Human ovarian cancer intrinsic mechanisms regulate lymphocyte activation in response to immune checkpoint blockade

Human breast cancer cell line MDA-MB-231 and human ovarian cancer cell lines OVCAR3, OVCAR4, OVCAR8, PEO1, PEO4, PEA1, PEA2, Ovsaho, Kuramochi and SKOV3 were maintained in IMDM medium (Thermo Fisher Scientific, Massachusetts, USA) supplemented with 10% heat-inactivated FBS and 1% PenStrep (Sigma-Aldrich, Poole, UK), referred as ‘assay medium’ below. Green fluorescent protein (GFP)-expressing MDA-MB-231 cell line was purchased from AntiCancer Inc., and maintained as above with the addition of 1 mg/ml G418 (Thermo Fisher Scientific, Massachusetts, USA). All cell lines have been authenticated by Eurofins Genomics (Ebersberg, Germany).

In experiments using primary EOC patient samples, ascites were obtained from treatment naïve patients with written consents in accordance with approved ethical permission by the West London Research Ethics Committee (Reference 12/WA/0196). Primary EOC tumour cells were isolated from ascites by gradient centrifugation with Ficoll-Paque Plus (GE Healthcare, Chalfont St. Giles, UK) and maintained in RPMI-1640 media (Sigma-Aldrich, Poole, UK) with 20% FBS (Sigma-Aldrich, Poole, UK), l-glutamine 200 mM, penicillin 10,000 units, streptomycin 10 mg/mL solution (Sigma-Aldrich, Poole, UK), 34 ng/ml insulin (Sigma-Aldrich, Poole, UK) and 2.2 mM sodium pyruvate (Sigma-Aldrich, Poole, UK). All cells were cultured at 37 °C with 5% CO₂.

Leukocyte cones were obtained from anonymous healthy blood donors, in accordance with the Human Tissue Act (NHS Blood Transfusion Unit, approval number M153). In order to isolate primary human peripheral blood mononuclear cells (PBMCs), blood products were first diluted in 50 ml PBS and laid carefully on top of 10 ml Lymphoprep reagent (StemCell Technology, Vancouver, Canada). The tubes were then centrifuged at 800 g for 20 min with the brakes off. Next, PBMCs were harvested from the interface above the Lymphoprep and washed twice in 50 ml PBS. In some cases, the cells were incubated in 5 ml ACK lysis buffer (Thermo Fisher Scientific, Massachusetts, USA) at room temperature for 3 min to remove the remaining red blood cells. Primary human monocytes were then depleted using EasySep CD14+ selection kit II (StemCell Technology, Vancouver, Canada) following the manufacturer’s instructions. The resulted primary lymphocytes were then resuspended in 3 ml PBS containing 1.5 μM CellTrace CFSE or CellTrace Violet Cell Proliferation Dye (Thermo Fisher Scientific, Massachusetts, USA), incubated at room temperature for 5 min and washed twice with 10 ml PBS. The ‘assay-ready’ lymphocytes were frozen at up to 100 million cells per vial and stored in liquid nitrogen tanks.

To set up the TICS assay, human cancer lines or primary ascetic cells were seeded at indicated concentrations in 100 μl assay medium in a 96-well flat bottom plate and incubated at 37° overnight to allow cell adherence. On the same day, ‘assay-ready’ primary human lymphocytes were quickly thawed at 37 °C in a water bath, washed twice in 15 ml PBS, resuspended in assay medium and incubated at 37 °C with 5% CO₂ overnight. Next, lymphocytes were resuspended at 3 million cells per ml and added to the wells containing tumour cells in 100 μl assay medium. To block the human PD-1/PD-L1 pathway, a human IgG4 isotype control (BioLegend, California, USA) or nivolumab (Bristol-Myers Squibb, New York, USA), a human IgG1 isotype control (NIP228, AstraZeneca, Cambridge, UK), a PD-1 blocking antibody (clone LO115, AstraZeneca, Cambridge, UK) or durvalumab (MEDI4736, AstraZeneca, Cambridge, UK) were added at 10 μg/ml. Activation of different immune cell subsets was analysed by flow cytometry on day 6, and soluble IFNγ levels in the supernatants were analysed by Human IFNγ MAX Deluxe ELISA kit (BioLegend, California, USA).

