Discovery of a pre-vein progenitor that requires VEGF/ERK inhibition to complete vein differentiation

Abstract

Despite substantial insight into mechanisms underlying arterial blood vessel development, multiple aspects of vein development remain elusive, including vein-determining extracellular signals and cell-fate trajectories^1-10. One might expect arteries and veins to arise simultaneously during development, as both are needed for a functional circulatory system. Nevertheless, arteries often precede veins in vivo, as exemplified by the first intraembryonic blood vessels^11-17. Here we present a model of vein differentiation that answers longstanding questions in the field. By reconstituting human vein endothelial cell (EC) differentiation from mesoderm in vitro, we discovered that vein development unfolds in two steps driven by opposing signals. First, VEGF is necessary to differentiate mesoderm into "pre-vein" ECs-a newly-defined intermediate state-and to endow endothelial identity. Second, once cells have acquired pre-vein EC identity, VEGF/ERK inhibition is necessary to specify vein ECs. Pre-vein ECs co-expressed certain arterial (SOX17) and venous (APLNR) markers and harbored poised chromatin at future venous genes. However, VEGF/ERK inhibition was necessary to activate poised venous genes (e.g., NR2F2), and for pre-vein ECs to complete venous differentiation. Intersectional lineage tracing supported a pre-vein intermediate step in vivo: early Sox17+ Aplnr+ ECs also formed veins in mouse embryos. We leveraged this developmental knowledge for disease modeling by differentiating human pluripotent stem cells into artery and vein ECs, and comparing their responses to Ebola and Andes viruses under biosafety-level-4 containment. Artery and vein ECs responded divergently to the same virus, thus revealing that developmentally specified cell identity impacts viral infection. Taken together, we propose a two-step model for vein development wherein VEGF first differentiates mesoderm into pre-vein ECs, but subsequent VEGF/ERK inhibition generates vein ECs. VEGF activation is thought to be broadly essential for vascular development^6,18, and thus our discovery that VEGF/ERK inhibition specifies vein identity has potential implications for understanding current therapies that either activate or inhibit VEGF signaling^19,20.

Extended Data Figure 1:. Cellular diversity during early steps of hPSC differentiation

A) Summary of hPSC differentiation approach. i: inhibition. B) Summary of gene expression, chromatin state, and cell-surface marker profiling conducted as part of this study. Dotted lines indicate that profiling was not conducted on a given cell-type. C) Gene expression on days 0, 1, and 2 of H1 hPSC differentiation, as detected by scRNAseq. D) scRNAseq of day-0 H1 hPSCs that were differentiated into day-1 primitive streak, day-2 lateral mesoderm, day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs. None of cell populations shown here were FACS purified and consequently, mesenchymal cells were detected alongside artery, pre-vein, and vein ECs on days 3–4 of differentiation. The same day-0 hPSC, day-1 primitive streak, and day-2 primitive streak scRNAseq datasets were also used in Fig. 1A; however, FACS-purified downstream EC populations are shown in Fig. 1A. E) scRNAseq of H1 hPSC-derived mesenchymal cells that arose alongside day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs. Differentially expressed genes that distinguish these three different types of mesenchyme are shown. Diff: differentiation.

Extended Data Figure 2:. Characterization of pre-vein endothelial cells in vitro

A) Bulk RNA-seq of the indicated H1 hPSC-derived cell-types. FACS was used to purify CD144+ pre-vein and vein ECs for RNA-seq. i: inhibition. B) qPCR of the indicated H1 hPSC-derived cell-types. qPCR data normalized to reference gene YWHAZ (i.e., YWHAZ levels = 1.0). Error bars: standard deviation (SD). C) scRNAseq of H1 hPSC-derived day-3 pre-vein EC populations, showing the percentage of cells that expressed APLNR and/or SOX17 mRNAs. D) scRNAseq of FACS-purified CD144+ DLL4+ CD73lo/− day-3 artery ECs, CD144+ day 3 pre-vein ECs, and CD144+ DLL4− CD73hi day-4 vein ECs generated from H1 hPSCs. In each of these in vitro cell-types, expression of in vivo arterial and venous markers reported by Hou et al., 202214 and Su et al., 201833 is shown

