Main
VEGF is frequently targeted in clinical settings to suppress tumour angiogenesis, metastasis and retina disease6. However, resistance to VEGF-blocking antibodies arises over time, which limits the effectiveness of such treatments4,5,7,8. By contrast, resistance to loss of VEGF signalling has not been reported during physiological vascular development.
Embryos or postnatal tissues lacking activity of the VEGF signalling pathway in all cells fail to develop vessels3. VEGFR2 (encoded by the Kdr (also known as Vegfr2) gene in mice) is the most important VEGF receptor and has been shown to be essential for angiogenesis in all vascular development contexts studied so far1,9. Germline mutants for this gene completely lack blood vessels and die very early during embryonic development2, and mouse embryos chimeras formed by wild-type and Vegfr2-knockout (Vegfr2KO) progenitor cells cannot form endothelial cells (ECs)10. Conditional deletion of Vegfr2 with Tie2–Cre after endothelial differentiation11, or in postnatal tissues with Cdh5–CreERT2, also strongly impairs angiogenesis12, without any evidence of genetic compensation or resistance.
Despite a large body of literature showing that VEGFR2 is essential for endothelial differentiation, proliferation, migration and survival, previous studies involving deletion of Vegfr2 in all ECs (Vegfr2iDEC) during embryonic or postnatal organ development could only analyse its effect on angiogenesis for a relatively short period of time. Its essential role in vascular development and hence in tissue development and organ function causes severe impairment of angiogenesis and premature lethality of Vegfr2iDEC embryos and postnatal animals; thus, such studies have been unable to determine whether ECs can adapt to loss of VEGFR2 over time or make certain organ vascular beds in the complete absence of VEGF–VEGFR2 signalling.
Here we used iFlpMosaics13 and iSuRe–HadCre14 to mosaically induce and longitudinally track Vegfr2KO ECs, enabling us to assess the role of this gene during the entire embryonic and postnatal vascular development of different organs. We found that despite its essential role during the initial wave of embryonic and postnatal angiogenesis, VEGF–VEGFR2 signalling is dispensable for many subsequent angiogenic processes originating from veins. Vegfr2KO ECs expand and out-compete wild-type ECs in venous vessels, which serve as a continuous source of these mutant cells for angiogenesis in most organs. Single-cell RNA sequencing (scRNA-seq) showed that Vegfr2KO ECs adapt over time through upregulation of ligands for other VEGF receptors and through increased expression of venous-specific genes, particularly Nr2f2. Targeting these adaptive and resistance-associated molecular mechanisms prevents expansion of ECs following prolonged loss of VEGFR2 signalling.
Vegfr2 KO ECs can form new vessels
Given the essential role of VEGFR2 in early angiogenesis and embryonic development, we induced mosaic loss of VEGFR2 with iFlpMosaics13 to investigate whether mutant cells could adapt and form new vessels over time. This technology enables FlpO-dependent ratiometric induction and tracking over long periods of time of cells expressing MTomato-2A-Cre (knockout for any floxed gene) or MYFP (wild type). To induce these mosaics, we used the Apln-FlpO allele, which is expressed in the first angiogenic and sprouting ECs at embryonic day 8.5 (E8.5) and throughout vascular development. In female embryos, the X-chromosome-located Apln–FlpO allele is mosaically silenced15 and expressed only in a fraction (23–50%) of ECs, generating relatively few mutant and wild-type cells, which we found to be compatible with embryonic development (Fig. 1a and Extended Data Fig. 1a). Analysis of major organs of Apln-FlpO iFlpMosaic mice showed that ECs with full loss of VEGFR2 (Vegfr2KO-E8.5–P7) were able to survive and proliferate during embryonic development and until postnatal day 7 (P7), with only a relatively minor competitive disadvantage (2–7-fold) compared with wild-type cells (Fig. 1b). This competitive disadvantage was more pronounced in the lung and kidney endothelium and less so in brain ECs (Fig. 1b,c and Extended Data Fig. 1a–c). As Apln-FlpO is constitutively active and can drive recombination at later stages of embryonic development, we also induced Vegfr2 deletion using 4-hydroxytamoxifen (4-OHT) at E10.5 and E12.5 in embryos carrying Cdh5-CreERT2, iSuRe-HadCre and an iMb2-Mosaic16 control reporter that recombines independently of Vegfr2 deletion (Extended Data Fig. 2a–c). This resulted in a maximum of 1–6% recombined ECs, as opposed to 23–50% with Apln-FlpO. However, this recapitulated the trend in which brain ECs were least affected by VEGFR2 loss, followed by those in the heart, liver, lung and kidney (Extended Data Fig. 2b,c).
a, Schematic of iFlpMosaics induced by Apln-FlpO from E8.5 to P7. b, log2 ratio of MTomato+/MYFP+ within CD31+CD45− cells detected by FACS for the indicated organs. The absolute fold change of the ratio of Vegfr2KO ECs versus the ratio of control Tomato+ cells is indicated in red. c, Representative gating scheme for the ratiometric analysis (see also Extended Data Fig. 1a). d, Representative confocal micrographs of control and Vegfr2KO-E8.5–P7 retinas immunostained for Tomato, YFP and ERG (which labels EC nuclei). Heatmap, XY dot plot and charts showing the zonal distribution and percentage of Tomato+Cre+ cells when equal numbers of randomly selected YFP+ and Tomato+ cells were plotted (see method in Extended Data Fig. 3a). Right, log2 ratio of Tomato+/YFP+ cells detected in whole-mount retina immunostainings, showing an overall competitive disadvantage of Vegfr2KO cells in the mapped retinas. e, Representative confocal micrographs of retinas immunostained for ERG at P7, from pups injected with 4-OHT at P1. Heatmap, XY dot plot and charts showing distribution of MTomato+ERG+ (wild-type, Vegfr2HET or Vegfr2KO) cells in space in relation to MTomato−ERG+ (wild-type) cells. f, Bar plot contrasting the relative frequency of the indicated cells in the indicated areas and genotypes. g, log2 fold change of Vegfr2KO-E8.5–P7 and Vegfr2IECKO-P1–P7 cells relative to the respective control wild-type cells. Numbers indicate fold change (×) between the two groups. Data are presented as the mean ± s.d. Scale bars, 100 μm. A, arteries; AF, angiogenic front; dist., distance; freq., frequency; IntCap, intermediate capillaries; LE, leading edge; MA, mature area; V, veins; WT, wild type.
Source data
Vegfr2 KO ECs can form arteries
VEGFR2 activity is essential for NOTCH activation and the development of arteries17,18,19. Arteries are formed by tip ECs20, which have the highest level of VEGF–VEGFR2 signalling and guide vascular growth. Our high-resolution spatial zonation data with long-term iFlpMosaics showed that Vegfr2KO-E8.5–P7 ECs were significantly enriched in veins and surrounding capillaries, and avoided angiogenic front capillaries or tip ECs, but were able to form arteries (Fig. 1d; see also Extended Data Fig. 3a for the heatmapping method).
To complement these results, we intercrossed the Vegfr2 floxed allele with Cdh5-CreERT2 and iSuRe-HadCre14 alleles to ensure fast (24 h; Extended Data Fig. 3b,c) and effective mosaic genetic deletion and full loss of VEGFR2 protein and downstream targets in ECs (Cdh5 expressing), after a single injection of 4-OHT. Vegfr2IECKO-P1–P7 ECs were located mainly in veins and surrounding capillaries, as well as in the more proximal and mature segments of retina arteries (Fig. 1e). This indicates that when Vegfr2KO ECs are induced in the initial retina sprouting plexus at P1, they can make arteries forming from P1 to P3. However, owing to their inability to continue to sprout and form tip cells at the leading edge, they cannot participate in the subsequent arterial development process occurring from P3 to P7, which in the retina is dependent on incorporation of new tip cells21.
Notably, compared with Vegfr2IECKO-P1−P7 ECs, Vegfr2KO-E8.5−P7 ECs were more frequent at the leading edge and angiogenic front capillaries (Fig. 1f,g) and were still able to make arteries. This suggests that embryonic Vegfr2KO-E8.5−P7 ECs can adapt to long-term loss of VEGFR2 and contribute to venous vascular growth, tip cells and arteries at postnatal stages.
Vegfr2 KO ECs expand in veins
We next investigated how Vegfr2KO ECs could compete with wild-type cells and contribute to growing vascular networks over long periods of time, given the proposed important role of this pathway in EC proliferation1,22. During angiogenesis, there are two main locations for EC proliferation: the capillary angiogenic front and veins (Fig. 2a). We found that the surviving Vegfr2KO-E8.5−P7 ECs proliferated (as determined by Ki-67 detection) as well as wild-type cells in postnatal brain, heart and lung vessels (Fig. 2b).
a, Representative confocal micrograph of a control P7 retina immunostained for the proliferation marker Ki-67 and the EC nuclear marker ERG. The chart shows the frequency of Ki-67+ERG+ (pink pseudocolored) nuclei in different vascular zones. b, Representative confocal micrographs of the indicated organs of Vegfr2KO-E8.5–P7 mice at P7, immunostained for YFP, Tomato, ERG and Ki-67. Yellow arrows indicate YFP+Ki-67+ cells and magenta arrows Tomato+Ki-67+ ECs. c, Representative confocal micrographs of control and Vegfr2IECKO-P1–P20 retinas induced at P1 and analysed at P20. Heatmap, XY dot plot and chart showing that Vegfr2IECKO-P1–P20 ECs grow in veins and surrounding capillaries. d,e, Representative confocal micrographs of retinas immunostained for cleaved caspase 3 and respective quantification charts, showing increased apoptosis of Vegfr2IECKO-P4–P7 but not Vegfr2IECKO-P1–P7 ECs. f, Representative confocal micrographs of the indicated organs of control and Vegfr2IECKO-P4–P7 mice induced at P4 and analysed at P7. Arrows indicate recombined ECs (Tomato+) that are Ki-67+ (yellow) and Ki-67− (white), respectively. Chart shows the percentage of Tomato+ ECs positive for Ki-67. Data are presented as mean ± s.d. Scale bars, 100 μm (a,c–e), 50 μm (b,f). Cap., capillaries; CC3, cleaved caspase 3; Ctrl, control; DCap, distal capillaries; IntCap, intermediate capillaries; PCap, proximal capillaries. *Adjusted P < 0.05.
Source data
Venous ECs provide a continuous source of ECs for angiogenesis and these migrate against the blood flow to capillaries and then towards arterial vessels20,23,24. When we induced Vegfr2KO ECs for longer periods of time, from P1 to P20 (Vegfr2IECKO-P1–P20), we observed that mutant cells stayed in the vein stem, where they proliferated and gradually outcompeted wild-type cells, which moved outwards towards the VEGF-rich leading edge and adjacent capillaries (Fig. 2c). This preferential occupation of the venous stem niche by Vegfr2KO ECs allowed them to continue proliferating for longer periods of time and later to form most of the perivenous capillary vessels and the deeper vascular plexus that developed during a second wave of VEGF-independent angiogenesis from P7 to P20 (Extended Data Fig. 4a,b). This was consistent with previous findings showing that the tip cells that dive to form the deeper vascular plexus in the retina lack expression of the VEGFR2 signalling target gene Esm1; instead, their migration is dependent on other pathways such as Alk or TGFβ and WNT signalling25,26,27.
Together, these data show that VEGFR2 function is essential for the initial sprouting angiogenesis driven by high VEGF levels in hypoxic embryos and organs but not for the development of veins and perivenous capillaries or the subsequent steps of capillary angiogenesis and remodelling that occur after the first vascular plexus is formed in various organs. Thus, by forming veins, which are known to provide a continuous source of ECs for angiogenesis23, Vegfr2KO ECs can continue to proliferate and participate in subsequent steps of vascular development for long periods of time.
Acute effects and adaptation to VEGFR2 loss
We previously observed that the distribution of Vegfr2KO ECs in the retina plexus at P7 differed between animals with long-term (Vegfr2KO-E8.5–P7) and short-term (Vegfr2IECKO-P1–P7) mosaic deletion of Vegfr2. Vegfr2KO-E8.5–P7 ECs were more frequent at the leading edge and angiogenic front than Vegfr2IECKO-P1–P7 ECs (Fig. 1f,g). Vegfr2KO-E8.5–P7 ECs also proliferated normally (Fig. 2b). Therefore, we hypothesized that Vegfr2KO-E8.5–P7 ECs might be genetically rewired and adapt over time to loss of the most important receptor for VEGF.
