Spatially resolved transcriptomics deconvolutes histological prognostic subgroups in patients with colorectal cancer and synchronous liver metastases.

Abstract

165 Background: Up to 50% of patients with Colorectal Cancer (CRC) will metastasise to the liver (CRLM). KRAS-mt liver metastases particularly when co-mutated with TP53 are associated with poor prognosis. We have used the Glasgow Microenvironment Score (GMS), a histological score performed on H+E slides utilising immune and stromal components of the microenvironment to robustly stratify outcome in primary CRC. The aim of the current study was firstly to determine the utility of GMS in metastatic CRC and secondly to employ the NanostringTM GeoMx Digital Spatial Profiler (DSP), a state-of-the art analysis platform enabling spatial transcriptomic characterisation while maintaining tumour microenvironment (TME) topographical features, to interrogate the functional biology underlying the GMS. Methods: FFPE specimens from primary and metastatic lesions from 44 patients undergoing synchronous resection of CRLM underwent GMS, IHC and panel genomic assessment. Primary endpoints were recurrence-free survival (RFS) and cancer-specific survival (CSS). In addition to bulk transcriptomic assessment, 4 matched pairs from the cohort were selected for GeoMx analysis: 2 samples were GMS0 (high-immune) and 2 were GMS1 (low-immune) with an equal distribution of KRASmt and wt. After multiplex IF staining (PanCK, CD45, DAPI, αSMA), 48 regions of interest were selected and Cancer Transcriptome Atlas Transcriptomic outputs (2000 genes) were analysed using Pathway enrichment analysis with immune deconvolution of the transcriptome performed. Results: GMS0 (high-immune) was associated with improved RFS (p=0.0048) and CSS (p=0.0012) remaining an independent predictor of survival on multivariate analysis (HR 2.90, 95% C.I 1.18-7.16 P=0.021). GMS0 lesions were enriched for adaptive immune (NES=2.20 p.adj<0.0005) and IL-10 (NES=1.9 p.adj<0.0005) pathways specifically at the invasive edge. In contrast, a poor prognostic KRAS/TP53 lesion demonstrated profound immunosuppression, upregulated NOTCH signalling (NES=2.13 p.adj<0.0005) and neutrophil degranulation (NES=1.99 p.adj<0.0005). Topographical Immune-cell deconvolution demonstrated significantly higher populations of CD4 (p=0.05) and CD8 (p=0.0003) cells in GMS0 leading edges. Conclusions: We have demonstrated that spatial transcriptomic analysis using the Nanostring GeoMx tool can reveal potential novel mechanisms underlying biologically relevant histological and mutational subgroups (KRAS-mt) of CRC, providing potential therapeutic targets requiring further investigation. Future studies will apply this technology to pre and post treatment biopsy samples.