Glyco-Engineered Anti-Human Programmed Death-Ligand 1 Antibody Mediates Stronger CD8 T Cell Activation Than Its Normal Glycosylated and Non-Glycosylated Counterparts

Abstract

The programmed death 1 (PD-1)/programmed death-ligand 1 (PD-L1) axis plays a central role in suppression of anti-tumor immunity. Blocking the axis by targeting PD-L1 with monoclonal antibodies is an effective and already clinically approved approach to treat cancer patients. Glyco-engineering technology can be used to optimize different properties of monoclonal antibodies, for example, binding to FcγRs. We generated two glycosylation variants of the same anti-PD-L1 antibody: one bearing core fucosylated N-glycans in its Fc part (92%) and its de-fucosylated counterpart (4%). The two glycosylation variants were compared to a non-glycosylated commercially available anti-PD-L1 antibody in various assays. No differences were observed regarding binding to PD-L1 and blocking of this interaction with its counter receptors PD-1 or CD80. The de-fucosylated anti-PD-L1 antibody showed increased FcγRIIIa binding resulting in enhanced antibody dependent cellular cytotoxicity (ADCC) activity against PD-L1⁺ cancer cells compared to the "normal"-glycosylated variant. Both glycosylation variants showed no antibody-mediated lysis of B cells and monocytes. The non-glycosylated reference antibody showed no FcγRIIIa engagement and no ADCC activity. Using mixed leukocyte reaction it was observed that the de-fucosylated anti-PD-L1 antibody induced the strongest CD8 T cell activation determined by expression of activation markers, proliferation, and cytotoxicity against cancer cells. The systematic comparison of anti-PD-L1 antibody glycosylation variants with different Fc-mediated potencies demonstrates that our glyco-optimization approach has the potential to enhance CD8 T cell-mediated anti-tumor activity which may improve the therapeutic benefit of anti-PD-L1 antibodies.

Keywords: Fc part; T cell activation; anti-programmed death-ligand 1; antibody; defucosylation; glyco-engineering.

Figure 1

Glyco-engineered anti-human programmed death-ligand 1 (PD-L1) antibody shows an enhanced binding to FcγRIIIa. A competitive FcγRIIIa AlphaLISA was performed for the three anti-PD-L1 variants. Thereby the test antibody competes with antibody-conjugated acceptor beads for binding to FcγRIIIa-conjugated donor beads. The chemiluminescent signal as a result of close proximity of the donor and acceptor beads was plotted against increasing concentrations of αPDL1_NG (open circles), αPDL1_WT (gray triangles), and αPDL1_GE (black squares). Statistics: mean and SD were plotted in the graph. Data are representative of two independent experiments.

Figure 2

Glyco-engineered anti-human programmed death-ligand 1 (PD-L1) antibody and its normal and non-glycosylated counterparts show comparable antigen binding characteristics. The three anti-PD-L1 variants αPDL1_NG (open circles), αPDL1_WT (gray triangle), and αPDL1_GE (black squares) were tested for PD-L1 antigen binding and their capacity to block interaction with PD-L1 ligands in enzyme-linked immunosorbent assays (ELISA). (A) PD-L1 antigen binding ELISA. OD_450–620 values were plotted against increasing concentrations of test antibody to assess binding to plate-bound human PD-L1. (B) Competitive ELISA measuring binding of soluble programmed death 1 to plate-bound PD-L1 in presence of test antibody. OD4_450–620 values were plotted against increasing concentrations of test antibody. (C) Competitive ELISA measuring binding of soluble CD80 to plate-bound PD-L1 in presence of test antibody. OD_450–620 values were plotted against increasing concentrations of test antibody. Statistics: mean and SD were plotted in all graphs. Data are representative of two independent experiments.