Extended Data Figure 3:. Cell-surface markers of human arteriovenous identity

A) LEGENDScreen high-throughput flow cytometry of cell-surface markers in day-0 hPSCs, day-1 primitive streak, day-2 lateral mesoderm, day-3 artery ECs, and day-4 vein ECs. Day-3 artery and day-4 vein EC populations were pre-gated on the CD144+ EC subset before depicting marker expression B) LEGENDScreen high-throughput flow cytometry of cell-surface markers in day-0 hPSCs, day-1 primitive streak, day-2 lateral mesoderm, day-3 artery ECs, and day-4 vein ECs. Day-3 artery and day-4 vein EC populations were pre-gated on the CD144+ EC subset before depicting marker expression. C) scRNAseq of human Carnegie Stage 17 (CS17) fetal ECs. scRNAseq data were obtained from Calvanese et al., 202266. D) Left: Population heterogeneity of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs, before and after FACS purification based on the indicated cell-surface marker combinations. Proportions of cell clusters in scRNAseq data are shown. Right: The copyright holder for this preprint bioRxiv preprint doi: https://doi.org/10.1101/2025.10.11.681838; this version posted December 4, 2025. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license. scRNAseq of FACS-purified CD144+ DLL4+ CD73lo/− day-3 artery ECs and CD144+ DLL4− CD73hi day-4 vein ECs generated from H1 hPSCs. E) scRNAseq of FACS-purified CD144+ DLL4+ CD73lo/− day-3 artery ECs generated from H1 hPSCs. Left: FACS isolation strategy. Right: subclustering was performed to assess any potential population heterogeneity, and differentially expressed genes that distinguished cell subsets are shown. F) scRNAseq of FACS-purified CD144+ DLL4− CD73hi day-4 vein ECs generated from H1 hPSCs. Left: FACS isolation strategy. Right: subclustering was performed to assess any potential population heterogeneity, and differentially expressed genes that distinguished cell subsets are shown.

Extended Data Figure 4:. Generation and characterization of pre-vein endothelial cells in vitro

A) First, H1 hPSCs were differentiated into day-2 lateral mesoderm. Then, lateral mesoderm was then treated with different doses of VEGF pathway modulators (VEGF [0–100 ng/mL] or ERK inhibitor [ERKi; PD0325901, 500 nM]) alongside other pre-vein EC-inducing signals (TGFβ inhibitor + NOTCH inhibitor + BMP inhibitor + WNT inhibitor + Vitamin C) for 24 hours. Flow cytometry was conducted on day 3 of hPSC differentiation. This revealed that high VEGF for 24 hours is required during pre-vein EC specification to efficiently generate ECs by day 3. B) Flow cytometry of H1 SOX17-2A-mPlum; NR2F2-2A-GFP hPSCs differentiated into day-3 artery ECs or day-5 vein ECs. Day-4 vein ECs were maintained in the same vein induction media until day 5. Day 3–5 populations were pre-gated on the CD144+ EC subset before depicting marker expression. C) First, H1 hPSCs were differentiated into day-3 pre-vein ECs. Then, pre-vein ECs were then treated with different doses of VEGF pathway modulators (VEGF [0–100 ng/mL] or ERK inhibitor [ERKi; PD0325901, 500 nM]) alongside other vein EC-inducing signals (TGFβ inhibitor + NOTCH inhibitor + WNT agonist + Vitamin C) for 1–24 hours. qPCR was conducted on day 4 of hPSC differentiation, and expression is normalized to the sample with the highest expression. This revealed that ERK inhibition for 12 hours significantly upregulated venous marker expression. D) CUT&RUN profiling of H3K4me3 and H3K27me3 and bulk RNA-seq H1 hPSC-derived day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs.