For real-time monitoring of tumour cell killing, TICS assay was set up in the presence of 1 μM green caspase-3/7 cell apoptosis reagent (Essen Bioscience/Sartorius, Göttingen, Germany) and imaged using an IncuCyte ZOOM instrument.

All antibodies used for flow cytometry are summarised in Table S1. Briefly, cells were harvested and transferred to a 96-well v-bottom plate and then centrifuged at 800 g for 3 min. Cell pellets were washed twice in 200 μl PBS and resuspended in 20 μl PBS containing detection antibodies or cell viability dye. After a 30 min incubation at 4°, the cells were washed twice with 200 μl PBS and fixed in 1% paraformaldehyde. For CA-125 staining, following primary antibody incubation, cells were washed twice with 200 μl PBS and then incubated for 30 min with a goat anti-mouse PE-conjugated secondary antibody, prior to washing and fixing. BD LSRFortessa or BD FACSCanto II instruments (Becton Dickinson Immunocytometry Systems, California, USA) were used to acquire the data, and the results were analysed using FlowJo software version 6.4.7 (Tree Star Inc, Oregon, USA).

For knockdown experiments, siRNA targeting IFNγR (catalog number L-011057-00-0005), PD-L1 (catalog number L-015836-01-0005) or a non-specific siRNA control (catalog number D-001810-10-05) (GE Healthcare Dharmacon, Inc., Colorado, USA) were used. Tumour cells were seeded onto a six-well plate at a density of 2 × 10⁵ cells per well and cultured for 24 h. Before transfection, the cell culture medium was replaced with antibiotic-free medium. For each well to be transfected, the appropriate siRNA was diluted in Opti-MEM I Medium and complexed with Lipofectamine RNAiMAX (Thermo Fisher Scientific, Massachusetts, USA), following the manufacturer’s instructions, at a final siRNA concentration of 20 nM. After 24 h, the cells were trypsinised and seeded for TICS assay as described above.

The DNA methyltransferase inhibitor guadecitabine (SGI-110) was obtained from Astex Pharmaceuticals and reconstituted in its clinical diluent (65% propylene glycol, 25% glycerine, 10% dehydrated ethanol). Tumour cell lines were treated with the drug or vehicle at the indicated concentrations on day 1 and day 3. Cells were trypsinised and stained, as described above, for the quantification of surface expression of HLA-ABC and PD-L1 by flow cytometry on day 8. On the same day, to evaluate the effects of SGI-110 in TICS, vehicle or compound-treated cells were co-cultured with primary human lymphocytes as described above. In some experiments, blocking antibodies against antigen presentation molecules human HLA-ABC (clone W6/32, BioLegend) or HLA-DR (clone L243, BioLegend) were included.

In order to investigate the biological mechanisms of the PD-1/PD-L1 axis during human tumour–lymphocyte interaction, we have established and optimised a tumour-immune co-culture system, where unsorted healthy donor-derived primary human lymphocytes were co-cultured with human cancer cell lines (Figure S1a). The primary lymphocytes were labelled with a fluorescent dye, which shows reduced intensity as the cells proliferate. Surface markers CD38 and CD56 were used in combination with the cell trace reagents to identify activated T cells or NK cells in the proliferating cell subsets, respectively. Monocytes were depleted prior to the assay to improve reproducibility and reduce donor-to-donor variability. Utilising the mixed lymphocyte population allowed simultaneous analyses of activation and proliferation of CD4+ T cells, CD8+ T cells and NK cells as well as detection of immune modulatory factors released into the supernatant (Figure S1b). We first characterised the effects of PD-1 or PD-L1 blockade in the TICS using a human breast cancer cell line MDA-MB-231, due to its high expression of PD-L1 and HLA-ABC on the cell surface (Fig. 1a). As shown in Figure S2a, the magnitude of αPD-1-driven enhancement in lymphocyte activation and IFNγ levels was dictated by the cancer cell-to-lymphocyte ratio. Treatment with a Food and Drug Administration (FDA)-approved PD-L1 blocking antibody, durvalumab, resulted in significantly increased activation of CD8+ T cells (p = 0.0204), CD4+ T cells (p = 0.0195) and NK cells (p = 0.0011), as well as increased levels of soluble IFNγ (p = 0.0039, Fig. 1b) in 11 lymphocyte donors. Similarly, PD-1 blockade also increased activation of CD8+ T cells using multiple donor lymphocytes (Figure S2b). Interruption of antigen presentation machinery with blocking antibodies against HLA-ABC or HLA-DR abolished the activation of CD8+ or CD4+ T cells upon PD-L1 blockade, respectively (Figure S2c). Of note, there was a strong correlation between CD8+ and CD4+ T cell activation levels among donors in the isotype-treated groups (R²= 0.9075) as well as in the αPD-L1-treated group (R²= 0.69, Fig. 1c). In contrast, activation of NK cells showed weak correlation with CD8+ T cells (Fig. 1c) or CD4+ T cells (Figure S2d), indicating the non-overlapping regulation of different lymphocyte subsets during cancer–immune interaction.

