A human CD137×PD-L1 bispecific antibody promotes anti-tumor immunity via context-dependent T cell costimulation and checkpoint blockade

Abstract

Immune checkpoint inhibitors demonstrate clinical activity in many tumor types, however, only a fraction of patients benefit. Combining CD137 agonists with these inhibitors increases anti-tumor activity preclinically, but attempts to translate these observations to the clinic have been hampered by systemic toxicity. Here we describe a human CD137xPD-L1 bispecific antibody, MCLA-145, identified through functional screening of agonist- and immune checkpoint inhibitor arm combinations. MCLA-145 potently activates T cells at sub-nanomolar concentrations, even under suppressive conditions, and enhances T cell priming, differentiation and memory recall responses. In vivo, MCLA-145 anti-tumor activity is superior to immune checkpoint inhibitor comparators and linked to recruitment and intra-tumor expansion of CD8 + T cells. No graft-versus-host-disease is observed in contrast to other antibodies inhibiting the PD-1 and PD-L1 pathway. Non-human primates treated with 100 mg/kg/week of MCLA-145 show no adverse effects. The conditional activation of CD137 signaling by MCLA-145, triggered by neighboring cells expressing >5000 copies of PD-L1, may provide both safety and potency advantages.

Fig. 1. Screening of CD137 x PD-1/PD-L1 bAb combinations identifies MCLA-145.

a Left panel activity of CD137xPD-L1 and CD137xPD-1 bispecifics in a CD137 NFκB-luc reporter assay (fold induction of luciferase signal) Right panel experiment repeated with the addition of an Fc binding antibody; b Left panel levels of IL-2 released by activated T cells after 3 days of incubation with eight candidate CD137xPD-L1 bAb combinations in a dose titration in the presence of CHO-PD-L1 cells, Right panel, experiment repeated in the presence wildtype CHO cells; c levels of IL-2, IFNγ and TNFα released by activated T cells after 3 days of incubation with the three best performing CD137xPD-L1 bAbs and control antibodies in a dose titration in the presence of CHO-PD-L1 cells (bAb001 = MCLA-145).

Fig. 2. MCLA-145 binds with high affinity and to a unique epitope on CD137.

a Summary of SPR and cell-based affinity measurements for binding to PD-L1; b Summary of SPR and cell-based affinity measurements for binding to CD137; c SPR affinity analysis showing dual binding of human PD-L1 and CD137 to immobilized MCLA-145. Green curves: human PD-L1 injected first followed by injection of human CD137 or running buffer. Red curves: human CD137 injected followed by human PD-L1 or running buffer; d surface homology model of the CD137 ectodomain in complex with CD137L (PDB: 6MGP), the CRD domains 1–4 are shaded in yellow, green, purple and cyan respectively, CD137L is depicted as a gray ribbon, the backbone and residues of urelumab* (dark red) and utomilumab* (orange) epitopes are highlighted, critical binding residues for MCLA-145 as determined by alanine scanning are shown as solid blue spheres; e binding of labeled CD137L to CHO-CD137⁺ cells was measured by flow analysis in a titration series of MCLA-145 and control antibodies f NFAT activity of PD-1 reporter cells cocultured with CHO-PD-L1⁺ and incubated with a titration of MCLA-145 and control antibodies.

Fig. 3. Effects of MCLA-145 activity on primary T cell function.

