Cardio-oncology-relevant programmes in the cervical cancer tumour microenvironment: an exploratory multi-omics and...

Bevacizumab and immune-checkpoint inhibitors (ICIs) improve outcomes in advanced cervical cancer but may cause cardiovascular toxicity (CVT). Public cervical-cancer transcriptomic datasets do not contain cardiac tissue or adjudicated CVT outcomes; consequently, tumour-microenvironment (TME) expression cannot by itself establish a cardiac susceptibility mechanism. We reanalysed the available data to ask the narrower question of whether cervical tissues contain angiogenic, endothelial, inflammatory, T-cell-exhaustion, and TGF-β/fibrotic programmes that are biologically relevant to cardio-oncology.

All nine single-cell RNA-sequencing specimens in GSE208653 were analysed: two HPV-negative healthy cervix, two HPV-positive histologically normal cervix, two high-grade squamous intraepithelial lesions, two squamous-cell carcinomas, and one adenocarcinoma. After prespecified quality control, 74,457 cells were integrated with Harmony and annotated into broad lineages. Prespecified module scores and four ligand-receptor expression-product scores were aggregated by biological specimen; malignant versus non-malignant differences were evaluated with exact specimen-label permutation tests and Benjamini-Hochberg correction. Four GSE208654 Visium count matrices were analysed descriptively at spot level. Coexpression among the five programmes was assessed in 304 TCGA-CESC primary tumours. qRT-PCR summary data from HUVEC (ATCC PCS-100-010) and primary human aortic endothelial cells (HAEC; ATCC PCS-100-011) were reinterpreted as endothelial pharmacology, not cardiac validation.

The scRNA-seq atlas contained 32 clusters spanning epithelial, immune, endothelial, and stromal lineages. None of the five specimen-level programmes differed significantly between malignant and non-malignant specimens after multiple-testing correction (all q ≥ 0.119). Likewise, none of the VEGFA-KDR, IL1B-IL1R1, TGFB1-TGFBR2, or LGALS9-HAVCR2 expression-product scores was significant (all q ≥ 0.143). The four spatial specimens showed heterogeneous spot-level expression but, with one specimen per histology and no slide coordinates in the supplied matrices, did not support group-level spatial inference. TCGA tumours showed coordinated intratumoural expression, strongest between angiogenesis and TGF-β/fibrosis (Spearman ρ=0.707; q = 2.24 × 10⁻46), but TCGA-CESC contained no CVT endpoint. VEGFA increased KDR, ICAM1, VCAM1, and EDN1 mRNA in HUVEC, and bevacizumab attenuated these changes; IL-1β induced inflammatory transcripts in HAEC, and anakinra attenuated IL6, TNF, and CXCL8.

Cervical tissues contain cardio-oncology-relevant transcriptional programmes, but the available data neither demonstrate cardiac expression nor identify bevacizumab- or ICI-induced CVT susceptibility. The endothelial experiments show expected pathway responsiveness and pharmacological blockade, not treatment-induced cardiotoxicity. Prospective analyses must compare CVT-positive and CVT-negative treated patients using matched clinical outcomes and, ideally, blood or cardiac-relevant biospecimens.