Single-cell RNA-sequencing reveals key markers in endothelial populations

Previously conducted single-cell RNA-sequencing on the LinnegCD34+ compartment of the aortae of three C57Bl/6 mice was re-analyzed to illuminate highly expressed genes in endothelial and mesenchymal clusters of interest (Supplementary Fig. 1a) [25]. To better capture the transcriptomic heterogeneity among endothelial subpopulations, we applied a higher resolution when performing unsupervised clustering. More distinctive LinnegCD34+ cell subclusters emerged, allowing the analysis of gene expression within previously delegated clusters (Fig. 1a).

Fig. 1

Single-cell RNA-sequencing and flow cytometry reveal endothelial protein C receptor (PROCR) and platelet derived growth factor – A (PDGFRA) as key markers in endothelial populations. a UMAP plot shows different subclusters of endothelial (LinnegCD34+) cells from murine aorta (n = 3). Endovascular progenitors (EVP), differentiated (D) and mesenchymal (M) cell clusters were highlighted for our clusters of interest. b Expression levels of genes of interest overlaid on the UMAP plot highlighting the cluster specificity of their expression. The expression levels are shown as log2(counts + 1)-transformed values, and the gradient represents low (purple) to high (yellow) values. c Heat map showing top 30 markers (15 UP; 15 DOWN) for EVP (Cluster 4) vs. M (Cluster 1/2/3) vs. D. d Dot plot showing relative expression of genes of interest across clusters. The size of dots represents the percentage of cells in each cluster that have non-zero expression of each gene. The color gradient indicates mean expression level of all cells in each cluster. e Representative flow cytometry dot plots showing the gating strategy used to isolate the endothelial hierarchy. Endothelial cells were gated as Lineage (Lin)negVE-cadherin+. From this population, EVP cells were gated as the CD31neg/loCD34+ population whereas D cells were gated as the CD31+CD34+ population. f, g Representative flow cytometry dot plots showing expression of Procr, CD157, Abcg2-YFP, and Sox18-YFP in (f) EVP and (g) D cells. h Quantification of expression of cell surface markers (i) Procr (*** p = 0.0008; n = 5), (ii) PDGFRα (*** p = 0.0005; n = 5), (iii) CD157 (n = 2 descriptive observation), (iv) Abcg2 (*p = 0.0490; n = 3), and (v) Sox18 (*,p = 0.0452; n = 3) in EVP and D cells i Representative flow cytometry dot plots showing alternative gating strategy where live cells are first gated as PROCR+PDGFRA+, followed by LinnegVE-cadherin+ and finally gated as EVP and D cells based on CD31 and CD34 expression. j Quantification of percent of EVPs gated using gating strategy in (i) (****, p < 0.0001, n = 3). Data are presented as mean ± SD. Statistical analysis was performed using paired t-test

Clusters 0, 10, and 12 were characterized as mature differentiated endothelial (D) cells based on expression of pan-endothelial markers classically used to define the endothelial compartment including Pecam1 (CD31) and Cdh5 (CD144), confirmed using SingleR labeling analysis (Fig. 1a and b (i-ii), Supplementary Fig. 1b). Clusters 1, 2, and 3 showed upregulation of mesenchymal markers leading to their designation as mesenchymal (M) clusters, whereas cluster 4 showed expression of both mesenchymal and endothelial markers, leading to the designation of this cluster as the putative endovascular progenitor (EVP) population (Fig. 1a). Neither the putative mesenchymal or EVP groups possessed expression of hematopoietic cell markers (confirmed with SingleR, data not shown).

