CD146 is expressed on blood capillaries in normal human skin

To investigate the distribution of CD146 on the dermal microvasculature of normal human skin in situ, we analyzed its expression by double immunofluorescence staining with two specific endothelial markers, CD31 or Plasmalemma vesicle-associated protein (PLVAP, also known as PAL-E, PV-1 and FELS) (Fig. 1). We observed co-localization of CD146 with the majority of CD31 positive capillaries (Fig. 1A-C), while only a small fraction of CD31 positive capillaries were negative for CD146 (Fig. 1A, white arrow, Fig. 1B, inset). A high density of CD146/CD31 double positive capillaries was detected in the superficial capillary plexus of the papillary dermis (Fig. 1A, B), although CD146 was also observed in large vessels of the deep plexus of the reticular dermis (Fig. 1C). In addition, CD146 positive cells were found around CD31 positive endothelial cells surrounding the capillary wall (Fig. 1A-C), indicating that CD146 was expressed in both endothelial and perivascular cells in human skin.

Fig. 1

CD146 expression in normal human skin. A, B, C Representative immunofluorescence images of CD146 and CD31 of normal human skin. Co-localization between CD146 and CD31 is observed on the majority of capillaries, in both upper and lower dermis. A small fraction of CD31 positive capillaries results negative for CD146 (arrow and inset). Single positive cells for CD146 are observed around CD31 positive endothelial cells. Scale bar: A, B: 100 µm, C: 50 µm, inset: 20 µm. D Double immunofluorescence analysis of CD146 and PLVAP shows that CD146 co-localizes with PLVAP-positive endothelial cells and is expressed in the surrounding perivascular cells. Scale bar: 50 µm. E Triple immunofluorescence staining for CD146, PLVAP and CD31 confirms the co-localization between CD146, PLVAP and CD31. A fraction of CD31+CD146−PLVAP− is observed, probably corresponding to lymphatic capillaries (asterisk). (n = 3 independent donors). Scale bar: 50 µm. White dashed lines indicate the epidermal-dermal junction

Fig. 2

Lymphatic capillaries are negative for CD146 expression. A, B, C Lymphatic and blood capillaries are discriminated as CD31 and Podoplanin/Lyve1/Prox1 double positive and CD31 positive Podoplanin/Lyve1/Prox1 negative, respectively. Representative immunofluorescence images show no expression of CD146 on lymphatic capillaries positive for Lyve1 (A), Podoplanin (B) and Prox1 (C, inset). CD146 expression is detected only on blood capillaries and their related perivascular cells. (n = 3 independent donors). Scale bar: 50 µm. White dashed lines indicate the epidermal-dermal junction

The observed small fraction of CD146-negative capillaries (Fig. 1A, arrows, Fig. 1B, inset) suggested that this adhesion molecule is not uniformly expressed in the total skin microcirculation. Since CD31 does not discriminate between blood and lymphatic capillaries, we investigated the expression of CD146 using the specific blood vessel marker PLVAP [22, 23]. First, we confirmed the expression of PLVAP on blood capillaries, which is mutually exclusive with that of the lymphatic capillary markers, Podoplanin, Lyve1 and Prox1 (Supplementary Fig. S1A-D). Indeed, these markers can be interchangeable to discriminate and/or isolate lymphatic endothelial cells (LECs) [18, 24]. We showed that CD146 localized in both PLVAP-positive endothelial cells and surrounding perivascular cells (Fig. 1D, inset). Triple immunofluorescence staining confirmed the co-expression of CD146 with CD31 and PLVAP (Fig. 1E) and highlighted CD31+CD146−PLVAP− cells corresponding to lymphatic capillaries (Fig. 1E, asterisk). The individual channels of each staining are reported in Supplementary Fig. S1E.

Further, we definitely excluded the expression of CD146 on lymphatic skin capillaries. Lymphatic and blood capillaries were identified as CD31 and Podoplanin/Lyve1/Prox1 double-positive and CD31 positive Podoplanin/Lyve1/Prox1 negative, respectively. All the lymphatic capillaries alternatively stained for Podoplanin, Lyve1 or Prox1 showed no CD146 expression, which was instead evident only in blood capillaries and in their surrounding perivascular cells (Fig. 2A-C, inset). The single channels are reported in Supplementary Fig. S2.

