IS and outside it (Fig. 15). These data indicate that expression of PD-L1 on target cells increases PD-1 localization to the IS while decreasing CD28 levels. In addition, PD-1 blocking mAb reduces the relative amount of PD-1 in the IS and restores CD28 levels, thereby increasing the relative ratio of CD28 to PD-1.

It was then tested whether PSMAxCD28 could enhance the effects of PD-1 blockade on T cell activation-induced tumor cell killing. For this purpose, the prostate cancer line 22RV1 was used. 22RV1 cells endogenously express PSMA and have been genetically modified to express PD-L1 (22RV1/PD-L1). Due to the lack of TCR/CD3 stimulation by the allogeneic T cell response, PSMAxCD3 (US Patent No. 10179819) was used to provide the primary stimulus. In co-cultures of human peripheral blood mononuclear cells (PBMCs containing human T cells) and 22RV1/PD-L1 cells, PSMAxCD3 alone induced approximately 40% tumor cell killing (Fig. 16, diamonds, EC50 was 8E-10). Addition of PD-1 mAb to PSMAxCD3 increased tumor cell killing to approximately 55% (Fig. 16, solid triangles, EC ₅₀ was 4E-10). Interestingly, the combination of PSMAxCD28 and PSMAxCD3 similarly increased the tumor cell killing rate to approximately 55% with increased efficiency (Fig. 16, solid circles, EC ₅₀ was 6E-11), suggesting that PSMAxCD28 is able to overcome PD-1-mediated /PD-L1 inhibition. Interestingly, the triple combination of PSMAxCD28, mAb to PD-1 and PSMAxCD3 provided the strongest tumor cell killing at approximately 70%, indicating a synergistic effect of this combination (Fig. 16, solid squares, EC50 was 7E-11). As expected, neither PSMAxCD28 alone, nor the anti-PD-1 mAb, nor the combination of PSMAxCD28 plus anti-PD-1 mAb induced any tumor cell killing (Fig. 16, open symbols). In agreement with this, a maximum increase in ZFNy release was observed when treated with the triple combination (Fig. 17).

The effect of the combination of PSMAxCD28 and anti-PD-1 mAb on the activation of primary human T cells in vitro was then determined. A mixed lymphocyte culture reaction (MLR) was used to reproduce physiological PD-L1 expression and TCR/CD3 stimulation. In unidirectional MLR, incompatibility of allogeneic determinants leads to T cell activation, which can be quantified by cytokine production. Here, T cells from healthy donors were incubated with DU145/PSMA cells, a genetically modified prostate cancer cell line endogenously expressing PD-L1 and overexpressing PSMA, and the indicated antibodies (FIG. 18). In the presence of DU145/PSMA cells and T cells, bispecific PSMAxCD28 resulted in a dose-dependent increase in IL-2 release of approximately 3-4 fold compared to the IgG4 isotype control (FIG. 18, circles). Similarly, addition of 20 nM mAb to PD-1 in the MLR assay also increased IL-2 release by approximately 3-4 fold compared to the IgG4 isotype control (FIG. 18, triangles). The combination of PSMAxCD28 and 20 nM anti-PD-1 mAb markedly enhanced PSMAxCD28 bispecific-induced activation, with maximal IL-2 levels being increased approximately 20-fold compared with the isotype control (Figure 18, squares), suggesting that the bispecific PSMAxCD28 in combination with a PD-1 blocking mAb potently and synergistically activates T cells in the presence of tumor cells with endogenous levels of TCR/CD3 activation and PD-L1 inhibition.

Overall, these results demonstrated that PSMAxCD28 bispecific can effectively enhance the ability of anti-PD-1 mAb to promote T cell activation under TCR/CD3 signaling (driven by CD3 bispecific or allogeneic response), resulting in increased cytokine release and tumor cell killing. , expressing PSMA and PD-L1 in vitro.

Overexpression of the natural CD28 ligand on tumor cells acts synergistically with anti-PD-1 mAb treatment to induce CD8 T cell-mediated long-lasting antitumor immunity in vivo.

To determine whether the interaction of CD28 with its natural ligand(s) could enhance the antitumor efficacy of anti-PD-1 mAbs in vivo, MC38 tumor cells were genetically modified to overexpress CD86, one of the co-stimulatory ligands for CD28 (data not shown). The combination of MC38/CD86 cells and anti-PD-1 mAb treatment significantly inhibited tumor growth (Figure 19A), resulting in complete tumor regression associated with a sustained survival benefit (Figure 19B) compared to negative control MC38 transfected cells control empty vector (MC38/EV). Depletion of CD8+ T cells during treatment resulted in complete reversal of the antitumor efficacy achieved by combining anti-PD-1 mAb therapy with MC38/CD86 cells, suggesting a requirement for CD8+ T cells (Figure 19C). Of note, tumor-free mice initially implanted with MC38/CD86 cells and treated with anti-PD-1 mAb rejected a second parental MC38 tumor implanted more than 60 days after implantation of the primary tumor, indicating the presence of a memory T cell response ( Fig. 19D). Therefore, these data demonstrate that the synergistic effect of constitutive expression of CD28 ligand and anti-PD-1 antibody therapy can lead to durable CD8-dependent antitumor immunity in vivo.