To demonstrate the direct link between lymphocyte activation and cancer cell death in real-time, we have employed an IncuCyte live imaging platform. In vitro proliferation of the GFP-transfected human MDA-MB-231 cells was efficiently controlled by primary human lymphocytes in a dose-dependent manner (Fig. 1d). Using a fluorescent dye that detects active caspase 3/7 during cell death, we have confirmed that this inhibition was due to apoptosis of cancer cells mediated by activated primary lymphocytes rather than non-specific factors such as nutrient deprivation (Figure S3a). Indeed, either blocking PD-1 or PD-L1 enhanced cancer cell death in the presence of primary human lymphocytes. The effects were most evident after 3 days (Fig. 1e, Movie 1 and 2). Although the absolute counts of apoptotic cancer cells differ among lymphocyte donors in the TICS, the relative increases induced by an anti-PD-1 blocking antibody were highly consistent, as compared to the controls (Figure S3b).

Because nivolumab has demonstrated disease control in a small cohort of OC patients [17], we sought to measure the lymphocyte activation primed by human OC cell lines in response to nivolumab in TICS. As shown in Fig. 2a, established human OC cells expressed varying levels of surface HLA-ABC and PD-L1. Using 3 lymphocyte donors in TICS, we have demonstrated the differential baseline release of immune stimulatory cytokine IFNγ as well as magnitude of response to nivolumab with selected OC cell lines (Fig. 2b). The MDA-MB-231 cancer cell line was included as a benchmark control. Of note, the HLA-ABC negative cell line OVCAR8 failed to elicit robust IFNγ release in TICS (Fig. 2b) and prime CD8+ T cell activation (data not shown), highlighting the necessity of TCR-HLA engagement for cancer cell recognition. Importantly, we have demonstrated strong correlations between IFNγ levels and activation of CD8+ T cells across lymphocyte donors and OC cell lines, both at the baseline (R²= 0.5723) and in presence of nivolumab (R²= 0.696) (Fig. 2c). In order to assess the dynamics of IFNγ release in TICS, we performed time-course measurements of IFNγ using OVCAR4 and PEA1 cell lines. Nivolumab-induced enhancement of IFNγ levels was detectable on day 3 in the PEA1 co-culture and was maximal on day 6. In contrast, nivolumab-induced IFNγ was not detected until day 6 in the OVCAR4 co-culture (Fig. 2d). Based on these data, we decided to use soluble IFNγ on day 6 as the readout for further experimental analysis.

PD-L1 expression and defects in the IFNγ pathway have been reported to influence the efficacy of immune checkpoint blockade. To demonstrate the functional impact of the PD-L1 and IFNγ pathways, we used siRNA to knockdown PD-L1 and IFNγ receptor in PEA1 cells (Fig. 3a). To normalise the donor-to-donor variability of IFNγ concentrations at baseline, all values are shown as relative changes to the condition where cells were treated with the control siRNA and control isotype antibody. As expected, PD-L1 knockdown resulted in a 2.06-fold increase in baseline IFNγ release (2.06 ± 0.06, n = 4, Fig. 3b). IFNγ receptor siRNA did not clearly change the baseline IFNγ release in TICS (1.60 ± 0.43, n = 4, Fig. 3b). Moreover, nivolumab treatment substantially increased IFNγ release in cells treated with the control siRNA, but this effect was abolished when IFNγ receptor was silenced (p = 0.0464, n = 4).