To assess the early response of ECs to loss of VEGFR2, we first analysed apoptosis markers in growing retina vessels, as VEGFR2 has been shown to regulate EC survival. Vegfr2IECKO-P4–P7 mice showed a very significant increase in the number of apoptotic (cleaved caspase 3 positive) ECs (Fig. 2d). This apoptosis was a reaction to the acute loss of VEGFR2, as Vegfr2KO-E8.5–P7 or even Vegfr2IECKO-P1–P7 mice did not show any apoptotic ECs (Fig. 2e), suggesting adaptation of the surviving cells over time. When the acute loss of Vegfr2 was induced in all ECs, we observed an immediate depletion of Vegfr2KO ECs from the VEGF-high angiogenic front, but not from veins or more mature vessels (Extended Data Fig. 4c). Therefore, even though many ECs in the retina plexus showed upregulation of the apoptosis marker in the first days after Vegfr2 loss (Fig. 2d), only the angiogenic ECs that were exposed to high VEGF levels at the time of Vegfr2 deletion died (Extended Data Fig. 4c).
We also found that Vegfr2IECKO-P4–P7 ECs were significantly less proliferative than control cells only in the heart and kidney (Fig. 2f), whereas the proliferation of Vegfr2KO-E8.5–P7 ECs was similar to that of controls (Fig. 2b). This suggests that Vegfr2KO ECs adapt over time and are later able to proliferate in these organs. In the newborn retina, deletion of Vegfr2 did not affect the proliferation of surviving ECs in any segment (Extended Data Fig. 4d), indicating that in this organ VEGFR2 mainly controls EC migration and survival, not proliferation.
Together, our data suggest that ECs lacking VEGF signalling during active angiogenesis suffer an acute VEGFR2 withdrawal effect, followed by a period of adaptation to the loss, after which they can form new vessels. ECs that are more distant from the VEGF source, and therefore have less VEGF signalling, adapt better to the loss and can expand over time, particularly in veins and perivenous vessels.
To further confirm the insensitivity of more mature ECs to loss of VEGFR2 signalling, we deleted Vegfr2 in several adult quiescent vascular beds for 2 weeks. Only the quiescent sinusoidal liver endothelium showed a strong sensitivity and loss of Vegfr2KO ECs, whereas only the kidney endothelium showed significantly lower proliferation (Extended Data Fig. 5), suggesting that only fenestrated or discontinuous mature endothelial vessels present in these organs remain sensitive to VEGFR2 loss.
Acute organotypic transcriptional changes
Given the observed differences in the biology of Vegfr2IECKO-P4–P7 and Vegfr2KO-E8.5–P7 ECs, we performed scRNA-seq of these cells and respective control Tomato+ cells to identify possible differences in their transcriptional profiles and potential mechanisms of adaptation to loss of VEGFR2 signalling in distinct organ vascular beds.
Although VEGFR2 is a central regulator of angiogenesis and widely regarded as essential for EC biology, Vegfr2IECKO-P4–P7 ECs showed only modest changes in their transcriptome (Fig. 3 and Extended Data Fig. 6), even though VEGFR2 protein and downstream signalling were lost already within the first 24 h following 4-OHT administration (Extended Data Fig. 3b,c). In general, the expression of genes related to sprouting or tip cell biology was decreased, whereas that of genes related to apoptosis and vascular permeability was increased in all organs analysed (Fig. 3a,b). However, considerable differences were observed among organs.
a,b, Heatmaps displaying the average logarithmic fold change (log2FC) of all (a) or top (b) DEGs in mutant ECs for the indicated filter conditions (see Extended Data Fig. 6a for the single-cell method). c,j, Violin plot and dot plots showing the expression levels of the last exon (exon 30) and the deleted floxed exon (exon 3), demonstrating Vegfr2 deletion efficiency, in brain (c) and heart (j). Negative numbers in red indicate the fold change between conditions. d,k, Uniform manifold approximation and projection (UMAP) showing the identified EC clusters and histobar plots showing their proportions per condition in brain (d) and heart (k). e,l, Dot plots showing the top cluster marker genes in brain (e) and heart (l). f,m, UMAP spatial distributions of control and Vegfr2IECKO-P4–P7 ECs in brain (f) and heart (m). Bar plots show log2 ratios of Vegfr2IECKO-P4–P7 over control cells per EC cluster. g,n, Volcano plots depicting the top DEGs and other selected and biologically significant genes when comparing control and Vegfr2IECKO-P4–P7 ECs in brain (g) and heart (n). h,o, Top significantly upregulated and downregulated hallmark pathways in Vegfr2IECKO-P4–P7 ECs compared with control ECs in brain (h) and heart (o). Highlighted hallmarks match the colour code used in g and n. i,p, Heatmaps showing the average log2FC of apoptosis- and tip cell-related genes in Vegfr2IECKO-P4-P7 ECs compared with control ECs in brain (i) and heart (p). For volcano plots and heatmaps (a,b,g,i,n,p), statistical analyses used Seurat’s FindMarkers function, with Wilcoxon rank-sum test and Bonferroni correction for multiple testing. For hallmark pathways (h,o), statistical analyses used the multilevel Monte Carlo approach implemented in fgseaMultilevel (R package fgsea) and Benjamini–Hochberg false discovery rate correction for multiple testing. *Adjusted P < 0.05.
In the brain, clusters of proliferating (S–G2–M) ECs expressing venous capillaries markers were increased after Vegfr2 deletion for 3 days (Fig. 3c–f). This was consistent with higher clonal expansion of Vegfr2KO ECs from P1 to P20 in this organ (Extended Data Fig. 7a). Clusters comprising capillary and activated or angiogenic ECs (with high expression of Apln, Adm, Mcam, Lamb1 and Angpt2) were decreased (Fig. 3d–f). In total, only 569 genes were significantly deregulated (Fig. 3g). Gene set enrichment analysis (GSEA) revealed an increase in the expression of E2F and MYC target genes, as well as G2M, DNA repair and mitotic spindle genes, confirming the increase in EC proliferation (Fig. 3h and Extended Data Fig. 7a). The expression of most tip cell marker genes was decreased in Vegfr2IECKO-P4–P7 brain ECs (Fig. 3i and Extended Data Fig. 6a–f).
In contrast to those in the brain, Vegfr2IECKO-P4–P7 capillary ECs in the heart were significantly less proliferative (Fig. 3j–m and Extended Data Fig. 7b), matching the 54% decrease in Ki-67 percentage observed in sections (Fig. 2f). The proportion of mutant ECs found in bigger vessels (arteries and veins) was increased 2- to 4-fold (Fig. 3k–m), suggesting a proportional fast depletion of Vegfr2KO capillary ECs in only 3 days. Unlike in the brain, GSEA of Vegfr2KO heart ECs showed the expected decrease in expression of E2F and MYC target genes, as well as decreases in expression of G2M, mitotic spindle and tip cell genes (Fig. 3n–p), in accordance with the decrease in proliferation and sprouting of Vegfr2IECKO-P4–P7 heart ECs. Vegfr2KO heart ECs also showed upregulation of genes related to apoptosis and epithelial-to-mesenchymal transition (Fig. 3n–p), as has been shown previously for ECs losing ERK signalling28.
In contrast to the results for brain and heart, we found no significant difference in the proportions of proliferating cells among control and Vegfr2IECKO-P4–P7 ECs in lung (Extended Data Fig. 6g–j), as shown by immunostaining for Ki-67 and clonal analysis (Fig. 2f and Extended Data Fig. 7c). The proportion of the vein cluster was only slightly increased, and the capillary cluster was slightly decreased. Despite the lack of significant changes in cluster proportions, GSEA showed increased TNF signalling through NF-κB, hypoxia and reactive oxygen species signalling and a decrease in the interferon response (Extended Data Fig. 6k,l). In lung, there was only a small trend of decreased expression of tip cell marker genes (Extended Data Fig. 6m), and several genes related to apoptosis were upregulated (Extended Data Fig. 6l,m).
In the kidney, similar to the heart, Vegfr2 deletion for 3 days led to an increase in the relative proportion of ECs of bigger vessels (arteries and veins), suggesting fast depletion of a fraction of capillary ECs in only 3 days (Extended Data Fig. 6n–q). Proliferation and clonal expansion of kidney capillary ECs were also reduced (Extended Data Figs. 6o,q and 7d), as shown by immunostaining for Ki-67 (Fig. 2f). GSEA showed the expected decrease in expression of E2F, G2M and mitotic spindle genes and an increase in the expression of genes related to p53, xenobiotic metabolism and apoptosis (Extended Data Fig. 6r,s), indicative of cellular stress in kidney ECs after short-term Vegfr2 deletion from P4 to P7. Similar to brain and heart ECs, the expression of tip cell marker genes was also decreased in kidney Vegfr2IECKO-P4–P7 ECs (Extended Data Fig. 6t).
Adaptation to VEGFR2 loss by venousization
Following our analysis of Vegfr2IECKO-P4–P7 ECs, we performed scRNA-seq of Vegfr2KO-E8.5–P7 ECs of distinct organs to identify potential differences in gene expression related to long-term adaptation of cells to loss of VEGFR2 (Extended Data Fig. 8a). In contrast to Vegfr2IECKO-P4–P7 ECs, Vegfr2KO-E8.5–P7 ECs exhibited predominantly upregulation rather than downregulation of genes, including marked upregulation of several venous-enriched genes (compare Figs. 3a,b and 4a).
a, Heatmaps displaying the average log2FC of all or top DEGs in mutant ECs given the indicated filter conditions. b,i, Violin plot and dot plots showing the expression levels of the last exon 30 and deleted floxed exon 3, demonstrating Vegfr2 deletion in heart (b) and lung (i) ECs. Note that in lung ECs compensatory alternative splicing and exon 30 mRNA reads can exist without a functional protein37. c,j, UMAP showing the identified EC clusters and histobar plots showing their proportions per condition in heart (c) and lung (j). d,k, UMAP spatial distribution of control and Vegfr2KO-E8.5–P7 ECs in heart (d) and lung (k). Bar plots show the log2 ratios of Vegfr2KO-E8.5–P7 compared with control cells per EC cluster. e,m, Volcano plots depicting the top DEGs and other biologically significant genes when comparing control and Vegfr2KO-E8.5–P7 ECs in heart (e) and lung (m). f,n, Top significantly deregulated hallmark pathways in Vegfr2KO-E8.5–P7 ECs compared with control ECs in heart and lung. Highlighted hallmarks match the colour code used in e and m. g, Heatmap showing the average log2FC of tip-cell-related genes for the indicated comparisons. h,l, Heatmaps showing the average log2FC of venous gene expression across EC clusters in heart (h) and lung (l). o,p, Confocal micrographs of flat mount retina (o) and heart (p) cryosections from P7 mice and quantification of signal intensity per nuclei. q, log2 ratios of Tomato+ and YFP+ ECs obtained by FACS analysis of CD31–APC+ ECs from Apln-FlpO iFlpMosaic mice at P7, with the floxed alleles indicated (see representative FACS plots in Extended Data Fig. 9e). r, Representative confocal images and related quantification data showing the relative proportions of Tomato+Cre+ and YFP+ cells in veins and other vessels. Scale bars, 25 μm (r,p) and 100 μm (o). For volcano plots and heatmaps (a,e,g,h,l,m), statistical analyses used Seurat’s FindMarkers function, with Wilcoxon rank-sum test and Bonferroni correction for multiple testing. For hallmark pathways (f,n), statistical analyses used the multilevel Monte Carlo approach implemented in fgseaMultilevel (R package fgsea) and Benjamini–Hochberg false discovery rate correction for multiple testing. Prol., proliferative. *Adjusted P < 0.05.
Source data
In contrast to brain Vegfr2IECKO-P4–P7 ECs, brain Vegfr2KO-E8.5–P7 ECs had increased frequencies of venous capillaries (Extended Data Fig. 8b–d) and marked upregulation of the key regulator of vein specification Nr2f2 (Extended Data Fig. 8e). Similarly to Vegfr2IECKO-P4–P7 ECs, brain Vegfr2KO-E8.5–P7 ECs had decreased frequencies of activated and sprouting ECs and lower expression levels of several tip cell genes (Extended Data Fig. 8c–e). The number of differentially expressed genes (DEGs), particularly upregulated genes, was much higher (1,621 versus 569 DEGs) in Vegfr2KO-E8.5–P7 ECs; this is likely to represent a long-term adaptation response to loss of VEGFR2 function (compare Extended Data Fig. 8e with Fig. 3g). Whereas in Vegfr2IECKO-P4–P7 brain ECs, genetic pathways related to EC proliferation and cellular activity (E2F targets, MYC, oxidative phosphorylation and glycolysis) were paradoxically increased, in Vegfr2KO-E8.5–P7 ECs they were decreased (compare Extended Data Fig. 8e,f with Fig. 3g,h). This suggests that brain Vegfr2KO-E8.5–P7 ECs adapt to the long-term loss of VEGFR2 signalling through a decrease in cellular activity and upregulation of Nr2f2, which probably decreases their dependence on VEGF signalling for growth and survival.
