Immune Landscape of EBV associated Gastric Cancer: Analysis from a Western, Academic Institution

Abstract

Background:

Epstein-Barr virus associated gastric cancer (EBVaGC) may be a meaningful biomarker for potential benefit from immunotherapy. Further investigation is needed to characterize the immune landscape of EBVaGC. We assessed our institutional frequency of surgically treated EBVaGC and analyzed the immunologic biomarker profile and tumor infiltrating lymphocyte (TIL) phenotypes of a series of EBVaGC compared to non-EBVaGC cases.

Methods:

Available tissue samples from all patients with biopsy-confirmed gastric adenocarcinoma who underwent resection with curative intent from 2012-2020 at our institution were collected. In-situ hybridization was used to assess EBV status; multiplex immunohistochemistry was performed to assess MMR status, PD-L1 expression, and phenotypic characterization of TILs.

Results:

Sixty-eight samples were included in this study. EBVaGC was present in 3/68 (4%) patients. Among all patients, 27/68 (40%) had positive PD-L1 expression; two of three (67%) EBVaGC patients exhibited positive PD-L1 expression. Compared to non-EBVaGC, EBV positive tumors showed 5-10-fold higher density of TILs in both tumor and stroma and substantially elevated CD8+ T cell to Tregulatory cell ratio. The memory subtypes of CD8+ and CD4+ T cells were upregulated in EBVaGC tumors and stromal tissue compared to non-EBVaGC.

Conclusions:

The incidence of surgically resected EBVaGC at our center was 4%. EBVaGC tumors harbor elevated levels of TILs, including memory subtypes, within both tumor and tumor-related stroma. Robust TIL presence and upregulated PD-L1 positivity in EBVaGC may portend promising responses to immunotherapy agents. Further investigation into routine EBV testing and TIL phenotype of patients with gastric cancer to predict response to immunotherapy may be warranted.

Introduction

Epstein-Barr Virus associated gastric cancer (EBVaGC) is a defined molecular subtype of gastric adenocarcinoma.[1, 2] It has been reported to comprise 8-10% of all gastric cancers according to The Cancer Genome Atlas.[3] EBVaGC possesses a distinct immunological profile characterized by increased immune cell infiltration and elevated PD-L1 expression.[4, 5] Considering this, some studies suggest that EBVaGC could potentially garner benefit from immune checkpoint inhibitors (ICIs).[6] As such, the National Comprehensive Cancer Network (NCCN) has suggested EBVaGC to be a potentially meaningful biomarker for personalized treatment strategies, yet no formal recommendation for routine clinical testing has been made.[7]

Characterization of the tumor infiltrating lymphocyte (TIL) phenotypes of EBVaGC has been understudied. Specific T cell and myeloid cell populations have previously been shown to infiltrate EBVaGC tumors, however, comprehensive TIL analysis of memory-type subsets as well as density and spatial distributions between tumor and tumor-related stroma are lacking.[8] Reports in other malignancies have shown that the these facets of the immune microenvironment are predictive of response to immunotherapy.[9-11] In this study, we aimed to assess the frequency of EBVaGC at a western, academic center and comprehensively describe the immunologic profile of EBVaGC, with emphasis on TILs and biomarker profile, compared to a cohort of non-EBVaGC.

Methods

Patient cohort

After obtaining institutional review board consent, all adult patients with biopsy proven diagnosis of gastric adenocarcinoma who ultimately underwent a resection with curative intent from 2012-2020, either endoscopic or surgical, at our institution and had available formalin fixed, paraffin embedded (FFPE) tissue samples for histologic analysis were included in this study. A retrospective review of a prospectively maintained, clinically-oriented database of patients was conducted. Review of the patient electronic health record was performed for missing data. After patients were identified, additional FFPE slides were requested from areas of invasive tumor at least 2 mm in diameter. Slides were reviewed during creation, and the areas of invasive disease were determined by a board-certified gastrointestinal pathologist (author JK).