CD146 is expressed on pericytes in normal human skin

To ascertain that the single positive cells for CD146 detected in close proximity to CD31+ blood capillaries correspond to pericytes, we performed triple immunofluorescence co-staining for CD146, CD31, and different specific markers of perivascular cells: NG2, desmin, and αSMA [10] (Fig. 3). A clear co-localization between CD146 and NG2 was observed all around CD31 positive blood capillaries, where they seemed to continuously invest the capillary wall (Fig. 3A). CD146 and NG2 appeared to be ubiquitously expressed on all perivascular cells of the blood capillaries both in the upper superficial capillary plexus (Fig. 3A) and deep vascular plexus of the reticular dermis (Fig. 3B). On the other hand, a co-localization between desmin and CD146 was observed only in the pericytes of the lower dermis (Fig. 3C), while desmin expression was almost undetectable in the upper dermis (data not shown). Among the well-known markers of perivascular cells, αSMA is mostly expressed around venules and arterioles, but is usually not detected around capillaries [10, 25]. We observed a co-localization between CD146 and αSMA (Fig. 3D) indicating that CD146 is expressed in perivascular cells around capillaries, venules and arterioles. The single channels are reported in Supplementary Fig. S3.

Fig. 3

CD146 expression on pericytes in normal human skin. A, B, C, D Representative immunofluorescence images of human skin sections stained for CD146, CD31 and different markers of perivascular cells. Pericyte-like cells double positive for CD146 and NG2 localize around the entire blood capillaries, in both upper (A) and deep (B) vascular plexus. NG2−/CD146+/CD31+ blood endothelial cells and NG2+/CD146+/CD31− pericytes can be clearly distinguished in blood capillaries (B). Whereas a small population of desmin/CD146-double positive cells can be observed only on a small fraction of blood capillaries (C), co-localization between CD146 and αSMA is extensively observed in most of the blood capillaries (D). (n = 3 independent donors). Scale bar: 50 µm. White dashed lines indicate the epidermal-dermal junction

CD146 is expressed on blood endothelial cells in vitro

To perform a profound analysis of CD146 expression on endothelial cells, we performed complete digestions of the dermal fraction of human foreskins to isolate single cells. The single cell suspensions were then analyzed by flow cytometry for CD31, Podoplanin and CD146 expression (Fig. 4A). CD31 and podoplanin were used as markers to discriminate between blood endothelial cells (CD31+Podoplanin-) and lymphatic endothelial cells (CD31+ Podoplanin+). CD146 expression was then investigated among LECs and BECs gates. The results showed that CD146 expression was mainly detected on freshly isolated BECs (57,3 ± 7,4%), while almost no expression of CD146 was observed on freshly isolated LECs (2,6 ± 0,7%) (Fig. 4A). The detailed gating strategy is described in Supplementary Fig. S4A.

Fig. 4

Expression of CD146 on cultured HDMECs and sorted BECs and LECs. A Flow cytometry analysis of freshly isolated HDMECs from human foreskin dermis. The hierarchical gating strategy involved the consecutive exclusion of debris, doublets and dead cells using Zombie Aqua staining. CD31 and Podoplanin were used to discriminate between BECs (CD31+Podoplanin−) and LECs (CD31+Podoplanin+). CD146 expression was detected only on freshly isolated BECs (57,3 ± 7,4%), while freshly isolated LECs were almost completely negative for CD146 expression (2,6 ± 0,7%). n = 4 independent skin donors. B In vitro cultured HDMECs at passage 1 (P1) were stained for CD146, CD31 and Prox1. Two subpopulations of BECs (CD31+/ Prox1−) and LECs (CD31+/ Prox1+) are identified in HDMECs (panel I). CD146 expression is detected on a small fraction of CD31+ cells (panel II), while Prox1+cells lack the expression of CD146 (panel III). Triple immunofluorescence staining shows the expression of CD146 only in CD31+Prox1− cells, corresponding to BECs (panel IV). Cell nuclei are stained with Hoechst (blue). Scale bar: 100 μm. (n = 3 independent donors). C Representative immunofluorescence images of BECs and LECs at passage 1 separated by FACS and cultured in vitro. CD31 and Prox1 staining confirm the purity of the two cell populations. Whereas CD146 expression is detected only on BECs, LECs lack the expression of CD146. Cell nuclei are stained with Hoechst (blue). Scale bar: 100 μm. (n = 3 independent donors). D Western blot analysis of sorted BECs and LECs shows the expression of CD146 only on BECs. Detection of Prox1 only on LECs confirmed the purity of the populations. Fibroblasts (FBs) were used as negative control. Equal loading was assessed with anti-GAPDH antibody. (n = 3 independent donors)

