Endothelial cell cycle state determines propensity for arterial-venous fate

Abstract

During blood vessel development, endothelial cells become specified toward arterial or venous fates to generate a circulatory network that provides nutrients and oxygen to, and removes metabolic waste from, all tissues. Arterial-venous specification occurs in conjunction with suppression of endothelial cell cycle progression; however, the mechanistic role of cell cycle state is unknown. Herein, using Cdh5-CreER^T2;R26FUCCI2aR reporter mice, we find that venous endothelial cells are enriched for the FUCCI-Negative state (early G1) and BMP signaling, while arterial endothelial cells are enriched for the FUCCI-Red state (late G1) and TGF-β signaling. Furthermore, early G1 state is essential for BMP4-induced venous gene expression, whereas late G1 state is essential for TGF-β1-induced arterial gene expression. Pharmacologically induced cell cycle arrest prevents arterial-venous specification defects in mice with endothelial hyperproliferation. Collectively, our results show that distinct endothelial cell cycle states provide distinct windows of opportunity for the molecular induction of arterial vs. venous fate.

Fig. 1. Endothelial cell cycle state during retinal vascular development.

a, b Cdh5-CreER^T2;R26FUCCI2aR transgenic mouse allows for tamoxifen-induced endothelial cell-specific expression of the FUCCI reporter, which uses dynamic degradation of mVenus-hGem(1/110) and mCherry-hCdt1(30/120) in different cell cycle states to distinguish FUCCI-Negative, FUCCI-Red G1, and FUCCI-Green S/G2/M cell cycle states. c Representative confocal image of P6 retinas from Cdh5-CreER^T2;R26FUCCI2aR mice immunostained with IB4 and anti-Erg1/2/3 (scale bar = 150 μm), 1–4) magnified images of vein and artery branch sections (scale bar = 50 μm). d Representative confocal image of P15 retinas from Cdh5-Cre ER^T2;R26FUCCI2aR mice immunostained with IB4 and anti-Erg1/2/3 (scale bar = 200 μm), 5–8) magnified images of vein and artery branch sections (scale bar = 100 μm). e Quantification of cell cycle states in vascular regions of P6 retinal vasculature (mean +/− SD, # shows statistical significance between vein and artery, n = 4 retinas, 5 to 10 images per retina). f Representative confocal image of tip cells from Cdh5-Cre ER^T2;R26FUCCI2aR mice immunostained with IB4 and anti-Erk1/2/3 (scale bar = 25 μm). White arrows indicate FUCCI-Negative state and red arrows indicate FUCCI-Red G1 state. Quantification of cell cycle states in endothelial tip cells (n = 4 retinas, 20–25 tip cells per retina). g Cell cycle states quantified in veins and arteries of P15 retinal vasculature (mean +/− SD, # shows statistical significance between vein and artery, n = 4 retinas, 5 to 10 images per retina). Source data are provided as a Source data file. Statistical comparison of means by two-way ANOVA post hoc Tukey, p-values represented as */# < 0.05, **/## < 0.01, ***/### < 0.001, ****/#### < 0.0001.

Fig. 2. Gene expression analysis of developing retinal endothelial cells in FUCCI cell cycle states.

a All significantly variable genes within different cell cycle states from P6 R26FUCCI2aR retinal endothelial cells from RNA sequencing, each column of heatmap shows Z-score of a single gene with enrichment in each cell cycle state highlighted by highest expression. b GO term enrichment analysis of genes upregulated in FUCCI-Negative state, FUCCI-Red G1 state or FUCCI-Green S/G2/M states (statistical analysis reported from GO enrichment pipeline as p-value). c–f Expression of venous and arterial genes in retinal endothelial cells in different cell cycle states quantified at P6 and P15 time points by qPCR (mean +/− SEM, n = 3–7, statistical test one-way ANOVA post hoc Tukey for c and d and two-sided t-test for e and f), each sample isolated from separate litters and normalized to FUCCI-Negative expression value within the same sample. Source data are provided as a Source data file. Represented p-values as * <  0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Fig. 3. Single-cell RNA sequencing analysis of developing retinal endothelial cells.

a UMAP dimensionality reduction plot and clustering with labeled endothelial cell populations. b Dot plot showing expression of genes related to venous, capillary, arterial, flow (Klf2), migration (Cdc42, Rac1, Rhoa), tip, and cell cycle regulator functions within clusters. c Module scoring within clusters of genes related to TGF Signaling and BMP Signaling (box = 25th−75th percentile, center = median, whiskers = 10th−90th percentile, statistical test one-way ANOVA post hoc Tukey, cells in each cluster left-to-right: n = 238, 312, 516, 343, 60. 955, 283, 486, 89). d, Linear regression analysis of each cell comparing Venous vs. Arterial Score to TGF Signaling Score or BMP Signaling Score (statistical test of simple linear regression). e, f (left panels) Representative confocal images of P6 retinal vein and artery immunostained for SMAD1 or SMAD2/3. Hoechst labels cell nuclei, outlined by dashed red lines. (right panels) Quantification of nuclear localization of SMAD1 or SMAD2/3 (mean +/− SEM, nuclear localization calculated as nuclear fluorescent intensity normalized to total intensity in the vessel × 100%, n = 3 retinas, 5–10 images per retina, statistical test two-sided t-test). Source data are provided as a Source data file. Represented p-values as * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Fig. 4. Differentiation trajectory analysis using PHATE with cell cycle scoring.