Definitions and immunologic profile characterization

EBV status of resected whole tissue specimens was determined by in situ hybridization (iSH) detection of EBV-encoded small RNA (EBER)-positive tumor cells (ARUP Laboratories, Salt Lake City, UT). Assessment of mismatch repair (MMR) protein expression was performed via immunohistochemistry (IHC) analysis of MLH1, PSM2, MSH2, and MSH6 proteins (Leica; Wetzlar, Germany); deficiency (dMMR) was defined as loss of >95% of any one of the protein expressions in tumor cells. Programmed Death-Ligand 1 (PD-L1) expression was measured via the Combined Positivity Score (CPS), defined as the number of positive PD-L1 stained cells via IHC divided by the total number of tumor cells multiplied by 100; values greater than 1 were considered positive expression (Leica; Wetzlar, Germany). TIL populations were characterized by IHC staining of associated cell surface (CD, cluster of differentiation) or intranuclear markers using the Vectra-7-tumor infiltrating lymphocyte kit (PerkinElmer, Waltham, MA). The included TILs and markers are as follows: B cells/CD220+, CD8 T cells/CD8+, CD4 T cells/CD4+, T regulatory (T_reg) cells/forkhead box P3 (FOXP3)+, CD8 memory T cells/CD8+/CD45RO+, CD4 memory T cells/CD4+/CD45RO+, memory Treg cells/CD4+/FOXP3+/CD45RO+, memory B cells/CD220+/CD45RO+, epithelial malignant cell/pan cytokeratin. TIL density was defined as the number of above stained immune cells per mm² designated within tumor or stroma.

Multiplexed Immunohistochemistry

IHC was performed using an autostainer and then slides reviewed using image processing software following a previously employed protocol²²: Vectra 3.0 Automated Quantitative Pathology Imaging System (PerkinElmer) was used with the Bond RX autostainer (Leica). Slides were deparaffinized, heat treated in epitope retrieval solution 2 (ER2) antigen retrieval buffer for 20 minutes at 93 C (Leica), blocked in antibody (Ab) Diluent (PerkinElmer), incubated for 30 min with the primary antibody, 10 minutes with horseradish peroxidase-conjugated secondary polymer (anti-mouse/anti-rabbit, Perkin Elmer), and 10 minutes with horseradish peroxidase-reactive OPAL fluorescent reagents (Perkin Elmer). Slides were washed between staining steps with Bond Wash (Leica) and stripped between each round of staining with heat treatment in antigen retrieval buffer. After the final staining round, the slides were heat-treated in antigen retrieval buffer, stained with spectral 4′,6-diamidino-2-phenylindole (PerkinElmer), and cover slipped with Prolong Diamond mounting media (ThermoFisher; Waltham, MA). Whole slide scans were collected using the 10× objective at a resolution of 1.0 μm. Then 10 regions of interest identified by a gastrointestinal subspecialty trained board-certified pathologist (author JK) were scanned for multispectral imaging with the 20× objective at a resolution of 0.5 μm. The multispectral images were analyzed with inForm software (PerkinElmer) to unmix adjacent fluorochromes; subtract autofluorescence; segment the tissue into tumor regions and stroma; segment the cells into nuclear, cytoplasmic, and membrane compartments; and to phenotype the cells according to cell marker expression.

Statistical analysis

Parametric and nonparametric data are presented in means with standard deviations and medians with interquartile range, respectively. Categorical variables are expressed as absolute and relative frequencies (count and number). Categorical variables were compared using Chi-squared test; for continuous variables, parametric data was analyzed via Student’s T test and non-parametric data with Mann-Whitney U test. All statistical analyses were performed in IBM SPSS version 28.0. Quantification of IHC staining of MMR, PD-L1, and TIL densities was completed with inform Imaging Analysis Software (Akoya Biosciences, Marlborough, MA). Statistical significance was considered p≤0.05.

Results

Patient Demographics

Eighty unique tissue samples were identified, 68 of which were sufficient for histologic analysis. Most patients were male (59%, n=40), Caucasian race (67%, n=46), and mean age 63 years at the time of diagnosis (28-87 years). All tumors were adenocarcinoma in origin. In the total cohort, most patients harbored clinical stage T3/3b tumors (50%, n=34) with N0-N1 disease (91%, n=60). Nearly 75% of patients received neoadjuvant chemotherapy (NAC) (71%, n=48), with combination folinic acid, fluorouracil, and oxaliplatin (FOLFOX) being the most common regimen (37%, n=25); only 9% of patients who received NAC received FOLFOX plus docetaxel (FLOT) (n=6). Neoadjuvant radiation (NAR) was given in 4% of patients (n=3). Of note, no patients received neoadjuvant immunotherapy Surgical resection in the form of total gastrectomy or subtotal gastrectomy was performed in 93% of patients (n=63), the remaining five patients underwent endoscopic resection. Perioperative details are outlined in Table 1.