Extended Data Figure 5:. Comparison of methods to differentiate hPSCs into ECs

A) scRNA-seq of hPSCs differentiated into ECs using various protocols, depicting the number of genes detected per single cell as a quality control metric. B) Summary of protocols to differentiate hPSCs into ECs^,– used to generate the scRNA-seq datasets described in Fig. 3A. C) scRNA-seq of differentiated hPSC populations described in Fig. 3A. Subclustering was performed to assess population heterogeneity, and differentially expressed genes that distinguished cell subsets are shown. Clusters are annotated and colored as described in Fig. 3A.

Extended Data Figure 6:. Comparison of methods to differentiate hPSCs into ECs

A) scRNA-seq of differentiated hPSC populations described in Fig. 3A, depicting expression of pan-endothelial, arterial, venous, and mesenchymal marker genes. Clusters are annotated and colored as described in Fig. 3A.

Extended Data Figure 7:. Improved generation and expansion of artery and vein ECs, while preserving arteriovenous identity

A) Schematic of differentiation protocol, which includes prolongation of lateral mesoderm induction from 24 hours (V1 protocol) to 40 hours (V2 protocol). B) qPCR of lateral mesoderm and vascular marker expression in H1 hPSC-derived day-1 primitive streak cells challenged with lateral mesoderm induction media for up to 48 hours. 0 hours corresponds to day-1 primitive streak cells, prior to addition of lateral mesoderm induction media. Cells were lysed every 4 hours and qPCR was performed. For each gene, expression is normalized to the sample with the highest value, which was set to “1.0”. Error bars: SEM. C) Effect of lateral mesoderm induction duration on the subsequent production of CD144+ artery and vein ECs from the H1 and WTC11 hPSC lines, as measured by flow cytometry. Duration of lateral mesoderm induction was varied from 24 hours (V1 protocol) to 48 hours, with all other parameters of the differentiation protocol unchanged. Error bars: SEM. D) Schematic of CellSTACK system for large-scale differentiation of hPSCs into ECs. E) Yield and percentage of CD144+ DLL4+ day-3.67 artery ECs generated in the large-scale CellSTACK differentiation system, depending on initial hPSC seeding density. Yield and differentiation purity were respectively measured by flow cytometry and cell counting. F) Yield and percentage of CD73+ DLL4+ day-4.67 vein ECs generated in the large-scale CellSTACK differentiation system, depending on initial hPSC seeding density. Yield and differentiation purity were respectively measured by flow cytometry and cell counting. G) Strategy for expanding hPSC-derived artery and vein ECs. Artery ECs are cultured in EGM2 medium, and vein ECs are cultured in EGM2 medium + TGFβ inhibitor + NOTCH inhibitor + PKA activator. i: inhibitor. H) qPCR of H1 hPSC-derived artery ECs that were thawed in EGM2 medium and cultured for 6 days. ROCK inhibitor (Thiazovivin, 2 μM) was added for the first day of thawing to improve cell survival, and was subsequently removed in later days. Forskolin (3 μM) was added as indicated. For each gene, expression is normalized to the sample with the highest value, which was set to “1.0”. Error bars: SEM. I) qPCR of H1 hPSC-derived vein ECs that were thawed in EGM2 medium + SB505124 (2 μM, TGFβ inhibitor) + RO4929097 (1 μM, NOTCH inhibitor) and cultured for 6 days. ROCK inhibitor (Thiazovivin, 2 μM) was added for the first day of thawing to improve cell survival, and was subsequently removed in later days. Forskolin (3 μM) was added as indicated. For each gene, expression is normalized to the sample with the highest value, which was set to “1.0”. Error bars: SEM. J) qPCR of H1 hPSC-derived vein ECs that were thawed in EGM2 medium + SB505124 (2 μM, TGFβ inhibitor) + RO4929097 (1 μM, NOTCH inhibitor) + Forskolin (10 μM, adenylate cyclase agonist) and cultured for 6 days. ROCK inhibitor (Thiazovivin, 2 μM) was added for the first day of thawing to improve cell survival, and was subsequently removed in later days. As a control, H1 hPSC-derived artery ECs prior to expansion were also analyzed. qPCR data are normalized to the sample with the highest expression, which was set to “1.0”. Error bars: SEM. K) qPCR of H1 hPSC-derived vein ECs that were thawed in EGM2 medium + SB505124 (2 μM, TGFβ inhibitor) + RO4929097 (1 μM, NOTCH inhibitor), in the presence or absence of Forskolin (3–10 μM, adenylate cyclase agonist) and cultured for 6 days. qPCR data are normalized to the sample lacking forskolin, which was set to “1.0”. Error bars: SEM. L) qPCR of pan-endothelial, arterial, and venous marker expression over the course of 6-day expansion of SUN004.2 CAG-mScarlet hPSC-derived artery and vein ECs. Artery ECs were expanded in EGM2 medium, whereas vein ECs were expanded in EGM2 medium + SB505124 (2 μM, TGFβ inhibitor) + RO4929097 (1 μM, NOTCH inhibitor) + Forskolin (10 μM, adenylate cyclase agonist). qPCR data normalized to reference gene YWHAZ (i.e., YWHAZ levels = 1.0). Error bars: SEM.