- 64 045943

PSMAxCD28 acts synergistically with anti-PD-1 mAb treatment to induce antitumor immunity in a syngeneic tumor model.

The results described above were then extended to demonstrate the antitumor efficacy of treatment using the anti-TSAxCD28 bispecific antibody alone or in combination with an anti-PD-1 mAb in syngeneic tumor models. Using the established syngeneic MC38 tumor model in C57BL6 as described herein, the hPSMA gene (pLVX.EFla.hPSMA) was genetically introduced into MC38 cells to generate the tumor-specific antigen MC38/hPSMA as described herein. To avoid the possibility that mice would spontaneously reject these otherwise syngeneic tumors expressing the administered human tumor antigen, PSMA was genetically humanized in these mice. In addition, the CD3y-5-e and CD28 genes were also humanized using VelociGene technology as described previously (D. M. Valenzuela et al., High-throughput engineering of the mouse genome coupled with high-resolution expression analysis. Nat Biotechnol 21, 652- 659 (2003); W. T. Poueymirou et al., F0 generation mice fully derived from gene-targeted embryonic stem cells allowing immediate phenotypic analyzes. Nat Biotechnol 25, 91-99 (2007)), so that bispecifics recognize host T cells (via hCD3 or hCD28), as well as human tumor antigens in both healthy tissues and tumors (for example, hPSMA), simulating a real clinical situation (Y. Kinoshita et al., Expression of prostate-specific membrane antigen in normal and malignant human tissues World J Surg 30, 628-636 (2006). In this model, the combination of bispecific PSMAxCD28 and anti-PD-1 mAb provided the best tumor growth control, resulting in a sustained survival benefit.

Acute treatment with a combination of anti-CD28 X anti-PSMA and anti-PD1 antibody synergistically inhibits tumor growth.

In the immediate treatment study, anti-CD28 x anti-PSMA, anti-PD1, or rat isotype control IgG2a were administered as monotherapy or in combination by intraperitoneal injection on days 0, 7, and 14 at a dose of 5 mg/kg. Tumor growth dynamics were monitored using calipers to measure X and Y diameters. Tumor volume (X*Y*(X/2)) was calculated. Mice were sacrificed when tumors grew to a size greater than 2000 mm ³ .

As shown in FIG. 20A-20E, anti-CD28 x anti-PSMA antibodies suppressed tumor growth and increased survival in mice when given alone, but also acted synergistically with anti-PD-1 antibodies to promote tumor rejection and increase survival even more than when given alone. antibodies to PD-1.

Treatment with a combination of CD28xPSMA and PD1 induces long-lasting antitumor immunity to tumor reimplantation.

Additionally, tumor-free mice implanted with a combination of bispecific PSMAxCD28 and anti-PD-1 mAb rejected a second parental tumor cell line, MC38, that was implanted more than 60 days after the initial tumor implantation, demonstrating the formation of immunological memory ( Fig. 21A), which is consistent with the observations described herein. These results suggest that the endogenous antigen-specific TCR signal (signal 1) is generated by the peptide-MHC complex toward implanted MC38/PSMA tumor cells. It has been shown that MC38 tumor cells express high levels of reactivated endogenous retroviral proteins, such as p15E, and that intratumoral T cells in C57BL6 mice respond to this p15E antigen (J. C. Yang, D. Perry-Lalley, The envelope protein of an endogenous murine retrovirus is a tumor-associated T-cell antigen for multiple murine tumors. J Immunother 23, 177-183 (2000); H. J. Zeh, 3rd, D. Perry-Lalley, M. E. Dudley, S. A. Rosenberg, J. C. Yang, High avidity CTLs for two self-antigens demonstrate superior in vitro and in vivo antitumor efficacy. J Immunol 162, 989-994 (1999). Consistent with this finding, it was found that combination therapy of PSMAxCD28 and anti-PD-1 mAb induced peripheral T cell reactivity to at least one endogenous P15E antigen, confirming the formation of antitumor immunological memory (Fig. 21B). Notably, in similar experiments where treatment with PSMAxCD28 in combination with PSMAxCD3 induced tumor rejection of MC38/hPSMA, these tumor-free mice failed to reject recurrent tumors (Figure 21C). Overall, these data demonstrate that CD28 bispecifics act synergistically with anti-PD-1 mAbs in this PSMA-expressing syngeneic tumor model and can enhance endogenous TCR/CD3-dependent T cell responses.