Further, we investigated the expression profile of CD146 on in vitro cultured HDMECs to verify if its expression could be also retained in vitro (Fig. 4B). Immunofluorescence staining for CD31 and the transcription factor Prox1 allowed us to discriminate between BECs (CD31+/ Prox1−) and LECs (CD31+/ Prox1+). As expected, CD31 stained all the endothelial cells, while Prox1 was detected only in a fraction of these cells, showing the presence of a mixed population of BECs and LECs in HDMECs (Fig. 4B panel I). Furthermore, we demonstrated that CD146 expression can be retained in vitro and in particular detected only in a population of HDMECs (Fig. 4B, panels II and III), which we showed to be negative for the lymphatic marker Prox1 (Fig. 4B, panel III). The triple co-immunofluorescence for CD31, CD146, and Prox1 conclusively indicated the expression of CD146 only on BECs (Fig. 4B, panel IV). We further validated our results on pure BECs and LECs separated by FACS (Supplementary Fig. S4B), to investigate CD146 expression on the single population without the interference of the other one (Fig. 4C). CD31 and Prox1 staining was performed to confirm the purity of the two cell populations (Fig. 4C). Whereas all the BECs resulted positive for CD146 expression, CD146 was not detected in LECs (Fig. 4C). These results were confirmed by western blot analysis, which displayed a specific band corresponding to the molecular weight of CD146 only in BECs (Fig. 4D). Prox1 protein expression was detected only in LECs, confirming the specificity of this marker. Meanwhile, CD31 expression was detected in both BECs and LECs, with lower expression levels in LECs (Fig. 4D), which is in agreement with a previous observation [23]. Fibroblasts (FBs) were used as negative control (Fig. 4D).

Cultured CD146-positive pericytes displayed markers of perivascular cells

After our observations of CD146-positive pericytes surrounding blood capillaries in human skin (Fig. 3), we further characterized CD146 as a marker of human skin pericytes at single cell level. To this end, single cell suspensions freshly isolated from the human dermis were analyzed by flow cytometry (Fig. 5A) for CD146, CD90 and CD31 expression. Since CD90 represents a marker of both pericytes and fibroblasts, we could discriminate in our analysis of freshly isolated dermal cells two subpopulations attributable to pericytes (CD146+CD90+, 15,5 ± 2,1%) and fibroblasts (CD146−CD90+, 52,5 ± 2,6%) (Fig. 5A). A small population of CD146 + CD31 + cells (3,1 ± 0,7%) was gated among the entire live cell population (Fig. 5A). Thus, these results indicated that CD146 can be detected in both pericytes and endothelial cells. The detailed gating strategy is described in Supplementary Fig. S5.