In the heart, we confirmed previous histological analysis showing that heart Vegfr2KO-E8.5–P7 ECs proliferated at levels similar to those of control cells and markedly higher than those of Vegfr2IECKO-P4–P7 ECs (compare Fig. 3k,m with Fig. 4b–d). They also had many more DEGs (1,750 DEGs versus 242 DEGs), particularly upregulated genes (compare Fig. 3n with Fig. 4e). GSEA showed increased epithelial–mesenchymal transition, KRAS signalling and angiogenesis, and less expression of metabolic-pathway-related genes (Fig. 4e,f), as seen in the brain. Expression of tip or activated EC genes was not as compromised in Vegfr2KO-E8.5–P7 ECs compared with Vegfr2IECKO-P4–P7 ECs (Fig. 4g), suggesting some adaptation over time to the loss of VEGFR2. Notably, Vegfr2KO-E8.5–P7 ECs were four times more frequent in veins (Fig. 4c,d), similar to our observations in the retina (Fig. 1d–f), confirming that mosaic loss of Vegfr2 causes mutant ECs to expand more in veins. In comparison with Vegfr2IECKO-P4–P7 ECs, Vegfr2KO-E8.5–P7 ECs showed strong upregulation of most venous-specific genes, including Nr2f2, Fbln2, Igf1 and Vwf (Fig. 4e). This was not caused only by the increased proportion of venous ECs in these mutants; capillary and angiogenic ECs also showed much higher expression of these venous marker genes, indicating venousization of mutant Vegfr2KO-E8.5–P7 capillary ECs over time (Fig. 4h,p).
In the lung, the same adaptation and venousization of Vegfr2KO-E8.5–P7 ECs was observed. Mutant cells were four times more frequent in veins (Fig. 4i–k), and even capillary or CAR4+ ECs showed significantly upregulated expression of lung venous genes including Vegfc, Nr2f2, Vwf and Fbln2 (Fig. 4l,m and Extended Data Fig. 9a,b). Similar to findings in the other organs, lung Vegfr2KO-E8.5–P7 ECs had significantly more DEGs than lung Vegfr2IECKO-P4–P7 ECs (3,116 vs 502), particularly upregulated genes (compare Extended Data Fig. 6k with Fig. 4m). Volcano plots and GSEA revealed upregulation of many venous-enriched and epithelial–mesenchymal transition genes, as well as oestrogen response genes. Lung Vegfr2KO-E8.5–P7 ECs had lower expression of E2F targets and metabolism genes, as seen in the other organs (Fig. 4m,n).
Notably, deletion of Vegfr2 for 2 weeks in adult organ blood vessels did not induce venousization of adult Vegfr2KO ECs; instead, it decreased expression of venous genes (Extended Data Fig. 8g–u), indicating that upregulation of Nr2f2 and venousization may represent a long-term adaptative response to VEGFR2 loss that occurs only during active angiogenesis.
The scRNA-seq data indicating venousization of Vegfr2KO-E8.5–P7 ECs over time were confirmed by immunostaining for the master regulator of venous development NR2F2 (also known as COUP-TFII) (Fig. 4o,p and Extended Data Fig. 9c,d). To understand the venousization requirement for long-term adaptation of Vegfr2KO ECs, we also analysed Apln-FlpO iFlpMosaic Vegfr2Flox/Flox, Nr2f2Flox/Flox mice. When ECs lacked both VEGFR2 and NR2F2 (also known as COUP-TFII), their capacity to expand was significantly reduced, particularly in veins (Fig. 4q,r and Extended Data Fig. 9e). This suggests that endothelial venousization is a key mechanism of adaptation to long-term loss of VEGFR2 signalling during angiogenesis that enables mutant cells to expand in veins and derived vessels over time.
In addition to analysing embryonic and postnatal angiogenesis, we tested the capacity of Vegfr2KO ECs to form tumour vessels over a period of 2 weeks. When Vegfr2 deletion was induced a few days before tumour implantation, Vegfr2IECKO-day-3–adult ECs formed tumour vessels mainly at the tumour border and failed to penetrate deeper into the tumour, where they were 5.2-fold less frequent than control ECs (Extended Data Fig. 10a). By contrast, Vegfr2KO-E8.5–adult ECs, which had undergone venousization throughout embryonic and postnatal development, could better reach inside the tumour and were only 2.6-fold less frequent than control ECs (Extended Data Fig. 10b). These data show that although VEGFR2 signalling promotes tumour angiogenesis, ECs can progressively adapt to its absence and sustain tumour angiogenesis.
Differential compensation by VEGFR1 and VEGFR3
In addition to our findings with respect to venousization, we observed significant upregulation of expression of other VEGF receptor ligands in Vegfr2KO-E8.5–P7 ECs (Extended Data Fig. 11a), including PGF and VEGFC, which bind and activate VEGFR1 and VEGFR3, respectively1. In the absence of VEGFR2, VEGFR1 is the only other receptor that can bind VEGFA. However, all studies so far have shown that VEGFR1 kinase activity is too weak29,30, whereas its extracellular soluble form (sVEGFR1 or sFLT1) acts as a strong VEGF decoy receptor31. Therefore, when VEGFR2 is present, VEGFR1 is considered to mainly antagonize VEGFA–VEGFR2 signalling and angiogenesis. However, we found that Vegfr1 expression was much higher in ECs in the brain, the organ least sensitive to VEGFR2 loss (Extended Data Fig. 11b). We also found that during postnatal angiogenesis, sVegfr1 showed 2–5-fold lower expression than the membrane full-length Vegfr1 (flVegfr1) mRNA form, except in brain ECs (Extended Data Fig. 11c), suggesting a potential signalling role for flVEGFR1.
To evaluate whether VEGFR1 could compensate for loss of VEGFR2, we induced mosaic codeletion of Vegfr1 and Vegfr2 in ECs with iSuRe-HadCre from P1 to P7 (Vegfr1/2IECKO-P1–P7). Compared with Vegfr2KO ECs, Vegfr1/2KO ECs showed a further decrease in ability to migrate and form retina capillaries, but not veins or arteries (Fig. 5a). In veins, Vegfr1/2KO ECs could clonally expand (Extended Data Fig. 11d). Brain and heart Vegfr1/2KO ECs proliferated less than Vegfr2KO ECs at P7 (Extended Data Fig. 11e,f). Deletion of both Vegfr2 and Vegfr1 had a stronger effect on endothelial and pup survival (Extended Data Fig. 11g–i), confirming the requirement for VEGFR1 signalling when VEGFR2 is absent in capillaries. Mosaic deletion of Vegfr1 alone also led to a decrease in EC migration towards the VEGF-rich front (Extended Data Fig. 11j). In contrast to deletion of Vegfr1 in all ECs, mosaic loss of Vegfr1 in single ECs did not increase VEGF signalling (as assessed by ESM1 expression), probably owing to normal levels of sFLT1 present in the mostly wild-type microenvironment (Extended Data Fig. 11k).
a,b, Representative confocal micrographs of retinas from the indicated genetically distinct groups of mice induced at P1 and collected at P7, with cellular heatmaps, XY dot plot and related charts. c, Fold change in the contribution of Vegfr2/3KO-P1–P7 ECs relative to Vegfr2KO-P1–P7 ECs for artery cells. d, Representative confocal micrograph of a Vegfr2/3IECKO-P1–P7 retina immunostained for Ki-67 and ERG. Magnified insets show vein and AF. Yellow arrows indicate Tomato+Ki-67+ cells. e, High-magnification confocal micrographs of retinas stained for VEGFR3 and ERG, and violin plot indicating VEGFR3 immunostaining intensity in control and Tomato+Cre+ (gene KO) ECs normalized to background levels. f, Representative confocal micrographs of retinas from mice induced at P1 and collected at P7, with cellular heatmaps, XY dot plot and related charts. g, Proportional stacked bar plot contrasting the relative frequency of Tomato+Cre+ (KO for the indicated genes) cells in distinct retinal vascular regions at P7. h, Bar plot showing fold change in the frequency of Tomato+Cre+ ECs in different vascular regions and mutant groups in relation to Vegfr2KO. Asterisks indicate complete loss of ECs (zero) at LE. i, Representative confocal micrographs of retinas from P7 mice with the indicated alleles (n = 6 pups). The iFlpMosaicH2B-Reporter transgenic allele enabled stochastic labelling of a small fraction of individual nuclei with distinct fluorescent proteins, showing that individual Vegfr1/2/3IECKO-P1–P7 ECs clonally expand in veins. Data are presented as mean ± s.d. Scale bars, 100 μm (a,b,d,f,i) and 10 μm (e).
Source data
In addition to VEGFR1 and VEGFR2, ECs express VEGFR3. Previous studies have shown that VEGFR3 is strongly expressed by capillary and venous ECs; however, this receptor cannot bind VEGF (VEGFA), only VEGFC and VEGFD, which are key molecules for lymphangiogenesis but less relevant to blood vessel angiogenesis. Vegfr2IECKO-E8.5–P7 ECs showed very strong upregulation of VEGFC (Extended Data Fig. 11a), which could signal to VEGFR3 in the absence of VEGFR2. Loss of VEGFR3 in growing retina ECs has been shown to induce angiogenesis, suggesting that VEGFR3 negatively regulates VEGFR2 signalling32. However, single Vegfr3IECKO-P1–P7 ECs also had decreased ability to migrate and reach the leading edge (Extended Data Fig. 11l–n). Arterial ECs and smooth muscle cells express high levels of VEGFC33,34,35 a ligand that can bind both VEGFR2 and VEGFR3. Compared with Vegfr2IECKO-P1-P7 ECs, Vegfr2/3IECKO-P1-P7 ECs had a very similar zonal distribution in capillaries and veins (Fig. 5b) but a 3.4-fold lower frequency in arteries (Fig. 5c); this suggests an important role for VEGFC to VEGFR3 signalling in arterial mobilization or survival, but only when VEGFR2 signalling is absent. Loss of both VEGFR2 and VEGFR3 did not compromise the proliferation of surviving ECs in any vascular segments (Fig. 5d).
We also observed that Vegfr2KO ECs had decreased VEGFR3 expression only at the angiogenic front, where VEGFA levels are highest. By contrast Vegfr1/2KO ECs showed strong downregulation of VEGFR3 in all vascular segments, except in arteries (Fig. 5e). Vegfr1/2/3IECKO-P1-P7 ECs behaved similarly to Vegfr1/2IECKO-P1-P7 ECs, except for a pronounced decrease in formation of arterial vessels (Fig. 5f–h). Taken together, our findings show that in the absence of VEGFR2 signalling, VEGFR1 is required to sustain VEGFR3 expression and growth and migration of capillary ECs. In arteries, VEGFR3 continues to be expressed after loss of VEGFR1 and VEGFR2 and regulates development of arteries when the other VEGFA receptors are missing. Notably, vein ECs can survive and clonally expand in the complete absence of all VEGF receptor signalling (Fig. 5i).
Discussion
Our study challenges the longstanding view that VEGFR2 signalling is indispensable for endothelial proliferation, migration and sprouting during angiogenesis1,22. By using iFlpMosaics to induce ratiometric genetic mosaics of wild-type and Vegfr2KO ECs, we demonstrated that these mutant cells are capable of surviving, proliferating and competing with wild-type ECs during the entire embryonic and postnatal organ growth period (Extended Data Fig. 12). Contrary to previous studies suggesting that VEGFR2 is required for all forms of vascular differentiation and development10,12,36, our findings indicate that VEGFR2-independent mechanisms support angiogenesis in certain vascular beds, particularly the long-term development of venous and perivenous capillaries.
We identified significant organ-specific differences in adaptation to acute loss of VEGFR2 signalling (Extended Data Fig. 12a), thereby extending previous findings37,38. Whereas angiogenic retina ECs were strongly affected, brain ECs were the least affected, retaining high angiogenic potential, partially because of VEGFR1 compensation (Extended Data Fig. 12b). By contrast, postnatal heart, kidney and liver ECs were more sensitive to acute withdrawal of VEGFR2, showing partial depletion of Vegfr2KO ECs in capillaries, as well as a partial reduction in proliferation. The postnatal lung showed a distinct response, with no major changes in EC proliferation or cluster composition but instead a shift in transcriptional programs towards inflammation and metabolic adaptations. These findings highlight the heterogeneity of endothelial responses across different vascular beds and suggest that organ-specific factors influence EC resilience to perturbations in VEGF signalling.
