How can the identity of pluripotent stem cells be determined more reliably during their differentiation? Researchers from the IUF – Leibniz Research Institute for Environmental Medicine and Heinrich Heine University Duesseldorf have systematically validated marker genes for this purpose and developed a web-based application for the automated classification of early cell fates. The study has been published in Stem Cell Reports.

A key step in early human development is the establishment of cell identities, which takes place in successive phases. During this process, pluripotent stem cells pass through transient and partially overlapping states before committing to a lineage. These changes in gene expression are one of the major challenges when working with human pluripotent stem cells (hPSCs): existing markers cannot always distinguish transient changes in gene expression from true lineage commitment.

To determine the developmental state of cells more accurately, the research team Jochen Dobner, Alessandro Prigione and Andrea Rossi established a curated panel of 67 validated markers: 20 for the undifferentiated state, 17 for endoderm, and 15 each for ectoderm and mesoderm. Starting from 172 candidate genes identified by RNA sequencing, the team experimentally validated markers by qPCR in multiple independent stem cell lines and combined them with markers from their previous work.

Using the marker panel, publicly available RNA-seq datasets could be separated into undifferentiated cells and the three germ layers. The marker signatures also mapped onto the corresponding cell populations in reference datasets from human and macaque embryos at gastrulation stages. For a subset of markers, the results were confirmed at the protein level using comprehensive proteomic profiling of the reference cell line KOLF2.1J. Researchers can thus use the panel to assess the differentiation state of their cultures more reliably, even in heterogeneous datasets.

To make these results available to the research community, the authors developed DeepDiff. This freely accessible web application provides an overview of gene expression across 33 early differentiation states of hPSCs. It also gives access to protein-level data for KOLF2.1J, which closes a gap in available resources, along with a correlation analysis linking transcript and protein levels and a machine learning-based classifier for the automated annotation of bulk RNA-seq data.

Availability – https://jdobner.shinyapps.io/diff_sets/