Transcriptomic characterization and construction of M2 macrophage-related prognostic and immunotherapeutic signature in ovarian metastasis of gastric cancer

Abstract

Background: Ovarian metastasis (OM) poses a major threat to the outcome of gastric cancer (GC) patients. Recently, immunotherapy emerged as a novel promising therapeutic strategy to treat late-stage GC, whereas its efficacy is influenced by tumor immune microenvironment (TIME). M2 macrophage, a key subset within TIME, plays dual immunosuppressive and pro-tumorigenic roles in cancer progression and is recognized as a potential therapeutic target. However, molecular mechanisms underlying OM remain elusive and the TIME-related prognostic and immunotherapeutic index for these patients is yet to establish.

Methods: Differential expressed genes (DEGs) between paired normal mucosa, primary GC and OM of patients from Fudan University Shanghai Cancer Center (FUSCC) cohort (n = 6) were identified by transcriptome sequencing, followed by the functional annotation of enriched hallmark pathways of DEGs between them. CIBERSORT was used to profile the relative expression level of 22 immune cell subsets in normal tissues, primary and metastatic tumors, followed by weighted gene coexpression network analysis (WGCNA) uncovering immune cell-correlated gene sets. The intersected genes between DEGs and M2 macrophage-related genes were processed by least absolute shrinkage and selection operator (LASSO) regression analysis to construct a predictive signature, M2GO, which was further validated by training set and test set of The Cancer Genome Atlas-Stomach Adenocarcinoma (TCGA-STAD), GSE62254 and GSE84437 cohorts. GC patients were divided into M2GO-high and -low subgroup according to the optimal cutoff value of the M2GO score. Furthermore, the clinical, molecular and immune features between M2GO-high and -low subgroups were analyzed. Clinical cohorts of immunotherapy were used to validate the predictive value of M2GO in regard to immunotherapy effectiveness.

Results: Transcriptomic sequencing and follow-up analyses of triple-matched normal tissues, primary and ovarian metastatic tumors identified distinctive sets of DEGs and enriched immune-, cancer- and metastasis-related pathways between them. Of note, M2 macrophage, a major immunosuppressive and pro-tumorigenic component within TIME, was significantly up-regulated in OMs. WGCNA and LASSO regression were applied to establish a novel OM- and M2 macrophage-related predictive signature, M2GO, based on M2 macrophage-related prognostic genes including GJA1, MAGED1 and SERPINE1. M2GO served as an independent prognostic factor of GC patients. Comprehensive molecular and immune characterization of M2GO-based subgroups uncovered their distinctive features in terms of enriched functional pathways, tumor mutation burden, key immune checkpoints, major regulators of natural immune cGAS-STING pathway, infiltrated subsets of immune cells and tumor immune exclusion/dysfunction (TIDE) score. Notably, the M2GO score was significantly lower in responsive group than non-responsive group (P < 0.05) in clinical cohort of metastatic GC patients undergoing immunotherapy.

Conclusion: Transcriptomic characterization of paired normal mucosae, primary and ovarian metastatic tumors revealed their unique molecular and immune features. Follow-up analyses established a novel OM- and M2 macrophage-related signature, M2GO, which served as a promising prognostic and immunotherapeutic biomarker to distinguish the clinical outcome, molecular and immune features of GC patients and predict their differential responses to immunotherapy.

Keywords: Gastric cancer; Immunotherapy; M2 macrophage; Ovarian metastasis; Prognostic signature; Transcriptome sequencing.

Fig. 1

Schematic layout of this study

Fig. 2

Comparative analyses of transcriptome sequencing of triple-matched samples of ovarian metastatic GC patients in FDUSCC. A–C Volcano plot of DEGs between normal mucosae, primary GCs and the ovarian metastatic tumors. D The intersected genes between three sets of samples illustrated by Venn diagram. E–F The enriched cancer hallmark signaling pathways of the up-regulated and down-regulated DEGs of primary tumors when compared to normal tissues, respectively. G–H The enriched cancer hallmark signaling pathways of the up-regulated and down-regulated DEGs of ovarian metastatic tumors when compared to normal tissues, respectively. I–J The enriched cancer hallmark signaling pathways of the up-regulated and down-regulated DEGs of ovarian metastatic tumors when compared to primary tumors, respectively. DEG, differentially expressed gene. FDUSCC, Fudan University Shanghai Cancer Center