Extended Data Figure 8:. Existence and fate of Sox17+ Aplnr+ endothelial cells in the mouse embryo

A) scRNAseq of E9.5 mouse embryo ECs. scRNAseq data were obtained from Chen et al., 2024. B) scRNAseq of Carnegie Stage 12 (CS12) human embryo ECs. scRNAseq data were obtained from Calvanese et al., 2022. C) Whole-mount Erg and Sox17 immunostaining of E8.75 mouse embryo. DA: dorsal aorta (arrowhead). VV: vitelline vein (arrowhead). ISV: intersomitic vessel (arrowhead). Arrows: solitary Sox17+ Erg+ ECs. Ant: anterior. Post: posterior. Scale: 100 μm. D) Whole-mount Aplnr mRNA staining of E10 Sox17^Cre; R26^zsGreen mouse embryos^, by HCR3 in situ hybridization. Scale: 100 μm. E) E12.5 Sox17^Cre; R26^mTmG mouse embryos^, were sectioned and stained for Vegfr2, SMAα, and GFP proteins. Scale: 200 μm. F) 4-hydroxytamoxifen (4OHT) was administered in utero to E8.75 Aplnr-CreER; R26-tdTomato mouse embryos^,, which were then isolated at E9.75 and immunostained for Sox17 and Erg proteins. Bottom row arrows: cardinal vein. Scale: 200 μm (top row), 50 μm (bottom row). G) 4-hydroxytamoxifen (4OHT) was administered in utero to E8.75 Aplnr-CreER; R26-tdTomato mouse embryos^,, which were then isolated at E9.75 and immunostained for Sox17 and Erg proteins. Quantification of Erg+ ECs in the dorsal aorta and cardinal vein that co-expressed tdTomato (indicative of Aplnr-CreER activity) and Sox17. Error bars: SD. **P<0.01.

