Tumor cell-intrinsic programs of metabolic adaptation and immune evasion in colorectal cancer liver metastases (CRCLMs).

Abstract

3554 Background: CRCLMs account for the majority of colorectal cancer (CRC)-related mortality and represent a key site of immunotherapy resistance. The tumor cell-intrinsic molecular programs enabling adaptation to the liver microenvironment and immune evasion remain poorly defined. Methods: We performed spatial whole transcriptomic analysis of 189 patient-matched primary CRC and CRCLM samples from two independent cohorts (Cohort 1 = 47; Cohort 2 = 141). Samples were represented using tissue microarrays and analyzed using the GeoMx Digital Spatial Profiling (DSP) platform. Two ROIs (tumor cell [CK] and stroma [DAPI]) were analyzed per sample, with analyses focused on tumor cells. Differential expression analyses were conducted independently within each cohort. Results were integrated using Fisher’s combined probability method for batch-aware meta-analysis. Expression of selected transcripts in tumor and matched non-tumor tissue in TCGA and associated survival outcomes were evaluated using the GEPIA2 webtool. Gene set enrichment analysis (GSEA) was applied to identify consistently altered biological pathways. Exploratory LASSO regression using normalized expression values identified gene sets distinguishing liver metastases from primary tumors. Results: We identified multiple genes consistently enriched in malignant cells from CRCLM compared to primary CRC. Prioritized genes include MMP10 (logFC = 0.41, p = 0.007), KRT17 (logFC = 0.87, p < 0.001) and PLA2G2A (logFC = 0.72, p = 0.002), all showing higher tumor expression relative to normal tissue, association with unfavorable survival, and translational relevance (Table 1). GSEA revealed enrichment of acute-phase response (NES = 1.86, FDR = 0.003), complement activation (NES = 1.81, FDR = 0.002) and xenobiotic metabolism (NES = 1.71, FDR = 0.006) in liver metastases, together with reduced antigen processing and presentation (NES = -1.78, FDR = 0.012). Exploratory LASSO regression yielded cohort-specific gene signatures with limited gene-level overlap but convergence on shared biological programs distinguishing CRCLM from primary CRC, including epithelial and metabolic pathways. Of note, PLA2G2A was among the selected features in an independent cohort, reinforcing its association with CRCLM. Conclusions: We identified tumor cell-intrinsic programs in CRCLMs consistent with cellular metabolic adaptations and immune evasion with potential relevance in disease progression and treatment resistance. Our analyses also identified genes, including MMP10, KRT17, and PLA2G2A, as potential therapeutic vulnerabilities in CRCLMs for future investigation. Prioritized liver metastasis-enriched genes. Gene Translational Relevance MMP10 Enzymatic target; inhibitor scaffolds exist KRT17 Tumor-specific biomarker/pathway proxy PLA2G2A Enzymatic target; pathway inhibition feasible