Cluster 4 maintained key endothelial marker expressions such as Cdh5, Pecam1 or CD34 although at a lower level compared to differentiated endothelial cells (Supplementary Fig. 1). Among candidate progenitor genes studied, Cd157 (Cluster 10), Sox18 (Cluster 0, Cluster 10), and Abcg2 (Cluster 0, Cluster 12) showed significant upregulation (adj p < 0.001) in the differentiated endothelial cell clusters (Fig. 1b (iii-v)) while Procr (Cluster 4) and Pdgfrα showed expression in EVP cluster 4, with mesenchymal marker Pdgfrα being most upregulated in the mesenchymal M cell clusters (Cluster 1, Cluster 2; Fig. 1b (vi-viii)). Heat map analysis of top 30 markers in EVP cluster 4 showed upregulation of genes expressed in endothelial cells (Fig. 1c). Pathway analysis and dot plots of genes of interest conducted on cluster 4 to characterize differentially expressed (DE) genes showed enrichment of angiogenic, TGFB- and Wnt-signaling pathways, and genes related to both vascular maintenance and development (Sema3c, Tmem100, Mfap5) and mesenchymal populations (Igfbp5, Igfbp6, Pcolce2), potentially indicating that the population maintains an identity balanced between endothelial and mesenchymal states (Fig. 1d, and Supplementary Fig. 1c-f). Other major clusters comprised non-endothelial cell populations, including some remaining hematopoietic cells not depleted during the flow sort particularly T cells (C5; Cd3g, Cd3d, Cd3e), B cells (C6; Cd19, Cd79a) but also epithelial cells (C7; Krt19, Krt18, Krt7), pericytes (C8; Rgs5, Pdgfrb), platelets (C9; Pf4, Gp1bb), and cardiomyocytes (C11; Myl7, Myoz2) (Fig. 1a).

EVPs highly express PROCR and PDGFRA

To narrow a true EPC population, flow cytometry was performed on adult C57BL/6 mouse aortae using the markers identified in above single-cell RNA sequencing analysis in conjunction with the previously used markers to characterize EVPs [25]. From total aorta cells, the endothelial hierarchy was segregated based on cell surface marker profiles and the original gating strategy outlined in Patel et al. [13]. Putative EVPs were segregated as LinnegVE-cadherin+CD34+CD31neg/lo, and D cells as LinnegVE-cadherin+CD34+CD31+ (Fig. 1e). EVP and D cells were then further evaluated for expression levels of putative endothelial progenitor markers PROCR [16], PDGFRA [20], and CD157 [22] (Fig. 1f-g, Supplementary Fig. 2a). Procr and Pdgfra were 1.4-fold (p < 0.001) and 3.3-fold (p < 0.001) more frequently expressed in EVPs as compared to D cells, respectively, whereas CD157 was 5.9-fold (descriptive observation.) more highly expressed in D cells than in EVPs (Fig. 1f-h, Supplementary Fig. 2b). In addition to these markers, two additional mouse strains were used based on the studies described above to test for differences in Abcg2 [33] and Sox18 [13] between EVP and D cells.

Flow Cytometry analysis was performed on aortae of adult Abcg2-Ires-CreERT2/ROSA-EYFP [33] and Sox18-Cre ERT2/ROSA-EYFP [13] mice treated with tamoxifen for 5 consecutive days. Characterization of EVP and D populations based on YFP expression showed that Abcg2 and Sox18 were 2.3-fold (p < 0.05) and 4.3-fold (p < 0.05) more frequently expressed in D cells compared to EVPs, respectively (Fig. 1f-h).

These findings suggested that PDGFRA and PROCR were additional markers that could enrich progenitors within the already described EVP population. Interestingly, among EVP cells, an average of 78.04% were PROCR+ and 82.28% were PDGFRA+, suggesting that these markers may allow refining of the progenitor definition. Importantly, an alternative gating strategy on live aortic cells co-expressing both PROCR and PDGFRA showed that an average of 91.43% (** p = 0.0012) were LinnegVE-cadherin+, and from here a further 94.97% (**** p < 0.0001) were CD34+CD31neg/lo, corresponding to EVPs (Fig. 1i; n = 3). This demonstrates the powerful ability of PROCR and PDGFRA co-expression alone to label the same population as the classic EVP gating strategy to a high degree of confidence. Similarly, the large overlap between PROCR and PDGFRA among EVPs allowed us to use only one marker at a time to examine functional characteristics.

