henhaobio
Sign in

DevelopmentDevelopmental BiologyPeer-reviewed journal article

Human vein-to-artery endothelial cell fate transition is driven by VEGF/ERK activation and PI3K inhibition

Zhainib A Amir-UgokweiD, A. L. Pyke, Emily TrimmiD, Meenakshi ChakrabortyiD, Xiaochen FaniD, Lay Teng AngiD, Kyle M. LohiD, Kristy Red‐HorseiD

PDFDOIPreprint
In-depth full-text digest available · Jump to digest ↓

Abstract

Artery endothelial cells (ECs) arise through different pathways, including differentiation from mesodermal cells (vasculogenesis) or from already established vein or capillary plexus ECs (angiogenesis), the latter being most common during embryonic development and regeneration. Understanding the vein-to-artery (v2a) transition could improve revascularization therapies, but progress is limited by a lack of human models. Here, we develop a human pluripotent stem cell (hPSC) differentiation protocol that models the v2a EC conversion. Comparing v2a and mesoderm-to-artery (m2a) transcriptomes with publicly available single cell RNA sequencing (scRNA-seq) data from human embryos showed they reflected angiogenesis- and vasculogenesis-derived artery ECs, respectively. This reductionist system revealed that VEGF activation alongside PI3K inhibition was sufficient for vein ECs to acquire arterial identity within 48 hours. We model a critical step in vascular development and define the minimal signals required for artery differentiation from veins, providing a framework to promote this conversion in revascularization or therapeutic contexts.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (20 May 2026) · CC BY 4.0

A stem cell-derived human vein-to-artery endothelial conversion model mirrors angiogenesis-derived embryonic arteries; VEGF plus PI3K inhibition alone yields 92% DLL4+CXCR4+ endothelium by 48 hours.

Most organs are vascularized by angiogenesis, in which established vein endothelial cells can switch to artery identity, yet most human pluripotent stem cell protocols model vasculogenesis (mesoderm-to-endothelium), leaving no tractable human system for this switch. This work adds a defined vein-to-artery step and pinpoints its minimal signals, relevant to revascularization approaches where VEGF-based therapy has given limited durable benefit.

Key results

  1. ShownVein endothelial cells made from human embryonic stem cells switch on arterial markers and shut off venous genes within 48 hours in the six-factor mesoderm-to-artery (m2a) medium, with SOX17 induction preceding DLL4.

    Flow cytometry: ~50% DLL4+ and ~100% CXCR4+ at 24 h, ~100% DLL4+CXCR4+ at 48 h. qPCR: SOX17 at m2a levels by 24 h, DLL4 reaching m2a levels at 48 h, NR2F2 and APLNR nearly off by 24 h. Bulk RNA-seq: 1345 and 2590 differentially expressed genes vs vein ECs at 24 and 48 h (adjusted p < 0.05, |log2FC| > 2). Similar flow results in hiPSC-derived cells. · An in vitro model of human vein-to-artery…; Vein-to-artery endothelial cells lose vein…

  2. ShownVEGF plus the PI3K inhibitor GDC-0941 in base medium is enough to arterialize vein endothelial cells to the population frequency reached with full m2a medium, while VEGF alone gives only partial conversion.

    CD144+DLL4+CXCR4+ fraction: VEGF alone 10% at 24 h and 50% at 48 h vs 55%-90% in m2a medium; VEGF + GDC-0941 reached 92% at 48 h vs 90% with m2a. At 48 h, DLL4 protein was 1.3-fold higher and CXCR4 1.7-fold lower than in m2a-treated v2a cells; vein maintenance medium gave minimal marker expression; iPSC results similar. · VEGF activation and PI3K inhibition are…; Vein or primed endothelial cell arterialization

  3. ShownWithdrawing single m2a components shows VEGF is needed for endothelial survival and PI3K inhibition for arterial marker induction, whereas WNT and BMP inhibition, required for mesoderm-to-artery differentiation, are dispensable for vein-to-artery conversion.

    At 24 h: without VEGF only ~10% of cells were CD144+ and almost none DLL4+; without GDC-0941, DLL4+ ECs fell from ~60% to ~15%, DLL4 intensity halved and DLL4 transcripts dropped by about half by qPCR; removing the TGFβ agonist had a milder effect; omitting XAV939 (WNT inhibitor) or DMH1 (BMP inhibitor) caused no significant change. · Defining the extracellular signals that drive…

  4. SuggestedGene set variation analysis (GSVA) against human embryonic scRNA-seq signatures from Hou et al. aligns m2a cells with vasculogenesis-derived arteries and v2a cells with angiogenesis-derived arteries; some arterial genes are upregulated in v2a but not in m2a cells.

    Vasculogenic artery (AEC1) score: m2a mean z-score 1.4; v2a 24 h and 48 h -0.8 and 0.4. Angiogenic artery (AEC2) score: v2a 0.6 (24 h) and 1.0 (48 h) vs m2a -0.1. Only vein ECs scored high for the venous signature. ESM1, IGFBP3 and JAG2 were elevated in v2a; TOX2, CXCL12, JAG2, VEGFC and EPAS1 were induced in v2a but not m2a. · Vein-to-artery endothelial cells lose vein…; Artery and Vein GSVA Scoring

  5. ShownPrimed endothelial cells, the intermediate stage of vein differentiation, also arterialize in m2a medium (p2a). They induce DLL4 faster than vein cells and score closer to angiogenesis-derived than to vasculogenesis-derived arteries.

