Single-cell and bulk RNA sequencing reveal ligands and receptors associated with worse overall survival in serous ovarian cancer

Abstract

Background: Serous ovarian carcinoma is the most frequent histological subgroup of ovarian cancer and the leading cause of death among gynecologic tumors. The tumor microenvironment and cancer-associated fibroblasts (CAFs) have a critical role in the origin and progression of cancer. We comprehensively characterized the crosstalk between CAFs and ovarian cancer cells from malignant fluids to identify specific ligands and receptors mediating intercellular communications and disrupted pathways related to prognosis and therapy response.

Methods: Malignant fluids of serous ovarian cancer, including tumor-derived organoids, CAFs-enriched (eCAFs), and malignant effusion cells (no cultured) paired with normal ovarian tissues, were explored by RNA-sequencing. These data were integrated with single-cell RNA-sequencing data of ascites from ovarian cancer patients. The most relevant ligand and receptor interactions were used to identify differentially expressed genes with prognostic values in ovarian cancer.

Results: CAF ligands and epithelial cancer cell receptors were enriched for PI3K-AKT, focal adhesion, and epithelial-mesenchymal transition signaling pathways. Collagens, MIF, MDK, APP, and laminin were detected as the most significant signaling, and the top ligand-receptor interactions THBS2/THBS3 (CAFs)-CD47 (cancer cells), MDK (CAFs)-NCL/SDC2/SDC4 (cancer cells) as potential therapeutic targets. Interestingly, 34 genes encoding receptors and ligands of the PI3K pathway were associated with the outcome, response to treatment, and overall survival in ovarian cancer. Up-regulated genes from this list consistently predicted a worse overall survival (hazard ratio > 1.0 and log-rank P < 0.05) in two independent validation cohorts.

Conclusions: This study describes critical signaling pathways, ligands, and receptors involved in the communication between CAFs and cancer cells that have prognostic and therapeutic significance in ovarian cancer. Video abstract.

Keywords: Cancer-associated fibroblasts; Cell–cell communication; Organoids; Serous ovarian cancer; Single-cell RNA-sequencing.

Fig. 1

RNA-Sequencing analysis in cells from malignant effusions (baseline), 2D, and tumor-derived organoids (TDO) compared with normal ovarian tissues. A Top 1000 highly variable genes (highest standard deviation) across all samples. Rows and columns grouped by each condition (baseline, 2D, TDO, or normal tissue) were clustered based on the Euclidean distance between normalized counts values. In one patient, two TDO were obtained from distinct collections and days in culture (11 and 35 days), and are represented by an arrow and asterisks, respectively, at the bottom of the heatmap. AS: ascites and PE: pleural effusion. B Two-dimensional (2D) cell culture system presents a set of genes enriched for epithelial-mesenchymal transition (MSigDB Hallmark 2020), collagen fibril and extracellular matrix (ECM) organization (GO – Biological Process), and Focal Adhesion and PI3K-AKT-mTOR-signaling pathways (WikiPathway 2021). C Unsupervised PCA of RNA-Seq data (normalized counts) showing the segregation of ovarian cancer 2D cells and normal tissues from the cluster generated by ovarian cancer-derived organoids and baseline

Fig. 2

Single-cell transcriptomics re-analysis confirm the heterogeneity of cancer-associated fibroblasts (CAFs) in ovarian cancer malignant ascites. A Single-cell RNA-sequencing (scRNA-seq) data visualization of four CAFs populations using t-distributed stochastic neighbor embedding (tSNE). B Violin plots show the expression levels of canonical CAFs markers in CAFs populations, including myofibroblastic cancer-associated fibroblasts (myCAF1-2) and inflammatory cancer-associated fibroblasts (iCAF1-2). C Heatmap shows the pathway enrichment analysis (EnrichR [30, 31]) for gene sets of cluster-specific CAFs markers. Top 25 pathway terms (lowest adjusted P-value) selected from the WikiPathways 2021 and MSigDB Hallmark 2020 libraries available in EnrichR [30, 31]. D Violin plots show the expression levels of ovarian cancer CAFs markers among the four CAFs populations

Fig. 3

Cancer-associated fibroblasts (CAFs) share common outgoing communication patterns for the PI3K-AKT signaling pathway and focal adhesion in malignant effusions from ovarian cancer patients. Outgoing patterns for signaling A, cell B, and communication C of multiple cell types in malignant abdominal fluids of ovarian cancer patients obtained with CellChat [35] from scRNA-seq data [17]. The asterisks in C indicate pathways not associated with CAFs and epithelial cells communication (pattern 1, cluster 2). This communication pattern 1 (cluster 2) was used to identify pathways for gene sets of CAFs ligands D and epithelial cell receptors E for these ligands. In both D and E, the focal adhesion and PI3K-AKT-mTOR are among the top altered pathways. The top 10 pathway terms (lowest adjusted P-value) were selected from WikiPathways 2021 and MSigDB Hallmark 2020 libraries available in EnrichR [30, 31]