Table 1.

Demographic & Clinicopathologic Data

Variable	Overall cohort n = 68	EBVaGC n = 3
Baseline characteristics
Gender, n (%)
Female	28 (41)
Male	40 (59)	3 (100)
Race, n (%)
Wliite	46 (67.6)	2 (66.7)
Black/African-American	5 (7.3)	1 (33.3)
Asian	8 (11.7)
Other	9 (13.2)
Age, years (median, IQR)	65 (53-73)	55.8
BMI (mean, SD)	26 (5.2)	24.8 (3.9)
Smoker, n (%)
No	29 (43)	1 (33.3)
Former	9 (13)	1 (33.3)
Current	30 (44)	1 (33.3)
Hypertension, n (%)	27 (39.7)	1 (33.3)
GERD, n (%)	29 (42.6)	2 (66.7)
Barrett’s esophagus, n (%)	3 (4.4)	0
Diabetes mellitus, n (%)
No	52 (76.4)	1 (33.3)
Yes, insulin independent	10 (14.7)	1 (33.3)
Yes, insulin dependent	6 (8.8)	1 (33.3)
Clinicopathologic characteristics
Histology, n (%)
Adenocarcinoma	68 (100)	3 (100)
Intestinal	26 (38.2)	1 (33.3)
Diffuse/Signet-ring	41 (60.3)	2 (66.7)
Other (neuroendocrine features)	1 (1.5)
Location, n (%)
Stomach	61 (90)	3
Greater curvature	9 (13)	1 (33.3)
Lesser curvature	16 (26)	2 (66.7)
Pre-pylorus	21 (31)	-
Fundus	4 (6)	-
Linitis plastica	5 (7)	-
Gastroesophageal junction	7 (10)	-
Clinical staging, n (%)
T1-2	22 (32.3)	-
T3-4	44 (64.7)	3 (100)
N0	38 (55.9)	2 (66.7)
N1	22 (32.4)	1 (33.3)
N2	4 (5.9)	-
N3	2 (2.9)	-
Perioperative therapy
Neoadjuvant chemotherapy, n (%)	48 (71)	3 (100)
FOLFOX	25 (52)	1 (33.3)
FLOT	6 (12.5)
Other	17 (35.4)	2 (66.7)
Neoadjuvant chemoradiation, n (%)	3 (4.4)	0
Neoadjuvant treatment response, n (%)
Near complete response (CRS 3)	4 (8.3)	-
Moderate response (CRS 2)	22 (45.8)	2 (66.6)
Absent or minimal response (CRS 1)	22 (45.8)	1 (33.3)
Unknown	2 (4.2)	-
Resection type, n (%)
Surgical	63 (92.7)	3 (100)
Subtotal gastrectomy	40 (58.8)	2 (66.7)
Total gastrectomy	23 (41.2)	1 (33.3)
Endoscopic	5 (7.3)	-
EMR	3 (60)
ESD	2 (40)
Adjuvant therapy
Chemotherapy, n (%)	36 (45.6)	2 (66.7)
FOLFOX	16 (44)	1 (50)
FLOT	1 (2.8)	1 (50)
Other	15 (41.7)	-
Unknown	2 (5.6)	-
Radiation, n (%)	5	0
Immunotherapy, n (%)	3	0
Anti-PD-1	3 (100)

Demographic & clinicopathologic characteristics of cohort n, number of patients; IOR, interquartile range; SD, standard deviation; BMI, body mass index; GERD, gastroesophageal reflux disease; FOLFOX, leucovorin, fluorouracil, oxaliplatin; FLOT, fluorouracil, leucovorin, oxaliplatin, docetaxel; EBRT, external beam radiation therapy; CRS, neoadjuvant chemotherapy response score; EMR, endoscopic mucosal resection; ESD, endoscopic submucosal dissection

Frequency of Epstein-Barr Associated Gastric Cancer and Comparison to Other Clinically Relevant Tumors Markers