Given the significant overlap of three distinct strategies to identify progenitor cells in the endothelium, we henceforth called this population a refined endothelial progenitor cell (rEPC) population and proceeded to its functional analysis. To avoid ambiguity, we explicitly define rEPCs as non‑hematopoietic, endothelial‑restricted progenitors, distinct from classical EPCs described in Asahara et al. The use of single-cell RNA-seq suggested that cluster 4 was representative of this cell population. Differential gene expression defining this cluster included the expression of both major endothelial and mesenchymal genes as seen in the analyses of top differentially expressed genes from this cluster including Procr and Pdgfra (Supplementary Fig. 1c, 1e).

rEPCs show increased endothelial colony formation capacity in vitro and increased engraftment potential in vivo

To analyze the functional capacity of rEPCs versus other EVP and D cell populations, PROCR+ EVPs (rEPCs), PROCRneg EVPs, PROCR + D cells, and PROCRneg D cells were sorted from adult C57Bl/6 aorta and cultured in Matrigel for colony forming assay (Fig. 2a). The percentage of colonies formed in each condition based on number of wells plated with all experiments normalized to 10 cells per well was calculated (n = 14; * p < 0.05). rEPCs possessed the greatest colony formation capacity with a mean of 18% of wells plated per mouse displaying colonies, followed by 8% of PROCRneg EVP wells (Fig. 2c i-ii; *, p < 0.05). Among the colonies formed, two major morphologies were seen at day 12: a classic endothelial morphology (Fig. 2b (i)) with positive expression of endothelial marker as revealed by Griffonia (Bandeiraea) Simplicifolia Lectin I (GSL I, BSL I) (Isolectin) (IF; Fig. 2b (iii)), and an elongated morphology (Fig. 2b (ii)) that was not labeled by Isolectin (Fig. 2b (iv)).

Fig. 2

Refined endothelial progenitor cells (rEPCs) show increased endothelial colony formation capacity in vitro and engraftment potential in vivo compared to other populations. a Endothelial cells from C57Bl/6 aorta were FACS sorted based on cell surface expression of PROCR as depicted in graphical representation. Briefly, the live lineage (Lin)negV-CADH+CD34+ cells gate first from which CD31neg/loPROCR+ rEPC, CD31neg/loPROCRneg endothelial vascular progenitors (EVPs), CD31+PROCR+ differentiated (D) cells, and CD31+PROCRneg D cells were sorted. b Representative brightfield images (i-ii) depicting endothelial (i) and elongated (ii) cell morphology types at day 12 (4x magnification; scale bar = 500 μm) and immunofluorescence (IF) staining (iii-iv) of endothelial (iii) and elongated mesenchymal like (iv) colonies on day 12 showing expression of Isolectin BSL-I (10x magnification; scale bar = 150 μm). c Percentage of colonies formed in each condition based on number of wells plated with all experiments normalized to 10 cells per well (n = 14; * p < 0.05). d Graphical representation depicting experimental procedure for in vivo collagen gel engraftment assay. e Representative images of collagen gels containing (i) rEPCs or (ii) PROCRneg EVPs FACS-sorted from CAG-EGFP mice aortae collected following 7 days of implantation in NOD-scid Il2rynullB2mnull (NSG) mice (scale bars = 500 μm). f Percentage of GFP+ area of each gel upon collection measured via IF (* p < 0.05). g Representative IF images of sections from (i) rEPC and (ii) PROCRneg EVP collagen gels collected after 7 days and stained with DAPI, GFP, CD34 and Isolectin showing colocalization of GFP, Isolectin and CD34 (scale bars = 250 μm) and (iii) rEPC showing colocalization of GFP, ERG and CD34 (60x). a and d created with BioRender.com. Data are presented as mean ± SD. Statistical analysis was performed using Friedman one-way ANOVA (c) and unpaired t-test (f)

rEPCs formed exclusively isolectin + colonies with endothelial morphology, while PROCRneg EVPs formed isolectin negative colonies with an elongated morphology. The ability of rEPCs to give rise to endothelial cells reflect rEPCs as true progenitor cells. Importantly, D cells (PROCR+ or PROCRneg) never formed endothelial colonies based on positive staining for Isolectin.

