Integrating tumor-infiltrating iymphocytes (TILs) in the biomarker landscape of gastroesophageal adenocarcinoma: Going beyond PD-L1.

Abstract

4098 Background: Immune checkpoint inhibitors (CPI) are standard of care in perioperative and metastatic gastro-esophageal adenocarcinoma (GEA). PD-L1 scoring remains heterogeneous, with multiple assays and substantial interobserver variability. In breast cancer, immunotherapy responses occur even in PD-L1–negative tumors when tumor-infiltrating lymphocytes (TILs) are present. Exploratory analyses from CM649 suggest that immune infiltration and T-cell signatures may better capture CPI benefit than PD-L1 alone, particularly with ipilimumab–nivolumab. We evaluated TILs and their relationship with established biomarkers in a large GEA cohort. Methods: We performed a single-center retrospective study including localized and metastatic GEA patients diagnosed between 2019–2023. Histopathology included MMR status, HER2 (IHC/SISH), PD-L1 CPS, and CLDN18.2 (≥75% tumor cells). TILs were blindly reviewed by an expert pathologist using international TILs-WG guidelines (in press). Baseline characteristics and follow-up were collected until August 2024. Results: A total of 230 patients were included (140 localized, 90 metastatic). MMR, PD-L1 CPS, CLDN18.2, and HER2 status were available for all. By Lauren classification, 83 tumors were diffuse (36%), 131 intestinal (57%), and 17 mixed (7%). TILs were dichotomized at < 20% and < 10%. TIL density differed significantly across Lauren subtypes (p = 0.0007): diffuse tumors showed markedly lower infiltration (76% with TILs < 20%) compared with intestinal tumors (60% with TILs ≥20%). Using the < 10% cutoff, diffuse tumors again showed lower immune infiltration (78% vs. 22%; p < 0.001). PD-L1 CPS categories ( < 1, 1–4, 5–9, ≥10) did not differ by subtype. Notably, in CPS < 1 tumors, 28/48 (58%) were TIL-high (≥10%), whereas CPS ≥10 tumors showed strong concordance with TIL-high status (25/30, 83%). HER2-positive tumors had a higher proportion of TIL-high cases than HER2-negative tumors (OR 3.55; 95% CI 1.64–8.39; p = 0.0004). CLDN18.2 positivity showed no association with TILs (p = 0.44; OR 0.80; 95% CI 0.46–1.37). Conclusions: Using international TILs-WG guidelines, TIL assessment is feasible in routine GEA pathology. TIL density varied significantly across Lauren subtypes, with diffuse tumors showing consistently low immune infiltration. PD-L1 CPS did not reflect these differences and showed substantial discordance with TILs, particularly in PD-L1–negative but TIL-high tumors, suggesting potential CPI sensitivity. HER2 positivity, but not CLDN18.2, was associated with higher TIL levels. These findings support further evaluation of TILs alongside PD-L1 CPS/TAP score as improved predictive biomarkers for anti-PD-1 or anti-PD-1/CTLA-4 therapy.