Of the 68 patients included in this study, 4% were confirmed to have EBVaGC (n=3), 12% were dMMR (n=8) and 40% had positive PD-L1 expression (n=27) with a median CPS of 2.5 (1.65-4.25) (Figures 1&2). Among patients with EBVaGC, all were male and two of three had tumors in the proximal stomach; and all received NAC. In our cohort, deficient MMR status was not significantly associated with positive PD-L1 expression. PD-L1 positivity, but not dMMR status, was significantly associated with higher degree of TIL infiltration (p=0.047, data not shown). Among the three EBVaGC, all were classified as pMMR; two had PD-L1 positive expression.

Figure 1.

In situ hybridization detecting EBER1/2. (a) negative EBV; (b) positive EBV staining. 2.5x magnification.

Figure 2.

Representative images of multiplex immunofluorescent staining of gastric adenocarcinoma tissue samples. (a & b) Low and high tumor immune cell infiltrate, respectively. Only conventional CD8+ and CD4+ T cell stains represented. (c & d) Proficient and deficient mismatch repair protein expression, respectively. MLH1 and PSM2 protein expression shown; (d) absence of MLH1 and PSM2 staining in tumor cells indicated by yellow arrows (e & f) Negative and positive PD-L1 expression, respectively. All images captured at 40x magnification.

Characterization of Tumor Infiltrating Lymphocytes

Description of the TIL populations within tumor tissue, tumor related stroma, and tumor to stroma ratio of the entire cohort are displayed in Figures 2&3, Table 2. TIL infiltration within these tumor tissues and tumor related stroma were starkly higher compared to the rest of the cohort. In both tumor tissue and tumor related stroma, the median count of CD8+ and CD4+ T cells, and their memory subtypes, were 5-10-fold greater in EBVaGC (Figure 3, Table 2). This contrasts with T_reg cell densities (both classic and memory subtype) which were similar between the two groups and even greater in stroma from non-EBVaGC. The tumor to stroma ratio of CD8+ T cells was three times higher in EBVaGC compared to all other subtypes. The ratio of CD8+ T cells to Treg cells in tumor tissue, a marker of activated anti-cancer immunity, was nearly double in EBVaGC compared to non-EBVaGC. Interestingly, the EBVaGC tumor that was PD-L1 negative had a profound immune cell infiltration with, in reference to the entire cohort, the highest overall TIL density within tumor tissue. B cells were largely absent in all analyzed gastric tumors. Additional details regarding TIL profiles are listed in Table 2. All three EBVaGC received NAC without radiation. Two of the three EBVaGC had final pathologic diagnosis of ypT1aN0 tumors while the one EBV+/PD-L1 negative tumor had a pathologic stage of ypT3N1 disease. None of the patients with EBVaGC received postoperative immunotherapy (Table 1). There were no differences in oncologic outcomes between the overall cohort, EBVaGC, and MMR deficient or PD-L1 positive subtypes (Table 3).

Figure 3.

Scatter plot representing density of respective tumor infiltrating lymphocyte phenotype per subject in tumor or tumor-related stroma. Dashes represent medians. Conv, conventional; mem, memory.

Table 2.