Fig. 3

Identification of M2 macrophage-related genes. A Differential distribution of 22 subtypes of immune cells between normal mucosa, primary and ovarian metastatic tumors calculated and analyzed by CIBERSORT algorithm. B WGCNA recognized immune cell-associated modules and identified the “dark turquoise” as the most related module to the M2 macrophage. C Correlation between Module Membership in “dark turquoise” module and gene significance for M2 macrophage. D The intersected genes in “dark turquoise” module and DEG illustrated by Venn plot. WGCNA, weighted correlation network analysis

Fig. 4

Construction of M2-related genes prognostic signature for ovarian metastatic GC patients in the TCGA-STAD dataset. A Univariate cox analysis identified 55 prognostic genes in the illustration of Forest plot. B–C Construction of the prognostic signature, M2GO, consisting of 3 genes (GJA1, MAGED1 and SERPINE1) by LASSO regression algorithm. D–F M2GO could effectively stratify risk in training, test and overall set as patients with high M2GO score showed poor prognosis in the TCGA-STAD and vice versa. G–I ROC analysis of training, test and overall set demonstrated the prognostic value of M2GO considering the AUC at 1, 3 and 5 years. TCGA-STAD, The Cancer Genome Atlas-Stomach adenocarcinoma

Fig. 5

Clinical characterization of M2GO. A–C External validation of M2GO using GSE62254 (ARCG cohort) to prove the prognostic value of M2GO in all, male and female population, respectively. D–F External validation of M2GO using GSE84437 (Yonsei cohort) to prove the prognostic value of M2GO in all, male and female population, respectively. G Age, gender, grade, pathological stage and M2GO signature were associated with the prognosis as indicated by univariate cox analysis. H M2GO signature, age and pathological stage served as independent prognostic factors as indicated by multivariate Cox analysis. I Clinicopathological factors were comparable between M2GO-low (low-risk) and M2GO-high (high-risk) subgroups. J M2GO score-based grouping was independent of pathological TNM staging system (P = 0.847), serving as an independent classification

Fig. 6

Molecular features of M2GO-high and -low subgroup. A–B GSEA analysis revealed enriched signaling pathways in low- and high-risk subgroup, respectively. C–D Distribution of the top commonly mutated genes in different subgroups. The low-risk subgroup had a higher overall mutation rate than the high-risk subgroup (93.53% vs 83.33%). GSEA, Gene Set Enrichment Analysis

Fig. 7

Immune features of M2GO-high and -low subgroup. A mRNA expressions of PDCD1, TBK1 and IRF3 were higher in the low-risk subgroup than in the high-risk subgroup. B Differential distribution of 22 subtypes of immune cells between high- and low-risk subgroups in the TCGA-STAD cohort according to the CIBERSORT algorithm. C M2GO-based subtyping was correlated with IS-based immunophenotyping. D Proportion of CIN, EBV, MSI and GS subtype by TCGA molecular classification in high- and low-risk subgroups. E Proportion of MSS/TP53-, MSS/TP53 + , MSI and EMT subtype by ACRG molecular classification in high- and low-risk subgroups. F Distribution of IS-based immunophenotyping and TCGA molecular classification in high- and low-risk subgroups illustrated by Sankey diagram. ACRG, the Asian Cancer Research Group. EBV, Epstein–Barr virus. MSI, microsatellite instability. GS, genomically stable. CIN, chromosomal instability. MSS, microsatellite stability. EMT, epithelial–mesenchymal transition

Fig. 8

Predictive value of M2GO in immunotherapy efficacy of GC patients. A TMB level was higher in the low-risk subgroup than that in the high-risk subgroup. B TMB was negatively correlated with the M2GO risk score. C–E TIDE score, Exclusion score and Dysfunction score were significantly lower in the low-risk subgroup than that in the high-risk subgroup as indicated by TIDE algorithm. F Differential proportion of MSI-H, MSI-L and MSS in high- and low-risk subgroups. G–H M2GO effectively predicted immunotherapy response in Kim's cohort. G. The proportion of responsive patients in high-risk subgroup was 2.8 times less than those who were in the low-risk subgroup (86% vs 61%, P = 0.11, chi-square test). H. Non-responsive patients have a significantly higher M2GO score than responsive patients (P < 0.05, t test). I Comparison of M2GO score between EBV-positive and -negative subgroup. TMB, tumor mutation burden. TIDE, Tumor immune dysfunction and exclusion. MSI-H, microsatellite instability-high. MSI-L, microsatellite instability-low