a PHATE dimensionality reduction plot with clusters of endothelial cell populations. b Linear regression of relative G1 score versus Arterial, Venous, TGF Signaling, and BMP Signaling Scores in each cell. c Cell cycle score of FUCCI-Negative, FUCCI-Red G1 and FUCCI-Green S/G2/M within clusters determined by genes upregulated in P6 R26FUCCI2aR retinal endothelial cell bulk RNA sequencing datasets (box = 25th−75th percentile, center = median, whiskers = 10th−90th percentile, statistical test one-way ANOVA post hoc Tukey). d Cell cycle score FUCCI-Negative, FUCCI-Red G1 and FUCCI-Green S/G2/M of individual cells in PHATE dimensionality reduction plot. Source data are provided as a Source data file. Represented p-values as * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Fig. 5. Endothelial cell cycle-dependent regulation of TGF-β/BMP pathway.

a, b HUVEC-FUCCI reporter distinguishes S/G2/M, early G1 and late G1 cell cycle states, visualized by fluorescent imaging and FACS. c Differential expression of selected proteins within HUVEC-FUCCI cell cycle states by mass spectrometry analysis. In bulk RNA sequencing of early G1 and late G1 HUVEC-FUCCI (n = 3), d top 1000 significantly varying genes, e Signaling Pathway GO Term analysis (statistical analysis reported from GAGE pipeline as p-value), and f Volcano plot of fold-change against the log10(q-value) (TGF-β/BMP signaling pathway genes highlighted, upregulated in orange and downregulated in blue). g Western blot analysis of TGFBR1, pSMAD3 and pSMAD1/5/9 expression in early G1 and late G1 HUVEC-FUCCI. Each blot represents an independent experiment. h Quantification of TGFBR1, pSMAD3, and pSMAD1/5/9 in early G1 and late G1 HUVEC-FUCCI normalized to ACTIN, SMAD3, or SMAD5 respectively (mean +/− SD, n = 7, statistical test two-sided t-test). Source data are provided as a Source data file. Represented p-values as * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Fig. 6. Endothelial cell cycle-dependent arterial–venous specification via TGF-β/BMP signaling.

a Schematic of the TGF-β/BMP signaling pathway. b, c Representative Western Blot and quantification (mean +/− SD, n = 3) for SMAD proteins of lysates from SMAD4 co-immunoprecipitation after TGF-β1/BMP4-treated early G1 and late G1 HUVEC-FUCCI. d qRT-PCR analysis of DNA regions near EFNB2 and EPHB4 binding to SMAD4 complexes by chromatin-immunoprecipitation (mean +/− SD, n = 4). e TGF-β1/BMP4 induction of arterial and venous genes in early G1 and late G1 HUVEC-FUCCI (mean +/− SD, n = 4–6). f TGF-β1/BMP4 induction of EFNB2 and EPHB4 after NEG, SMAD2/3, or SMAD1/5 siRNA knockdown in early G1 and late G1 (mean +/− SD, n = 4). Source data are provided as a Source data file. Statistical comparison of means by two-way ANOVA post hoc Tukey, represented p-values as * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Fig. 7. Rescue of arterial–venous development defects with pharmacological CDK4/6 inhibition.

a CDK4/6i treatment timeline. b Representative confocal images of P6 retinas from WT, Cx37-KO, and Cx37-KO + CDK4/6i treated immunostained with IB4 and anti-αSMA (scale bars = 200 μm), quantified for: c vascular density (mean +/− SD), d αSMA vascular coverage (mean +/− SD). e P6 retinas from WT, Cx37-KO, and Cx37-KO + CDK4/6i treated mice immunostained with anti-Nrp2 and IB4 (scale bar = 150 μm). f Quantification of Nrp2 intensity in retinal veins and arteries normalized to IB4 intensity (mean +/− SD). g P6 retinas from WT, Cx37-KO, and Cx37-KO + CDK4/6i treated mice immunostained with anti-Sox17 and IB4 (scale bar = 150 μm). h Quantification of Sox17+ cells in retinal arterial branches (mean +/− SD). i, j (left panels) Representative confocal images of P6 retina from R26p-FUCCI2, R26p-FUCCI2 + Cx37-KO, and R26p-FUCCI2 + Cx37-KO + CDK4/6i treatment immunostained with IB4 and Hoechst (scale bars = 50 μm). Vessels outlined in dotted white lines, cell cycle state highlighted with colored stars. (right panels) Quantifications of cell cycle states in plexi above venous blood vessels and arterial blood vessels (mean +/− SEM). Source data are provided as a Source Data file. Statistical comparison of means by one-way ANOVA post hoc Tukey (c, d, f, h) or two-way ANOVA post hoc Tukey (i, j), represented p-values as * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Fig. 8. Cell cycle state determines propensity for arterial–venous specification hypothesis.

Early G1 state allows for greater induction of venous specification by BMP signaling, while late G1 state allows for greater induction of arterial specification by TGF-β signaling.