Immunologic and TIL phenotype profiles

Characteristic	Overall cohort n = 68	EBVaGC n =3	dMMR n = 8	PD-L1 positive n = 27
MMR status, n (%)
Proficient	60 (88.3)	3 (100)	-	25 (92.6)
Deficient	8 (11.7)			2 (7.4)
(−)PSM2	8 (100)
(−)MLH1	2 (25)
(−)MSH2	1 (12.5)
(−)MSH6	0
PD-L1 status, n (%)
Negative	41 (60.3)	1 (33.3)	6 (75)	-
Positive	27 (39.7)	2 (66.7)	2 (25)
CPS	2.5 (1.6-4.1)	7.5 (4.1-10.9)	5.5 (4.4-6.6)	2.5 (1.6-4.1)
Tumor infiltrating lymphocyte profile – tumor tissue
CD8+ T cells, cells/mm2
Classic (CD8+)	8.6 (3.4-37.1)	43.5 (17.5-690.4)	6.9 (1.1-33.0)	19.0 (7.7-44.1)
Memory (CD8+/CD45RO+)	1.8 (0.8-9.0)	9.4 (8.0-376.4)	2.0 (0.6-5.4)	5.2 (1.2-17.0)
CD4+ T cells, cells/mm2
Conventional (CD4+)	3.4 (0.8-8.0	30.8 (1.8-277.7)	0.8 (0.3-6.9)	5 (1.8-17.9)
Memory (CD4+/CD45RO+)	1.8 (0.4-5.0)	8.3 (1.1-92.8)	0.6 (0.2-3.3)	3.0 (0.8-8.3)
Treg cells, cells/mm2
Conventional (CD4+/FOXP3+)	0.4 (0.1-1.7)	0.7 (0.4-3.7)	0.1 (0.01-0.8)	1.1 (0.3-2.4)
Memory (CD4+/CD45RO+)	0.2 (0.04-1.1)	0.7 (0.2-1.28)	0.1 (0.01-0.4)	0.4 (0.06-1.2)
B cells, cells/mm2
Conventional (CD220+)	0.02 (0.003-0.2)	0.2 (0.1-0.17)	0.01 (0-0.05)	0.05 (0.01-0.3)
Memory (CD220+/CD45RO+)	0	0.01 (0-0.009)	0	0.002 (0-0.03)
All TIL (CD8+, CD4+, B cell)	13.6 (5.5-49.6)	74.4 (19.5-833.2)	10.8 (1.4-41.1)	23.7 (10.0-74.4)
CD8:Treg ratio	23.5 (6.7-54.6)	50.1 (6.5-993.6)	46.0 (12.6-61.1)	26.9 (7.2-50.1)
Tumor infiltrating lymphocyte profile – tumor related stroma tissue
CD8+ T cells, cells/mm2
Classic (CD8+)	4.9 (1.6-19.0)	33.1 (3.1-48.4)	5.0 (1.3-29.7)	10.3 (1.9-29.1)
Memory (CD8+/CD45RO+)	3.3 (1.0-11.6)	19.3 (2.7-39.3)	3.0 (0.9-12.1)	7.2 (0.9-15.2)
CD4+ T cells, cells/mm2
Classic (CD4+)	23.2 (6.3-53.1)	143 (23.7-162.5)	15 (2.8-31.5)	44.3 (20.0-106.8)
Memory (CD4+/CD45RO+)	9.7 (2.4-29.0)	48.2 (11.4-101.8)	8.9 (1.9-21.3)	15.8 (3.4-48.2)
Treg cells, cells/mm2
Classic (CD4+)	1.3 (0.2-3.4)	1.1 (0.6-5.9)	0.4 (0.1-2.2)	2.3 (0.8-7.5)
Memory (CD4+/FOXP3+)	0.5 (0.1-1.8)	0.6 (0.3-2.3)	0.3 (0.02-1.1)	0.9 (0.3-4.0)
B cells, cells/mm2
Conventional (CD220+)	1.1 (0.4-6.7)	1.1 (0.5-1.65)	0.3 (0.01-1.4)	1.2 (0.5-19.0)
Memory (CD220+/CD45RO+)	0.1 (0.02-1.2)	0.2 (0.2-0.3)	0.02 (0.01-1.0)	0.2 (0.05-3.2)
All TIL (CD8+, CD4+, B cell)	36.1 (8.6-74.9)	178.4 (27.9-212.1)	25.1 (4.3-64.6)	56.5 (23.1-177.3)
CD8:Treg ratio	3.6 (2.3-10.3)	3.0 (2.9-52.1)	11.4 (4.3-27.6)	3.6 (1.9-7.5)
Tumor infiltrating lymphocyte profile – tumor to stroma ratio
CD8+ T cells	1.9 (0.6-57)	57 (1.3-7.1)	1.5 (0.6-2.5)	1.9 (0.5-10.1)
CD4+ T cells	0.2 (0.04-0.5)	0.2 (0.1-0.8)	0.1 (0.03-0.3)	0.2 (0.04-0.8)
Treg cells	0.4 (0.06-1.6)	0.6 (0.3-0.9)	0.4 (0.07-0.8)	0.3 (0.06-2.1)
B cells	0.01 (0-0.1)	0.1 (0.04-0.2)	0.003 (0-0.1)	0.01 (0.001-0.1)

Immunologic profiles and tumor-infiltrating lymphocyte (TIL) phenotypes in overall cohort, EBVaGC, dMMR GC, and PD-L1 positive patients, grouped by location (tumor & tumor related stroma. MMR, mismatch repair protein; CPS, combined positivity score; Treg, T regulatory cell. All data shown as either n (%) or median (IQR).