The long-term adaptation of Vegfr2KO ECs was a notable observation. Although these cells initially had a competitive disadvantage during early embryonic and postnatal sprouting angiogenesis stages, they ultimately expanded and outcompeted wild-type cells in venous and perivenous regions, indicating that alternative pathways support their survival and proliferation in this niche. Veins are known to provide a continuous source of ECs for capillary angiogenesis and later arterialization20,23, and this venous dominance of Vegfr2KO ECs allows them to compete with wild-type cells over time. Our scRNA-seq analysis of different vascular beds further showed that ECs with prolonged loss of VEGFR2 exhibited a significant shift towards a venous transcriptional signature (venousization), marked by upregulation of many venous genes, including Nr2f2, a master regulator of veins development24,39. Vegfr2KONr2f2KO ECs were significantly less able to expand, particularly in veins, showing that venousization is a key adaptation mechanism that enables Vegfr2KO ECs to persist and contribute to vascular development (Extended Data Fig. 12c,d).
In addition to venousization, our findings demonstrate the compensatory roles of VEGFR1 and VEGFR3 in supporting survival and growth signalling when VEGFR2 is absent (Extended Data Fig. 12b). Previously, VEGFR1 has mainly been reported to negatively regulate VEGFR2 signalling during angiogenesis, because its soluble secreted version sFLT1 or VEGFR1 is a decoy receptor of VEGF1,22. In our mosaic Vegfr1KO and Vegfr1/2KO experiments, sFLT1 levels in the microenvironment were normal; this led us to find that membrane or full-length VEGFR1 acted as a positive regulator of angiogenesis and partially compensated for VEGFR2 loss, supporting both EC survival and migration. The mosaic single loss of VEGFR3 also showed that this receptor was a positive regulator of EC migration towards the leading edge. This was consistent with results obtained using VEGFR3-blocking antibodies40; however, our findings were contrary to results from pan-endothelial Vegfr3KO genetics experiments32 in which the receptor was lost from all ECs. In this case, VEGFC and/or VEGFD bioavailability in the whole body increases, thereby increasing VEGFC–VEGFR2 signalling, whereas in low-frequency Vegfr3KO mosaics this does not occur. In addition, Vegfr2/3KO ECs mobilized less to the angiogenic front and had decreased arterialization when compared with Vegfr2KO ECs. The ability of these alternative VEGF receptors to sustain EC survival, migration and proliferation at the single-cell level indicates potential functional plasticity and compensation within the VEGF signalling network, as suggested previously by studies using classical full-tissue conditional genetics12,32,37.
An important implication of our study is that EC sensitivity to VEGFR2 loss varies depending on the vascular context and the time since the loss. We found that acute loss of VEGFR2 during active angiogenesis resulted in significant EC depletion in the first 3 days, particularly depletion of tip cells and other ECs at the angiogenic front, where VEGF signalling is highest. However, Vegfr2KO ECs in VEGF-low areas and those that had undergone long-term adaptation were able to persist and even contribute to a substantial portion of capillary vessels, particularly those around veins. This suggests that ECs experiencing prolonged VEGFR2 deficiency undergo venousization in VEGF-low areas, enabling them to survive and then grow in the absence of canonical VEGF signalling. Our tumour models showed that Vegfr2KO ECs could form tumour vessels, particularly if they had previously undergone venousization during embryonic angiogenesis. The use of ratiometric and trackable iFlpMosaics was key to obtaining these findings.
Our work may also provide biomarkers and insights into resistance mechanisms to anti-VEGF therapies5,41,42,43. Adaptation of ECs to VEGFR2 loss by engaging alternative venous-related angiogenic pathways may also occur in tumours. If this is confirmed, therapeutic strategies that simultaneously block the compensatory and venous-related pathways identified here may be necessary to overcome resistance to anti-VEGF treatments.
In conclusion, our study reshapes our understanding of VEGFR2-dependent angiogenesis by demonstrating that ECs possess intrinsic mechanisms to adapt to long-term absence of VEGFR2 signalling. The discovery of endothelial venousization and compensatory VEGF receptor signalling provides a new framework for investigation of alternative modes of vascular development and resistance to anti-VEGF therapies.
Methods
Mice
All mouse husbandry and experiments were conducted using protocols approved by local animal ethics committees and authorities (Comunidad Autónoma de Madrid and Universidad Autónoma de Madrid CAM-PROEX 164.8/20, PROEX 293.1/22 and PROEX-288.7/24). The mouse colonies were maintained in racked individual ventilation cages according to current national legislation. The mice had dust- and pathogen-free bedding and sufficient nesting and environmental enrichment material for development of species-specific behaviour. All mice had ad libitum access to food and water in environmental conditions of 45–65% relative humidity, temperatures of 21–24 °C and a 12 h/12 h light–dark cycle. For preservation of animal welfare, mouse health was monitored with an animal health surveillance programme that followed the recommendations of the Federation of European Laboratory Animal Science Associations for specific-pathogen-free facilities
We used Mus musculus lines mostly on the C57BL6 genetic background, with some having a fraction of 129SV, or B6CBAF1 or DBA genetic backgrounds. Mice were backcrossed to C57Bl6 for several generations. To generate mice for analysis, we intercrossed mice aged between 7 weeks and 30 weeks. We did not anticipate any influence of mouse sex on our data, and both female and male mice were analysed. The initial sample size was determined taking into account the ‘3 Rs’ for animal experimentation and the expected experimental variability based on protocols (published and our own; refs. 13 and 14). The final sample size was defined a posteriori after statistical analysis of the data. Sample size and related statistical analysis methods are indicated in the figure legends or source data. No randomization or blinding was used, and the animals or tissues were selected for analysis on the basis of their genotype, the detected Cre- or FlpO-dependent recombination frequency and the quality of multiplex immunostaining.
The following mouse lines were used and intercrossed: Tg(Cdh5–CreERT2)44, Tg(iSuRe–HadCre)14, Vegfr2flox (ref. 45), Vegfr1flox (ref. 46), Vegfr3flox (ref. 47) (for this work, we also generated a new Vegfr3flox allele in the Vegfr1/2flox background, as indicated below), Nr2f2flox (ref. 48), iMb2–Mosaic16, Apln–FlpO15 and the iFlpMosaic lines13 (R26-iFlpMTomato-2A-Cre/MYFP and Tg-iFlpChromatin–Mosaic).
To generate a new Vegfr3flox allele combined with Vegfr1/2flox we injected Cas9, donor DNAs containing LoxP sites and guide RNAs (TGTTAAAGACACGCATACCT/AGG and AGATGGGGTGAATAGCTCCG/TGG) into mouse eggs of the Vegfr1/2flox iSuRe–HadCre Cdh5–CreERT2 line. Founders containing LoxP sites flanking exon 1 in the same chromosome were screened by PCR, and inducible Vegfr3 deletion was confirmed by PCR and immunostaining for the VEGFR3 protein (Extended Data Fig. 11l,m).
For induction of CreERT2 activity in adult mice, 1 g tamoxifen (Sigma-Aldrich, P5648_1G) was dissolved in 50 ml corn oil (stock tamoxifen concentration, 20 mg m−1), and aliquots were stored at −20 °C. Animals received intraperitoneal injections of 100 µl of this stock solution (total dose: 2 mg tamoxifen per animal at 80 mg kg−1), as indicated in the figures. To activate recombination in pups, we injected 4-OHT at the indicated stages at a dose of 40 mg kg−1. For double (VEGFR1/2, VEGFR2/3) or triple (VEGFR1/2/3) mutants, 4-OHT was administered to pups at 15 mg kg−1. For activation of recombination during embryogenesis, pregnant dams were induced with 0.25 mg 4-OHT supplemented with 0.125 mg progesterone at E10.5 and E12.5. All mouse lines and primer sequences required for genotyping are provided in Supplementary Table 1.
Tumour model
Lewis carcinoma cells were obtained from ATCC and cultured in DMEM (Gibco) containing 10% fetal bovine serum (FBS), 1× penicillin–streptomycin and 2 mM glutamine at 37 °C, 5% CO2. The cell medium was changed every other day, and cells were passaged every 3–4 days. Cells were routinely screened for mycoplasma contamination. Cells were used for tumour inoculation when they were at 75–80% confluency. Cells were trypsinized and resuspended in Matrigel solution (Biosciences). Age-matched (15–20 weeks) and sex-matched mice were injected subcutaneously with 5 × 105 tumour cells suspended in Matrigel (Corning, 354234). Mice were euthanized after 15 days, and tumours (less than 2 cm3) were extracted, fixed and processed for immunostaining.
Immunostaining of whole retinas
For immunostaining of mouse retinas, eyes were dissected from mouse pups and fixed by incubation with agitation for 20 min in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) (diluted from a stock of 16% PFA; EMS 15710). After two washes in PBS (5 min each), retinas were microdissected from the eyes; 4 incisions were made to enable later flat mounting, and the retinas were refixed in 4% PFA for 45 min with agitation. After two washes in PBS, retinas were blocked and permeabilized by incubation for 1 h in PBST (0.3% Triton X-100, 3% FBS and 3% donkey serum in PBS). Samples were then incubated with primary antibodies (Supplementary Table 2) diluted in PBST overnight at 4 °C with agitation. The next day, retinas were washed 5 times for 20 min in PBST diluted 1:2. Then, they were incubated in the same solution for 2 h at room temperature with Alexa-conjugated secondary antibodies (Supplementary Table 2). After washing (3 × 15 min) in PBS containing 0.15% Triton X-100, retinas were washed (2 × 15 min) in PBS and flat mounted in Fluoromount-G (SouthernBiotech). For combination of rabbit anti-ERG-647 with other antibodies produced in rabbit, retinas were incubated with the Alexa 594 AffiniPure Fab Fragment Donkey Anti-Rabbit secondary antibody, then with 5% rabbit serum for 30 min at room temperature, and finally with rabbit anti-ERG-647.
Immunofluorescence of tissue cryosections
For collection of organs and histological analysis, mice were euthanized in CO2 chambers, and 10 ml of PBS was injected into the left ventricles, followed by perfusion with 3.7–4% formaldehyde (ITW Reagents, 252931) at pH 7. The explanted organs were postfixed for 2 h in 4% PFA (Thermo Scientific, 043368.9M) in PBS at 4 °C with gentle rotation. After three washes in PBS for 10 min each, the organs were stored overnight in 30% sucrose (Sigma) in PBS. The organs were then embedded in optimal cutting temperature compound (Sakura) and frozen at −80 °C. Cryosections of 30 µm were cut on a cryostat (Leica), washed three times for 10 min each time in PBS, and blocked and permeabilized in PBS containing 10% donkey serum (Millipore), 10% FBS and 1% Triton X-100. The primary antibodies (Supplementary Table 2) were diluted in the same buffer and incubated with sections overnight at 4 °C. This step was followed by three 10-min washes in PBS and incubation for 2 h with conjugated secondary antibodies (Supplementary Table 2) and 4,6-diamidino-2-phenylindole (DAPI) in PBS at room temperature. After three washes in PBS, the sections were mounted with Fluoromount-G (SouthernBiotech). For combination of rabbit anti-ERG-647 with other antibodies produced in rabbit, retinas were incubated first with the other primary rabbit antibodies, followed by Alexa 594 AffiniPure Fab Fragment Donkey Anti-Rabbit secondary antibody, then rabbit epitopes blocked with 5% rabbit serum for 30 min at room temperature, and finally incubated with rabbit anti-ERG-647.
Image acquisition and analysis
For confocal scanning, immunostained organ sections or whole-mount retinas were imaged at high resolution with a Leica SP8 Navigator confocal microscope fitted with a ×10 or ×40 objective. Individual fields or tiles of large areas were acquired. All images shown are representative of the results obtained for each group and experiment. Animals were dissected and processed under identical conditions. Phenotypes or signal intensities were compared using images obtained with the same laser excitation and confocal scanner detection settings. We used ImageJ/Fiji v.1.54f to threshold, select and quantify the objects in confocal micrographs.
For arteriovenous mapping of ERG-labelled control and mutant ECs (Extended Data Fig. 3a), a semiautomatic Fiji script (available at https://github.com/RuiBenedito/Benedito_Lab) was developed. First, six regions of interest were manually defined by the user: artery, vein, mature capillaries between the established arteries and veins, intermediate capillaries, angiogenic front and leading edge. ERG labelling was used to segment endothelial nuclei and classify each nucleus according to reporter or marker positivity (determined by the percentage of pixels within each nucleus above a given intensity threshold level), and the XY coordinates of nuclei were extracted. Then, arterial, venous, distal and proximal arteriovenous boundaries were defined by the user to determine the relative position of each segmented nucleus to each arteriovenous field boundary; this enabled compilation of data extracted from different arteriovenous fields into a single XY dot plot or heatmap.
