Ex vivo modelling of PD-1/PD-L1 immune checkpoint blockade under acute, chronic, and exhaustion-like conditions of T-cell stimulation

Abstract

Blockade of PD-1/PD-L1 interactions is proving an exciting, durable therapeutic modality in a range of cancers whereby T cells are released from checkpoint inhibition to revive their inherent anti-tumour activity. Here we have studied various ways to model ex vivo T cell function in order to compare the impact of the clinically utilised anti-PD-1 antibody, pembrolizumab (Keytruda) on the activation of human T cells: focussing on the release of pro-inflammatory IFNγ and anti-inflammatory IL-10 to assess functionality. Firstly, we investigated the actions of pembrolizumab in an acute model of T-cell activation with either immature or mature allogeneic dendritic cells (DCs); pembrolizumab enhanced IFNγ and IL-10 release from purified CD4+ T-cells in the majority of donors with a bias towards pro-inflammatory cytokine release. Next, we modelled the impact of pembrolizumab in settings of more chronic T-cell activation. In a 7-day antigen-specific response to EBV peptides, the presence of pembrolizumab resulted in a relatively modest increase in both IFNγ and IL-10 release. Where pembrolizumab was assessed against long-term stimulated CD4+ cells that had up-regulated the exhaustion markers TIM-3 and PD-1, there was a highly effective enhancement of the otherwise exhausted response to allogeneic DCs with respect to IFNγ production. By contrast, the restoration of IL-10 production was considerably more limited. Finally, to assess a direct clinical relevance we investigated the consequence of PD-1/PD-L1 blockade in the disease setting of dissociated cells from lung and colon carcinomas responding to allogeneic DCs: here, pembrolizumab once more enhanced IFNγ production from the majority of tumour preparations whereas, again, the increase in IL-10 release was modest at best. In conclusion, we have shown that the contribution of PD-1-revealed by using a canonical blocking antibody to interrupt its interaction with PD-L1-to the production of an exemplar pro- and anti-inflammatory cytokine, respectively, depends in magnitude and ratio on the particular stimulation setting and activation status of the target T cell. We have identified a number of in vitro assays with response profiles that mimic features of dissociated cell populations from primary tumours thereby indicating these represent disease-relevant functional assays for the screening of immune checkpoint inhibitors in current and future development. Such in vitro assays may also support patient stratification of those likely to respond to immuno-oncology therapies in the wider population.

Figure 1

Pembrolizumab increases IFNγ and IL-10 levels in co-cultures of CD4⁺ T cells and allogeneic dendritic cells. Ability of pembrolizumab (anti-PD-1; 1.0 μg/ml) to modify IFNγ (A,D) or IL-10 (B,E) levels in co-cultures of CD4⁺ T cells and allogeneic immature (A–C) or mature (D–F) dendritic cells relative to isotype control. (A,B,D,E) indicate responses of individual donors (each colour represents an individual donor), (C,F) show percentage change in response in the presence of pembrolizumab relative to isotype control (mean + SEM, n = 8). (G) Shows impact of pembrolizumab upon CD4⁺ T cell proliferation in the presence of allogeneic immature or mature dendritic cells. Data presented as percentage change relative to isotype (mean + SEM, n = 8). The impact of pembrolizumab was statistically significant compared to isotype control (* one sample T-test or ^ one sample Wilcoxon test). There was a significant difference between IFNγ and IL-10 (+ 2-tailed T-test).

Figure 2

Pembrolizumab increases IFNγ and IL-10 levels in cultures of PBMCs stimulated with EBV peptides. Ability of pembrolizumab (anti-PD-1; 1.0 μg/ml) to modify IFNγ (A) or IL-10 (B) levels in cultures of PBMCs stimulated with EBV peptides. (A,B) indicate responses of individual donors (each colour represents an individual donor). (C) indicates percentage change in response in the presence of pembrolizumab relative to isotype control (mean + SEM, n = 12). The impact of pembrolizumab was statistically significant compared to isotype control (*one sample T-test).

Figure 3

Exhausted CD4⁺ T cell phenotype. Representative (n = 6) flow cytometry plots (A) demonstrating expression of cell surface markers (LAG-3, PD-1, TIGIT, TIM-3) indicating an exhaustive cell phenotype on CD4⁺ T cells characterised at day 0 (unstimulated, upper panels) and after 14 days stimulated with PHA (5.0 μg/ml, lower panels). Impact of PHA (5.0 μg/ml)-induced exhaustion upon expression of cell surface markers (PD-1, TIM-3, LAG-3, TIGIT, CTLA-4) on CD4⁺ T cells. Data presented as percentage of expressing CD4⁺ T cells (B), and MFI of CD4⁺ T cells (C), (mean + SEM, n = 4–10). The impact of PHA treatment on surface marker expression was significant (+ 2-tailed T-test, # Mann–Whitney U Test).

Figure 4

Exhausted CD4⁺ T cell function. Ability of pembrolizumab (α-PD-1; 1.0 μg/ml) and ipilimumab (α-CTLA-4; 1.0 μg/ml) to modify IFNγ (A,D) or IL-10 (B,E) levels in co-cultures of CD4⁺ T cells either unstimulated (clear bars) or pre-stimulated for 14 days with PHA (5.0 μg/ml; black bars) and allogeneic immature (A–C) or mature dendritic cells (D–F) relative to isotype control (mean + SEM, n = 9–13). (C,F) show percentage change in response in the presence of pembrolizumab relative to isotype control, (mean + SEM, n = 9–13). (G) Shows impact of pembrolizumab upon exhausted CD4⁺ T cell proliferation in the presence of allogeneic immature or mature dendritic cells. Data presented as percentage change relative to isotype (mean + SEM, n = 4). There was a significant difference between PHA-treated (exhausted) CD4⁺ T cells and unstimulated cells, and a significant difference between exhausted CD4⁺ T cells treated with isotype or treated with pembrolizumab (# Mann–Whitney U-Test). The impact of pembrolizumab was statistically significant compared to isotype control (*one sample T-test or ^one sample Wilcoxon test). There was a significant difference between IFNγ and IL-10 (+ 2-tailed T-test, # Mann–Whitney U test).

Figure 5

Response of dissociated tumour cells. Ability of pembrolizumab (anti-PD-1; 1.0 μg/ml) to modify IFNγ (A,D) or IL-10 (B,E) secretion by dissociated tumour cells from lung (A–C) or colon (D–F) tumours stimulated with allogeneic mature dendritic cells. (A,B,D,E) indicate responses of individual tumours (each colour represents an individual tumour), (C,F) indicate the percentage change in response in the presence of pembrolizumab relative to isotype control (mean + SEM; n = 11 [lung]; n = 4 [colon]). The impact of pembrolizumab was statistically significant compared to isotype control (^one sample Wilcoxon test). The difference between IFNγ and IL-10 was statistically significant (Mann–Whitney U test).

Figure 6

Ratio of IFNγ/IL10 responses across models. Graph showing the pembrolizumab-induced change in the ratio of IFNγ/IL-10 across the different in vitro models used in this study. Each point represents a donor/patient response. Mean ± SEM, n = 8 for allogeneic iDC or mDC/ T cell, n = 12 for EBV, n = 9–13 for iDC or mDC / Exhausted T cell, n = 11 for lung tumour and n = 4 for colon tumour.