Given the functional relevance of HLA-ABC and PD-L1 during the establishment of immune synapses, we hypothesised that enhanced expression of these molecules could improve lymphocyte activation in TICS. DNA methyltransferase inhibitors (DNMTi) have been reported to increase T cell-mediated cancer cell killing; one of the proposed mechanisms is through upregulation of MHC-I molecules [18]. Indeed, SGI-110, a DNMT inhibitor, upregulated surface expression of HLA-ABC and PD-L1 on four human ovarian cancer cell lines (Fig. 3c). As a result, ovarian cancer cell lines Ovsaho and Kuramochi, pre-treated with DNMTi, induced increased release of IFNγ in TICS (Fig. 3d).

Currently, the standard of care therapy for patients with high-grade serous carcinoma (HGSC) is platinum or taxane-based chemotherapy. In order to explore the impact of acquired chemotherapy resistance on baseline lymphocyte activation and response to PD-1 blockade, we have tested in the TICS two HGSC cell line pairs that were established from ascites at different treatment stages (Fig. 4a). The PEA1/PEA2 pair was established from a patient prior to chemotherapy (naïve) and at the point of relapse after 6 months (chemo relapse), respectively [19, 20]. The PEO1 cell line was established from a patient following treatment and chemo-sensitive relapse, while the PEO4 cell line was established following the development of clinical resistance [20, 21].

In TICS, treatment naïve PEA1 cells induced activation of CD8+ T cells, which was significantly enhanced by nivolumab (p = 0.031, Fig. 5b). Strikingly, the paired, chemotherapy-relapsed, PEA2 cell line did not induce strong T cell activation and failed to respond to PD-1 blockade (Fig. 3b). The reduced cancer-driven lymphocyte activation of PEA2 cells was confirmed using a clinically approved PD-L1 blocking antibody (data not shown). This lack of immune activation and response to PD-1 blockade in chemotherapy-resistant ovarian cancer cells were confirmed in the PEO1 and PEO4 cell lines, both of which were established after the patient has relapsed from chemotherapy (Fig. 4b). Measuring cell surface expression levels of HLA-ABC and PD-L1 in the pairs showed that expression of neither protein can explain anti-PD-1 response levels. However, PEA1 showed the highest HLA-ABC expression among the cell lines (Fig. 4c, d), which could in part contribute to its higher level of immune activation.

To rule out that these functional differences on immune activation are skewed by cancer-to-lymphocyte ratios, we have co-cultured a fixed number of primary human lymphocytes with increasing numbers of cancer cells. While tumour cell numbers played a role in the activation of CD8+ T cells co-cultured with the PEA1 cell line, the lack of immune response to PD-1 blockade in PEA2 cells (Fig. 4d) and the PEO1/PEO4 (Fig. 4e) pair were consistently seen in all ratios tested.

To confirm that treatment naïve primary OC cells could elicit lymphocyte activation in response to PD-1 blockade, we have obtained purified cancer cells from patient ascites. After gradient centrifugation, three out of four samples were enriched for cells expressing ovarian cancer cell markers EpCAM or CA-125 (Figure S4). In contrast, only low levels of macrophages were present (5.31 ± 2.36%) and less than 1% monocytes or T cells (data not shown) were detected (Figure S4). Next, we confirmed the expression of surface HLA-ABC and PD-L1 on the primary OC cells. Regardless of cell composition, all four samples showed detectable levels of HLA-ABC and PD-L1 (Fig. 5a). When tested in TICS, cancer cells isolated from patients 1, 2 and 3 induced detectable levels of soluble IFNγ, which were further enhanced after nivolumab treatment (Fig. 5b). Notably, cells isolated from patient 4 lacked EpCAM or CA-125 positive cancer cells, which did not show consistent enhancement of IFNγ in response to nivolumab (Fig. 5b).