Clinicopathologic features of complement activation signatures in colorectal cancer.

Abstract

3573 Background: Activation of the complement cascade pathway is associated with pro-oncogenic inflammation and immune-suppressing, myeloid-derived M2 macrophages for many solid tumors. Most colorectal cancers (CRC) are microsatellite stable that do not respond to immunotherapy. The role of complement activation (CA) in the “immune cold” CRC phenotype remains poorly detailed. We sought to identify molecular annotations of CRC subpopulations enriched for CA to guide future therapeutic strategies. Methods: CRC tumors from 207 patients with stages II-IV CRC at MDACC underwent bulk RNA sequencing. Transcriptomes were analyzed per GSEA “Hallmark Complement” gene set to assign a normalized enrichment score (NES) for CA to each patient and considered “complement high” (“CH”; N = 103) or “complement low” (“CL”; N = 104) if the complement NES score was above or below the median. Associations between CA and clinical and pathologic characteristics - e.g., demographics, mutation status, and Consensus Molecular Subtype (CMS) - were evaluated by chi-squared analysis. Single cell RNA (scRNA) sequencing was performed on a separate cohort of CRC primary tumors (N = 85) and liver metastases (N = 60) to compare CA among different cell types using a Wilcoxon’s test. To assess for an association between CA and response to immunotherapy in a previously annotated clinical trial of patients with MSS, BRAF V600E metastatic CRC (NCT04017650), we evaluated pretreatment biopsies by bulk RNA sequencing and compared transcriptomic differences in CA between responders versus non-responders to encorafenib, cetuximab, and nivolumab (E+C+N). Results: CH CRC featured a higher prevalence for MSI-H CRC (21.3% vs 4.2%; p = .005), CMS1 (30.1% vs 6.7%; p < .001), and CMS4 (26.2% vs 7.7%; p < .001) relative to CL CRC. CMS2 was more common among CL CRC (54.8% vs 14.6%, p < .001). CH CRC was associated with BRAF V600E mutations (29.7% vs 9.2% for CL, p = .002) but not with KRAS/NRAS mutations or RAS/BRAF wild-type CRC (p = n.s. for both). On scRNA analysis, CA scores were highest in myeloid cells and lowest for B cells (p < .0001). Among patients with MSS, BRAF V600E CRC, CH signature was associated with non-response to E+C+N (fold-change 3.0 relative to responders, p = .046). Conclusions: Association of CH status with MSI-H CRC is a novel finding that warrants further study in understanding differential patterns of benefit to immune checkpoint blockade. CH CRC, associated uniquely with BRAF V600E CRC, was distributed bimodally across the immune-activated CMS1 and the immune-suppressing CMS4 CRC, similar to known transcriptomic heterogeneity of BRAF V600E CRC. Our data suggest high CA, linked to immune-suppressing myeloid cell subpopulations, as a negative predictive biomarker for response to immunotherapy in MSS BRAF V600E CRC and support broader study of complement-targeting agents to improve treatment for selected patients with CRC.