Comprehensive immunophenotyping of gastric adenocarcinoma identifies an inflamed class of tumors amenable to immunotherapies

Abstract

Background: Gastric adenocarcinoma (GAC) imposes a considerable global health burden. Molecular profiling of GAC from the tumor microenvironment perspective through a multi-omics approach is eagerly awaited in order to allow a more precise application of novel therapies in the near future.

Methods: To better understand the tumor-immune interface of GAC, we identified an internal cohort of 82 patients that allowed an integrative molecular analysis including mutational profiling by whole-exome sequencing, RNA gene expression of 770 genes associated with immune response, and multiplex protein expression at spatial resolution of 34 immuno-oncology targets at different compartments (tumorous cells and immune cells). Molecular findings were validated in 595 GAC from the TCGA and ACRG external cohorts with available multiomics data. Prediction of response to immunotherapies of the discovered immunophenotypes was assessed in 1039 patients with cancer from external cohorts with available transcriptome data.

Results: Unsupervised clustering by gene expression identified a subgroup of GAC that includes 52% of the tumors, the so-called Inflamed class, characterized by high tumor immunogenicity and cytotoxicity, particularly in the tumor center at protein level, with enrichment of PIK3CA and ARID1A mutations and increased presence of exhausted CD8+ T cells as well as co-inhibitory receptors such as PD1, CTLA4, LAG3, and TIGIT. The remaining 48% of tumors were called non-inflamed based on the observed exclusion of T cell infiltration, with an overexpression of VEGFA and higher presence of TP53 mutations, resulting in a worse clinical outcome. A 10-gene RNA signature was developed for the identification of tumors belonging to these classes, demonstrating in evaluated datasets comparable clinical utility in predicting response to current immunotherapies when tested against other published gene signatures.

Conclusions: Comprehensive immunophenotyping of GAC identifies an inflamed class of tumors that complements previously proposed tumor-based molecular clusters. Such findings may provide the rationale for exploring novel immunotherapeutic approaches for biomarker-enriched populations in order to improve GAC patient's survival.

Keywords: Biomarker; Gastric Cancer; Immunotherapy.

Figure 1. Biomarkers and genetic alterations in GAC. (A) Alluvial plot representing the prevalence and distribution of companion biomarkers (MMR, HER2, PD-L1 and CLDN18.2) assessed by immunohistochemistry and associated with active targeted therapies in GAC. (B) Heatmap representing the frequency and type of recurrent somatic mutations in GAC ranked by their prevalence and sorted by MMR class. MMRp (normal protein MLH1/MSH2/MSH6/PMS2 expression), MMRd (loss of protein MLH1/MSH2/MSH6/PMS2 expression), HER2+ (3+or 2+/FISHamp), PD-L1+ (CPS≥5), CLDN18.2+ (≥75% moderate-high expression), TripleNeg (negative for HER2, PD-L1 and CLDN18.2). GAC, gastric adenocarcinoma; MMRd, deficient mismatch repair.

Figure 2. Landscape of tumor-infiltrating immune cells in GAC. (A) Illustrative example of the immune infiltration in intestinal and diffuse histological subtypes assessed by CD45 immunostaining. (B) Differential expression of immune cell markers by multiplex spatial protein profiling of immune cells (identified in the image in purple by CD45 immunofluorescence) depending on the region of the tumor (invasive margin vs tumor center). (C) Immune subpopulation abundance in tumors estimated by gene-expression deconvolution and their association with known histologies and biomarkers of GAC. Samples from the same molecular class were represented with a normalized enrichment score that is proportional to the size of the circle. (D) Protein expression at the immune compartment of tumors of immune-oncology targets including checkpoint molecules and T cell and myeloid cell regulators and their association with known histologies and biomarkers of GAC. GAC, gastric adenocarcinoma.

Figure 3. Identification and immunophenotyping of the GAC Inflamed class. (A) Solutions for gene-expression unsupervised clustering using non-negative matrix factorization are shown for k=2 to k=5 classes, with two being the number of classes with the highest cophenetic coefficient. (B) Gene signatures measuring biological variables crucial to the tumor-immune interaction (antigen availability, structural barriers to immune infiltration, inhibitory signaling, inhibitory metabolism, pro-immune signaling, killing of tumor cells, tumor receptiveness to immune signaling, tumor proliferation, and apoptosis) upregulated or downregulated in the inflamed and non-inflamed classes of GAC. Samples from the same molecular class were represented with a normalized enrichment score. (C) Immune subpopulation abundance inferred by gene-expression deconvolution in the inflamed and non-inflamed classes of GAC. Samples from the same molecular class were represented with a normalized enrichment score that is proportional to the size of the circle. (D) Spatial protein profiling of antitumoral immune response markers in the immune compartment (CD45+) of the inflamed and non-inflamed classes of GAC according to different regions of interest (invasive margin and tumor center). GAC, gastric adenocarcinoma.

Figure 4. Clinical and molecular characterization of the GAC Inflamed class. Heatmap representing clinical, pathological, and molecular features (companion biomarkers, recurrent somatic mutations, and tumor immune evasion signatures of the inflamed and noninflamed classes of GAC. Each column represents one patient. MMRd (loss of protein MLH1/MSH2/MSH6/PMS2 expression), HER2+ (3+ or 2+/FISHamp), PD-L1+ (CPS≥5), CLDN18.2+ (≥75% moderate-high expression). CAF, cancer-associated fibroblasts; GAC, gastric adenocarcinoma; MDSCs, myeloid-derived suppressor cells; MMRd, deficient mismatch repair; TAM, tumor-associated macrophages.

Figure 5. Validation and clinical utility of the GAC inflamed classifier. (A) Identification of the inflamed class of GAC through a 10-gene RNA signature in the TCGA and ACRG cohorts and validation of histological and mutational correlates. (B) Heatmap representing the association of radiological responses to pembrolizumab in a phase II trial for advanced GAC with biomarkers potentially predicting sensitivity to ICIs, including the Inflamed class of GAC. Each column represents one patient. (C) Predictive capability of response to immunotherapies measured by AUC of gene-expression signatures (Inflamed class of GAC, T-cell inflamed signature and the TIDE score²³) in external cohorts of melanoma, renal cell carcinoma and urothelial carcinoma. *Post ipilimumab progression. **Ipilimumab naïve. CPS, combined positive score; EBV, Epstein-Barr positive; GAC, gastric adenocarcinoma; ICIs, immune checkpoint inhibitors; MSI, microsatellite instability.