Extended Data Figure 9:. Triple CRISPRi knockdown of SOXF transcription factors impairs both human artery and vein EC differentiation in vitro

A) qPCR of day-3 artery ECs derived from control and single SOX7-, SOX17-, or SOX18-knockdown H1 CRISPRi hPSC lines. Gene expression normalized to levels in control hPSC-derived artery ECs, which was set as “1.0”. “X” indicates that qPCR data were not shown, because gene expression was under 2% of YWHAZ in control samples. Statistics: unpaired t-test. **P<0.01. B) qPCR of day-3 artery ECs derived from control and single SOX7-, SOX17-, or SOX18-knockdown H1 CRISPRi hPSC lines. Gene expression normalized to levels in control hPSC-derived artery ECs, which was set as “1.0”. Statistics: unpaired t-test. Error bars: SEM. n.s.: not significant. *P<0.05. **P<0.01. C) qPCR of day-3 artery ECs generated from H1 control vs. SOXF TKD hPSCs. Gene expression normalized to control artery ECs. Percentages indicate remaining gene expression in SOXF TKD artery ECs, relative to controls. Statistics: unpaired t-test. Error bars: standard error of the mean (SEM). D) Left: scRNA-seq of day-3 artery ECs, day-3 pre-vein ECs and day-4 vein ECs generated from either H1 control or SOXF TKD CRISPRi hPSCs. Right: PECAM1+ ECs and PECAM1− mesenchymal cells were detected. The same scRNAseq datasets are shown here as in Fig. 4C, except that mesenchymal cells were computationally excluded in Fig. 4C. E) qPCR of day-3 artery ECs and day-4 vein ECs generated from H1 control vs. SOXF TKD CRISPRi hPSCs. Gene expression normalized to reference gene YWHAZ (i.e., YWHAZ = 100%). Statistics: unpaired t-test. Error bars: SEM. *P<0.05. **P<0.01. F) Differentially expressed genes between day-3 pre-vein ECs generated from H1 control vs. SOXF TKD CRISPRi hPSCs are colored. G) Gene Set Enrichment analysis (GSEA) of scRNA-seq data from H1 CRISPRi control vs. SOXF TKD hPSCs differentiated into day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs. Color represents the log₁₀-transformed adjusted P-value. H) Left: OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 pre-vein ECs and day-4 vein ECs. Right: scRNA-seq of day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs generated from H1 control vs. SOXF TKD hPSCs. SOXF genes are required for CDH5 (VE-CADHERIN) expression in pre-vein ECs.

Extended Data Figure 10:. Chromatin hallmarks of human arteriovenous identity

A) OmniATAC-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs. Left: genetic loci preferentially accessible in either artery or vein ECs are colored. Right: transcription factor motifs enriched in artery- or vein-accessible chromatin elements. B) Expression of AP1, ETV, SOX, NR2F, and MAF family transcription factors at the mRNA level, as shown by scRNAseq of FACS-purified CD144+ DLL4+ CD73lo/− day-3 artery ECs and CD144+ DLL4− CD73hi day-4 vein ECs generated from H1 hPSCs. C) OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs. D) OmniATAC-seq and CUT&RUN analysis of the DLL4 −12kb enhancer in H1 hPSC-derived day-3 artery ECs, with transcription factor motifs labeled. E) LacZ staining of E8.5-E16.5 mouse embryos bearing a Dll4 −12 kB enhancer transgene driving LacZ expression. F) OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs. G) Top: image of a 3 day-post-fertilization (dpf) zebrafish bearing a Cxcr4 +135 kB enhancer driving GFP expression, together with a kdrl:HRAS-mCherry transgene to label ECs. Bottom: ATAC-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs, highlighting the CXCR4 +135 kB enhancer whose ortholog was tested in the zebrafish transgenic assay. H) OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs. I) LacZ staining of E11.5 mouse embryo bearing a Sema3g −13 kB enhancer transgene driving LacZ expression (VISTA Enhancer Browser ID hs2179)^,. Scale: 100 μm. J) Image of a 2 day-post-fertilization (dpf) zebrafish bearing a Sema3g −13 kB enhancer driving GFP expression. K) OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs. L) OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 artery ECs and day-4 vein ECs.