Table 3.

Oncologic Outcomes

Outcome	Overall cohort n = 68	EBVaGC n =3	dMMR n = 8	PD-L1 positive n = 27	P value
Death, n (%)
No	36 (52.9)	2 (66.7)	4 (50)	13 (48.1)	0.240
Yes	32 (47.1)	1 (33.3)	4 (50)	14 (51.9)
Distant recurrence, n (%)
No	24 (35.3)	3 (100)	3 (37.5)	10 (37.0)	0.150
Yes	22 (32.3)	-	5 (62.5)	8 (29.6)
Peritoneum	9 (40.9)		2 (40)	3 (37.5)
Liver	6 (27.2)		1 (20)	1 (12.5)
Lung	2 (9.1)		1 (20)	1 (12.5)
Other	5 (22.7)		1 (20)	3 (37.5)
Unknown	17 (25)		-	7 (25.9)
OS, mos (median, IQR)	22.2 (13.4-33.8)	27.9	21.0 (15.2-35.7)	22.8 (10.9-35.9)	0.210
DRFS, mos (median, IQR)	15 (11.0-21.9)	-	13.2 (8.0-18.0)	14.3 (11.9-30.2)	0.427
Follow-up, mos (median, IQR)	39.4 (19.3-54.8)	43.3 (30.6-55.9)	36.0 (16.3-52.5)	40.0 (20.9-55.9)	0.225

Oncologic outcome data for overall cohort, EBVaGC, dMMR status, and PD-L1 positive patients. OS, overall survival; DRFS, distant recurrence free survival; mos, months. All data shown as either n (%) or median with interquartile range (IQR).

Compared to dMMR tumors, EBV positive tumors exhibited higher median TIL counts in tumor tissue and tumor stroma. EBV positive tumors also had a higher degree of TIL infiltration compared to EBV negative/PD-L1 positive tumors. Overall, the disparity in TIL count between EBV positive and PD-L1 positive tumors was not as striking as EBVaGC to the cohort of dMMR patients. One notable exception is the dMMR cohort had equal or higher CD8+-to-Treg ratios when compared to EBVaGC tumor and stromal tissue, respectively (Table 2). Of note, among non-EBVaGC patients, the density of CD8+ T cells (14.2 vs 4.2; p=0.019) and overall TIL density (19.1 vs 7.1; p=0.04) were significantly higher in those who underwent NAC compared to those who did not. Approximately half of patients who underwent NAC had a moderate chemotherapy response score (CRS 2) and the other half had a poor response (CRS 1). Among those with EBVaGC, two patients had a moderate response whereas one had poor response. (Table 1). There were no significant differences between CRS based on EBV, MMR, or PD-L1 status/positivity (p>0.05).

Discussion

In the present report, we detail our institutional frequency of surgically treated EBVaGC and its immune biomarker profile at a Western, academic, quaternary referral center. These subjects were compared to subjects with non-EBVaGC. EBVaGC was found in just three of 68 patients (4%). Our findings that EBVaGC harbors high PD-L1 expression on tumor cells and induces a robust immune cell infiltration within tumor tissue aligns with previously reported literature. In addition, our work builds upon the current knowledge of the EBVaGC tumor microenvironment (TME) by demonstrating a high TIL infiltration within not only the tumor but the tumor related stroma, upregulation of memory T cell subtypes within the TME, and presence of an immunosuppressive component with marginally elevated T_reg cells in EBVaGC. Taken together, our results add to the mounting evidence that EBVaGC is a unique molecular entity of gastric cancer that may potentially benefit from ICIs.