To distinguish the potency of rEPCs and PROCRneg EVPs, these populations were challenged with a more stringent in vivo assay where collagen gels containing no cells (control), 100 rEPCs, PROCRneg EVPs, or total D cells, respectively, from CAG-EGFP mice were transplanted into the dorsal flanks of NOD-scid-Il2rynullB2mnull (NSG) recipients (Fig. 2d). Whole mount images of collagen gels after 7 days revealed that rEPCs had the highest engraftment potential with 9/15 gels engrafting and an average of 5.7% GFP+ area per plug (Fig. 2e (i), 2f), while only 2/10 PROCRneg EVP gels engrafted with an average of 0.8% GFP+ area per plug (Fig. 2e (ii), 2 F; p < 0.05). D cells were never able to engraft (0/6 gels), showing 0% GFP+ area, identical to the results of the gels containing no cells. IF staining was conducted on sections of these gels to further characterize the cells that had engrafted. GFP+ rEPCs co-expressed endothelial markers CD34, ERG and Isolectin while no overlap of these endothelial markers was seen with GFP+ cells from collagen gels containing PROCRneg EVPs (Fig. 2g). These findings more robustly pointed to key functional differences between EVP and D cells as reported [34, 35], but more remarkably between rEPCs expressing PROCR and EVPs devoid of PROCR.

rEPCs form a niche in the thoracic aorta and display increased clonogenic capacity

In order to confirm that aortic endothelial cells express PROCR and to find their anatomical distribution in situ, aortae from Cdh5- CreERT2/ROSA-EYFP mice, treated with tamoxifen to label endothelial cells with YFP, were harvested for ex vivo analysis. Immunofluorescence staining showed greater co-expression of PROCR and YFP in the thoracic aorta (68.18%; Fig. 3a, c and d) as compared to the abdominal aorta (21.10%; Fig. 3b and d; **, p = 0.005, n = 5).

Fig. 3

Refined endothelial progenitor cells (rEPCs) form a niche in the thoracic aorta displaying increased clonogenic capacity. a–c Aortae collected from Cdh5-CreERT2/ROSA-EYFP mice, opened and rolled lengthwise. a–b Representative images of sections of (a) thoracic and (b) abdominal aorta. c Section of thoracic aorta from A zoomed to 60x magnification and abdominal aorta from b; white arrows indicate regions of overlap between DAPI, PROCR, and YFP. d Quantification showing percent of PROCR+ length in abdominal and thoracic aorta (** p = 0.005; n = 5). e Representative brightfield image of colony grown from thoracic aorta of Zs-Green/ROSA-EYFP mice in Matrigel following 12 days; scale bar = 500 μm. f Quantification of number of wells that grew colonies from thoracic and abdominal aorta (**, p = 0.0052, n = 13). Data are presented as mean ± SD. Statistical analysis was performed using paired t-test

Next, we investigated if spatial difference of PROCR expression in the aorta [36] also correlated with a spatial difference in functionality in terms of clonogenic capacity. LinnegCD31loCD34+YFP+ EVPs from both the thoracic and abdominal aortae of Cdh5- CreERT2/ROSA-ZsGreen mice were FACS-sorted and plated in Matrigel® to compare the colony forming capacity between the two populations without biasing based on PROCR expression. Interestingly, 7/13 wells containing YFP+ EVPs from the thoracic aorta formed classical endothelial colonies while 0/13 wells from the abdominal aorta formed colonies (Fig. 3e-f; **, p = 0.0052, n = 3) confirming that the thoracic aorta is enriched for rEPCs.