Extended Data Figure 11:. Infection of hPSC-derived artery and vein ECs by Ebola and Andes viruses

A) Bulk RNA-seq of hPSC-derived artery ECs 6, 12, 24, and 48 hours after infection with Ebola, Andes, and Sendai viruses, or left uninfected (mock control). Error bars: SEM. B) Replication of Andes and Ebola viruses in hPSC-derived artery and vein ECs, as assayed by qPCR for intracellular viral genomes. C) Bulk RNA-seq of artery and vein ECs 48 hours after infection with Sendai, Andes, and Ebola viruses. Genes that are differentially expressed between infected artery and vein ECs, but are not differentially expressed between uninfected artery and vein ECs are correspondingly colored. Red: genes more highly induced by viral infection in hPSC-derived artery relative to vein ECs, and which do not show substantially different expression in uninfected artery vs. vein ECs. Purple: genes more highly induced by viral infection in hPSC-derived vein relative to artery ECs, and which do not show substantially different expression in uninfected artery vs. vein ECs. D) Number of genes that are differentially expressed in hPSC-derived artery vs. vein ECs upon infection with a given virus.

Figure 1:. A roadmap for human arteriovenous differentiation reveals pre-vein endothelial cells

A) scRNA-seq of day-0 H1 hPSCs, day-1 primitive streak, day-2 lateral mesoderm, CD144+ FACS-purified day-3 artery ECs, CD144+ FACS-purified day-3 pre-vein ECs, and CD144+ FACS-purified day-4 vein ECs. B) Differentially expressed genes that distinguish each hPSC-derived cell-type, as detected by scRNAseq. scRNA-seq was performed on day-0 H1 hPSCs, day-1 primitive streak, day-2 lateral mesoderm, CD144+ FACS-purified day-3 artery ECs, CD144+ FACS-purified day-3 pre-vein ECs, and CD144+ FACS-purified day-4 vein ECs. C) scRNA-seq of day-2 lateral mesoderm, CD144+ FACS-purified day-3 artery ECs, CD144+ FACS-purified day-3 pre-vein ECs, and CD144+ FACS-purified day-4 vein ECs. D) scRNA-seq of day-2 lateral mesoderm, CD144+ FACS-purified day-3 artery ECs, CD144+ FACS-purified day-3 pre-vein ECs, and CD144+ FACS-purified day-4 vein ECs. Gene expression is depicted in log_e unique molecular identifier (UMI) counts. E) Combined immunostaining for SOX17 protein and HCR3 in situ hybridization for APLNR mRNA in the indicated H1 hPSC-derived cell-types. Scale: 200 μm. F) Flow cytometry of H1 CRISPRi-expressing hPSCs, day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs. Day 3–4 populations were pre-gated on the CD144+ EC subset before depicting marker expression.

Figure 2:. Two separable steps of vein differentiation driven by temporally dynamic VEGF/ERK activation, followed by inhibition