Therapeutic treatment with a combination of anti-CD28 x anti-PSMA and anti-PD1 antibody synergistically inhibits tumor growth.

Similarly, in a delayed treatment protocol, the combination of PSMAxCD28 and anti-PD-1 mAb 10 days after implantation inhibited the growth of established MC38/hPSMA tumors (Fig. 22A), and also resulted in a significant increase in survival (Fig. 22C) and a decrease in tumor volumes (Fig. 22C). Fig. 22D). Interestingly, tumor-targeted combination therapy selectively increased intratumoral cytokine levels, as demonstrated by IFNγ (FIG. 22B). In these same mice, no induction of splenic or systemic cytokines was observed (FIGS. 23A and 23B). Besides,

- 65 045943 expression profiling of markers of T cell activation showed that intratumoral PD-1 expression increased after treatment with PSMAxCD28 (Fig. 24), which supports this combination treatment. To further characterize responding T cell subsets following combination treatment, tumor-infiltrating CD8+ T cells were profiled at day 17 post tumor implantation using high-dimensional flow cytometry. CITRUS (cluster identification, characterization, and regression analysis) was used to independently stratify statistically significantly different T-cell clusters. It is important to note that PD-1 blockade was found to cause an expansion of effector (CD44 ^beta CD62L ^low ) CD8' T cells (cluster C1) expressing high levels of activation/exhaustion markers (PD-1, TIM3, LAG3, Ki67) (Fig. 25). However, only combination treatment was able to stimulate the expansion of intratumoral CD8+ T cells (cluster C2) with a memory cell-like phenotype (high Tcf1, EOMES, CD62L, medium CD122 and CD127) and a less exhausted phenotype (low PD-1, LAG3, TIM3, CD38, KLRG1, higher CD5) (Fig. 25) (M. Philip et al., Chromatin states define tumor-specific T cell dysfunction and reprogramming. Nature 545, 452-456 (2017)). These data demonstrate that combination therapy with bispecific PSMAxCD28 and anti-PD-1 antibody induces durable antitumor immunity associated with intratumoral activation of T cells with a memory cell-like phenotype and a survival benefit.

TSAxCD28 alone or in combination with anti-PD-1 mAb therapy does not induce systemic T cell activation compared with an anti-CD28 superagonist in cynomolgus monkeys.

Data from an earlier clinical study showed that bivalent CD28-activating antibodies, termed CD28 superagonists (CD28-SA), activated a broad spectrum of T cells and resulted in significant toxicity associated with cytokine release syndrome (CRS) in a group of healthy volunteers (G. Suntharalingam et al., Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. N Engl J Med 355, 1018-1028 (2006)). Exploratory studies were conducted in genetically modified triple-humanized mice and cynomolgus monkeys to evaluate the tolerability of TSAxCD28 bispecifics alone or the potential for synergistic pharmacological effects in combination with anti-PD-1 mAbs. Three monkeys per treatment group received a single dose (10 mg/kg) of PSMAxCD28 alone or in combination with anti-PD-1 mAb (REGN2810) (10 mg/kg) by intravenous infusion (combination treatment groups received sequential infusions) (Table 20 , Fig. 26A-26C and 27).

In addition, three monkeys per group received a single dose (0.1 mg/kg) of a bispecific anti-PSMAxCD3 antibody and a single dose (10 mg/kg) of a superagonist anti-CD28 antibody as described herein (Table 20, FIG. 26A -26C and 27). Toxicity assessment was based on clinical observations, qualitative indicators of feed intake, body weight, vital signs (body temperature, heart rate, pulse oximetry and respiratory rate), laboratory diagnostics and pathological anatomy at the end of the experiment. Blood samples were collected for cytokine analysis and FACS immunophenotyping. PSMAxCD28 alone or in combination with PD-1 was well tolerated, and all animals survived throughout the study period. There were no clinical observations associated with the study drug (data not shown). There were no changes in organ weights or any macroscopic changes at necropsy after sacrifice (data not shown). In addition, no significant cytokine release, margination or activation of T cells was observed (Table 23, Figs. 26A-26C). In contrast, significant cytokine release, lymphocyte margination, and T cell activation were observed in monkeys treated with CD28 superagonist alone (FIGS. 29A and 29B). Abundant infiltration of immune cells was observed in the kidneys, brain, and seminal vesicles of animals treated with the CD28 superagonist. In contrast, no significant treatment-related histological changes were observed in animals treated with PSMAxCD28 alone or in combination with PD-1 (data not shown).

Table 23. No significant clinical observations or cytokine release were observed with anti-PSMAxCD28 alone or in combination with anti-PD-1 antibody in an exploratory single-dose toxicity study in monkeys.