The data collected in Fiji were post-analysed in R (v.4.5.2) for generation of graphs representing the spatial distribution of the cells of interest. For the XY plots and heatmaps, samples were randomly subsetted to enable us to display equal numbers of control and mutant cells (MYFP+ versus MTomato+ in the Apln–FlpO, iFlpMosaic background, and MTomato− versus MTomato+ in the Cdh5–CreERT2, iSuRe–HadCre background); this was to make clear the zonation differences between control and mutant cells. For visualization with XY dot plots, 2,000 randomly selected cells of each group were plotted; for the heatmap generation, the maximal number of cells (determined by the cell type (control or mutant) with fewer cells) was used. The percentage of each cell type in each arteriovenous region was calculated given its spatial distribution per field (Extended Data Fig. 3a).
We generated Fiji macro Script_Reporter_MarkerDetection_Mapping.ijm for spatial mapping of ERG and/or quantification of EC colocalization with reporters (MYFP or MTomato) and other markers (Ki-67) as in Figs. 1, 2, 5. R scripts were used for data quantification and heatmapping (Script_Spatial Heatmap; Figs. 1, 2, 5 and Extended Data Figs. 3, 4, 11) and are available at GitHub (https://github.com/RuiBenedito/Benedito_Lab). Fiji, Adobe Photoshop CC 24.1.0 and Adobe Illustrator CC v.27.1.1 were used for downstream image processing, analysis and illustration.
To quantify the clonal expansion of single ECs expressing the iSuRe–HadCre and the Tg-iFlpChromatin–Mosaic alleles, we segmented and analysed images with MTomato+H2B-Cherry+, MTomato+H2B-EGFP+ or MTomato+HA-H2B-Cerulean+ signals with the Clone_Identification_Quantification script, deposited at GitHub (https://github.com/RuiBenedito/Benedito_Lab/).
Flow cytometry and FACS
Postnatal and adult tissues were dissociated before fluorescence-activated cell sorting (FACS). Enzymatic digestion was carried out at 37 °C for 20 min using a solution containing 2.5 mg ml−1 collagenase type I (Thermo Fisher), 2.5 mg ml−1 dispase II (Thermo Fisher) and 50 μg ml−1 DNase I (Roche). Following dissociation, samples were pelleted by centrifugation (450g, 4 °C, 5 min). Pellets were gently resuspended in 1× blood lysis buffer (BioLegend, 420301) and incubated on ice for 5 min for lysis of erythroid cells. After lysis, cells were passed through a 70-μm cell strainer and centrifuged again (450g, 4 °C, 5 min). Cells were resuspended in antibody incubation solution (PBS without Ca2+ or Mg2+, supplemented with 2% dialysed FBS; Biowest, X0515), with primary antibodies (APC rat anti-mouse CD31 and APC-Cy7 anti-CD45; Supplementary Table 2) only or with the addition of Hashtag oligo (HTO) antibodies (Supplementary Table 2) when scRNA-seq was performed. Samples were then incubated at 4 °C with gentle agitation for 20 min.
After immunolabelling, cells were centrifuged (450g, 4 °C, 5 min), washed with antibody incubation solution, centrifuged again and resuspended in sorting buffer (10% FBS in PBS without Ca2+ or Mg2+). Stained cells were maintained on ice until acquisition. DAPI (5 mg ml−1) was added immediately before FACS analysis. Cells were analysed using an LSRFortessa cell analyser or sorted using a FACS Aria Cell Sorter (BD Biosciences). For an example of the FACS gating strategy, see Extended Data Fig. 1a. BD FACSDiva v.8.0.1 and FlowJo v.10 were used for data acquisition and analysis.
Cell isolation for transcriptomic analysis
For RNA-seq analysis, APC rat anti-mouse CD31 (which labels ECs), APC-Cy7 rat anti-CD45 (which labels blood cells) and HTO antibodies (BD Biosciences) (Supplementary Table 2) were used together to hashtag and distinguish the different organ Tomato+CD31+ ECs when loaded on the same 10x Genomics port (Extended Data Figs. 6a and 8a). After antibody incubation, samples were washed twice with antibody incubation solution, centrifuged (450g, 5 min, 4 °C) and resuspended in sorting buffer (10% FBS in PBS without Ca2+ or Mg2+). Stained cells were maintained on ice until acquisition. Immediately before FACS analysis, DAPI (5 mg ml−1) was added, and the cell suspension was passed slowly through a 70-µm cell strainer attached to a 1-ml syringe for removal of aggregates. Live, DAPI-negative, CD31+CD45− and endogenous-reporter-positive cells were sorted into cold 1.5-ml microcentrifuge tubes containing 300 µl of sorting buffer. Cells were collected using a 100-µm nozzle at 20 PSI in high-purity mode with a flow rate of fewer than 3,000 events per second to preserve cell viability and minimize contamination. After sorting, cells were pelleted at 500g for 5 min at 4 °C and resuspended in 30–40 µl cell-capture buffer (Ca2+- and Mg2+-free PBS supplemented with 0.04% ultrapure bovine serum albumin; Thermo Fisher, AM2616). Viability and cell concentration were determined using a Countess 3 Automated Cell Counter (Thermo Fisher Scientific) before loading of the 10x Genomics port.
Next-generation sequencing and library preparation
Single-cell suspensions were loaded on to the 10x Genomics Chromium Controller system for encapsulation into emulsion droplets. The 10x Genomics Chromium single-cell 3′ RNA-seq library preparation and sequencing were carried out at the CNIC Genomics Unit according to the manufacturer’s protocols. Each port was loaded with 20,000 to 50,000 cells. Final libraries were sequenced on an Illumina NextSeq 2000 platform, with software v.1.7.2.
scRNA-seq data analysis
scRNA-seq data were aligned and quality controlled by the CNIC Bioinformatics Unit using Cell Ranger (v.8 and v.9). Clustering, cell type identification and downstream analysis were performed by I.G.-G. The following pipeline was applied for transcript alignment, quantification, quality control and cell type classification.
Preprocessing and alignment
Reference transcriptomes were generated using the Ensembl GRCm38 mouse genome, build version 98 (sep2019.archive.ensembl.org). Genes metadata were obtained from the corresponding Ensembl BioMart archive. To enable downstream analysis of gene deletion efficiency and soluble versus full-length isoform abundance, we appended Vegfr2 (also known as Kdr) exon 3, Kdr exon 30, Vegfr1 (also known as Flt1) exon 13b and Flt1 exon 30 to the reference transcriptome as custom metadata features. Gene annotations were obtained from the corresponding Ensembl BioMart archive. Raw sequencing reads from gene expression libraries, hashtag oligonucleotides and targeted transcript sequences were aligned and quantified using Cell Ranger (v.8 and v.9).
Downstream analyses were performed using R (v.4.5.0) and the Seurat package (v.5.5.1; https://satijalab.org/seurat/)49. Cells were filtered for removal of low-quality profiles and technical artifacts using the following criteria: 2,000–30,000 total unique molecular identifiers (UMIs), 800–6,000 detected genes, mitochondrial transcript content less than 25%, fewer than 65% of reads mapping to the top 50 expressed genes and fewer than 0.1% haemoglobin transcripts. Cells with fewer than 15 hashtag UMIs were excluded to improve sample assignment accuracy. To enrich for endothelial populations and exclude hematopoietic contamination, we also required cells to express Cdh5 (normalized expression greater than 0.1) and no or very low expression of Ptprc (normalized expression less than 0.1). Cells identified as doublets on the basis of hashtag antibody signals were removed. Demultiplexing by organ was performed using sample-specific hashtag oligonucleotide signals (Supplementary Table 2) and after confirmed by organ-specific EC markers, and all downstream analyses were conducted separately for each organ ECs.
Mitochondrial genes, haemoglobin genes and transgenes were removed before clustering to avoid biasing of normalization and dimensionality reduction. Filtered count data were log-normalized and scaled before analysis. Dimensionality reduction was performed using principal component analysis, and batch effects across independent ports were corrected with the Harmony algorithm50.
Clustering, doublet removal and cell type identification
We clustered cells by constructing a shared nearest neighbour graph and applying the Louvain community detection algorithm to identify distinct cell populations. The optimal number of principal components for dimensionality reduction was determined using an elbow plot, and clustering resolution parameters were selected on the basis of the EC complexity observed in each organ.
Clusters and cells were classified using the SingleR function from the SingleR package v.2.10.0 (ref. 51). Putative doublets were identified as cell clusters that had high total UMI or gene counts or unusual gene coexpression or were located in principal component analysis outliers. Cells classified as singlets were retained and reclustered. Final cluster identities were manually curated on the basis of known EC subtypes and literature-based markers. The FUNCTION_Dotplot_Markers custom function in R was used to find and plot top marker genes per cluster to determine cell identities.
Quantification of exon deletion
To assess the rate of exon deletion in Vegfr2flox/flox samples, we computed the number of UMI reads aligned to each exon, including the floxed region and the terminal 3′ exon or untranslated region (which is typically overrepresented in 10x Genomics 3′ libraries). Only reads with valid barcodes, UMIs and matching strand orientation were retained.
Read counts for floxed exons were normalized to total UMI counts per cell. These values were added to the Seurat metadata. Deletion efficiency was visualized using dot plots and violin plots, comparing expression of the whole transcript (including the 3′ untranslated region). These visualizations were generated using the custom R function FUNCTION_violin_dotplot_deletion_FC.R, which combines violin plots of single-cell expression distributions with dot plots displaying total average expression numbers and signed fold change between conditions.
Calculation of sVegfr1/flVegfr1 ratio
To obtain the ratio between the soluble (sVegfr1) and full-length (flVegfr1) isoforms, the normalized reads from exon 13b (the last exon in the soluble form) were divided by the normalized reads of exon 30 (the last exon in the full-length version).
Comparison of mutant and control sample ratios
We quantified differences in cell type composition between control and mutant samples for each organ by comparing relative proportions across conditions. For UMAP visualizations (both split and integrated), equal downsampling of the two conditions was performed to ensure identical cell numbers, thereby avoiding sampling bias in visual comparisons.
Differential gene expression between conditions was visualized using volcano plots. These were generated using R functions (FUNCTION_VolcanoPlot) developed in-house that built on the EnhancedVolcano package (v.1.26.0)52. For volcano plots and heatmaps, statistical analyses were performed using Seurat’s FindMarkers function, with Wilcoxon rank-sum test and Bonferroni correction for multiple testing. GSEA was performed using FGSEA with a custom wrapper function (FUNCTION_Fgsea_TableFgsea) based on the FGSEA package. For hallmark pathways, statistical analyses were performed using the multilevel Monte Carlo approach implemented in fgseaMultilevel (R package fgsea v.1.34.2)53 and Benjamini–Hochberg false discovery rate correction for multiple testing.
Genes used for differential gene expression analysis were prefiltered using FUNCTION_FilterLowExpressionGenes_onlyTable; this removes genes with low and highly heterogeneous expression patterns across cells, which are often inconsistently detected between datasets and potentially reflect technical artefacts such as ambient RNA contamination. The function acts as an empirical, distribution-based filter to improve robustness of downstream differential expression analysis. Differential expression heatmaps were generated using a custom function (FUNCTION_Find_PlotGenesHeatmap_FC_Metadata), implemented in-house and based on the pheatmap package.
Statistics and reproducibility
Most numerical data shown in charts were first compiled and processed with Microsoft Excel 2019 and 2024 and then analysed and plotted with GraphPad Prism v.10.1.0. All bar graphs show mean ± s.d. Experiments were repeated with independent animals, as stated in the source data file or figure legends. Comparisons between two sample groups with a Gaussian distribution were performed using unpaired two-tailed Student’s t-tests. Comparisons among more than two groups were made by one-way or two-way analysis of variance followed by multiple comparison tests. Data points were analysed and plotted with GraphPad Prism. No randomization or blinding was used, and the animals or tissues were selected for analysis on the basis of their genotype, the detected Cre- or FlpO-dependent recombination frequency and the quality of multiplex immunostaining. For mosaic analysis, only samples with less than 50% recombination were selected. The sample sizes were chosen according to the observed statistical variation and published protocols.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
RNA-seq data can be viewed at the Gene Expression Omnibus under accession code GSE331295. Unprocessed FACS raw data files and original microscopy images of the data are available upon request. All other data supporting the findings in this study are included in the main article and associated files. Source data are provided with this paper.
Code availability
All custom scripts and functions used in this study are available at GitHub (https://github.com/RuiBenedito/Benedito_Lab/).