A) Summary of the present study. B) First, H1 hPSCs were differentiated into day-2 lateral mesoderm. Then, lateral mesoderm was then treated with different doses of VEGF pathway modulators (VEGF [0–100 ng/mL] or ERK inhibitor [ERKi; PD0325901, 500 nM]) alongside other pre-vein EC-inducing signals (TGFβ inhibitor + NOTCH inhibitor + BMP inhibitor + WNT inhibitor + Vitamin C) for 24 hours. Subsequently, cells were subject to vein EC differentiation for 24 hours. qPCR was conducted on day 4 of hPSC differentiation. In heatmaps, expression is normalized to the sample with the highest expression in either panel B or C. This revealed that high VEGF for 24 hours is required during pre-vein EC specification to subsequently generate vein ECs by day 4. Statistics: Wilcoxon rank sum test. Error bars: standard deviation (SD). *P<0.05. C) First, H1 hPSCs were differentiated into day-3 pre-vein ECs. Then, pre-vein ECs were then treated with different doses of VEGF pathway modulators (VEGF [0–100 ng/mL] or ERK inhibitor [ERKi; PD0325901, 500 nM]) alongside other vein EC-inducing signals (TGFβ inhibitor + NOTCH inhibitor + WNT agonist + Vitamin C) for 24 hours. qPCR was conducted on day 4 of hPSC differentiation. In heatmaps, expression is normalized to the sample with the highest expression in either panel B or C. This revealed that ERK inhibition for 24 hours is required to generate day-4 vein ECs, and that after cells acquire endothelial identity, VEGF/ERK is dispensable for the continued expression of pan-EC markers. Statistics: Wilcoxon rank sum test. Error bars: SD. **P<0.01, *P<0.05, n.s.: not significant. D) First, H1 hPSCs were differentiated into day-3 pre-vein ECs. Then, pre-vein ECs were then treated with either VEGF (100 ng/mL) or ERK inhibitor (PD0325901, 500 nM) alongside other vein EC-inducing signals (TGFβ inhibitor + NOTCH inhibitor + WNT agonist + Vitamin C) for 1–24 hours, followed by qPCR. This revealed that ERK inhibition for 12 hours significantly upregulated venous marker expression. E) Bulk-population RNA-seq of FACS-purified CD144+ day-3 pre-vein ECs and day 4-vein ECs generated from H1 hPSCs. Differentially expressed genes are colored purple. F) CUT&RUN profiling of H3K4me3 and H3K27me3 and bulk RNA-seq H1 hPSC-derived day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs.

Figure 3:. Comparison of differentiation methods suggests that temporally dynamic VEGF modulation is crucial to impart venous identity

A) scRNA-seq of hPSCs differentiated into ECs using various protocols, with the number of days of differentiation indicated. scRNA-seq datasets were subclustered at the same resolution to identify population heterogeneity. Clusters were annotated by marker expression. The proportion of ECs is indicated. scRNA-seq data were obtained from this study, Ang et al., 2022, Pan et al., 2024, McCracken et al., 2019, Paik et al., 2018, and Nikolova et al., 2025. B) scRNA-seq of differentiated hPSC populations described in Fig. 3A, and ECs were computationally isolated. An expression module score of arterial markers defined in vivo by Hou et al., 2022 is shown. Statistics: Wilcoxon rank sum test. **P<0.01. C) scRNA-seq of differentiated hPSC populations described in Fig. 3A, and ECs were computationally isolated. An expression module score of venous markers defined in vivo by Hou et al., 2022 is shown. Statistics: Wilcoxon rank sum test. **P<0.01.

Figure 4:. Sox17+ Aplnr+ endothelial cells exist in the early embryo, and contribute to vein endothelial cells

A) scRNAseq of E9.5 mouse embryo ECs. scRNAseq data were obtained from Chen et al., 2024. B) scRNAseq of Carnegie Stage 12 (CS12) human embryo ECs. scRNAseq data were obtained from Calvanese et al., 2022. C) Whole-mount Erg and Sox17 immunostaining of E8.5 mouse embryo. Ant: anterior. Post: posterior. DA: dorsal aorta (arrowhead). VV: vitelline vein (arrowhead). Arrow: single migrating Erg+ Sox17+ ECs. Scale: 100 μm. D) Whole-mount Erg and Sox17 immunostaining of E8.5 mouse embryo. E) Percentage of Erg+ ECs in the E9.5 dorsal aorta (DA) or cardinal vein (CV) that express Sox17. Error bars: SD. F) Left: whole-mount Erg immunostaining of E9.5 mouse embryo. Right: E9.5 mouse embryo sectioned and immunostained for Sox17 and Erg proteins, alongside in situ staining for Aplnr mRNA. Arrow: cardinal vein. Scale: 200 μm. G) sox7 and aplnrb staining of 27-hour post fertilization zebrafish embryos by HCR3 in situ hybridization. Lateral view of the mid-trunk region. Scale: 50 μm. H) 4OHT was administered in utero to E9.5 Sox17-Cre; Aplnr-DreER; RLTG mouse embryos^,,, which were then isolated at E12.5, cryosectioned, and immunostained for Erg and GFP proteins. Arrows: GFP+ vein ECs. Scale: 100 μm. Error bars: SD.