rEPCs form differentiated endothelial cells in vivo in homeostasis and injury

PROCR and mesenchymal marker PDGFRA showed a high degree of overlap in rEPCs in flow cytometry and scRNA-sequencing results. Therefore, Pdgfrα-MerCreMer/Rosa-EYFP mice were used to trace the fate of rEPCs. Animals were administered with tamoxifen to label PDGFRA-expressing cells permanently with YFP and trace this population in tissues of interest over time. Although the YFP+ cells could contain fibroblasts and other populations of mesenchymal origin, flow cytometry analysis of a whole adult homeostatic aorta after a short pulse of tamoxifen, confirmed that the LinnegPDGFRA(YFP)+ subpopulation of aorta largely consisted of EVPs rather than differentiated endothelial D cells (Fig. 4a and 91% compared to 3%, n = 4, p < 0.0001).

Fig. 4

Refined endothelial progenitor cells (rEPCs) from Pdgfra-MerCreMer/Rosa-YFP differentiate into mature endothelial differentiated (D) cells in homeostatic aorta. a (i-ii) Representative flow cytometry dot plots of PDGFRα-MerCreMer/Rosa-YFP aorta cells showing gating strategy to identify CD34+CD31neg endovascular progenitor (EVP) and CD34+CD31+ D cells from Lineage (Lin)negYFP+ population (iii) Percent of EVP and D cells in the LinnegYFP+ fraction of the adult homeostatic aorta (****, p < 0.0001, n = 4). b Representative image showing immunofluorescence staining of YFP+ rEPCs cultured for 12 days from PDGFRα-MerCreMer/Rosa-YFP mice (scale bar = 500 μm). c Experimental schematic for aortic lineage tracing where mice are injected with tamoxifen for 5 days prior to commencement of experiment (day (D)0) and collection of tissues at specified timepoints (denoted with a red X). d Representative immunofluorescence staining of aorta at D1 (i) and D84 (ii) (large image zoomed in image scale bars = 250 μm, higher magnification images = 10 μm) post cessation of tamoxifen. e Representative flow cytometry dot plots showing LinnegYFP+ compartment of homeostatic aorta of 4 week old mice injected with tamoxifen changes in proportions of EVPs (e (ii-iv) shown in red oval) and D cells (e (ii-iv) shown in black oval) between D1 and D84. f Quantification of EVP and D cells in the LinnegYFP+ fraction of the aorta between D1 and D84 (** p < 0.01, *** p < 0.001; n = 4). Data are presented as mean ± SD. Statistical analysis was performed using unpaired t-test (a (iii) and ordinary one-way ANOVA (f)

To confirm that the population being traced in this model was indeed the same endothelial population as previous studies, YFP+PROCR+ EVP colonies were cultured from both Cdh5-CreERT2/ROSA-EYFP and Pdgfrα-MerCreMer/Rosa-EYFP aortae. The colonies formed from both models showed no difference morphologically or phenotypically and expressed endothelial markers Isolectin and ERG in vitro (Fig. 4b).