- 66 045943

Table 23

Absolutely Activi precise Prolife Rovann Molecules a Description Dose (mg/k g) 1 day wedge isches would be nab l-ya number of T cells (U/μl) Rating ΚΪ67+ T cells (U3/μl) 1st I cells (U/µl) -ICOS CRP (mg/dl) Plasma cytokines 5 hours after dosing (pg/ml) cues _{etc etc} Etc 96 observable -I 2-4 days barely th tes 5 hours barely th tes 96 barely th tes h 24 hours IL-6 IL- 8 IL- 10 IF N- T TN F-a IL -2 IL -4 IL- 5 T T T 4.1 5 0.1 1 0.2 4 1.21 0.4 2 0.8 2 13.6 4449 53 1 287 8 15 9 B.L. Q* BL Q BL Q BL Q bsl5780 D antiPSMAxCD 3 0.1 arching 7.4 1 0.1 5 0.8 8 1.69 0.4 0 0.6 4 14.9 1123 9 39 10 221 40 18 00 BL Q 35 7 BL Q 15 2 5.6 8 0.0 1 0.4 2 3.87 0.6 8 3.7 5 15.4 1052 9 77 6 247 84 eleven 76 BL Q BL Q BL Q BL Q bs16429 anti- 1 - - 6.5 7.2 0.6 0.41 o, s 0.3 0.5 BLQ B.L. B.L. B.L. B.L. B.L. B.L. B.L. D PSMAxCD 8 2 2 3 2 Q Q Q Q Q Q Q 28 - - 4.9 1 5.0 9 0.4 2 0.23 0.2 6 0.1 7 0.5 BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q - - 4.5 0 4.0 4 0.6 3 0.39 0.5 1 0.4 8 1.1 BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q bs16429 D anti- PSMAxCD 10 - - 6.0 3 4.6 4 5.7 3 4.5 3 0.3 3 0.3 8 0.44 0.60 0.3 3 0.5 8 0.2 2 0.5 0 0.2 0.2 BLQ BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q - - Yu, 06 7.1 6 0.8 8 0.49 0.8 4 0.4 8 1.3 BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q REGN23 29 (TeGener o) superagon ist k CD28 10 - - 7.0 2 7.0 6 i, 87 0.2 6 0.1 9 0.6 6 0.6 0 0.7 9 1.1 3 2.57 1.55 3.61 0.5 1 0.4 3 0.4 8 1.9 4 1.1 1 2.0 9 15.4 15 15.3 764 2907 5226 BL Q 10 95 53 24 336 4 171 55 791 8 BL Q 33 3 15 09 BL Q BL Q 163 BL Q 23 8 52 2 BL Q BL Q 34 2 eleven 69 61 5 19 8 bs16429 anti- - - 3.7 6 1.9 0 o, s 8 0.37 o, s 4 0.2 8 10.8 BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q D+REGN 2810 PSMAxCD 28 +PD-1 10+10 - - 3.6 6 2.0 7 0.5 4 0.52 0.2 2 0.2 0 1 BLQ BL Q BL Q BL Q 195 BL Q 32 3 BL Q - - 3.5 1.3 o, s 0.40 0.1 0.1 I,9 BLQ B.L. B.L. B.L. B.L. B.L. B.L. B.L. 4 4 9 2 0 Q Q Q Q Q Q Q bsl6431 D+REGN 2810 anti- PSMAxCD 28 +PD-1 10+10 3.6 3 3.5 3 4.1 6 4.2 3 1.9 0 3.1 6 o,z 6 o,z 5 0.4 2 0.42 0.19 0.32 0.2 4 0.1 3 0.2 6 o, z 8 0.1 5 o,z 3 5.1 2.9 0.2 BLQ BLQ BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q BL Q B.L. Q 187 BL Q BL Q BL Q 33 3BL Q 32 3 BL Q BL Q BL Q - - 6.2 1 5.6 2 0.6 0 0.58 0.3 2 0.3 9 0.8 BLQ BL Q BL Q BL Q 151 BL Q 33 5 BL Q mAbl015 4P3 EGFRVIII Ab - - 7.2 2 7.7 2 0.8 4 0.77 0.3 9 0.4 1 AND BLQ BL Q BL Q BL Q 267 BL Q 37 8 BL Q - - 3.5 5 4.6 1 0.5 1 0.75 0.2 6 0.4 4 0.2 BLQ BL Q BL Q BL Q BL Q BL Q BL Q BL Q

*BLQ: Below lower limit of quantitation

Consistent with the above studies in cynomolgus monkeys, tumor-bearing or tumor-free naïve triple humanized mice (hCD3/hCD28/hPSMA) administered PSMAxCD28, alone or in combination with PD-1, did not exhibit increased cytokine levels (Fig. 27 and figs 29A and 29B). On the contrary, administration of a CD28 superagonist induced significant