Figure 5:. SOXF transcription factors are required for human vein EC specification in vitro

A) scRNA-seq of day-2 lateral mesoderm, CD144+ FACS-purified day-3 artery ECs, CD144+ FACS-purified day-3 pre-vein ECs, and CD144+ FACS-purified day-4 vein ECs generated from H1 hPSCs. Arrows indicate that pre-vein ECs express SOX7, SOX17, and SOX18. B) H1 CRISPRi-expressing hPSCs were transduced with sgRNAs targeting SOX7, SOX17, and SOX18 (SOXF triple knockdown [TKD]), and then subsequently differentiated into artery and vein ECs. C) scRNA-seq of day-3 artery ECs and day-4 vein ECs generated from H1 control vs. SOXF TKD hPSCs. Mesenchymal cells were computationally excluded. D) scRNA-seq of day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs generated from H1 control vs. SOXF TKD hPSCs. Mesenchymal cells were computationally excluded. A transcriptional module score computed from a panel of in vivo-defined arterial marker genes in vivo is shown. Statistics: Wilcoxon rank sum test. **P<0.01. E) scRNA-seq of day-3 artery ECs and day-4 vein ECs generated from H1 control vs. SOXF TKD hPSCs. Mesenchymal cells were computationally excluded. F) scRNA-seq of day-3 artery ECs, day-3 pre-vein ECs, and day-4 vein ECs generated from H1 control vs. SOXF TKD hPSCs. Mesenchymal cells were computationally excluded. A transcriptional module score computed from a panel of in vivo-defined venous marker genes in vivo is shown. Statistics: Wilcoxon rank sum test. **P<0.01. G) scRNA-seq of day-3 artery ECs and day-4 vein ECs generated from H1 control vs. SOXF TKD hPSCs. Mesenchymal cells were computationally excluded. Differentially expressed genes between control vs. SOXF TKD ECs are colored. H) OmniATAC-seq, CUT&RUN, and bulk RNA-seq of H1 hPSC-derived day-3 pre-vein ECs and day-4 vein ECs.

Figure 6:. hPSC-derived artery and vein ECs respond differently to Ebola and Andes viruses

A) Experimental summary. h: hour. B) Replication of Ebola and Andes viruses in hPSC-derived artery and vein ECs, as assayed by qPCR for viral genomes in the culture media. Statistics: unpaired t-test. Error bars: SEM. **P<0.01. n.s.: not significant. C) Bulk RNA-seq of interferon and interferon-stimulated gene expression in hPSC-derived artery and vein ECs 6, 12, 24, and 48 hours after infection with Sendai, Andes, or Ebola viruses. Fold change relative to uninfected cells is depicted. D) IFNβ protein secretion by hPSC-derived artery ECs after 24 or 48 hours of infection by Ebola, Andes, or Sendai viruses, or left uninfected (mock control), as measured by ELISA. Statistics: unpaired t-test with Welch correction. **P<0.01. E) Bulk-RNA-seq of hPSC-derived artery ECs 6, 12, 24, and 48 hours after infection with Ebola, Andes, or Sendai viruses, or left uninfected (mock control). Error bars: SEM. F) Bulk RNA-seq of inflammatory cytokine gene expression in hPSC-derived artery and vein ECs 6, 12, 24, and 48 hours after infection with Sendai, Andes, or Ebola viruses. Fold change relative to uninfected cells is depicted. G) Summary of the present study. H) Bulk-RNA-seq of hPSC-derived artery ECs 6, 12, 24, and 48 hours after infection with Ebola, Andes, or Sendai viruses, or left uninfected (mock control). Error bars: SEM.