Fig. 4

Cancer-associated fibroblasts (CAFs) receive a specific communication pattern from malignant effusion cells associated with the differentiation pathway and EGFR tyrosine kinase inhibitor resistance. Incoming patterns for signaling A, cell B, and communication C of multiple cell types in malignant abdominal fluids of ovarian cancer patients obtained with CellChat [35] from scRNA-seq data [17]. The asterisks in C indicate pathways not associated with CAFs and epithelial cells communication (communication pattern 1, cluster 2). This communication pattern 1 (cluster 2) was used to identify pathways for gene sets of cells ligands D and CAFs receptors for these ligands E. In both D and E, the focal adhesion-PI3K-AKT-mTOR are among the top altered pathways. The top 10 pathway terms (lowest adjusted P-value) were selected from WikiPathways 2021 and MSigDB Hallmark 2020 libraries available in EnrichR [30, 31]

Fig. 5

Top interactions between cancer-associated fibroblasts (CAFs) and epithelial cells have multiple ligands and receptors associated with the PI3K-AKTsignaling pathway and focal adhesion. A Top ligand-receptor interactions (Commun. Probability > 0.10) between CAFs and epithelial cells. Heatmaps show the EnrichR pathway enrichment analysis for gene sets that define all CAFs ligands B and epithelial cell receptors C. The top 10 pathway terms (lowest adjusted P-value) were chosen from WikiPathways 2021 and MSigDB Hallmark 2020 libraries available in EnrichR [30, 31]. D PI3K-AKT ligand-receptor interactions between CAFs and epithelial cells. E Top ligand-receptors interactions (Commun. Probability > 0.10) between epithelial cells and CAFs. Heatmaps showing the EnrichR pathway enrichment analysis for gene sets that define all epithelial cells ligands F and CAFs receptors G. The top 10 pathway terms (lowest adjusted P-value) were chosen from WikiPathways 2021 and MSigDB Hallmark 2020 libraries available in EnrichR [30, 31]. Commun. Prob., communication probability

Fig. 6

Establishment of in vitro system models based on tumor-derived organoids and culture enriched with cancer-associated fibroblasts (eCAFs) to validate the expression profile of ligand and receptor genes associated with the PI3K-AKT signaling pathway in ovarian cancer. Gene expression of CAFs ligands A and epithelial cells receptors (B) for these ligands. Genes were ordered based on a marker selection (signal to noise) that highlights the differences in the expression profile (normalized counts) of eCAFs A or normal tissues B compared to the other three conditions. C Top 50 highly variable genes from the PI3K-AKT signaling pathway across all samples. Genes were ordered based on a marker selection (signal to noise) that highlights the differences in the expression profile (normalized counts) between normal tissue and malignant cells. D Expression profile (normalized counts) of ligands and receptors of the PI3K-AKT signaling pathway across all samples. Rows (for ligand or receptor genes) were ordered based on a marker selection (signal to noise) that highlights the differences in the expression profile (normalized counts) of normal tissues when compared to the other three conditions

Fig. 7

Increased expression of genes involved in PI3K-AKT signaling in cancer-associated fibroblasts and cancer epithelial cells from ovarian cancer patients is associated with worse overall survival. A Expression levels [log2(norm_count + 1)] of ligands and receptors of the PI3K-AKT signaling pathway in ovarian cystadenocarcinoma (n = 419) of The Cancer Genome Atlas (TCGA) compared to normal ovarian tissues (n = 88) tissues of Genotype-Tissue Expression (GTEx). Rows (for ligand or receptor genes) were ordered based on a marker selection (signal to noise) that highlights the differences in the expression profile [log2(norm_count + 1)] of GTEx tissues when compared to the TCGA tissues. Twenty genes exhibited increased expression levels and were selected to test their association with overall survival. B Forest plots for these twenty ligand and receptor genes of the PI3K-AKT signaling pathway in two cohorts of patients with ovarian cancer (TCGA, n = 557 and GSE9891, n = 264) available on the Kaplan–Meier (KM) plotter [37]. The hazard ratio (HR) with 95% confidence intervals (CI) determined by Cox proportional hazards model was used to evaluate the association between gene expression values and overall survival. The statistical significance was determined by a log-rank test. C Representative Kaplan–Meier curves of overall survival for patients with ovarian cancer (TCGA and GSE9891) based on the expression of IGF2, COL1A2, IGF1, and FGFR2