ESTRO 2024 - Abstract Book

S5434

Radiobiology - Tumour biology

ESTRO 2024

Purpose/Objective:

The treatment response to standard-of-care neoadjuvant chemoradiotherapy (nCRT) for locally advanced rectal cancer (LARC) is heterogeneous. The Neoadjuvant Rectal (NAR) score quantifies nCRT response, using the degree of tumour downstaging through nCRT and nodal status after nCRT, and is reproducibly associated with long-term survival outcomes (1). Stratification of patients using the NAR score identifies patients who are potentially eligible for watch-and-wait organ preservation strategies (low NAR <8), benefitted from nCRT but require a resection (intermediate NAR 8 – 16) and did not benefit from nCRT (high NAR >16). This work will explore whether low NAR and high NAR tumours are biologically different pre-treatment and, thus, whether the clinical outcome of nCRT can be predicted.

Material/Methods:

Transcriptome microarrays were used to quantify RNA expression within pre-treatment biopsies from LARC patients in publicly available GSE87211and the S:CORT consortium Grampian dataset. NAR scores were calculated using the patient's TNM stage as determined by the pre-treatment MRI staging and pathological staging of resection specimens. RStudio was used for data analysis using packages: ‘limma’, ‘fgsea’, ‘GSVA’, ‘MCPcounter’, ‘CRISclassifier’, ‘dorothea’, ‘CMSclassifer’. The MSigDB Hallmark signature collection was used for gene set enrichment analysis (2). A LARC cohort from the Beatson West of Scotland Cancer Centre BWoSCC was used for cell-level validation using multiplex immunofluorescence and tumour-stromal percentages of H&E staining.

Results:

Gene expression from 48 low NAR and 52 high NAR cases from the Grampian cohort and 55 low NAR and 60 high NAR cases from GSE87211 was compared. Differential gene expression analysis identified n=123 differentially expressed genes (padj < 0.05; log2 fold-change > 0.3 / < -0.3) between high NAR and low NAR tumours in the Grampian database, but these were not consistent in GSE87211. Shown in the Figure, 28/50 MSigDB Hallmarks were enriched for the high NAR tumours in both cohorts (p < 0.25) and ‘Hallmark Epithelial Mesenchymal Transition’ had the highest normalised enrichment score for high NAR in both cohorts (padj < 0.001; NES 2.9 – 3.3). 3/50 MSigDB Hallmarks, all annotated as immune-related, were enriched for the low NAR tumours in both cohorts (p < 0.25). Additionally, in both cohorts, the Microenvironment Cell Population counter estimated higher ‘Cytotoxic Lymphocytes’ in the low NAR group (p < 0.05) and higher stromal cells (‘Endothelial Cells’ and ‘Fibroblasts’; p < 0.05) in the high NAR group. This finding was validated using CD8 densities, which were higher in the low NAR tumours of the BWoSCC. DoRothEA transcriptional regulon analysis inferred higher transcriptional activity within the high NAR group compared to the low NAR group. The popular colorectal cancer molecular subtyping classifiers CMS and CRIS could not classify 40 – 60% of biopsy samples.