Pluripotent stem cell-derived epithelium misidentified as brain microvascular endothelium requires ETS factors to acquire vascular fate

Abstract

Cells derived from pluripotent sources in vitro must resemble those found in vivo as closely as possible at both transcriptional and functional levels in order to be a useful tool for studying diseases and developing therapeutics. Recently, differentiation of human pluripotent stem cells (hPSCs) into brain microvascular endothelial cells (ECs) with blood-brain barrier (BBB)-like properties has been reported. These cells have since been used as a robust in vitro BBB model for drug delivery and mechanistic understanding of neurological diseases. However, the precise cellular identity of these induced brain microvascular endothelial cells (iBMECs) has not been well described. Employing a comprehensive transcriptomic metaanalysis of previously published hPSC-derived cells validated by physiological assays, we demonstrate that iBMECs lack functional attributes of ECs since they are deficient in vascular lineage genes while expressing clusters of genes related to the neuroectodermal epithelial lineage (Epi-iBMEC). Overexpression of key endothelial ETS transcription factors (ETV2, ERG, and FLI1) reprograms Epi-iBMECs into authentic endothelial cells that are congruent with bona fide endothelium at both transcriptomic as well as some functional levels. This approach could eventually be used to develop a robust human BBB model in vitro that resembles the human brain EC in vivo for functional studies and drug discovery.

Keywords: blood–brain barrier; cellular identity; endothelial cells; induced pluripotent; single-cell RNA sequencing.

Fig. 1.

Metaanalysis of global high-throughput gene expression profiles reveal Epi-iBMECs possess an epithelial transcriptomic signature. (A) Schematic diagrams for differentiation of hPSCs to Epi-iBMECs highlighting changes (marked in red) implemented since its initial description as well as differentiation of hPSCs into generic endothelial cells (iECs). (B) Heatmap showing Pearson correlation coefficients between previously reported Epi-iBMEC transcriptomes and Epi-iBMECs generated in the current study. Epi-iBMECs generated by our group are molecularly equivalent to those reported in the literature. (C) Principal component analysis plot approximating relative relationship of 109 distinct cell samples across 22 library preparations from previously published and newly generated bulk RNA-sequencing data. (D) Volcano plots depicting gene ontology of biological processes using the top 100 positive loading genes of PC1 demonstrating expression of genes involved in epithelial cell processes. (E) Volcano plots depicting gene ontology of biological processes using the top 100 negative loading genes of PC1, demonstrating expression of genes involved in endothelial cell processes.

Fig. 2.

Characterization of endothelial and epithelial cell gene expression profiles in primary and hPSC-derived cells. (A) Heatmap illustrating expression of the top 100 most significant negative and positive loading genes of PC1 generated from unsupervised analysis of all bulk RNA gene expression profiles analyzed in Fig. 1C. (B) Violin plots of key endothelial cell genes acquired from top positive loading genes of PC1 of the bulk RNA expression profile analysis, demonstrating significant differences in expression among endothelial cell and both Epi-iBMEC as well as epithelial cell populations. (significance indicates *P value <0.05). (C) Violin plots of key epithelial cell genes acquired from top negative loading genes of PC1 of the bulk RNA expression profile analysis, demonstrating significant differences in expression among endothelial cell and both Epi-iBMEC as well as epithelial cell populations. (significance indicates *P value <0.05).

Fig. 3.

Single-cell RNA sequencing resolves transcriptomic composition of iBMECs and demonstrates rescue of a vascular identity upon transduction of three transcription factors. (A) Schematic diagram for the generation of rECs highlighting the induction of ETV2, ERG, and FLI1 expression at day 6 of the Epi-iBMEC differentiation as well as FACS isolation of CDH5⁺PECAM1⁺ cells at day 12 followed by expansion on 0.1% gelatin with serum-free EC media. (B) sc-RNA expression profiles displayed on a UMAP plot illustrate 30 distinct cell samples used in analysis of primary ECs, iECs, and Epi-iBMECs at various days of differentiation and hPSCs as well as choroid plexus cells. rECs are highlighted to illustrate that the endothelial transcriptomic signature of Epi-iBMECs is rescued upon induction of ETV2, ERG, and FLI1. (C) Heatmap emphasizing differences in expression levels of both endothelial- and epithelial-specific genes in all cell samples from Fig. 2B. Genes were derived from PC1 of unsupervised bulk RNA analysis depicted in Fig. 1C. The rEC sample (outlined) shows restoration of both EC-specific gene expression to levels comparable to those of EC controls. (D) Confocal microscopy and flow cytometry of control ECs (HUVEC and BMEC) for PECAM1 (green), CDH5 (red), EPCAM (purple), and DAPI (blue). Representative plots of n = 5 biological replicates. (Scale bars, 50 µm.) (E) Confocal microscopy and flow cytometry of control hPSC-derived iECs for PECAM1 (green), CDH5 (red), EPCAM (purple), and DAPI (blue). Representative plots of n = 5 biological replicates. (Scale bars, 50 µm.) (F) Confocal microscopy and flow cytometry of control hPSC-derived Epi-iBMECs for PECAM1 (green), CDH5 (red), EPCAM (purple), and DAPI (blue). Representative plots of n = 5 biological replicates. (Scale bars, 50 µm.) (G) Confocal microscopy and flow cytometry of rECs for PECAM1 (green), CDH5 (red), EPCAM (purple), and DAPI (blue). Representative plots of n = 5 biological replicates. (Scale bars, 50 µm.) (H) Violin plots of the samples from the sc-RNA sequencing analysis confirming expression of PECAM1 and CDH5 in all endothelial cell samples to be far higher than Epi-iBMEC samples. EPCAM is shown to be expressed at a much higher level in Epi-iBMEC samples compared to all endothelial cell samples.

Fig. 4.

Transduction of ETV2, ERG, and FLI1 promotes functional properties characteristic of an endothelial cell in Epi-iBMECs. (A) Confocal images for E-Selectin (blue), PECAM1 (green), and CDH5 (red) illustrate an increase in E-Selectin (CD62E) protein expression at the cell membrane upon addition of 100 ng mL⁻¹ tumor necrosis factor alpha (TNFα) in HUVEC and rEC samples but not in Epi-iBMEC samples. n = 5 biological replicates (Scale bars, 50 µm.) (B) Flow cytometry plots for E-Selectin and PECAM1 confirm increased E-Selectin protein expression in HUVEC and rEC samples upon stimulation with 100 ng mL⁻¹ TNFα in contrast to Epi-iBMECs. (C) Schematic illustration of the in vivo tubulogenesis assay for the angiogenic potential of Epi-iBMECs (derived from H1, H6, and IMR90-4 at day 16 of differentiation), rECs and HUVECs in Matrigel plugs. Cells were mixed into Matrigel and s.c. injected into NSG-SGM3 mice. Plugs were excised 5 d postinjection, cryosectioned, and analyzed for vessel formation. (D) Confocal microscopy images for DAPI (blue), EpCAM (purple), Agglutinin (green), and CDH5 (red) depicting in vivo formation of vessel-like structures in HUVEC and rEC plugs. White arrows in Epi-iBMEC panels show groups of cells forming EpCAM⁺ cell clusters. n = 10 mice per experimental cell line. (E) Quantifications of in vivo tubulogenesis assay with the AngioTool. The graph on the Top shows the total number of vessel junctions in each sample, whereas the one on the Bottom shows the total length of all vessels in each sample. n = 3 biological replicates, n = 10 mice per cell line tested. (significance indicates *P value <0.05).