Multi-omics analysis of PD-1 inhibitor combination in neoadjuvant treatment of pMMR/MSS rectal cancer: Insights from the PANFORTE trial.

Abstract

e14613 Background: Programmed cell death-1 (PD-1) antibody monotherapy has shown limited effectiveness for proficient mismatch repair or microsatellite stable (pMMR/MSS) colorectal cancer. However, recent studies indicate that combining PD-1 inhibitor with traditional neoadjuvant therapy can enhance therapeutic efficacy in locally advanced rectal cancer (LARC), though the synergistic effects remain inconsistent, and the underlying mechanisms are still unclear. Therefore, we employed a comprehensive multi-omics approach to identify molecular characteristics associated with tumor responses to neoadjuvant therapy of combination PD-1 inhibitor and FOLFOXIRI in MSS/pMMR LARC based on PANFORTE trial (NCT06099951). Methods: This study included 42 patients with pMMR/MSS LARC who were treated with a combination of FOLFOXIRI and serplulimab for 4 cycles. Tumor volume reduction (TVR) rates were assessed using ultrasonic imaging after 4 treatment cycles. Whole-exome sequencing (WES), RNA sequencing, and single-cell transcriptomics were utilized for the analysis of biopsy specimens to identify molecular characteristics associated with treatment responses. Furthermore, five machine learning models were evaluated using RNA sequencing data to predict TVR. Results: After 4 cycles of combination therapy, the median TVR rate reached 72.2% (ranging from 21.9% to100%). WES analysis revealed that TVR rates positively correlated with tumor mutation burden (TMB) and HLA evolutionary divergence (HED) scores. Further analysis of HLA supertypes indicated that patients with HLA-A03 benefited more from the therapy, while those with HLA-A24 benefited less. RNA sequencing data showed positive correlations between tumor regression rates and estimated scores of macrophages, pericytes, memory B cells, CD8+ naive T cells, NKT cells, and NK cells, while negative correlations were observed with monocytes, myocytes, and CD8+ effector memory T cells. Among the machine learning models tested, the 16-gene ridge regression model including HLA-J showed high predictive accuracy (R² = 0.785, RMSE = 0.095, MAE = 0.074). Consistent with these results, single-cell sequencing data demonstrated a positive correlation between the abundance of CD8+ T cell infiltration, HLA signaling, MHC pathways and TVR rates. Conclusions: This study demonstrates the high efficacy of combining FOLFOXIRI and serplulimab as a neoadjuvant treatment for patients with pMMR/MSS LARC. Mechanistically, HLA-related signals across multi-omics analyses highlight the potential roles of neoantigen generation and presentation, immune cell composition, and the dynamics of signaling pathways in shaping intra-tumor heterogeneity and influencing treatment responses. The exploration of predictive models for TVR provides new insights into personalized treatment strategies.