- 67 045943 increase in the levels of ZFNy, TNFa, IL-2, IL-4 and IL-5 4 hours after dosing (Fig. 27 and Figs. 29A and 29B). Consistent with the results described above, the TSAxCD28 bispecific (as well as the parental non-superagonist bivalent anti-CD28 antibodies used to generate these bispecifics) were previously shown not to induce human T cell proliferation in the FDA-recommended in vitro application assay. dry and wet (R. Stebbings, D. Eastwood, S. Poole, R. Thorpe, After TGN1412: recent developments in cytokine release assays. J Immunotoxicol 10, 7582 (2013)) compared with the strong proliferation induced by CD28-SA (data not shown). Overall, these data suggest that TSAxCD28 bispecifics are largely devoid of immunogenicity.

Additionally, as previously described (Example 10, FIG. 28), treatment with anti-PSMAxCD28, alone or in combination with anti-PD1 antibody, did not increase serum cytokine levels in tumor-bearing mice, whereas treatment with anti-PSMAxCD3, alone or in combination with an anti-PD1 antibody, increased serum cytokine levels in tumor-bearing mice.

Discussion.

Herein, a novel tumor-targeted immunotherapy using TSAxCD28 bispecifics in combination with a PD-1 blocking monoclonal antibody is proposed and tested, where the immunotherapy induces long-term antitumor immunity and promotes sustained intratumoral T cell activation in animal tumor models. Toxicology studies in genetically humanized immunocompetent mice and cynomolgus macaques have demonstrated that these bispecifics do not exhibit toxicity alone or in combination with an anti-PD-1 monoclonal antibody, suggesting that this therapeutic approach may provide a well-tolerated, ready-to-use biological solutions with markedly enhanced, specific and synergistic antitumor activity.

Checkpoint inhibition with a PD-1 blocking monoclonal antibody is known to relieve restrictions on T cell activation, but the effectiveness of these approaches as monotherapy is often insufficient to achieve tumor clearance and durable antitumor responses in many cancers. Several approaches are currently being studied to improve response rates to PD-1 inhibition. Thus, identification of biomarkers to predict susceptibility to PD-1 mAbs (R. Cristescu et al., Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy. Science 362, (2018)), non-tumor-targeted combination therapies using PD-1 inhibition together with agonistic antibodies that trigger co-stimulatory receptors to improve T cell activation, or with chemotherapy or radiation therapy, are currently in preclinical and clinical studies (S. Hu-Lieskovan, A. Ribas, New Combination Strategies Using Programmed Cell Death 1/Programmed Cell Death Ligand 1 Checkpoint Inhibitors as a Backbone. Cancer J 23, 10-22 (2017); Y. K. Chae et al., Current landscape and future of dual anti-CTLA4 and PD1/PD-L1 blockade immunotherapy in cancer; lessons learned from clinical trials with melanoma and non-small cell lung cancer (NSCLC). J Immunother Cancer 6, 39 (2018); P. S. Chowdhury, K. Chamoto, T. Honjo, Combination therapy strategies for improving PD- 1 blockade efficacy: a new era in cancer immunotherapy. J Intern Med 283, 110-120 (2018)). However, the problem is that many of these combinations are often based on the availability of existing drugs and post hoc judgments about the rationality of combining therapies, rather than a true hypothesis-driven approach, which in some cases has led to poorer patient outcomes (M. J. Ahn, J. M. Sun, S. H. Lee, J. S. Ahn, K. Park, EGFR TKI combination with immunotherapy in non-small cell lung cancer. Expert Opin Drug Saf 16, 465-469 (2017)). It is clear that checkpoint inhibition and reactivation of the immune system provides the possibility of long-term remission in a number of patients (J. S. Weber et al., Nivolumab versus chemotherapy in patients with advanced melanoma who progressed after anti-CTLA-4 treatment (CheckMate 037): a randomized , controlled, open-label, phase 3 trial. Lancet Oncol 16, 375-384 (2015); S. L. Topalian et al., Survival, durable tumor remission, and long-term safety in patients with advanced melanoma receiving nivolumab. J Clin Oncol 32, 1020-1030 (2014); M. A. Postow, M. K. Callahan, J. D. Wolchok, Immune Checkpoint Blockade in Cancer Therapy. J Clin Oncol 33, 1974-1982 (2015); M. R. Migden et al., PD-1 Blockade with Cemiplimab in Advanced Cutaneous Squamous-Cell Carcinoma N Engl J Med 379, 341-351 (2018)), therefore options to further improve or increase T cell activity to ensure longer-lasting responses are needed. Herein, to improve the antitumor efficacy of anti-PD-1 mAbs, the concept of using the bispecific TSAxCD28 to enhance T cell activation and signaling was proposed. Thus, this new combination immunotherapy was tested using a tumor target (eg, PSMA), and it was demonstrated that co-stimulatory bispecific antibodies to CD28 act synergistically with mAbs to PD-1, not only generating sustained T cell activation, but also providing long-lasting antitumor responses without systemic toxicity. Therefore, this tumor-targeted combination therapy may provide a significant advantage over the non-targeted approaches described previously. The use of CD28-bispecific antibodies, which do not directly activate CD28, except in cases where it is clustered on the surface of tumor cells, makes it possible to provide co