a Effect of MCLA-145 on naïve CD8⁺ T cell priming. Graphs show the absolute number of dextramer+ antigen-specific CD8⁺ T cells from four donors after 10 days of co-culture with colored segments indicating the number of T naïve/memory stem cells (CD45RA + CCR7+), central memory T cells (CD45RA-CCR7+), effector memory T cells (CD45RA-CCR7-) and terminally differentiated effector T cells (CD45RA + CCR7−) as determined by flow cytometry (Dotted line represents control co-cultures without peptide priming). b Transactivation assay. EC₅₀ of human T cell cytokine production (IL-2, IFNγ, and TNFα) in the presence of CHO-PD-L1 cells (n = 7); c mixed lymphocyte reaction (MLR) measuring IFNγ production by human donor leukocytes (n = 3 independent donors) stimulated with allogeneic DCs (n = 1) in the presence of 10 µg/ml of test compounds, (error bars are SEM), IFNγ concentrations were significantly greater for MCLA-145 compared to urelumab*, atezolizumab* and Neg. Ctrl Ab (p < 0.001) and significantly less compared to ure* + atezo* (p < 0001) as determined by two-way ANOVA and Holm-Sidak’s multiple comparisons test; d SEB assays. Fold change in IL-2 or IFNγ production by PBMC (left panel, n = 3 independent donors, error bars are mean ± SD) or Whole blood (right panel, n = 7 independent donors, error bars are SEM) respectively after stimulation with SEB in the presence of increasing concentrations of antibodies; e Recall assay measuring fold induction of IFNγ and TNFα production by human CD4+ memory T cells re-stimulated with a viral peptide pool (CEFT) in the presence of 10 µg/ml antibodies, (n = 4 independent donors, error bars are SEM), *p < 0.0001 for IFNγ and p = 0.0012 for TNFα production (Neg. Ctrl Ab vs. MCLA-145) and p = 0.0026 for IFNγ and p = 0.0025 for TNFα production (MCLA-145 vs. ure* + atezo*) as determined by two-way ANOVA and Tukey’s test; f Treg assay measuring fold induction of TNFα and IFNγ production by activated leukocytes in the presence of autologous regulatory T cells (1:1 ratio) and in the presence of 10 µg/ml antibodies (n = 4 independent donors error bars are SD of the mean). *p = 0.0005 for IFNγ and p = 0.012 for TNFα production (Vehicle vs MCLA-145), as determined by one-way ANOVA and Tukey’s test; g PBMC:M2 suppression assay measuring IFNγ production by human leukocytes co-cultured with activated M2-polarized macrophages together with 10 µg/mL antibodies (n = 5 independent donors, error bars are SEM, *p < 0.0001 by one-way ANOVA and Tukey’s test).

Fig. 4. Ex vivo and in vivo activity of MCLA-145.

a Top panel, donut chart of relative proportion of T cell subsets including percentage of CD3+ cells in single-cell suspensions derived from five human endometrial tumors; Bottom panel, IFNγ levels in supernatant of each tumor explant 6 days after antibody incubation. Error bars: mean ± SD of three replicates, ND = not done, **p < 0.0001 in Tumor #1 and Tumor #5, *p = 0.0018 (Neg. Ctrl Ab vs. MCLA-145) and p = 0.0042 (MCLA-145 vs. atezolizumab*); Tumor #2, **p = 0.0007 (Neg. Ctrl Ab vs. MCLA-145); Tumor #4 *p = 0.0074 (MCLA-145 vs. atezolizumab*), determined by one-way ANOVA and Tukey’s test; b Ly95 T cells expressing an NY-ESO-specific TCR were transferred to NSG mice bearing A549-PD-L1^hi tumors and treated with indicated antibodies, end point data are shown, (n = 7 mice); c Stacked histogram, percentage of human CD3⁺ lymphocytes in total live cell population in blood after RBC lysis (open bar) and tumor (solid colored bar) in each of the treatment groups in (b); d Proportion of NY-ESO-1-specific T cells in tumors per treatment group (b), expressed as a percentage of CD3 + TILs, (n = 7, error bars are SEM), *p < 0.0216 (MCLA-145 vs. Ly95), determined by one-way-ANOVA and Tukey’s test; e Ly95 T cells expressing an NY-ESO-specific TCR were transferred to NSG mice bearing A549-PD-L1^hi or A549-PD-L1^null tumors and treated with indicated antibodies, end point data are shown. (n = 9 mice), **p < 0.0018 MCLA-145 treatment (A549-PD-L1^hi vs A549-PD-L1^null groups), determined by one-way-ANOVA and Tukey’s test. f Tumor volume in individual human CD34+ engrafted NSG mice following grafting with MDA-MB-231 cells and treatment with indicated concentrations MCLA-145 or reference antibody (n c= 9 mice per group). Tumor growth and moment of death are indicated for the individual mice per group, a representative experiment of two independent experiments is shown, *p < 0.05, **p < 0.005, ***p < 0.001 of treatment group vs negative control antibody, determined by a two-sided mixed-effects model and Dunnett’s test. g Proportion of CD8+ (upper panel) and CD4+ (middle panel) TILs in tumors from each treatment group, expressed as a percentage of the total population of tumor cells, bottom panel, proportion of PD-L1+ monocytes (bottom panel) in tumors per treatment group, expressed as a percentage of all monocytes (CD11b+ cells), (n = 7 mice per group, error bars are mean ± SD, *p = 0.0492 for CD8+ T cells and p = 0.0014 for PD-L1⁺ monocytes, determined by one-way ANOVA and Tukey’s test.

