Spatial transcriptomics and lethality-associated stromal remodeling of colorectal cancer.

Abstract

227 Background: Colorectal cancer (CRC) is surgically resectable and eminently screenable yet remains a lethal entity with high affinity to metastasise synchronously and metachronously. It has been proposed that cellular and extracellular components of the microenvironment contribute to metastatic potential. Dissociative profiling techniques such as bulk transcriptomic and single-cell RNA sequencing have contributed insight but limited therapeutic progress. Spatial transcriptomic (ST) assessment allows molecular profiling of tissue while preserving tissue architecture. Here we employ ST approaches to interrogate tumor compartments of primary resectable and matched synchronous CRC demonstrating stromal signatures with distinct collagen expression associated with outcome. Methods: 25 patients with primary resectable CRC and 4 patients with matched primary CRC and liver metastasis (CRLM) underwent single-cell spatial transcriptomics using the Nanostring CosMx Spatial Molecular Imager (SMI, 1000plex gene panel) (Discovery cohort). 71681 epithelial and 88806 stromal cells with intact spatial resolution were analysed. The spatial single-cell signatures were reconstituted in 89 patients using the Nanostring GeoMx Digital Spatial Profiler (Validation cohort). 3 GeoMx compartments were analysed: Epithelial (PanCK+); Stroma (aSMA+); Immune (PanCK-aSMA-). Results: CosMx demonstrated 2 distinct collagen signatures: COL1A1,COL1A2,COL3A1 associated with normal fibroblasts; COL9A2 associated with cancer-related stroma and lethal subtypes of epithelial cell. In the GeoMx validation cohort, 42 patients expressed the COL1A1 signature in the aSMA compartment and 47 did not (5 year recurrence free survival: 0.88 vs 0.49, p < 0.005) with different morphological patterns of each distinct aSMA group. In patients with favourable prognosis, chemokine high epithelial subtypes expressed CXCL5 and recruited COL9A2-, IL6+ neutrophils to the surrounding microenvironment. In contrast, epithelial cells in patients with poor prognosis expressed CXCL8 and recruited COL9A2+ neutrophils which were frequently in contact with SPP1+ macrophages. Conclusions: We have used ST approaches to interrogate cellular compartments of CRC demonstrating stromal subtypes that impact outcome. These insights could be used in the clinical setting to quantify COL9A2 expression in aSMA+ cells to prognosticate patients. Targeting of tumor collagen has been proposed to augment existing anti-cancer therapies and this work has demonstrated COL9A2 as a potential target in need of further investigation.