References
Lohela, M., Bry, M., Tammela, T. & Alitalo, K. VEGFs and receptors involved in angiogenesis versus lymphangiogenesis. Curr. Opin. Cell Biol. 21, 154–165 https://doi.org/10.1016/j.ceb.2008.12.012 (2009).
Article PubMed CAS Google Scholar
Shalaby, F. et al. Failure of blood-island formation and vasculogenesis in Flk-1-deficient mice. Nature 376, 62–66 (1995).
Article PubMed CAS Google Scholar
Carmeliet, P. et al. Abnormal blood vessel development and lethality in embryos lacking a single VEGF allele. Nature 380, 435–439 (1996).
Article PubMed CAS Google Scholar
Casanovas, O., Hicklin, D. J., Bergers, G. & Hanahan, D. Drug resistance by evasion of antiangiogenic targeting of VEGF signaling in late-stage pancreatic islet tumors. Cancer Cell 8, 299–309 https://doi.org/10.1016/j.ccr.2005.09.005 (2005).
Article PubMed CAS Google Scholar
Jain, R. K. et al. Biomarkers of response and resistance to antiangiogenic therapy. Nat. Rev. Clin. Oncol. 6, 327–338 https://doi.org/10.1038/nrclinonc.2009.63 (2009).
Article PubMed PubMed Central CAS Google Scholar
Apte, R. S., Chen, D. S. & Ferrara, N. VEGF in signaling and disease: beyond discovery and development. Cell 176, 1248–1264 https://doi.org/10.1016/j.cell.2019.01.021 (2019).
Article PubMed PubMed Central CAS Google Scholar
Wallsh, J. O. & Gallemore, R. P. Anti-VEGF-resistant retinal diseases: a review of the latest treatment options. Cells 10, 1049 https://doi.org/10.3390/cells10051049 (2021).
Article PubMed PubMed Central CAS Google Scholar
Zeng, Q. et al. Understanding tumour endothelial cell heterogeneity and function from single-cell omics. Nat. Rev. Cancer 23, 544–564 https://doi.org/10.1038/s41568-023-00591-5 (2023).
Article PubMed CAS Google Scholar
Potente, M., Gerhardt, H. & Carmeliet, P. Basic and therapeutic aspects of angiogenesis. Cell 146, 873–887 https://doi.org/10.1016/j.cell.2011.08.039 (2011).
Article PubMed CAS Google Scholar
Shalaby, F. et al. A requirement for Flk1 in primitive and definitive hematopoiesis and vasculogenesis. Cell 89, 981–990 (1997).
Article PubMed CAS Google Scholar
Milgrom-Hoffman, M. et al. The heart endocardium is derived from vascular endothelial progenitors. Development 138, 4777–4787 https://doi.org/10.1242/dev.061192 (2011).
Article PubMed PubMed Central CAS Google Scholar
Benedito, R. et al. Notch-dependent VEGFR3 upregulation allows angiogenesis without VEGF–VEGFR2 signalling. Nature 484, 110–114 https://doi.org/10.1038/nature10908 (2012).
Article PubMed CAS Google Scholar
Garcia-Gonzalez, I. et al. iFlpMosaics enable the multispectral barcoding and high-throughput comparative analysis of mutant and wild-type cells. Nat. Methods 22, 323–334 https://doi.org/10.1038/s41592-024-02534-w (2025).
Article PubMed CAS Google Scholar
Garcia-Gonzalez, I. et al. iSuRe-HadCre is an essential tool for effective conditional genetics. Nucleic Acids Res. 52, e56 https://doi.org/10.1093/nar/gkae472 (2024).
Article PubMed PubMed Central CAS Google Scholar
Luo, W. et al. Arterialization requires the timely suppression of cell growth. Nature 589, 437–441 https://doi.org/10.1038/s41586-020-3018-x (2021).
Article PubMed CAS Google Scholar
Pontes-Quero, S. et al. Dual ifgMosaic: a versatile method for multispectral and combinatorial mosaic gene-function analysis. Cell 170, 800–814 https://doi.org/10.1016/j.cell.2017.07.031 (2017).
Article PubMed PubMed Central CAS Google Scholar
Lawson, N. D., Vogel, A. M. & Weinstein, B. M. sonic hedgehog and vascular endothelial growth factor act upstream of the Notch pathway during arterial endothelial differentiation. Dev. Cell 3, 127–136 (2002).
Article PubMed CAS Google Scholar
Wythe, J. D. et al. ETS factors regulate Vegf-dependent arterial specification. Dev. Cell 26, 45–58 https://doi.org/10.1016/j.devcel.2013.06.007 (2013).
Article PubMed PubMed Central CAS Google Scholar
Herbert, S. P. et al. Arterial-venous segregation by selective cell sprouting: an alternative mode of blood vessel formation. Science 326, 294–298 https://doi.org/10.1126/science.1178577 (2009).
Article PubMed PubMed Central CAS Google Scholar
Xu, C. et al. Arteries are formed by vein-derived endothelial tip cells. Nat. Commun. 5, 5758 https://doi.org/10.1038/ncomms6758 (2014).
Article PubMed PubMed Central CAS Google Scholar
Pitulescu, M. E. et al. Dll4 and Notch signalling couples sprouting angiogenesis and artery formation. Nat. Cell Biol. 19, 915–927 https://doi.org/10.1038/ncb3555 (2017).
Article PubMed CAS Google Scholar
Simons, M., Gordon, E. & Claesson-Welsh, L. Mechanisms and regulation of endothelial VEGF receptor signalling. Nat. Rev. Mol. Cell Biol. 17, 611–625 https://doi.org/10.1038/nrm.2016.87 (2016).
Article PubMed CAS Google Scholar
Lee, H. W. et al. Role of venous endothelial cells in developmental and pathologic angiogenesis. Circulation 144, 1308–1322 https://doi.org/10.1161/CIRCULATIONAHA.121.054071 (2021).
Article PubMed PubMed Central CAS Google Scholar
Red-Horse, K. & Siekmann, A. F. Veins and arteries build hierarchical branching patterns differently: bottom-up versus top-down. Bioessays 41, e1800198 https://doi.org/10.1002/bies.201800198 (2019).
Article PubMed PubMed Central Google Scholar
Zarkada, G. et al. Specialized endothelial tip cells guide neuroretina vascularization and blood-retina-barrier formation. Dev. Cell 56, 2237–2251 https://doi.org/10.1016/j.devcel.2021.06.021 (2021).
Article PubMed PubMed Central CAS Google Scholar
Xia, C. H., Yablonka-Reuveni, Z. & Gong, X. LRP5 is required for vascular development in deeper layers of the retina. PLoS ONE 5, e11676 https://doi.org/10.1371/journal.pone.0011676 (2010).
Article PubMed PubMed Central Google Scholar
Ye, X. et al. Norrin, Frizzled-4, and Lrp5 signaling in endothelial cells controls a genetic program for retinal vascularization. Cell 139, 285–298 https://doi.org/10.1016/j.cell.2009.07.047 (2009).
Article PubMed PubMed Central CAS Google Scholar
Ricard, N. et al. Endothelial ERK1/2 signaling maintains integrity of the quiescent endothelium. J. Exp. Med. 216, 1874–1890 https://doi.org/10.1084/jem.20182151 (2019).
Article PubMed PubMed Central CAS Google Scholar
Waltenberger, J., Claesson-Welsh, L., Siegbahn, A., Shibuya, M. & Heldin, C. H. Different signal transduction properties of KDR and Flt1, two receptors for vascular endothelial growth factor. J. Biol. Chem. 269, 26988–26995 (1994).
Article PubMed CAS Google Scholar
Hiratsuka, S., Minowa, O., Kuno, J., Noda, T. & Shibuya, M. Flt-1 lacking the tyrosine kinase domain is sufficient for normal development and angiogenesis in mice. Proc. Natl Acad. Sci. USA 95, 9349–9354 (1998).
Article PubMed PubMed Central CAS Google Scholar
Wazan, L. E., Widhibrata, A. & Liu, G. S. Soluble FLT-1 in angiogenesis: pathophysiological roles and therapeutic implications. Angiogenesis 27, 641–661 https://doi.org/10.1007/s10456-024-09942-8 (2024).
Article PubMed CAS Google Scholar
Tammela, T. et al. VEGFR-3 controls tip to stalk conversion at vessel fusion sites by reinforcing Notch signalling. Nat. Cell Biol. 13, 1202–1213 https://doi.org/10.1038/ncb2331 (2011).
Article PubMed PubMed Central CAS Google Scholar
Partanen, T. A. et al. VEGF-C and VEGF-D expression in neuroendocrine cells and their receptor, VEGFR-3, in fenestrated blood vessels in human tissues. FASEB J. 14, 2087–2096 https://doi.org/10.1096/fj.99-1049com (2000).
Article PubMed CAS Google Scholar
Paavonen, K. et al. Vascular endothelial growth factors C and D and their VEGFR-2 and 3 receptors in blood and lymphatic vessels in healthy and arthritic synovium. J. Rheumatol. 29, 39–45 (2002).
PubMed CAS Google Scholar
Karkkainen, M. J. et al. Vascular endothelial growth factor C is required for sprouting of the first lymphatic vessels from embryonic veins. Nat. Immunol. 5, 74–80 (2004).
Article PubMed CAS Google Scholar
Zarkada, G., Heinolainen, K., Makinen, T., Kubota, Y. & Alitalo, K. VEGFR3 does not sustain retinal angiogenesis without VEGFR2. Proc. Natl Acad. Sci. USA 112, 761–766 https://doi.org/10.1073/pnas.1423278112 (2015).
Article PubMed PubMed Central CAS Google Scholar
Karaman, S. et al. Interplay of vascular endothelial growth factor receptors in organ-specific vessel maintenance. J. Exp. Med. 219, e20210565 https://doi.org/10.1084/jem.20210565 (2022).
Article PubMed PubMed Central CAS Google Scholar
Lee, S. et al. Autocrine VEGF signaling is required for vascular homeostasis. Cell 130, 691–703 (2007).
Article PubMed PubMed Central CAS Google Scholar
You, L. R. et al. Suppression of Notch signalling by the COUP-TFII transcription factor regulates vein identity. Nature 435, 98–104 (2005).
Article PubMed CAS Google Scholar
Tammela, T. et al. Blocking VEGFR-3 suppresses angiogenic sprouting and vascular network formation. Nature 454, 656–660 https://doi.org/10.1038/nature07083 (2008).
Article PubMed CAS Google Scholar
Loges, S., Schmidt, T. & Carmeliet, P. Mechanisms of resistance to anti-angiogenic therapy and development of third-generation anti-angiogenic drug candidates. Genes Cancer 1, 12–25 https://doi.org/10.1177/1947601909356574 (2010).
Article PubMed PubMed Central CAS Google Scholar
Bergers, G. & Hanahan, D. Modes of resistance to anti-angiogenic therapy. Nat. Rev. Cancer 8, 592–603 https://doi.org/10.1038/nrc2442 (2008).
Article PubMed PubMed Central CAS Google Scholar
Sharma, D., Zachary, I. & Jia, H. Mechanisms of acquired resistance to anti-VEGF therapy for neovascular eye diseases. Invest. Ophthalmol. Vis. Sci. 64, 28 https://doi.org/10.1167/iovs.64.5.28 (2023).
Article PubMed PubMed Central CAS Google Scholar
Sorensen, I., Adams, R. H. & Gossler, A. DLL1 mediated Notch activation regulates endothelial identity in mouse fetal arteries. Blood 113, 5680–5688 https://doi.org/10.1182/blood-2008-08-174508 (2009).
Article PubMed Google Scholar
Haigh, J. J. et al. Cortical and retinal defects caused by dosage-dependent reductions in VEGF-A paracrine signaling. Dev. Biol. 262, 225–241 https://doi.org/10.1016/S0012-1606(03)00356-7 (2003).
Article PubMed CAS Google Scholar
Ambati, B. K. et al. Corneal avascularity is due to soluble VEGF receptor-1. Nature 443, 993–997 https://doi.org/10.1038/nature05249 (2006).
Article PubMed PubMed Central CAS Google Scholar
Haiko, P. et al. Deletion of vascular endothelial growth factor C (VEGF-C) and VEGF-D is not equivalent to VEGF receptor 3 deletion in mouse embryos. Mol. Cell. Biol. 28, 4843–4850 https://doi.org/10.1128/MCB.02214-07 (2008).
Article PubMed PubMed Central CAS Google Scholar
Takamoto, N. et al. COUP-TFII is essential for radial and anteroposterior patterning of the stomach. Development 132, 2179–2189 https://doi.org/10.1242/dev.01808 (2005).