Few studies in the past have been able to identify a single Cre reporter system distinguishing progenitors from differentiated cells given the large overlap in markers. The large differential expression of Pdgfra in the aorta between rEPC and differentiated endothelial cells (D cells) provided a unique opportunity to demonstrate the lineage relationship between endothelial populations. Briefly, 4 week-old Pdgfrα-MerCreMer/Rosa-EYFP mice were administered with tamoxifen to label PDGFRA-expressing cells with YFP (Fig. 4c). Homeostatic aortas were then assessed from juvenile age to adulthood to trace the fate of YFP+ cells. Immunofluorescence staining of aorta at day 1 post-tamoxifen (D1) showed YFP+ cells in the intima co-expression of ERG and to some extent CD31 (Fig. 4d (i)). This further showed that at least a fraction of PDGFRA-expressing cells labelled by YFP are endothelial as in intimal position and not simply in the mesenchymal layers of the aorta. To examine whether these endothelial cells were rEPCs or of any other subpopulation, we performed flow-cytometry on D1 aorta revealing that all LinnegYFP+ cells resided in the rEPC with nearly no fully differentiated D cells (Fig. 4ei and ii, Supplementary Fig. 2c). The fate of D1 labelled pdgfra-expressing cells was further examined at D28 and D84. Co-expression of endogenous YFP with CD31 and ERG in the intima at D84 confirmed endothelial fate of these cells and could be identified in some areas of the aorta as patches interrupted by unstained cells, (Fig. 4d (ii)). This result was validated quantitatively using flow cytometry and the percentage of rEPCs between D1 and D84 ranged from an average of 70–88% of the LinnegYFP+ compartment, while the percentage of D cells increased significantly from 0.74% at D1 to 4.67% at D84 (Fig. 4e-f; n = 5, ** p < 0.01, *** p < 0.001) suggesting that some rEPCs labelled at D1 gave rise to D cells by D84.

To analyze this lineage relationship in the context of injury, Pdgfra-MerCreMer/Rosa-EYFP mice were treated with tamoxifen before performing full skin excisional wounds at D0 (Fig. 5a). The immunofluorescent staining of the wounds at D1 showed YFP expression on isolated cells in the center of the wounds with no expression of mature endothelial markers as expected (Fig. 5b). This was further confirmed by flow cytometry showing that YFP labelled cells were mostly rEPC (CD31low/negCD34+) or mesenchymal (CD31negCD34neg) and no D cells. Co-labelling of YFP with CD31 and ERG at D5 demonstrated that YFP+ cells have differentiated into mature endothelial cells (Fig. 5c).

Fig. 5

Refined endothelial progenitor cells (rEPCs) from Pdgfrα-MerCreMer/Rosa-YFP differentiate into mature endothelial differentiated (D) cells in an injury model of full-skin excisional wounds. a Experimental design for full-skin excisional wound lineage tracing where mice are injected with tamoxifen for 5 days prior to full-skin excisional wounding at day (D)0 and wound site tissue collected at specified timepoints (denoted with a red X). b-c Representative immunofluorescence stained images showing skin wound section from D1 (b) and D5 with endothelial markers (c). d Representative flow cytometry dot plots showing lineage (Lin)negYFP+ compartment of full-skin excisional wounds from adult mice changes in proportions of endovascular progenitors (EVPs) (d (ii-iv) shown in red oval) and D cells (e (ii-iv) shown in black) between D1 and D5. f Percent of EVP and D cells in the LinnegYFP+ fraction of the wounds between D1 and D5 (* p < 0.05, ** p < 0.01; n = 7). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA (e(i)) and Kruskal-Wallis test (e (ii))

Flow cytometry analysis at each time point confirmed that the percentage of rEPCs between D1 and D5 ranged from an average of 29.43–43.29% of the LinnegYFP+ compartment, while the percentage of D cells increased significantly from an average of 0.0% at D1 to 1.2% at D5 (Fig. 5d-e, Supplementary Fig, 2d, n = 7, * p < 0.05; ** p < 0.01). Overall, making use of PDGFRA expression as a reporter of rEPCs allowed tracing the fate of this population to demonstrate its contribution to differentiated endothelial cells both in homeostatic aorta and skin wounds.

