Comprehensive multi-omics analysis of resectable locally advanced gastric cancer: Assessing response to neoadjuvant camrelizumab and chemotherapy in a single-center, open-label, single-arm phase II trial

Abstract

Background: The current standard of care for locally advanced gastric cancer (GC) involves neoadjuvant chemotherapy followed by radical surgery. Recently, neoadjuvant treatment for this condition has involved the exploration of immunotherapy plus chemotherapy as a potential approach. However, the efficacy remains uncertain.

Methods: A single-arm, phase 2 study was conducted to evaluate the efficacy and tolerability of neoadjuvant camrelizumab combined with mFOLFOX6 and identify potential biomarkers of response through multi-omics analysis in patients with resectable locally advanced GC. The primary endpoint was the pathological complete response (pCR) rate. Secondary endpoints included the R0 rate, near pCR rate, progression-free survival (PFS), disease-free survival (DFS), and overall survival (OS). Multi-omics analysis was assessed by whole-exome sequencing, transcriptome sequencing, and multiplex immunofluorescence (mIF) using biopsies pre- and post-neoadjuvant therapy.

Results: This study involved 60 patients, of which 55 underwent gastrectomy. Among these, five (9.1%) attained a pathological complete response (pCR), and 11 (20.0%) reached near pCR. No unexpected treatment-emergent adverse events or perioperative mortality were observed, and the regimen presented a manageable safety profile. Molecular changes identified through multi-omics analysis correlated with treatment response, highlighting associations between HER2-positive and CTNNB1 mutations with treatment sensitivity and a favourable prognosis. This finding was further supported by immune cell infiltration analysis and mIF. Expression data uncovered a risk model with four genes (RALYL, SCGN, CCKBR, NTS) linked to poor response. Additionally, post-treatment infiltration of CD8+ T lymphocytes positively correlates with pathological response.

Conclusion: The findings suggest the combination of PD-1-inhibitor and mFOLFOX6 showed efficacy and acceptable toxicity for locally advanced GC. Extended follow-up is required to determine the duration of the response. This study lays essential groundwork for developing precise neoadjuvant regimens.

Trial registration: ClinicalTrials.gov NCT03939962.

Keywords: PD‐1 inhibitor; chemotherapy; gastric cancer; molecular markers; neoadjuvant therapies.

FIGURE 1

Study overview. (A) Study flow chart. (B) Overview of the study. The representative radiological (C) and pathological images (D) from responsive and nonresponsive patients.

FIGURE 2

Efficacy and safety of neoadjuvant treatment. (A) Response and duration for the patients with neoadjuvant treatment. OS (B) and PFS (C) Kaplan–Meier survival curve for patients with neoadjuvant treatment. Kaplan–Meier survival analysis depicting the probability of OS (D) and PFS (E) stratified by treatment response status. (F) Comparative efficacy of neoadjuvant treatment for TNM staging. (G) Comparison of HER2 status between patients with response and non‐response. (H) Pathological response subgroup analysis based on baseline characteristics.

FIGURE 3

Genomic characteristics of tumours pre‐neoadjuvant therapy. Whole‐exome sequencing analysis was performed on baseline samples from eight responders and 15 non‐responders. (A) Heatmap showing functional mutations present in two or more patients. Number of variants (B) and ratio of variants (C) in tumours from pre‐treatment biopsies. (D) Comparison of tumour‐associated signalling pathways alterations between patients with response and non‐response. (E) Heatmap showing the mutational profile of the Wnt/β‐Catenin pathway. Each column represents a different sample and the rows represent genes.

FIGURE 4

CTNNB1 mutation is associated with poor prognosis of GC patients. Kaplan–Meier curves of OS (A) and PFS (B) comparing the mutation status of CTNNB1 from GC patients in TCGA. The representative radiological (C) and pathological images (D) from CTNNB1‐mutation patients.

FIGURE 5

Transcriptome analysis and changes during neoadjuvant therapy. Heatmap showing DEGs between responders and non‐responders from pre‐ (A) and post‐treatment (B) biopsies, pre‐ and post‐treatment from responders (C) and non‐responders (D). KEGG function enrichment analysis is shown in panel below. (E) Venn diagrams showing the overlap in genes significantly changed in the responders group and non‐responders group pre‐ and post‐treatment. (F) Heatmap showing DEGs between pre‐ and post‐treatment and responders and non‐responders from pre‐treatment. (G) Risk curves and scatter plots for candidate risk genes in the TCGA‐GC cohort. (H) Kaplan–Meier analysis showing overall survival in low‐risk and high‐risk patient groups in the TCGA‐GC cohort.

FIGURE 6

Roles of the risk model in predicting immune phenotypes. (A, B) Correlations between risk score and the enrichment scores of immunotherapy‐predicted pathways in two independent cohorts (TCGA‐GC and GSE84437). (C) GO analysis of the DEGs between low‐risk and high‐risk patient groups visualized by ClueGO. (D, E) Comparison of the estimated proportion of ten lymphocytes between low‐risk and high‐risk patient groups in two independent cohorts (TCGA‐GC and GSE84437), respectively. (F) Kaplan–Meier curves show overall survival in the high‐risk and low‐risk subgroups after the PD‐L1 blockade immunotherapy in the IMvigor210 cohort. (G) Differences in neoantigen burden between high‐risk and low‐risk subgroups in the IMvigor210 cohort. (H) The proportion of patients in the IMvigor210 cohort with different responses to PD‐L1 blockade immunotherapy.

FIGURE 7

Immune cell subpopulation analysis. (A) Comparison of immune cell subpopulations assessed by the Quantiseq method using the RNA sequencing data including responders (n = 7) and non‐responders (n = 7). (B) Comparison of CD8 T cells in responders and non‐responders by immunofluorescence staining.