- 68 045943 stimulation only at the tumor site, avoiding the systemic toxicity of conventional CD28-activating antibodies (G. Suntharalingam et al., Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. N Engl J Med355, 1018-1028 ( 2006)), which is often observed in the case of a combination of CLTA-4 and PD-1 blockade (J. Larkin et al., Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. N Engl J Med 373, 23-34 (2015); D. B. Johnson et al., Fulminant Myocarditis with Combination Immune Checkpoint Blockade. N Engl J Med 375, 1749-1755 (2016); M. H. Pollack et al., Safety of resuming anti-PD-1 in patients with immune-related adverse events (irAEs) ) during combined anti-CTLA-4 and anti-PD1 in metastatic melanoma. Ann Oncol 29, 250-255 (2018)) or other costimulatory agonistic divalent antibodies (N. H. Segal et al., Results from an Integrated Safety Analysis of Urelumab, an Agonist AntiCD137 Monoclonal Antibody. Clin Cancer Res 23, 1929-1936 (2017)). Toxicology studies in genetically humanized immunocompetent mice, as well as cynomolgus monkeys, demonstrated that these bispecifics do not exhibit toxicity when used as monotherapy or in combination with anti-PD-1 mAbs. The safety profile, together with the increased antitumor efficacy of the anti-PSMAxCD28 bispecific antibodies of the invention with anti-PD-1 mAb in syngeneic models, suggests that this therapeutic modality is robust and may have broader application as a new class of immunotherapeutic combination.

To enhance T cell-mediated killing of tumor cells, tumor-targeting approaches are being developed (E. Dahlen, N. Veitonmaki, P. Norlen, Bispecific antibodies in cancer immunotherapy. Ther Adv Vaccines Immunother 6, 3-17 (2018)). Thus, CD3-based bispecific antibodies are a new class of antibodies that can effectively trigger T cell activation by binding the T cell to a tumor cell and activating TCR/CD3 (E. J. Smith et al., A novel, native-format bispecific antibody triggering T-cell killing of B-cells is robustly active in mouse tumor models and cynomolgus monkeys. Sci Rep 5, 17943 (2015)), thereby mimicking conventional signal 1. However, despite promising clinical efficacy, CD3-bispecific may be associated with cytokine release syndrome (CRS) due to direct T cell activation and lack of tumor specificity (S. L. Maude, D. Barrett, D. T. Teachey, S. A. Grupp, Managing cytokine release syndrome associated with novel T cell -engaging therapies. Cancer J 20, 119-122 (2014)). The present invention demonstrates for the first time that combination therapy with bispecific TSAxCD28 and anti-PD-1 mAb induces tumor-specific T cell activation associated with long-term anamnestic response in an immunocompetent mouse tumor model. Bispecific antibodies to TSAxCD28 show virtually no activity in the absence of signal 1, and blockade of PD-1 depends on the response of endogenous antigen-specific T cells to tumor peptides (W. Hugo et al., Genomic and Transcriptomic Features of Response to Anti-PD-1 1 Therapy in Metastatic Melanoma. Cell 165, 35-44 (2016); N. A. Rizvi et al., Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348, 124-128 (2015 ); J. M. Mehnert et al., Immune activation and response to pembrolizumab in POLE-mutant endometrial cancer. J Clin Invest 126, 2334-2340 (2016); D. T. Le et al., Mismatch repair deficiency predicts response of solid tumors to PD- 1 blockade. Science 357, 409-413 (2017). Therefore, signal 1 provided by endogenous tumor antigens is important for combination treatment of PSMAxCD28 and anti-PD-1 mAb. This is in contrast to CD3 bispecifics, which activate T cells regardless of their TCR specificity and therefore cannot generate long-term tumor-specific immunity. Indeed, it was found that although combination treatment of PSMAxCD3 and PSMAxCD28 achieved high antitumor efficacy, it did not generate a strong anamnestic response. In addition, MC38 tumor cells have been shown to express high levels of reactivated endogenous retroviral peptides such as p15E, and C57BL6 mice can generate endogenous T cells that recognize and respond to this neoepitope (J. C. Yang, D. Perry -Lalley, The envelope protein of an endogenous murine tumor retrovirus is an associated T-cell antigen for multiple murine tumors. JImrmmother 23, 177-183 (2000); H. J. Zeh, 3rd, D. Perry Lalley, M. E. Dudley, S. A. Rosenberg, J. C Yang, High avidity CTLs for two self-antigens demonstrate superior in vitro and in vivo antitumor efficacy. J Immunol 162, 989-994 (1999). In the MC38 model described herein, it was demonstrated that combination therapy of PSMAxCD28 and anti-PD1 mAb increased the number of T cells responding to this p15E neoantigen. Furthermore, in the present invention, through extensive profiling of tumor-infiltrating T cells, the combination of PSMAxCD28 and anti-PD-1 mAb was found to produce less dysfunctional CD8 T cells and promote a strong intratumoral memory T cell phenotype. Thus, CD28 bispecifics, together with PD-1 blockade, may enhance endogenous TCR/CD3-dependent T cell responses, inducing durable antitumor responses.