PROCR is expressed in human aorta scRNA-seq data and leads to increased clonogenic capacity in a human term placental model of ECFCs

To identify rEPC equivalent population in human tissues, publicly available human normal aorta single-cell RNA-sequencing data was reanalyzed [27]. Data from 3 normal aortae samples was re-clustered after filtering doublets and before removing hematopoietic clusters based on known marker expression and SingleR labeling analysis (Fig. 6a, Supplementary Fig. 3a-b). The remaining clusters were identified based on SingleR labelling as primarily mesenchymal (M). The human counterparts to the rEPC cluster specifically were identified as clusters 3 and 12 as they had high degree of overlap between top differentially expressed (DE) genes in these clusters and rEPC cluster 4 in the murine sc-RNA seq dataset (Fig. 6b, Supplementary Fig. 3b). These overlapping genes were found to be implicated in endothelial, mesenchymal, extracellular matrix, and cell cycle pathways, indicative of genes maintaining a population between endothelial and mesenchymal states (pathways evaluated using EnrichR) [37]. Other genes listed play essential roles in endothelial identity or regulation of mesenchymal transition and fibrosis. Markers of interest outlined above were then analyzed in remaining endothelial, M cell, and EVP-like clusters using FeaturePlots, Dot Plots of top DE genes, and pathway analysis (Fig. 6c, Supplementary Fig. 3c-e).

Fig. 6

PROCR is expressed in human aorta scRNA-seq data and leads to increased clonogenic capacity in a human term placental model of ECFCs. a Single-cell RNA-sequencing data from human control aorta, clustered and filtered to remove hematopoietic clusters and label primary populations (n = 3). b Categorization of overlapping genes from top 100 DE genes of clusters 3 and 12 from human normal aorta dataset and murine aorta sc-RNA seq endovascular progenitor (EVP) cluster 4. c Markers of interest shown across clusters. d Representative flow cytometry plots showing the gating strategy of PROCR+/neg endothelial colony forming cells (ECFCs) from human term placenta. e Brightfield images of colonies growing from cultured PROCRneg (i) or PROCR+ (ii) ECFCs (scale bar = 200 μm). f Quantification of percentage of (i) well with colonies and (ii) colony types formed from cultured PROCRneg or PROCR+ ECFCs. g Immunofluorescent staining of PROCR+ colony at passage (P)6. Data are presented as mean ± SD. Statistical analysis was performed using unpaired t-test (f (i)) and 2 way ANOVA (f(ii))

Major endothelial genes PECAM and CDH5 (Fig. 6c (i and ii)) were most highly expressed in endothelial (D cell) cluster 8 along with SOX18 and ABCG2 and (Fig. 6c (iii and iv)). CD34, PROCR, and PDGFRA were expressed in multiple clusters but were highest in EVP clusters 3 and 12 as well as D cell cluster 8 (Fig. 6c (v-vii)).

Upon confirmation that the expression of markers of interest in human control aorta using scRNA-seq resembled the expression seen in mouse models described previously, functional assays were conducted to investigate whether progenitor capacity was increased in human cells expressing these markers as seen in murine studies. Placental CD34+CD45neg cells were FACS-sorted as described previously [32], with the additional gating of PROCR+ vs. neg for each population of varying CD31 expression (negative, low, intermediate, and high) (Fig. 6d, Supplementary Fig. 3f) and cultured on collagen coated plates. In particular, colony formation of the CD31lo and CD31int populations were examined as both have been shown in the past to contain progenitors that give rise to endothelial colony forming cells (ECFCs) [9]. Given both the CD31 and CD144(Cdh5) exhibit highly overlapping expression patterns across human placental endothelial subsets, including endothelial progenitor [9], CD31 alone was used to sort different endothelial populations.

Across all donors, CD31intPROCR+ cells had 90% more colony forming capacity as compared to CD31intPROCRneg cells (15%) (Fig. 6f (i), n = 3 placentas). Interestingly, 80% of CD31intPROCR+ colonies continually expanded and reached the size of high proliferative potential (HPP; >1000 cells) colonies (Fig. 6F (ii)). These HPP colonies were then passaged and showed the ability to continually expand through P6 (Fig. 6g). Moreover, IF staining of these colonies confirmed the endothelial nature of this population with positive expression of CD31 and VE-cadherin (Fig. 6g). Contrarily, CD31intPROCRneg cells formed colonies which only grew to < 50 cells before dying, therefore classifying them as endothelial clusters (EC; Fig. 6e (ii), n = 3 placentas).