Article PubMed CAS Google Scholar
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 https://doi.org/10.1016/j.cell.2021.04.048 (2021).
Article PubMed PubMed Central CAS Google Scholar
Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 https://doi.org/10.1038/s41592-019-0619-0 (2019).
Article PubMed PubMed Central CAS Google Scholar
Aran, D. et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat. Immunol. 20, 163–172 https://doi.org/10.1038/s41590-018-0276-y (2019).
Article PubMed PubMed Central CAS Google Scholar
Blighe, K., Rana, S. & Lewis, M. EnhancedVolcano: publication-ready volcano plots with enhanced colouring and labeling. GitHub https://github.com/kevinblighe/EnhancedVolcano (2018).
Korotkevich, G., Sukhov, V. & Sergushichev, A. Fast gene set enrichment analysis. Preprint at bioRxiv https://doi.org/10.1101/060012 (2019).
Download references
Acknowledgements
We thank the members of the CNIC microscopy, genomics, cytometry and bioinformatics units for their expertise and support throughout the project. We also thank K. Kruse from MPI for Molecular Biomedicine for help with interpretation and presentation of single-cell data.
Funding
Research in R.B.’s laboratory was supported by the European Research Council (ERC) Consolidator Grant AngioUnrestUHD (101001814), the Ministerio de Ciencia, Innovación y Universidades (PID2020-120252RB-I00 and PID2023-148880OB-I00) and ‘La Caixa’ Foundation (HR22-00316 AngioHeart), all awarded to R.B. Research in T.M.’s laboratory was supported by the Knut and Alice Wallenbergs Foundation (2020.0057), the Swedish Research Council (2020-02692) and Jenny and Antti Wihuri Foundation. The CNIC is supported by the Instituto de Salud Carlos III (ISCIII), the Ministerio de Ciencia, Innovación y Universidades (MICIU) and the Pro CNIC Foundation. It is a Severo Ochoa Center of Excellence (grant CEX2020-001041-S funded by MICIU/AEI/10.13039/501100011033). Microscopy experiments were carried out at the CNIC Microscopy and Dynamic Imaging Unit, an ICTS-ReDib facility cofunded by MCIN (/AEI/10.13039/501100011033) and the ERDF ‘A way to build Europe’ (ICTS-2018-04-CNIC-16). Open access funding provided by Max Planck Society.
Author information
Author notes
These authors contributed equally: Irene Garcia-Gonzalez, Susana F. Rocha
Authors and Affiliations
Molecular Genetics of Angiogenesis Group, Centro Nacional de Investigaciones Cardiovasculares (CNIC), Madrid, Spain
Irene Garcia-Gonzalez, Susana F. Rocha, Maria S. Sanchez-Muñoz, Mariya Lytvyn, Aroa Garcia-Cabero & Rui Benedito
Department of Immunology, Genetics and Pathology, Uppsala University, Uppsala, Sweden
Filipa V. Oliveira & Taija Mäkinen
Wihuri Research Institute, Helsinki, Finland
Filipa V. Oliveira & Taija Mäkinen
Faculty of Medicine, University of Helsinki, Helsinki, Finland
Filipa V. Oliveira & Taija Mäkinen
Department of Functional Genetics, Max Planck Institute for Molecular Biomedicine, Münster, Germany
Ute Ipe & Rui Benedito
Endothelial Pathobiology and Microenvironment, Josep Carreras Leukaemia Research Institute, Barcelona, Spain
Mariona Graupera
CIBERONC, Instituto de Salud Carlos III, Madrid, Spain
Mariona Graupera
Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain
Mariona Graupera
Authors
- Irene Garcia-Gonzalez
- Susana F. Rocha
- Filipa V. Oliveira
- Maria S. Sanchez-Muñoz
- Mariya Lytvyn
- Aroa Garcia-Cabero
- Ute Ipe
- Mariona Graupera
- Taija Mäkinen
- Rui Benedito
Contributions
I.G.-G., S.F.R. and R.B. designed most of the experiments, interpreted results, assembled figures and wrote the manuscript. I.G.-G. performed confocal microscopy, FACS and scRNA-seq bioinformatic analysis, edited text and assembled figures. S.F.R. performed immunostainings, microscopy, FACS analysis, and image quantifications in Fiji and GraphPad; edited text and figures; and assembled figures. M.G. provided assistance with the Nr2f2 (also known as Coup-TFII) mouse model and related experiments. T.M. and F.V.O. bred mice and provided the samples for the Vegfr3KO and Vegfr2/Vegfr3KO phenotype analysis. M.L., M.S.S.-M., A.G.-C. and U.I. gave general technical assistance with experiments and genotyped the mouse colonies. All authors approved the final version of the manuscript.
Corresponding author
Correspondence to Rui Benedito.
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature thanks Kristy Red Horse and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Ratiometric analysis of the expansion and mobilization of Vegfr2KO and wild-type cells.
a, Representative FACs gating strategy and plots of CD31+ ECs collected from different organs (see quantitative data in Fig. 1b), showing YFP+ and Tomato+ cell frequencies, and their log2 ratio, in mice carrying the iFlpMTomato-Cre/MYFP allele in combination with the Apln-FlpO allele. b, Representative confocal images of cryosections from E13.5 and postnatal (P7) organs immunostained for Tomato, YFP, ERG and VEGFR2. Arrowheads point to Tomato+VEGFR2- ECS (white) and YFP+ VEGFR2+ ECs (yellow). Quantification chart showing depletion of VEGFR2 protein in Tomato+ ECs. c, Representative liver and heart confocal micrographs of Apln-FlpO, iFlpMTomato-Cre/MYFP -induced mosaic deletion of Vegfr2 (Vegfr2KO-E8.5-P7; n = 3 pups), showing that Tomato+ Vegfr2KO ECs locate preferentially to venous vessels (Endomucin+SMA+). Scale bars are 25 μm. Abbreviations: PV-Portal vein, CV-Central vein, A – artery and V- vein.
Source data
Extended Data Fig. 2 Pulse-chase analysis of Vegfr2KO ECs from E10.5 to P7.
a, Schematic showing the difference in the structure and Cre activity of standard Cre-reporters and iSuRe-HadCre. b, Schematic of the genetic approach and representative confocal micrographs of Vegfr2KO-E10.5-P7 ECs (orange arrowheads) and their frequency in the different organs at P7. c, Schematic of the genetic approach, FACS analysis and representative micrographs of Vegfr2KO-E10.5-P7 ECs (Tomato+,YFP+) and internal reporter control wildtype ECs (YFP+). Chart shows the ratio of mutant (Tomato+,YFP+) and wildtype (YFP+) cells in the distinct organs in comparison to the frequencies shown in the Brain. Data are presented as mean values +/− SD. For statistics see Source Data File 1. Scale bars are 50 μm.
Source data
Extended Data Fig. 3 Cellular digital mapping and heatmap generation.
a, Schematic of the iFlpMosaic and iSuRe-HadCre genetic systems used to induce and map genetic mosaics. Flow-chart depicting the main steps in image analysis used for the digital mapping of the identified cells, and the generation of heatmaps or XY dot plots, with an example of the analysis for a control retinal AV field with AplnFlpO, iFlpMTomato-Cre/MYFP or Cdh5-CreERT2, iSure-HadCre genetics. Definition of the 6 regions of interest, ERG-based segmentation (endothelial nuclei), nuclear-classification according to reporter or marker expression and relative distance of each nucleus to the AV limits are determined using a custom made semi-automatic FIJI script (steps i-iv). Once a minimum of 2000 cells per reporter group (Tomato+/−, YFP+) are analysed, data is then post-analysed in RStudio to determine the cell distribution of each reporter group in each region and the contribution of Tomato+ cells to each region when compared to control (YFP+ or Tomato- cells) (step v). XY dot plots (step vi) and heatmaps (step vii) are then generated. Note that dot plots (shown in vi) have a maximum of 2000 + 2000 mapped cells plotted as dots, to allow their spatial visualization, whereas heatmaps (shown in vii) are generated with maximal number of mapped cells (with equal cell number in control and mutant group).b,c, Representative confocal micrographs showing the loss of VEGFR2 protein (b), and its downstream target ESM1 (c), 24 h after 4-OHT administration. Data are presented as mean values +/−SD. For statistics see Source Data File 1. Scale bars are 25 μm.
Source data
Extended Data Fig. 4 Impact of VEGFR2 loss on ECS during different retinal angiogenesis waves.
a, Scheme depicting the experimental protocol and the development of the 3 retinal vascular layers from P1 to P20. Representative confocal micrographs of control and Vegfr2KO-P1-P20 retinas induced mosaically with Cdh5-CreERT2 and the iSure-HadCre reporter at P1 and analysed at P20. Charts with quantification of percentage of Tomato+ ECs and the EC density in the surface and deep vascular layer of the retina, showing that Vegfr2KO-P1-P20 cells contribute to the deeper layers (formed between P7 and P20) at the same percentage they are found in the initial surface layer (formed until P7), and similar to control cells. b, Scheme depicting the experimental protocol and the development of the 3 retinal vascular layers from P7 to P14. Representative Z-volume confocal micrographs of P14 retinas, after inducing full recombination with 4-OHT at P7. Even when induction is close to 100% (not mosaic), Vegfr2KO ECs can migrate from surface (P7) to form deeper retina layers from P7 to P14. c, Representative confocal micrographs of postnatal mice retinas, 3 days after inducing Vegfr2 deletion, immunostained for ERG (EC nuclei) and respective quantification chart. d, Representative confocal micrographs of control and Vegfr2KO-P4-P7 retinas immunostained for Ki67 and ERG. Magnified insets of vein showing Ki67+Tomato+ cells (yellow arrows) and Ki67-Tomato+ cells (white arrows), respectively. Chart with quantifications showing no significant difference in Ki67 positivity between control and Vegfr2KO-P4-P7 ECs. Data are presented as mean values +/− SD. For statistics see Source Data File 1. Scale bars are 100 μm. Abbreviations: A – artery and V- vein.
Source data
Extended Data Fig. 5 Impact of VEGFR2 loss on the proliferation of adult ECs.
Representative confocal micrographs of the indicated adult organs collected 2 weeks after induction with tamoxifen, and immunostained for Ki67 and ERG. Yellow arrows indicating Tomato+Ki67+ ECs. Chart with quantification of percentage of Ki67+,Tomato+ among all Tomato+ ECs. Data are presented as mean values +/− SD. For statistics see Source Data File 1. Scale bars are 25 μm.
Source data
Extended Data Fig. 6 scRNAseq analysis of Vegfr2KO-P4-P7 ECs.
a, Schematic of the genetics and cell sorting gating strategy to isolate the cells used to perform the scRNAseq contrasts shown in Fig. 3 and this figure. b,c, UMAP showing the different brain endothelial clusters and their proportions (b) and the expression of canonical tip cell marker genes in the activated cell cluster (c). d,e, Umaps and violin plots showing the expression of two genes known to be highly expressed by sprouting tip cells, and downregulated after Vegfr2 deletion in brain ECs. f, Violins showing a tip cell genes score based on the combined expression of several tip cell genes (PlexinD1, Kcne3, Esm1) in brain ECs. g, Violin plots showing expression of the terminal exon 30 (highly detected in the 3′10x genomics library sequencing) and the upstream deleted floxed exon 3, used to assess Vegfr2 deletion efficiency from P4 to P7 in lung ECs. Red negative values indicate fold changes between conditions. Green and red labels above the dots show the absolute average expression (not log). h, UMAP and accompanying histobar plot illustrating the distribution and proportion of lung endothelial cell clusters within each condition. i, Dot plot showing the expression of key marker genes for the different lung EC clusters. j, UMAP projection of the control and Vegfr2KO-P4-P7 lung ECs. The associated bar plot shows the log2 ratio of the number of Vegfr2KO-P4-P7 divided by the number of control cells per cluster. k, Volcano plot of showing top DEGs in lung ECs. Coloured genes correspond to highlighted pathways or processes of biological relevance. l, List of hallmark pathways deregulated in lung Vegfr2KO-P4-P7 ECs. m, Heatmap showing the average log2 fold change of apoptosis-related and tip-cell genes in lung ECs between Vegfr2KO (P4–P7) and control samples. n, Violin plots showing expression of the terminal exon 30 (detected in the 3′10x genomics library sequencing) and the deleted floxed exon 3, used to assess Vegfr2 deletion efficiency from P4 to P7. Note that in some cells there is still some residual Vegfr2 mRNA given the short-term deletion nature of the assay, or some contamination with mRNA from wild-type kidney ECs. Red negative values indicate fold changes between conditions. Green and red labels above the dots show the absolute average expression (not log). o, UMAP and accompanying histobar plot illustrating the distribution and proportion of kidney endothelial cell clusters within each condition. p, Dot plot showing the expression of key marker genes for the different clusters. q, UMAP projection of the control and Vegfr2KO-P4-P7 Kidney ECs. The associated bar plot shows the log2 ratio of the number of Vegfr2KO-P4-P7 divided by the number of control cells per cluster. r, Volcano plot of showing top DEGs. Colored genes correspond to highlighted pathways or processes of biological relevance. s, List of hallmark pathways deregulated in kidney Vegfr2KO-P4-P7 ECs. t, Heatmap showing the average log2 fold change of apoptosis-related and tip-cell genes in kidney ECs between Vegfr2KO (P4–P7) and control samples. For volcano plots and heatmaps (k, m, r, t), statistical analyses were performed using Seurat’s FindMarkers function, with the Wilcoxon rank-sum test and Bonferroni correction for multiple testing. For hallmark pathways (l, s), statistical analyses were performed using the multilevel Monte Carlo approach implemented in fgseaMultilevel (R package fgsea) and Benjamini–Hochberg (BH) FDR correction for multiple testing.
