Immune microenvironment of Epstein-Barr virus (EBV)-negative compared to EBV-associated gastric cancers: implications for immunotherapy

Abstract

Background: Gastric carcinomas (GC) are aggressive malignancies, and only ~15% of patients respond to anti-programmed cell death (ligand) 1 (PD-(L)1) monotherapy. However, Epstein-Barr virus (EBV)-associated GCs (~5-10% of GCs) often harbor PD-L1 and PD-L2 chromosomal amplifications and robust CD8+ T cell infiltrates, and respond at a high rate to anti-PD-1. The current study compares the tumor immune microenvironments (TiMEs) of EBV+ versus EBV(-) GCs.

Methods: Over 1000 cases of primary invasive GCs were screened to identify 25 treatment-naïve specimens for study (11 EBV+, 14 EBV(-)). Quantitative immunohistochemistry (IHC) was conducted for markers of immune cell subsets and co-regulatory molecules. Gene expression profiling (GEP) was performed on RNAs isolated from macrodissected areas of CD3+ T cell infiltrates abutting PD-L1+ stromal/tumor cells, using multiplex quantitative reverse transcriptase PCR for a panel of 122 candidate immune-related genes.

Results: IHC revealed that 17/25 GCs contained PD-L1+ stromal cells, with no significant difference between EBV+/- specimens; however, only 3/25 specimens (all EBV+) contained PD-L1+ tumor cells. CD8+ T cell densities were higher in EBV+ versus EBV(-) tumors (p=0.044). With GEP normalized to the pan-leukocyte marker PTPRC/CD45, EBV+ GCs overexpressed ITGAE (CD103, marking intraepithelial T cells and a dendritic cell subset) and the interferon-inducible genes CXCL9 and IDO1. In contrast, EBV(-) tumors overexpressed several functionally-related gene groups associated with myeloid cells (CD163, IL1A, NOS2, RIGI), immunosuppressive cytokines/chemokines (CXCL2, CXCR4, IL10, IL32), coinhibitory molecules (HAVCR2/TIM-3 and VSIR/VISTA), and adenosine pathway components (ENTPD1/ CD39 and NT5E/CD73). Notably, compared with EBV+ GCs, EBV(-) GCs also overexpressed components of the cyclooxygenase 2 (COX-2)/prostaglandin E2 (PGE2) pathway associated with cancer-promoting inflammation, including PTGS2/COX-2 (most highly upregulated gene, 32-fold, p=0.005); prostaglandin receptors PTGER1 (EP1; up 21-fold, p=0.015) and PTGER4 (EP4; up twofold, p=0.022); and the major COX-2-inducing cytokine IL1B (up 11-fold, p=0.019). Consistent with these findings, COX-2 protein expression trended higher in EBV(-) versus EBV+ GCs (p=0.068).

Conclusions: While certain markers of immunosuppression are found in the GC TiME regardless of EBV status, EBV(-) GCs, which are much more common than EBV+ GCs, overexpress components of the COX-2/PGE2 pathway. These findings provide novel insights into the immune microenvironments of EBV+ and EBV(-) GC, and offer potential targets to overcome resistance to anti-PD-(L)1 therapies.

Keywords: Gastric Cancer; Gene expression profiling - GEP; Immunotherapy; Tumor microenvironment - TME.

Figure 1. Quantitative IHC analysis of candidate immune-related markers in EBV+ and EBV(−) GC. (A) Representative photomicrographs are displayed for each marker with IHC in different EBV+ and EBV(−) specimens. Scale bar, 100 µm. (B) Densities of immune cell subsets and co-regulatory molecules in EBV+ versus EBV(−) GCs, determined by HALO image analysis. Some specimens with limited amounts of tissue were not tested for every marker. Bars indicate mean values and SEM. P values, Wilcoxon rank-sum test. CSF-1R, colony stimulating factor 1 receptor; EBV, Epstein-Barr virus; FOXP3, forkhead box P3; GC, gastric cancer; GITR, glucocorticoid-induced tumor necrosis factor receptor family-related protein; IDO-1, indoleamine 2,3-dioxygenase 1; IHC, immunohistochemistry; PD-1, programmed cell death 1.

