Back

Tumour architecture shapes polarized epithelial states that predict survival in high-grade serous ovarian cancer

Nersesian, S.; Abou-Hamad, J.; Durocher, E.; Akiki, G.; Domecq, C.; Southworth, A.; Deng, H.; Meunier, L.; de Ladurantaye, M.; Mes Masson, A.-M.; Tessier Cloutier, B.; Cook, D. P.

2026-06-01 cancer biology
10.64898/2026.05.27.727977 bioRxiv
Show abstract

Epithelial heterogeneity defines high-grade serous ovarian carcinoma (HGSC), yet principles that generate this diversity within and across tumours remain unclear. Integrating single-cell RNA sequencing (scRNA-seq) data from 13 studies (1,980,703 cells, 371 samples), we resolve a dominant axis of secretory cell polarization spanning proliferative, progenitor-like SecA cells and quiescent SecB cells expressing a mucosal injury response program. Targeted spatial transcriptomics across 8 whole HGSC tissues and a 97-patient tissue microarray shows this axis is spatially deterministic: tumour architecture shapes a hypoxic gradient along which SecB cells localize to avascular, luminal regions, where HIF/NF-{kappa}B-driven survival and glycolysis displace the mitogenic signalling of SecA. Within this niche, SecB cells rewire adhesion, ECM-remodelling, and immune-regulatory programs. Transcriptionally reprogrammed macrophages are enriched in this niche, while lymphocytes are excluded or dysfunctional. These cells assemble a coordinated multicellular niche poised for dissemination. SecB cells are enriched both in ascites and after chemotherapy, and are progressively lost in patient-derived organoids, only partially restored by IFN{gamma}, suggesting SecB is environmentally programmed rather than clonally fixed. SecB proportion independently predicts worse overall (HR = 1.31, p = 0.023) and progression-free survival (HR = 1.28, p = 0.011). Tumour architecture is thus a primary axis of malignant identity in HGSC, coupling microenvironment, cell state, and immune niche to clinical outcomes.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.