Extended Data Fig. 7 Clonal expansion analysis of control and Vegfr2KO ECs from P1 to P20.
a-d, Representative confocal micrographs of the indicated organs from mice induced at P1 and collected at P20. Besides the Cdh5-CreERT2 and the iSure-HadCre alleles, these animals have the iFlp-MosaicH2B-Cherry/GFP/Cerulean reporter allele (Garcia-Gonzalez et al.13), that enables sparse and multispectral labelling of nuclei within a fraction of the Tomato+/mutant cells, giving scClonal resolution. Charts with the quantification of clone size and its frequency. Individual nuclei labelled cells, or their clones, are depicted by dotted circles. Scale bars are 100 μm.
Source data
Extended Data Fig. 8 scRNAseq analysis of Vegfr2KO-E8.5-P7 and Vegfr2KO-2weeks ECs.
a, Schematic of the iFlpMosaic genetics and cell sorting gating strategy to isolate the cells used to perform the scRNAseq contrasts of Vegfr2KO- E8.5-P7 ECs shown in Fig. 4 and this figure (b-f). b, Violin plot and dot plots showing the expression levels of the last exon (exon 30) and the deleted floxed exon (exon 3) demonstrating Vegfr2 deletion efficiency. Negative numbers in red indicate the fold change between conditions. c, UMAP showing the identified EC clusters and histobar plots showing their proportions per condition. d, UMAP spatial distribution of control and Vegfr2KO-E8.5-P7 ECs, normalized by cell number. The bar plot shows the log2 ratios of Vegfr2KO-E8.5-P7 over control cells per EC cluster. e, Volcano plot depicting the top DEGs and also other biologically significant genes when contrasting control and Vegfr2KO-E8.5-P7 ECs. f, Top significantly up and downregulated hallmark pathways in Vegfr2KO-E8.5-P7 ECs when compared to control ECs. g, Schematic of the iSuRe-HadCre genetics and cell sorting gating strategy to isolate the cells from adult mice used to perform the scRNAseq contrasts shown in this figure. h-n, Adult heart ECs scRNAseq data comparison. o-u, Adult Lung ECs scRNAseq data comparison. h,o, Violin plots showing expression of the whole Vegfr2 mRNA. Red negative value indicate fold change between conditions. i,p, UMAP showing the identified EC clusters and histobar plots showing their relative proportions per condition. j,q, UMAP showing the spatial distribution of control and Vegfr2KO-2weeks ECs. The bar plot shows the log2 ratios of Vegfr2KO-2weeks ECs over control cells per EC cluster. k,r, Dot plot showing the top cluster marker genes. l,s, Volcano plot depicting the top DEGs and also other biologically significant genes when contrasting control and Vegfr2KO-2 weeks. m,t, Top significantly up and downregulated hallmark pathways in heart Vegfr2KO-2 weeks ECs when compared to heart control ECs. Highlighted hallmarks match colour code used in l and s. n,u, Heatmap showing the average Log2FC of the indicated genes in Vegfr2KO-2 weeks ECs when compared to control ECs. For volcano plots and heatmaps (e, l, n, s, u), statistical analyses were performed using Seurat’s FindMarkers function, with the Wilcoxon rank-sum test and Bonferroni correction for multiple testing. For hallmark pathways (f, m, t), statistical analyses were performed using the multilevel Monte Carlo approach implemented in fgseaMultilevel (R package fgsea) and Benjamini–Hochberg (BH) FDR correction for multiple testing.
Extended Data Fig. 9 The master regulator of venous development COUPTFII/NR2F2 is needed for the long-term expansion of Vegfr2KO-E8.5-P7 ECs.
a,b, Direct comparison between scRNAseq data of long- (Vegfr2KO-E8.5-P7) and short-term (Vegfr2KO-P4-P7) Vegfr2 deletion in heart and lung tissues showing increased venousization in long-term Vegfr2 deleted cells by dot or volcano plots. Note that the genetics of Vegfr2KO-P4-P7 and Vegfr2KO-E8.5-P7 ECs is different and hence the comparisons should be done with their independent respective genetic controls, as shown in Fig. 4h,l. c,d, Representative confocal micrographs of immunostained P7 mouse retinas and hearts (n = 3 pups) showing the upregulation of NR2F2/COUP-TFII specifically in Tomato+/Vegfr2KO-E8.5-P7 ERG+ ECs (white arrowheads) and not in control YFP+/ERG+ ECs (yellow arrowheads), and quantification chart of signal intensity per nuclei (white insets also shown in main Fig. 4o,p). e, Representative FACS gatings of CD31-APC+ ECs collected at P7 from the indicated organs (n = 3 pups), and corresponding to the chart numbers showed in Fig. 4q. Data are presented as mean values +/− SD. For statistics see Source Data File 1. Scale bars 100 μm (c) and 50 μm (d). For volcano plots (b), statistical analyses were performed using Seurat’s FindMarkers function, with the Wilcoxon rank-sum test and Bonferroni correction for multiple testing.
Source data
Extended Data Fig. 10 Vegfr2KO ECs can form tumor vessels.
a, Schematic of the iSure-HadCre, Cdh5-CreERT2 genetics used to analyse the ability of ECs with short-term (2–3 days before implanting the tumor) deletion of Vegfr2 to contribute to Lewis Lung carcinoma (LLC) tumor vessels. Quantifications show that upon short-term deletion, VEGFR2KO ECs(Tomato+ERG+) contribute less to intratumoral vessels when compared to vessels in extratumoral region of the tumor. b, Schematic of the iFlpMosaic genetics used to analyse the ability of ECs (ERG+ nuclei) with long-term (since embryogenesis) deletion of Vegfr2 to contribute to LLC tumor vessels. Quantifications show that upon long-term deletion, the disadvantage of VEGFR2KO ECs(Tomato+ERG+) to contribute to intratumoral vessels when compared to vessels in extratumoral region of the tumor, is smaller than upon short-term deletion. Note that the Apln-FlpO allele also drives the recombination of the iFlpMosaic allele in other cell types besides ECs, and hence the need to use ERG staining (labels the EC nuclei). Data are presented as mean values +/− SD. For statistics see Source Data File 1. Scale bar 250 microns.
Source data
Extended Data Fig. 11 Expression and functional mosaic genetics of angiogenic ECs lacking the function of VEGF receptors.
a, Table showing the average log2 Fold change in expression of VEGF ligands and receptors in Vegfr2KO-E8.5-P7 ECs when compared with Control E8.5-P7 ECs. b, Violins showing the expression of Vegfr1 mRNA in single ECs obtained at P7 from the indicated organs. c, Ratios of soluble and full-length Vegfr1 mRNA expression in the indicated organs single ECs (bulk analysis). d, Representative confocal micrographs of retinas from P7 mice with the indicated alleles. The iFlpMosaicH2B-Reporter allele enables the stochastic labelling of a small fraction of individual nuclei with distinct fluorescent proteins, showing that individual Vegfr1/2KO-P1-P7 ECs clonally expand in veins (n = 4 pups). e, Representative confocal micrographs of Vegfr1/2KO-P1-P7 retina immunostained for Ki67 and ERG. Magnified insets of vein and AF are shown. Yellow arrows indicate Tomato+Ki67+ cells. Chart with quantifications showing no significant difference in Ki67 positivity between control and Vegfr1/2KO-P1-P7 ECs in the indicated vascular areas (note that these mutant cells were rarely found at the angiogenic front, and never at the leading edge, and hence no data could be presented for this region). f, Representative confocal micrographs and quantifications of the % of Ki67+ ECs in the indicated organs of Vegfr1/2KO-P4-P7 mice. Arrows indicate Tomato+Ki67+ cells (yellow) or Tomato+Ki67- (white). g, Representative confocal micrograph of Vegfr1/2KO-P4-P7 retinas immunostained with ERG and respective quantification charts, showing a further decrease in EC/vascular density when compared to Vegfr2KO-P4-P7 retinas. h, Representative confocal micrographs of Vegfr1/2KO-P4-P7 retinas immunostained for cleaved caspase 3 (CC3), and respective quantification chart showing a similar increase in EC apoptosis to Vegfr2KO-P4-P7 ECs. i, Newborns survival plot showing that double deletion of Vegfr1 and Vegfr2 induced at P1, leads to 100% lethality of pups before P20. j, Representative confocal micrograph of retinas with mosaic deletion of Vegfr1 in ECs from P1 to P7. Heatmap, XY dot plot and chart showing the position and relative % of Tomato+ (KO) and Tomato negative (WT) cells. k, Representative confocal micrographs of retinas from Vegfr1 heterozygous floxed (flox/wt; n = 3 pups) and Vegfr1 homozygous mice, with complete (n = 6 pups) or mosaic deletion (n = 5 pups) in ECs induced at P1 with 4-OHT. Note the much higher expression of ESM1 in Tomato+ ECs when Vegfr1 is deleted in all/most ECs, but not when deleted mosaically. l, Vegfr3 floxed gene structure and the guide RNAs sequences used to produce it. m, Validation of Vegfr3 deletion in retina ECs by immunostaining (n = 3 pups). Note the presence of VEGFR3 protein in angiogenic front Tomato negative (EC nuclei ERG+) cells. n, Representative confocal micrograph of retinas with mosaic deletion of Vegfr3 in ECs from P1 to P7. Heatmap, XY dot plot and chart showing the position and relative % of Tomato+ (KO) and Tomato negative (WT) cells. Data are presented as mean values +/−SD. For statistics see Source Data File 1. Scale bar 100 microns. Abbreviations: LE – leading edge, AF – angiogenic front, IntCap – intermediate capillaries, MA – mature area, A – artery, V- vein.
Source data
Extended Data Fig. 12 Schematic of the main findings.
a, Single cell RNAseq and histological analysis of Vegfr2KO-P4-P7 ECs (gene deletion induced at P4 and analysis at P7) in distinct organs revealed common but also distinct transcriptional programs and functions for VEGFR2. b, Loss of VEGFR2 affects the migration of ECs towards VEGF produced by hypoxic tissues, during the first wave of angiogenesis. This leads to the accumulation of Vegfr2KO ECs in more mature vessels and veins, which provides for a continuous source of mutant ECs during subsequent waves of angiogenesis derived from veins. The additional loss of VEGFR1 compromises capillary angiogenesis, whereas the additional loss of VEGFR3 compromises arterialization. c, The comparison of scRNAseq data from ECs with short-term (P4 to P7) and long-term (embryonic to P7) deletion of Vegfr2 revealed that over time Vegfr2KO ECs undergo significant transcriptional changes and upregulate the expression of many venous genes (venousization), particularly Nr2f2. d, The upregulation of Nr2f2, and venousization, enables Vegfr2KO-E8.5-P7 ECs to compete with wildtype ECs over time, and form veins and many other venous-derived vessels, such as capillaries, tip cells and even arteries. Note that these cells adapt over time to the loss of VEGFR2 and become more capable of angiogenesis than Vegfr2KO-P1-P7 ECs. The additional loss of the master regulator of vein specification Nr2f2, prevents the venousization and growth of Vegfr2KO ECs.
Supplementary information
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and permissions
About this article
[Check for updates. Verify currency and authenticity via CrossMark]
Cite this article
Garcia-Gonzalez, I., Rocha, S.F., Oliveira, F.V. et al. Venousization enables angiogenesis in the absence of VEGFR signalling. Nature (2026). https://doi.org/10.1038/s41586-026-11224-w
Download citation
Received: 20 June 2025
Accepted: 29 September 2026
Published: 04 November 2026
Version of record: 04 November 2026
DOI: https://doi.org/10.1038/s41586-026-11224-w