The data herein demonstrate that PD-1 accumulates at the immune synapse when PD-L1 is expressed by target cells, and its accumulation is associated with a decrease in CD28 at the synapse, suggesting that PD-1 may inhibit T cells. preventing CD28 from localizing to the synapse. In addition, the present invention found that PD-1 blockade prevented synaptic localization of PD-1, while the accumulation of CD28 at the synapse was increased, allowing the TSAxCD28 bispecific to markedly enhance the ability of the mAb to target PD-1 stimuli.

-

Claims (23)

stimulate activation of T cells. This may be one mechanism by which a PD-1 blocking antibody promotes T cell activation. Overall, imaging the localization of PD-1 and CD28 at the immunological synapse following PD-1-PD-L1 interaction and/or PD-1 inhibition provides insight into the impact of PD-1 blockade on T cell activation, as well as the synergistic interaction between TSAxCD28 and mAb to PD-1 at the level of the immune synapse. Although anti-PD-1 mAbs represent an important new class of immunotherapy, further optimization of antitumor activity will undoubtedly be important in many settings. Similar to how CAR-T cell approaches used chimeric receptors that artificially activate both signal 1 and signal 2 to improve their antitumor activity (E. A. Zhukovsky, R. J. Morse, M. V. Maus, Bispecific antibodies and CARs: generalized immunotherapeutics harnessing T cell redirection. Cut Opin Immunol 40, 24-35 (2016); A. I. Salter et al., Phosphoproteomic analysis of chimeric antigen receptor signaling reveals kinetic and quantitative differences that affect cell function. Sci Signal 11, (2018)) This invention demonstrates the potential benefit of combining PD-1 inhibition with CD28 bispecifics (which provide signal 2) to enhance antitumor activity. This approach has several practical advantages over CAR-T cell therapy - that it does not require the labor-intensive preparation of cell therapy, which must be individually tailored to each patient, and does not require patients to first undergo lymphodepletion through toxic chemotherapy often associated with adverse events, as a result of which they cannot receive such cell therapy (S. H. June, R. S. O'Connor, O. U. Kawalekar, S. Ghassemi, M. S. Milone, CAR T cell immunotherapy for human cancer. Science 359, 13611365 (2018)). This bispecific approach provides opportunities for increased efficacy as well as improved safety due to the specificity of its action. Taken together, these data suggest that combining CD28-based bispecifics with clinically successful PD-1 mAbs such as cemiplimab may provide well-tolerated, off-the-shelf biologic solutions with markedly enhanced and synergistic antitumor activity. The present invention is not to be limited in scope by the specific embodiments described herein. Indeed, various modifications of the present invention in addition to those described herein will be apparent to those skilled in the art from the foregoing description and accompanying drawings. Such modifications are intended to be within the scope of the appended claims. CLAIM 1. A bispecific antibody comprising a first antigen binding domain that specifically binds human CD28 and a second antigen binding domain that specifically binds human PSMA, wherein the first antigen binding domain contains three heavy chain complementarity determining regions (D1-HCDR1, D1-HCDR2 and D1-HCDR3 ) contained in the first heavy chain variable region (D1-HCVR), and three light chain complementarity determining regions (LCDR1, LCDR2 and LCDR3) contained in the light chain variable region (LCVR), and where the second antigen binding domain contains three complementarity determining regions heavy chain (D2-HCDR1, D2-HCDR2 and D2-HCDR3) contained in the second heavy chain variable region (D2-HCVR), and three light chain complementarity determining regions (LCDR1, LCDR2 and LCDR3) contained in LCVR; where D1-HCDR1 contains the amino acid sequence SEQ ID NO: 12, D1-HCDR2 contains the amino acid sequence SEQ ID NO: 14, D1-HCDR3 contains the amino acid sequence SEQ ID NO: 16, D2-HCDR1 contains the amino acid sequence SEQ ID NO: 4, D2 -HCDR2 contains the amino acid sequence SEQ ID NO: 6, D2-HCDR3 contains the amino acid sequence SEQ ID NO: 8, LCDR1 contains the amino acid sequence SEQ ID NO: 20, LCDR2 contains the amino acid sequence SEQ ID NO: 22 and LCDR3 contains the amino acid sequence SEQ ID NO: : 24. 2. The bispecific antibody according to claim 1, wherein D1-HCVR contains the amino acid sequence of SEQ ID NO: 10 and LCVR contains the amino acid sequence of SEQ ID NO: 18. 