ESTRO 2025 - Abstract Book

S4020

Radiobiology - Tumour radiobiology

ESTRO 2025

Our goal was to establish a gene signature specific to cervical cancer, and unrelated to tumour stage or treatment strategy.

Material/Methods: Five cervical cancer cell lines (CaSki, HeLa, SiHa, SW-756, MS-751) were exposed to normoxia (21% O 2 ) and hypoxia (1% O 2 ) for 24h. The experiments were performed in triplicate. RNA was extracted and sequenced on the Illumina Hi-Seq platform. Differentially expressed genes (DEGs) were obtained using the DESeq2 package with log fold change (LFC) ≥0.5 and false discovery rate (FDR) <0.05. Training, internal validation and hazard modelling were conducted in The Cancer Genome Atlas data for uterine cervical cancer data (TCGA-CESC). The median gene expression of each gene was used as a threshold to divide the TCGA samples into high and low expression groups. The gene expression threshold was applied to a retrospective cohort of primary cervical cancer patients treated at our institution between 2013 and 2018. To curate the in-house cohort the diagnostic FFPE biopsy sections were cut for RNA extraction and H&E staining. Two consultant histopathologists reviewed the H&E slides for tumour percentage, histology, and grade. Finally, RNA was extracted and analysed using the Clariom-S platform to calculate hypoxia status. Results: Sixty-one DEGs in ≥4/5 cell lines were selected (FDR p<0.000001) and 55 mapped onto the TCGA-CESC. Principal component analysis (PCA; Fig 1a ) and heatmap ( Fig 1b ) separate cell-line clusters defined by the experimental oxygen conditions. The 55 genes were strongly hypoxia-associated when enriched in gene ontology. These all regulate recognised hypoxia-induced pathways, and most have been investigated for prognostic significance in various cancers.

External validation in the TCGA (n=275) and in-house curated (n=168) cohorts identified tumours classified as ‘hypoxic’ to have a significantly worse overall survival (p<0.0001 ; Fig 2a and b ). Hypoxia was prognostic in univariate (HR=2.7, p-value=2.6e-07) and multivariate (HR=1.7, p-value=0.0116) analyses in the TCGA cohort. In the in-house cohort, hypoxic status was significant in univariate analysis (HR=2.76, p-value=4.41894E-05), but not in the multivariate (HR=1.48, p-value=0.12.).