Skip to content

Atlas

Benign Fallopian Tube Epithelium Atlas

Generated from Single-cell sequencing of normal fallopian tube epithelial cells

The repertoire of serous ovarian cancer non-genetic heterogeneity (Hu et al., 2020), published in Cancer Cell, asks how much of the variability inside a serous ovarian tumor comes from cells switching between states rather than from mutations. That kind of variability helps tumors survive stress, but it is hard to measure, because genetic differences between cells blur it. The authors' answer was to start from the normal fallopian tube epithelium, where the disease is thought to begin, and use its cell states as a reference.

Explore the data: Open the dataset in the Discovery Lab TRE to run reproducible analyses where the data resides.

Dataset page · Open in the Discovery Lab TRE

Key Findings

  • Normal tube epithelium already holds distinct cell states. Sequencing about 6,000 epithelial cells, mostly from women without ovarian cancer, the authors identified six subtypes of fallopian tube epithelial cells.
  • Tumors carry a mix of those states. Using each subtype's gene signature to read tumor expression data, they found substantial variability within individual tumors.
  • The mix tracks survival. Grouping about 1,700 tumors by their subtype makeup separated patients by survival, a basis for more accurate prognosis.

The charts below come from the authors' deposited matrix of 1,857 cells from 5 donors. They show the epithelial cell states in that data, and that the programs defining them hold up in an independent atlas.

The Dataset

  • Cells: 1,857 benign fallopian tube epithelial cells, sequenced full length (Smart-seq2), one library per cell.
  • Donors: 5 donors across 6 plates.
  • Source: the authors' GEO count matrix (GSE139079).

Paper: Hu Z, Artibani M, Alsaadi A, et al. The Repertoire of Serous Ovarian Cancer Non-genetic Heterogeneity Revealed by Single-Cell Sequencing of Normal Fallopian Tube Epithelial Cells. Cancer Cell 37, 226–242 (2020). doi:10.1016/j.ccell.2020.01.003

Secretory and Ciliated Cells Make Up the Epithelium

The tube's lining is built from two kinds of epithelial cell: secretory cells, which are thought to give rise to serous ovarian cancer, and ciliated cells. In this data they are 64% and 28% of all cells. Each dot is a cell, placed so that cells with similar expression sit together, and colored by cell type.

Chart

Loading chart
1,857 cells colored by cell type: secretory, ciliated, and a few immune and stromal cells.

Secretory Cells Come in Several States

Even in this smaller sample, clustering separates the secretory cells into four groups and the ciliated cells into two. Secretory markers (PAX8, OVGP1) and ciliated markers (FOXJ1, CAPS) split the two lineages. Diversity among normal cells like this is what the authors, with about three times as many cells, resolved into six subtypes and used as a reference for reading tumors.

Chart

Loading chart
Average expression (color) and share of cells expressing (dot size) for 18 marker genes across the nine clusters.

The Same Programs Appear in an Independent Atlas

A reference is only useful if it holds up beyond one study. Comparing the same genes in this dataset and in the Postmenopausal Fallopian Tube and Ovary Atlas, from different patients, a different lab and a different sequencing method, shows strong agreement: genes that separate secretory from ciliated cells in one study do so in the other (rank correlation 0.82 across 7,065 genes).

Chart

Loading chart
Each point is a gene: how much more it is expressed in secretory than in ciliated cells, in each study. Eight marker genes are drawn larger.

Chart

Loading chart
Expression level in secretory cells agrees too (rank correlation 0.66 across 5,334 genes).
Discovery Lab

About These Charts

Clusters and cell types are data-driven: Leiden clustering on log1p CPM, with each cluster taking its best-scoring marker set. They are not the paper's published labels.

Counts are the authors' GEO matrix (Trim Galore, STAR, featureCounts on hg19). The cross-study charts join on gene symbol, and the two studies use different units, so agreement is reported as rank correlation.

These charts were built in the Discovery Lab TRE (count matrix d088b92e; tables 39c4c0a9, with the Lengyel pseudobulk from fbe13ac1).