Integrative single-cell and exosomal multi-omics uncovers SCNN1A and EFNA1 as non-invasive biomarkers and drivers of ovarian cancer metastasis

Abstract

Background: Ovarian cancer (OV) is the deadliest gynecologic malignancy owing to its late diagnosis and high metastatic propensity. Current biomarkers lack sufficient sensitivity and specificity for the detection of early-stage cancer. To address this gap, we integrated single-cell transcriptomic profiling of tumor tissues with analysis of circulating exosomal RNA, aiming to uncover candidate markers that reflect tumor heterogeneity and metastatic potential and that may serve as sensitive, non-invasive diagnostics.

Methods: We integrated single-cell RNA sequencing (scRNA-seq) data from primary tumors and metastatic lesions with bulk tissue transcriptomes and plasma-derived exosomal RNA sequencing (RNA-seq). Differentially expressed genes (DEGs) shared across tumor cells, metastatic subpopulations, and exosomes were identified through intersection analysis. Candidate genes were validated in clinical specimens using qPCR and immunohistochemistry. We then applied ten machine learning algorithm to exosomal transcriptomic data to evaluate diagnostic performance and identify the optimal classifier. Tumor cell differentiation states were evaluated using CytoTRACE, and intercellular communication was analyzed with CellChat.

Results: Intersection analysis highlighted 52 overlapping DEGs, of which SCNN1A and EFNA1 emerged as the top prognostic indicators. Both genes were significantly upregulated in tumor tissues, metastatic foci, and plasma exosomes (P < 0.01). The exosome-based Adaboost model had an area under the curve of 0.955 in an independent test cohort. Single-cell subcluster analyses revealed high SCNN1A/EFNA1 expression correlated with stem-like differentiation states and enriched pathways associated with immune evasion and adhesion. CellChat analysis demonstrated that highly differentiated tumor cells extensively engaged with fibroblasts and endothelial cells, implying their role in niche formation.

Conclusions: By coupling single-cell, bulk tissue, and exosomal transcriptomics, we elucidated the key molecular drivers of OV metastasis and established SCNN1A and EFNA1 as promising non-invasive biomarkers. This multi-omics framework provides an effective strategy for early detection and a better understanding of metastatic progression in OV.

Keywords: biomarker; exosome; metastasis; ovarian cancer; single-cell RNA sequencing.

Figure 1

Tumor tissue atlas: (A) UMAP plot of OV tissue scRNA-seq. (B) Marker gene distribution of major cell types, with annotations for cell type, classic markers, and origin ratio. (C) UMAP for TCGA bulk RNA-seq, GTEx, and OV exosome bulk RNA-seq, differentiating tumor (pink) and control (blue) samples. (D) Bar plot of DEGs with tumor-upregulated (pink) and downregulated (blue) DEGs, along with enriched pathway proportions (KEGG). (E, F) Single-cell sub-cluster UMAPs and DEGs bar plots demonstrating differences between the tumor and control groups.

Figure 2

Metastatic tissue atlas: (A, B) UMAP of brain metastasis and pleural effusion scRNA-seq samples. (C) DEG count and proportion plot highlighting the intersection of OV exosome upregulated DEGs and DEGs of differential cell types. (D, E) Heatmap of high-expression genes and bubble plot of GO functional enrichment analysis.

Figure 3

Biomarker discovery for OV metastasis: (A) UpSet plot for intersecting gene analysis. (B, F) Survival analysis of the key genes. (G) qPCR validation of key genes in the plasma exosomes. (H) Classification performance of ten machine-learning models, assessed by six metrics: AUC, accuracy, recall, precision, F1 score, and Cohen’s kappa.

Figure 4

IHC validation: (A, B) IHC staining for EFNA1 and SCNNs in different tissue sections. (C, D) Quantification of positive staining regions. *P < 0.05.

Figure 5

Tumor cell subcluster analysis: (A) UMAP visualization of primary OV cell subclusters. (B) Bubble plot demonstrating cytoTRACE-predicted differentiation potential across tumor subclusters, with the expression profiles of key genes EFNA1 and SCNN1A. (C) GO functional enrichment analysis across the tumor subclusters. (D, E) UMAP projections of brain metastasis-derived (D) and pleural effusion-associated (E) subclusters, paired with bubble plots indicating cytoTRACE scores and marker gene expression levels. Abbreviations: Brain metastasis-derived MTCs cluster: Brain MTCs, Pleural effusion-derived MTCs cluster: Ple MTCs.

Figure 6

Cell-cell interaction analysis: (A) Heatmap of cellular interactions in primary OV samples. (B, C) Interaction heatmaps for brain and pleural effusion metastasis samples. Brain metastasis-derived MTCs cluster: Brain MTCs, Pleural effusion-derived MTCs cluster: Ple MTCs.