    Nearly all CD144+ primed cells were DLL4+ at 24 h, with DLL4 levels above m2a and v2a cells; HEY1, DLL4 and CXCR4 rose and NR2F2, NT5E and NRP2 fell at 48 h vs untreated primed cells. GSVA at 48 h, p2a vs v2a: vasculogenic -0.5 vs -0.2, angiogenic 0.6 vs 0.8, venous -0.5 vs -0.7. · Primed endothelial cells also arterialize…

Methods, data and reproducibility

H1 embryonic and WTC11 induced pluripotent stem cells were taken through lateral mesoderm and primed endothelium to vein ECs, then given m2a medium, single-factor dropouts, or VEGF with or without GDC-0941 for 24 or 48 h; arterial identity was scored by flow cytometry, qPCR, CD34+ bulk RNA-seq and GSVA against embryo scRNA-seq signatures.

  • H1 human embryonic stem cells (WiCell, male) differentiated into artery, primed, vein, v2a and p2a endothelial cells · human · n = 2-5 independent experiments (Fig. 1)
  • WTC11 CDH5-GFP reporter induced pluripotent stem cells (Allen Institute for Cell Science/Coriell, male) used to replicate the arterialization assays · human
  • HUVECs (Lonza) used as a primary endothelial comparator; they co-expressed arterial DLL4 and venous CD73 · human
  • Bulk RNA-seq of CD34+ MACS-enriched stem cell-derived ECs: artery (m2a), vein, primed, v2a and p2a at 24 and 48 h · human · GEO GSE312764
  • Hou et al. embryonic endothelial scRNA-seq clusters (VeEC, AEC1 vasculogenic, AEC2 angiogenic) aligned to lineage-traced mouse trajectories; used as GSVA gene sets · human (integrated with mouse trajectories)
Methods & tools
Flow cytometry (Sony MA900; CD144, DLL4, CXCR4, CD73 antibodies), FlowJo v10.10.0, qPCR on QuantStudio 7 Flex (ΔΔCT; YWHAZ and CD144 normalization), CD34 microbead MACS enrichment (Miltenyi), FastQC, Trimmomatic, HTSeq, DESeq2, GSVA (Bioconductor, R), GRCh38 (hg38) reference with GENCODE v44 annotation, GraphPad Prism v10, dplyr, tidyverse, ggplot2; Fiji/ImageJ
Reagents & environment
FlowJo v10.10.0; GraphPad Prism v10; reads aligned to GRCh38 (hg38) with GENCODE v44 primary assembly annotation; GSVA Bioconductor package in R; versions of DESeq2, HTSeq, GSVA and R not stated.
Reproducibility gaps
Bulk RNA-seq data are in GEO (GSE312764; no URL given), but no analysis code or repository is shared. The Table S1 count matrix is filtered by fold-change and adjusted p-value thresholds, and RNA-seq package versions are not stated.
Limitations (6)
  • The system models only the fate-conversion step of angiogenesis (not sprouting) and maps to early, pre-organ-specific vasculature, so it does not fully capture in vivo arterialization or tissue-specific specialization. · stated by the authors
  • Shear stress, a known regulator of arterial fate, was not applied; the authors propose adding it with microfluidic devices. · stated by the authors
  • Only bulk RNA-seq profiled the v2a, p2a and m2a populations, so cell-to-cell heterogeneity was not resolved; single-cell or proteomic profiling is proposed. · stated by the authors
  • How VEGF activation and PI3K inhibition install arterial identity was not tested mechanistically, and the v2a-enriched genes ESM1, JAG2 and IGFBP3 were not genetically perturbed. · stated by the authors
  • The minimal VEGF plus GDC-0941 condition was read out only by flow cytometry for CD144, DLL4 and CXCR4; no transcriptome was reported, so whether it reproduces the angiogenesis-like program of m2a-treated v2a cells is untested. · noted in this summary
  • The GSVA z-scores quoted for v2a at 48 h differ between analyses (angiogenic 1.0 vs 0.8, vasculogenic 0.4 vs -0.2 across the v2a and p2a comparisons), possibly from rescaling across sample sets. Fig. 1 data rest on 2-5 independent experiments with unpaired t-tests. · noted in this summary

Read this if you study arteriovenous specification, need a defined human stem cell model of vein-to-artery endothelial conversion, or design revascularization strategies targeting VEGF and PI3K signaling.

Citations
1
FWCI
2.05
Field percentile
top 12%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISAng '25Weijts '18Yoshimoto '25Wang '20Chavkin '22Jerafi-Vide '21
83% semantic matchDevelopmental Biology
Lay Teng Ang, Sherry Li Zheng, Kevin J. Liu+27 more·bioRxiv·Oct 11, 2025· 6 citations