Fig. 5. Safety of MCLA-145 in a repeat dose non-human primate study.

a Schematic showing the design for the repeat dose safety study; b mean serum concentrations of MCLA-145 measured over one cycle (n = 10 animals for vehicle and 100 mg/kg, n = 6 animals for other treatment groups); c mean maximum serum concentration (C_max) and area under the curve (AUC) at 10 mg/kg (blue), 30 mg/kg (green) and 100 mg/kg (orange) of MCLA-145 measured 24 h after the first dose and 24 h after the fifth dose combined for males and females; n = 10 animals for vehicle and 100 mg/kg, n = 6 animals for other treatment groups, error bars are mean ± SD); d Mean serum concentrations of AST and ALT and e mean number of circulating neutrophils and platelets measured over the study period at the indicated doses (n = 10 animals for vehicle and 100 mg/kg, n = 6 animals for other treatment groups, error bars are SD). Reference mean (male and female bold blue and pink lines respectively) and mean +1 or −1 SD (male and female bold blue and pink lines respectively) as previously reported; f Number of circulating B (CD20⁺), NK (CD16⁺) and T cells (CD3⁺) with CD4⁺ and CD8⁺ T cell numbers and proportion of different subpopulations shown for the 100 mg/kg/week dose group or vehicle, mean ± SD of male and female groups combined (n = 10 animals).

Fig. 6. Colocalization of CD137 and PD-L1 mediates CD137 clustering.

Fold induction of CD3-stimulated Jurkat NFκB-luc-CD137 reporter activity after 24 hours of co-incubation with either a CHO-PD-L1 cells or b human tumor cells endogenously expressing PD-L1 at increasing concentrations of MCLA-145; numbers in brackets indicate PD-L1 binding sites per cell as determined by QIFIKIT analysis; c Schematic illustrating the VeraTag assay to measure CD137 clustering; d CD137-expressing activated Jurkat T cells and PD-L1-expressing CHO cells cocultured with 10 µg/mL test antibody, fixed and measured for CD137 proximity using two different CD137 detection antibodies; e Schematic illustrating the VeraTag assay to measure CD137 and PD-L1 proximity; f Cells cultured and prepared as in (d) to measure PD-L1 and CD137 in proximity, RPA relative peak area; confocal images of CD137 receptor internalization in cocultures of activated CD8+ T cells and CHO-PD-L1⁺ incubated with (g) MCLA-145, (h) urelumab* or (i) negative control antibody for 15 min, nuclear (DAPI) staining (blue), CD8 (green), CD137 (magenta). Single plane view, magnification ×20 (insets ×63). Data are representative of two repeat experiments for Fig. 6g–i.

Fig. 7. Schematic of the MCLA-145 mechanism of action.

MCLA-145 is shown binding to PD-L1 expressed on a tumor cell or APC (depicted in pink) and CD137 expressed on a T effector cell (depicted in green). Engagement of MCLA-145 results in clustering of CD137 molecules on the surface of the effector T cell and downstream activation of the NF-kB signaling pathway. In addition, engagement of PD-L1 by MCLA-145 blocks interaction with its receptor PD-1 reversing PD-1 dependent suppression of downstream signaling from the T cell receptor complex. The net effect of MCLA-145 binding in trans to tumor cell/APC and T effector cells is shown: upregulation of CD137 (on T cells) and PD-L1 (on tumor cells and APCs), the release of pro-inflammatory cytokines from effector T cells resulting in T cell differentiation and expansion and activation of effector T cells to promote cytotoxic attack of tumor cells.