3. The bispecific antibody according to claim 1, wherein D2-HCVR contains the amino acid sequence of SEQ ID NO: 2 and LCVR contains the amino acid sequence of SEQ ID NO: 18. 4. The bispecific antibody of claim 1, wherein D1-HCVR contains the amino acid sequence of SEQ ID NO: 10, D2-HCVR contains the amino acid sequence of SEQ ID NO: 2, and LCVR contains the amino acid sequence of SEQ ID NO: 18. 5. A bispecific antibody according to any one of claims 1 to 4, containing a first heavy chain and a paired light chain connected by disulfide bonds, where the first heavy chain contains HCVR and a heavy chain constant region containing C _H 1, CH2 and CH3 domains, and the pair the light chain comprises an LCVR and a light chain constant region, wherein the first heavy chain and the paired light chain comprise a first antigen binding domain; And - 70 045943 a second heavy chain and a paired light chain connected by disulfide bonds, where the second heavy chain contains HCVR and a heavy chain constant region containing CH1, CH2 and CH3 domains, and the paired light chain contains LCVR and a light chain constant region, where the second heavy chain chain and the paired light chain contain a second antigen-binding domain. 6. The bispecific antibody according to claim 5, wherein the first heavy chain or the second heavy chain, but not both, contains a CH3 domain containing the modification H435R (according to the EU numbering) and the modification Y436F (according to the EU numbering). 7. The bispecific antibody according to claim 5 or 6, wherein the heavy chain constant region of the first heavy chain and the heavy chain constant region of the second heavy chain are of the IgG1 isotype. 8. The bispecific antibody according to claim 5 or 6, wherein the heavy chain constant region of the first heavy chain and the heavy chain constant region of the second heavy chain are of the IgG4 isotype. 9. The bispecific antibody of claim 5, wherein the first heavy chain comprises the amino acid sequence of SEQ ID NO: 81 and the paired light chain contains the amino acid sequence of SEQ ID NO: 83, and the second heavy chain contains the amino acid sequence of SEQ ID NO: 82 and the paired light chain contains the amino acid sequence SEQ ID NO: 83. 10. A pharmaceutical composition comprising a bispecific antibody according to any one of claims 1 to 9 and a pharmaceutically acceptable carrier or diluent for use in the treatment of cancer expressing PSMA. 11. An isolated host cell containing nucleic acid molecules encoding HCVR and LCVR specified in any of claims 1-5. 12. An isolated host cell containing expression vectors carrying nucleic acid molecules encoding HCVR and LCVR specified in any of claims 1-5. 13. A method for producing a bispecific antibody according to any one of claims 1 to 9, comprising culturing the isolated host cell according to claim 11 or 12 under conditions allowing the production of a bispecific antibody, and isolating the bispecific antibody thus obtained. 14. The method of claim 13, wherein the method further comprises formulating the bispecific antibody into a pharmaceutical composition with a suitable carrier for use in the treatment of PSMA-expressing cancer. 15. A method of treating cancer expressing PSMA, comprising administering to a subject in need thereof a bispecific antibody according to any one of claims 1 to 9 or a pharmaceutical composition according to claim 10. 16. The method of claim 15, wherein the cancer is prostate cancer. 17. The method of claim 15 or 16, further comprising administering a second therapeutic agent or therapy. 18. The method of claim 17, wherein the second therapeutic agent or therapy comprises an antitumor agent, radiation therapy, an antibody-drug conjugate, a bispecific antibody conjugated to an anticancer agent, a checkpoint inhibitor, or combinations thereof. 19. The method of claim 17, wherein the second therapeutic agent is a checkpoint inhibitor. 20. The method of claim 19, wherein the checkpoint inhibitor targets PD-1 or CTLA4. 21. The method of claim 20, wherein the PD-1 inhibitor is selected from the group consisting of pembrolizumab, nivolumab, atezolizumab and cemiplimab. 22. The method of claim 20, wherein the CTLA4 inhibitor is ipilimumab. 23. The method of claim 17, wherein the second therapeutic agent is a bispecific antibody comprising a first antigen binding domain that binds to human CD3 and a second antigen binding domain that binds to human PSMA. -