Our institutional frequency of EBVaGC was 4%, somewhat lower than the 9% reported by The Cancer Genome Atlas, which analyzed gastric cancer tissue samples from centers in North America, Europe, and Asia.[2] Our reported incidence is also in contrast to the suggestion that EBVaGC positivity rates are higher in the Americas compared to East Asia.[12] Our EBVaGC rate may be comparatively lower for a few reasons. Our method of detection, currently considered the gold standard, was EBER1/2 in situ hybridization rather than genomic sequencing utilized by the TCGA which may lead to differences in positivity rates, although the concordance between these two methods has been previously validated.[13] The size of our patient cohort in addition to regional demographic differences and local referral patterns may be limiting in accurately reflecting the rate of EBV positivity seen at Western centers at-large. Furthermore, eradication of H. pylori with current treatment modalities may contribute to the lower rate of EBVaGC as H. pylori infection is a known risk factor for gastric carcinoma and EBVaGC.[14] Lastly, tobacco use is a known risk factor for EBVaGC. Tobacco use in our cohort is lower than that reported in historic studies (57% vs 64%), which may effectively lower the risk of EBV positive GC.[15]

It is well known that EBVaGC is characterized by a high immune cell infiltrate. Studies from the late 20^th century described gastric cancers harboring the EBV genome to have specific clinicopathologic as well as genetic, molecular, and immune composition compared with non-EBV gastric cancer.[16] Saiki et al. and Kijima et al. both revealed a preponderance of CD8+ T, CD57+ Natural Killer, and antigen presenting cells within EBVaGC specimens.[17, 18] Van beek et al. further characterized the TIL composition by demonstrating the CD8+ T cell population was of a cytotoxic nature with high expression of the cytotoxic protease, granzyme B.[8] However, precise spatial distribution of specific TILs within the tumor and tumor related stroma have not previously been characterized. We found that compared to non-EBVaGC including dMMR tumors, the elevated TIL count, namely CD8+ and CD4+ T cells, persisted in the stroma surrounding tumor tissue. These results confirm previously cited data that PD-L1+ immune cells infiltrate the center of EBVaGC tumors rather than staying at invasive margins of non-EBV gastric carcinomas.[5] Several studies have demonstrated that stromal TILs are important prognosticators for survival as well as response to anti-PD-1 therapy.[9, 19, 20] Thus, the presence of high TILs, both intratumoral and stromal, is a unique characteristic of EBVaGC that adds new perspective to its lymphoid dense nature.

In addition to spatial distribution of CD4+ and CD8+ T cells within tumor related tissues, we found that memory subtypes of these T cells were significantly upregulated within both tissue regions in EBVaGC compared to non-EBVaGC and dMMR tumors. This finding is significant in that memory subtypes are thought to be responsible for long-term response to malignancy and thus termed the “gold standard of anti-cancer immunity”.[21] In patients with advanced melanoma treated with ICIs, a higher proportion of memory T cells positively predicted response rates and overall survival.[10] Furthermore, effector and central memory T cells (T_EM and T_CM, respectively), subtypes of memory cells, have been shown to express high levels of PD-1, the target of promising ICI therapies. In both advanced melanoma and non-small cell lung cancer, T_CM was found to be a favorable predictor of response to anti-PD-1 therapy and survival.[11] Interestingly, T_EM cells have been shown to drive production of interferon-gamma (IFNγ), a known tumoricidal downstream cytokine, but also known to induce expression of PD-L1 response on tumor and immune cells, thereby regulating anti-tumor activity.[5] Collectively, the dense presence of memory T cells in EBVaGC tumors with high production of IFNγ, yet balanced by upregulation of PD-L1 response in the TME, support the potential efficacy of PD-L1/PD-1 blockade in EBVaGC.

Lastly, considering the immunosuppressive element in gastric carcinomas, we found the FOXP3+ T_reg cells were only slightly increased in EBVaGC tumor tissue compared to dMMR subtypes and non-EBVaGC. Few studies have characterized T_reg cells in the context of EBVaGC, most with the finding that T_reg cells are significantly upregulated in EBV-driven gastric cancer and execute immunosuppressive function via the transforming growth factor-beta (TGF-β) pathway.[10, 22] We add to current understanding that along with a robust cytotoxic inflammatory response, evidenced by an increased CD8+:T_reg cells ratio within tumor tissue, there is an associated immunosuppressive component in the TME of EBVaGC, although not strikingly so compared to non-EBVaGC among our cohort, potentially due to sample size limitations.