Figure 2. Semi-quantitative IHC analysis of immune cell subsets and checkpoint molecules in EBV+ versus EBV(−) GC. (A) While CD3 scores were similar in the two GC subtypes, the CD4:CD8 ratio was lower in EBV+ GC, indicating a higher density of CD8+ T cells. (B) There were no significant differences in the expression of LAG-3 among CD3+ T cells, or PD-L1 by tumor or stromal cells, in EBV+ versus EBV(−) GCs. PD-L1 expression was more prevalent on stromal cells than on tumor cells. Percent positive cells were estimated visually in 5% increments. Bars indicate mean values and SEM. P values, Wilcoxon rank-sum test. EBV, Epstein-Barr virus; GC, gastric cancer; IHC, immunohistochemistry; LAG-3, lymphocyte activating gene 3; PD-L1, programmed cell death ligand 1.

Figure 3. Expression of candidate immune-related genes in primary gastric cancer specimens using quantitative reverse transcriptase PCR. Heat map displays unsupervised clustering of ΔCt values normalized to immune cell content (Ct gene – Ct PTPRC) in EBV+ (n=10) and EBV(−) (n=14) GCs. Clustering on genes and samples was performed by Cluster 3.0 using city-block distance with average linkage. 122 immune-related genes and 18S are depicted. EBV, Epstein-Barr virus.

Figure 4. Differential expression of candidate immune-related genes in EBV+ (n=10) versus EBV(−) (n=14) primary GCs. Data were normalized to PTPRC expression using the ΔΔCt method. The vertical hatched lines represent a twofold expression difference. The horizontal hatched line represents p value=0.1, as determined by the Wilcoxon rank-sum test. EBV(−) GCs overexpressed genes associated with the cyclooxygenase 2/prostaglandin E2 pathway, including PTGS2, PTGER1, PTGER4, and IL1B. Two genes that failed to amplify in any specimen (FAM183B/CFAP144P1 and IL4) were excluded from this analysis. EBV, Epstein-Barr virus; GC, gastric cancer.

Figure 5. COX-2 protein expression in EBV(−) versus EBV+GCs, detected with IHC. (A) Left, COX-2 expression among tumor cells was quantified in 25% increments. There was a trend towards overexpression of COX-2 among tumor cells in EBV(−) versus EBV+ GC. P value from Wilcoxon rank-sum test (one-sided test, based on prior differential gene expression result). Right, representative H&E and COX-2 immunohistochemistry images from an EBV(−) GC. COX-2 was expressed by tumor cells (yellow star) and normal gastric epithelium (red arrow). Scale bar, 100 um. H&E, hematoxylin and eosin. (B) COX-2 and CD68 expression in serial sections of an EBV(−) GC. Expression was observed in adenocarcinoma cells (red arrow), CD68+ myeloid cells (blue arrow), and non-myeloid stromal cells. This specimen is devoid of normal gastric epithelium. Scale bar, 100 microns. COX-2, cyclooxygenase 2; EBV, Epstein-Barr virus; GC, gastric cancer.

Figure 6. PTGS2 (COX-2) expression in EBV(−) versus EBV+ GC and association with overall survival in TCGA. Analysis of GC transcriptional profiling from TCGA shows that EBV(−) tumors have (A) significantly higher PTGS2 expression and (B) significantly lower expression of a 13-gene inflammation score. In (A), gene expression is measured as log2TPM (transcripts per million). Standard box and whisker plot notation are used to represent the data distribution: thick horizontal line in the middle of the box for median value, box for IQR (25%–75%), and whiskers for non-outlier data range (1.5×IQR). (C) Patients whose GCs have the highest quartile of PTGS2 expression have significantly worse overall survival (OS; red line) than those whose tumors have the lowest quartile of PTGS2 expression (blue line; p=0.032). EBV, Epstein-Barr virus; GC, gastric cancer; IQR, interquartile range; OS, overall survival; TCGA, The Cancer Genome Atlas.