IFN-γ from lymphocytes induces PD-L1 expression and promotes progression of ovarian cancer

Abstract

Background: PD-L1 (programmed cell death 1 ligand 1) on tumour cells suppresses host immunity through binding to its receptor PD-1 on lymphocytes, and promotes peritoneal dissemination in mouse models of ovarian cancer. However, how PD-L1 expression is regulated in ovarian cancer microenvironment remains unclear.

Methods: The number of CD8-positive lymphocytes and PD-L1 expression in tumour cells was assessed in ovarian cancer clinical samples. PD-L1 expression and tumour progression in mouse models under conditions of altering IFN-γ signals was assessed.

Results: The number of CD8-positive cells in cancer stroma was very high in peritoneally disseminated tumours, and was strongly correlated to PD-L1 expression on the tumour cells (P<0.001). In mouse models, depleting IFNGR1 (interferon-γ receptor 1) resulted in lower level of PD-L1 expression in tumour cells, increased the number of tumour-infiltrating CD8-positive lymphocytes, inhibition of peritoneal disseminated tumour growth and longer survival (P=0.02). The injection of IFN-γ into subcutaneous tumours induced PD-L1 expression and promoted tumour growth, and PD-L1 depletion completely abrogated tumour growth caused by IFN-γ injection (P=0.01).

Conclusions: Interferon-γ secreted by CD8-positive lymphocytes upregulates PD-L1 on ovarian cancer cells and promotes tumour growth. The lymphocyte infiltration and the IFN-γ status may be the key to effective anti-PD-1 or anti-PD-L1 therapy in ovarian cancer.

Figure 1

PD-L1 expression on ovarian cancer cells is induced by IFN-γ. (A) PD-L1 expression in cell lines from the CCLE data set. The horizontal line at value 4.950718 represents the determined cutoff value, as mentioned in the Materials and Methods section. (B) Expression of IFNGR1 and IFNGR2 in cell lines of ovarian origin from the CCLE data set. The horizontal and vertical lines represent the determined cutoff values. (C) PD-L1 expression in ascites tumour cells from four cases of ovarian cancer. CD326-positive and 7-AAD-negative cells were gated as viable tumour cells. Open histogram with grey line: anti-PD-L1 antibody; filled histogram: isotype control. PD-L1-positive cases (lower panels) and -negative cases (upper panels) are shown. (D) Ascites cells designated by * in (C) were incubated with (upper panel) or without (lower panel) IFN-γ for 24 h, and PD-L1 expression is shown. Open histogram with grey line: anti-PD-L1 antibody; filled histogram: isotype control.

Figure 2

Interferon-γ signature score is correlated to lymphocyte infiltration in ovarian cancer clinical samples. The IFN-γ signature score (x axis) and the number of CD8-positive (A), CD4-positive (B) or both (C) lymphocytes (y axis) showed a significant, positive correlation in KOV75 cases.

Figure 3

Stromal CD8-positive cells are prominent in peritoneal dissemination and are related to PD-L1 expression on tumour cells in ovarian cancer clinical samples. (A) Number of intraepithelial and stromal CD8-positive cells in peritoneal dissemination. (B) Immunohistochemistry for PD-L1 and CD8. Representative cases are shown. The area marked ‘E' represents tumour epithelium, and the area marked ‘S' represents tumour stroma. Note the abundant stromal infiltration of CD8-positive cells, and dense PD-L1 staining on both cancer cells and immune cells. Bars=50 μm. (C) PD-L1 expression level and stromal CD8-positive cell number in a peritoneal dissemination sample. PD-L1 staining score 0–1 vs 2–3; P<0.001.

Figure 4

PD-L1 expression and lymphocyte infiltration vary between mouse subcutaneous and peritoneal tumours. (A) Representative images of immunohistochemistry for PD-L1 and CD8 for subcutaneous and peritoneal disseminated tumours of HM-1. Green bars: 25 μm; black bars: 50 μm. (B) CD8-positive cell number per high-power field (HPF) using immunohistochemistry (n=3). (C) The percentage of CD8-positive cells in the total cells of tumour tissue analysed by flow cytometry (n=4). (D) CD4-positive cell number per high-power field using immunohistochemistry (n=3). (E) The percentage of CD4-positive cells in the total cells in tumour tissue analysed using flow cytometry (n=4). NS=not significant.

Figure 5

Blocking the IFN-γ signal in the mouse peritoneal dissemination model resulted in lower PD-L1 expression and more TILs. (A) IFNGR1 expression on HM-1 cells was successfully depleted by shRNA. HM1-control (black line histogram), HM1-shIFNGR1 (blue line histogram) and matched isotype control (shaded histogram). (B) Upper panels: Immunohistochemistry for PD-L1 in mouse peritoneal disseminated tumours. Green bars: 25 μm. Lower panels: Expression of PD-L1 on tumour cells. Green line histogram: anti-PD-L1 antibody; filled histogram: isotype control. Representative data from four mice per group with similar results. (C) Western blotting analysis for mouse peritoneally disseminated tumours. Cont=HM1-control; sh=HM1-shIFNGR1. (D) CD8-positive cells in mouse peritoneally disseminated tumour. Upper panel: representative dot plots. The percentage of CD3-positive, CD8-positive cells in total tumour cells are shown in the box. Lower panel: bar graph (n=4). GAPDH=glyceraldehyde 3-phosphate dehydrogenase.

Figure 6

Blocking the IFN-γ signal in the mouse peritoneal dissemination model resulted in slower tumour progression and longer survival. (A) Photograph of mouse peritoneum from HM1-control- or HM1-shIFNGR1-injected group on day 19. Yellow arrowheads: peritoneally disseminated tumours on peritoneum; blue arrowheads: disseminated tumours on omentum. (B) Survival of HM1-shIFNGR1-injected mice (blue line) and HM1-control-injected mice (green line) (n=6).