Transcriptomic relationships among human fallopian tubes, ovaries, and high-grade serous ovarian cancer.

(A) Overview of human samples used for bulk RNA-seq, including cancer-free ovaries, cancer-free fallopian tube segments (isthmus, ampulla, fimbria), and high-grade serous ovarian cancer (HGSOC) tumors. (B) Principal component analysis of bulk RNA-seq data showing three distinct clusters corresponding to ovary, fallopian tube, and HGSOC. (C) Hierarchical clustering and heatmap of pairwise Pearson correlations identifying two HGSOC subgroups: fallopian tube–like (EOC_F) and ovary-like (EOC_O). (D) Heatmap of seven major gene sets differentially enriched among ovary, fallopian tube, and EOC_F samples. Representative genes for each set are shown. See also Data File S1.

Cellular composition and differentiation hierarchy of the human fallopian tube epithelium revealed by snRNA-seq.

(A) UMAP visualization of 34,267 epithelial nuclei from human fallopian tubes, identifying 12 epithelial subclusters. (B) Hierarchical clustering of subclusters grouped into five major epithelial categories. (C) Dot plot showing marker genes defining stem cells, progenitors, secretory cells, and ciliated cells. (D) RNA velocity fields projected onto UMAP indicate directional flow from stem cells toward either secretory or ciliated lineages. (E) PAGA graph depicting connectivity among epithelial states, placing stem cells at the root. (F–G) PHATE embeddings and pseudotime trajectories showing two major differentiation paths: secretory and ciliated. (H–I) Heatmaps of dynamic gene expression along secretory (H) and ciliated (I) trajectories. (J–K) Immunofluorescence localization of LGR5 and PGR in basal epithelial stem cells of the fallopian tube. Scale bars: 50 µm. See also Figure S2 and Data File S2–S4.

Cellular diversity and inter-patient heterogeneity in human HGSOC.

(A) UMAP plot of 15,322 nuclei from four HGSOC tumors identifying 13 cellular clusters. (B) Dendrogram showing three non-malignant clusters (macrophages, cancer-associated fibroblasts (CAFs), endothelial cells) and five major HGSOC cell states. (C) Annotation of cancer subtypes: proliferative, mesenchymal-like, epithelial-to-mesenchymal transition (EMT), ciliated-like, and immunoreactive (MHC). (D) Dot plot of representative marker genes across cancer subtypes. (E–F) UMAP embeddings highlighting strong inter-patient heterogeneity among malignant cells (G) Correlation matrix demonstrating transcriptional divergence among malignant subtypes. (H) Expression patterns of markers defining proliferative, EMT, mesenchymal-like, ciliated-like, and immunoreactive cancer cell states. See also Figure S3–S4 and Data File S5–S6.

Lineage relationships linking fallopian tube epithelial progenitors to HGSOC malignant states.

(A) Force-directed graph layout showing 45,295 combined epithelial and cancer nuclei, identifying 15 integrated clusters. (B) PAGA analysis reveals three major lineages from epithelial stem cells: secretory cells, ciliated cells, and malignant cells. (C) RNA velocity projection demonstrating early progenitors giving rise to either differentiated epithelial states or cancer trajectories. (D) Pseudotime analysis showing malignant clusters emerging as late branches from the progenitor axis. (E) Hierarchical clustering divides clusters into ciliated, secretory, and cancer cell compartments. (F–G) UMAP overlays illustrating that a subset of malignant cells co-cluster with epithelial progenitors, indicating shared transcriptional programs. (H–I) Pseudotime heatmaps depicting temporal gene expression patterns along immunoreactive (H) and mesenchymal-like (I) malignant trajectories. See also Data File S7–S8.

Transcription factor regulatory networks governing normal fallopian tube differentiation and HGSOC malignant transformation.

(A) Integrative analysis workflow combining normal FTE and HGSOC nuclei for SCENIC regulon inference. (B) Heatmap of regulon activity (AUC scores) identifying stem, progenitor, secretory, ciliated, and cancer-specific transcriptional programs. (C–D) Ranking of regulons reveals SPDEF as a key stem/progenitor regulator and NR2F6 as a highly specific cancer-associated regulator. (E) Gene regulatory network map showing lineage-specific TF modules. (F–G) NR2F6 knockdown reduces colony formation in ES2 and OVCAR3 cells. (H–I) NR2F6 knockdown decreases invasive capacity of ES2 and OVCAR3 cells in Transwell assays. Data presented as mean ± SEM. Student’s t test. p < 0.05; p < 0.01; p < 0.001.

Genome-wide DNA copy number variation (CNV) landscape in HGSOC.

(A) Overview of inferCNV workflow using FTE epithelial cells as the reference population. (B) Hidden Markov Model–based classification of copy number states (0x, 0.5x, 1x, 1.5x, 2x, ≥3x). (C) Genome-wide CNV heatmap showing chromosomal losses and gains across malignant cells. (D) CNV-driven gene expression biases, highlighting amplified oncogenes (S100A6, MUC16, COL1A1/2, CD74, WFDC2) and genes reduced through copy loss. See also Data File S9.

Collagen signaling networks orchestrating stromal–tumor communication in HGSOC.

(A–B) Ligand–receptor interaction networks showing CAFs as the dominant source of outgoing signals and major regulators of the tumor microenvironment. (C) Selected ligand–receptor interaction patterns across malignant and stromal cell types. (D) Collagen signaling network indicating extensive cross-talk between CAFs and malignant cells. (E) Violin plots of collagen-related gene expression (COL1A1, COL1A2, COL4A1/2, COL6A1/2/3) across major tumor and non-tumor cell subtypes. See also Figure S5.