The current evidence supporting the use of immune checkpoint inhibition in EBVaGC is nascent yet promising. In a prospective observational study by Xie et al., nine patients with metastatic EBVaGC (mEBVaGC) were treated with ICI as first to third-line therapy. Eight of nine patients, all of whom were PD-L1 positive (CPS >1) either demonstrated stable or partial response to immune checkpoint blockade.[23] In another study, 50% of a cohort with mEBVaGC either experienced partial or complete response (67% PD-L1 positivity) undergoing ICI therapy with a median progression free survival of over eight months.[24] Kim et al. in an observational study of six patients with stage IV EBVaGC had 100% response rate to anti-PD-1 therapy.[25] While results of ICI therapy from these small, highly selected cohorts offer a favorable outlook on the utility of immunotherapy in EBVaGC, others offer a more cautious prospectus. Sun et al. and Wang et al. report more moderate rates of partial or complete response to ICI ranging from 25-28%; another study found no patients to have partial or complete response to ICI but 67% with stable disease during the treatment period.[26-28] Importantly, many of the patients included in these studies with less encouraging results exhibited poor performance status and ascites from peritoneal metastases, both of which have specifically been cited to render anti-PD-1 therapy ineffective.[29] Additionally, the PD-L1 positivity of EBVaGC patients was lower in studies with less favorable response to ICI. PD-L1 positivity, reported to be present in nearly half of EBVaGC, seems to achieve more favorable responses to anti-PD-1 blockade compared to PD-L1 negative expression.[30] Thus, it is necessary to continue to investigate biomarkers and patient factors predictive of response to immunotherapy even within EBVaGC.

The limitations of this report are important to consider. As a retrospective study, there is potential for inherent selection bias. The incidence of EBVaGC among our cohort was low, thus, we were not able to make any definitive statistically significant conclusions. Further, we were not able to fully characterize all the immune components or biomarkers that have been demonstrated to be important in pathogenesis of gastric cancer or EBV variants – included in our analyses were subsets of the TME we believed to be particularly relevant and constructive to that which has been described previously. Lastly, approximately half of our patient cohort received neoadjuvant multimodal therapy. Of note, 90% of the subjects included in this study represented true gastric cancers, and only 10% involved the gastroesophageal junction, which impacts neoadjuvant treatment decision making. Although we found significantly higher CD8+ T cell and overall TIL density within tumor tissue of non-EBVaGC patients who received neoadjuvant therapy, the effect of NAC on the TME and TILs in gastric cancer is complex and has not been fully explored. This may in turn produce a potentially heterogenous sampling of TME and immunogenic profiles not fully generalizable to gastric cancer at-large.

In summary, analysis of this cohort of patients with nonmetastatic gastric cancer treated with multimodal therapy at an academic high-volume center affirms that EBVaGC is characterized by a strong immune infiltrate and positive PD-L1 expression. In addition, we showed that the robust immune cell presence persists within the EBVaGC tumors as well as the tumor related stroma with upregulation of memory T cells and marginally increased T regulatory cells. Our findings add to the growing literature describing the unique immune composition of EBVaGC and its potential positive response to current immune checkpoint inhibition. Dedicated prospective studies are warranted to assess the clinical benefit of routine EBV testing in gastric carcinoma and therapeutic intervention with immunotherapy.

Abbreviations:

EBVaGC

Epstein-Barr virus associated gastric cancer

PD-L1

programmed death-ligand 1

TIL

tumor infiltrating lymphocyte

IHC

immunohistochemistry

MMR

mismatch repair

ICI

immune checkpoint inhibitor

NCCN

National Comprehensive Cancer Network

FOLFOX

folinic acid, fluorouracil, oxaliplatin

FLOT

fluorouracil, leucovorin, oxaliplatin, docetaxel

NAC

neoadjuvant chemotherapy

NAR

neoadjuvant radiation

TME

tumor microenvironment

EBER

Epstein-Barr virus encoded small RNAs

IFNγ

interferon gamma

FOXP3

forkhead box P3

TGFβ

transforming growth factor beta

mEBVaGC

metastatic-EBVaGC

FFPE

formalin fixed paraffin embedded

iSH

in situ hybridization

TCGA

The Cancer Genome Atlas

GERD

gastroesophageal reflux disease

EBRT

external beam radiation therapy

EMR

endoscopic mucosal rection

ESD

endoscopic submucosal dissection