Figure 3

Glyco-engineered anti-human programmed death-ligand 1 (PD-L1) antibody induces strongest NK cell-mediated antibody dependent cellular cytotoxicity (ADCC) against PD-L1⁺ cancer cells, but none against B cells and monocytes. (A) The NK cell line KHYG-1-CD16aV as effector cells was incubated with europium-loaded PD-L1⁺ DU-145 cancer cells as target cells in an effector to target ratio of 10:1 in the presence of increasing concentrations of αPDL1_NG (white circles), αPDL1_WT (gray triangles), or αPDL1_GE (black squares) for 5 h to determine the lysis of target cells in an in vitro cytotoxicity assay. The percentage of specific target cell lysis is plotted against the antibody concentration used. The dashed line indicates % of lysis in the medium control. (B,C) The NK cell line KHYG-1-CD16aV as effector cells were incubated with calcein-labeled primary B cells (B) or monocytes (C) as target cells in an effector to target ratio of 10:1 in the presence of increasing concentrations of αPDL1_WT (gray bars) or αPDL1_GE (black bars) for 4 h to determine the killing of target cells in a flow cytometry based in vitro cytotoxicity assay. The relative frequency of dead 7-AAD⁺ of calcein⁺ target cells was plotted against the antibody concentration used. As a positive control (striped bars) either obinutuzumab (αCD20) was used to induce B cell lysis or staurosporine was used to induce monocyte lysis. Statistics: mean and SD were plotted in all graphs. Data are representative of two independent experiments. Significance was tested against the medium control (*p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001. Abbreviation: ns., not significant).

Figure 4

Glyco-engineered anti-human programmed death-ligand 1 (PD-L1) antibody induces strong CD8 T cell activation in a mixed leukocyte reaction. The three anti-PD-L1 variants αPDL1_NG (open circles), αPDL1_WT (gray triangles), and αPDL1_GE (black squares) were tested for their effect on T cell activation in a mixed leukocyte reaction (MLR). The medium control (black crosses) served as a negative control. T cells as responder cells were isolated from a single healthy donor (donor A). Monocyte-derived dendritic cells as stimulator cells were generated from different healthy donors. (A) IL-2 enzyme-linked immunosorbent assays on day 2 of MLR. The determined concentrations of IL-2 in the culture supernatants were plotted. (B) The activation status of CD8 and CD4 T cells in the MLR was determined on day 5 by flow cytometric analysis. The relative frequencies of CD25⁺ and CD137⁺ in CD8 and CD4 T cells were plotted. (C) The proliferation of CFSE-labeled CD8 T cells in the MLR was determined on day 5 by CFSE dilution measured by flow cytometric analysis. Representative plots of CD25 expression and CFSE signal intensity of CD8 T cells are shown. Statistics for (A,B): besides individual data points (n = 8 for IL-2 secretion, n = 16 for CD25 expression, and n = 7 for CD137), mean and SEM were plotted in all graphs (*p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001. Abbreviation: ns, not significant).

Figure 5

Glyco-engineered anti-human programmed death-ligand 1 (PD-L1) antibody induces increased CD8 T cell activation in presence of cancer cells. The three anti-PD-L1 variants αPDL1_NG (light gray bar), αPDL1_WT (dark gray bar), and αPDL1_GE (black bar) were tested for their effect on T cell activation (donor A) in allogeneic mixed leukocyte reaction (MLRs) in absence or presence of the cancer cell lines HSC-4 (horizontal stripes), ZR-75-1 (plaid), and Ramos (vertical stripes). MLR without addition of test antibody (medium; white bar) served as negative control. The relative frequencies of CD25⁺ in CD8 T cells were plotted. Statistics: mean and SD were plotted in all graphs. Data are representative of two independent experiments. Significance was tested against the medium control without presence of cancer wells (*p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001).

Figure 6

αPDL1_GE-dependent enhancement of CD8⁺ T cell activation results in increased cytotoxicity against cancer cells. T cells of two different donors (donor B and donor C) pre-activated in a mixed leukocyte reaction (MLR) for 5 days in the presence of αPDL1_NG (open circles), αPDL1_WT (gray triangles), and αPDL1_GE (black squares) were harvested and incubated with europium-loaded ZR-75-1 cells for 5 h to determine the lysis of target cells in an in vitro cytotoxicity assay. T cells isolated from MLR without addition of test antibody (medium; black crosses) served as negative control. The cytotoxicity assay was performed in absence and presence of a bispecific antibody binding to a tumor antigen on ZR-75-1 and to CD3 on T cells. The fold change in cytotoxicity compared to the medium control is plotted. Statistics: besides individual data points (n = 8), mean and SEM were plotted. Significance was tested against the medium control (*p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001).