Fig. 5

Characterization of CD146+ pericytes. A Flow cytometry analysis of a single cell suspension freshly isolated from human foreskin dermis. The hierarchical gating strategy involved the consecutive exclusion of debris, doublets, and dead cells using Zombie Aqua staining. The staining for CD146 and CD90 highlights the presence of two populations: pericytes (CD146+CD90+, 15,5 ± 2,1%) and fibroblasts (CD146−CD90+, 52,5 ± 2,6%). The staining with CD31 shows also the presence of a CD146+CD31+ cell population (3,1 ± 0,7%). n = 4 independent skin donors. B Representative immunofluorescence images of isolated CD146+ pericytes (n = 3 independent donors). The entire population of pericytes results positive for CD146, while HDMECs and FBs, used as control, displays a small population or no detection of CD146, respectively. Moreover, CD146 + pericytes results entirely positive for NG2, while only a small percentage is also positive for desmin. CD146 + pericytes are also positive for αSMA and CD90. No detection of these perivascular markers is observed on HDMECs, while FBs results positive for αSMA and CD90. Cell nuclei are stained with Hoechst (blue). Scale bar: 100 μm. C Western blot analysis of perivascular markers on CD146 + pericytes, HDMECs and FBs. CD146 + pericytes show protein expression of CD146, NG2, desmin, αSMA and CD90. HDMECs display higher protein expression of CD146 compared to pericytes. αSMA and CD90 are also detected on FBs. Equal loading was assessed with anti-GAPDH antibody. (n = 3 independent donors)

Since CD146 has been described as a suitable marker for pericyte isolation, in particular in skeletal muscle and non-muscle tissues [10, 26], and our data on human skin sections showed that among the different perivascular markers CD146 is the most ubiquitously and strongly expressed, we decided to use CD146 as marker for human dermal pericyte isolation. Isolated CD146+ pericytes were then cultured in vitro and characterized for specific marker expression (Fig. 5B). First, we confirmed CD146 expression on all isolated pericytes by immunofluorescence (Fig. 5B). HDMECs, used as comparison, showed only a small fraction of CD146 positive cells, as expected, while FBs resulted negative. NG2 was uniformly expressed on all pericytes, while only a small fraction of cells was positive for desmin (Fig. 5B). This is in agreement with our observation on human skin sections that desmin positive pericytes represent only a small fraction of the entire pericyte population. CD146+ pericytes were also positive for αSMA and CD90 (Fig. 5B). HDMECs lacked the expression of these markers, while fibroblasts used as control were positive only for αSMA and CD90. Western blot analysis confirms these results (Fig. 5C). Interestingly, despite detectable CD146 expression in both pericytes and HDMECs, the protein levels were higher in HDMECs compared to pericytes (Fig. 5C).

TNFα/IL-1β-mediated NFκB activation induces CD146 up-regulation in both pericytes and BECs

Next, we sought to investigate if the expression of CD146 could be modulated in pericytes and BECs by stimulation with the proinflammatory cytokines TNFα, IL-1β and IL-6.

To this aim, cells were stimulated with TNFα, IL-1β, or IL-6, for different time points and fluorescein diacetate–propidium iodide (FdA–PI) assay was performed to confirm that the treatments have not cytotoxic effect on cells (Supplementary Fig. S6).

Western blot analysis of CD146 expression showed a time-dependent up-regulation of CD146 protein levels after stimulation with TNFα or IL-1β in both pericytes (Fig. 6A) and BECs (Fig. 6B), while IL-6 treatment showed no effect on both pericytes and BECs (Fig. 6A and B). In addition, no induction of CD146 was observed in LECs stimulated with TNFα, IL-1β, or IL-6 (Supplementary Fig. S7).

Fig. 6

CD146 expression is up-regulated by stimulation with TNFα and IL-1β through NFκB activation. A, B CD146+ pericytes and BECs were left untreated or stimulated with TNFα, IL-1β, and IL-6 for 12 h, 24 h, and 48 h. Western blot analysis shows up-regulation of CD146 expression in response to TNFα and IL-1β treatment compared to unstimulated cells, while no changes in the protein levels of CD146 are observed after IL-6 stimulation. Parallel evaluation of p65 phosphorylation shows that only TNFα and IL-1β treatment induce p65 phosphorylation/activation in both pericytes and BECs. The equal loading was assessed using anti-GAPDH antibody for CD146 and anti-p65 for phospho-p65. For the densitometric analysis, the values from three independent experiments were normalized and expressed as fold increases and are reported as mean values ± standard deviations (SD). Unpaired Student’s t-test was performed, and significance levels are defined as * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. NS, not significant. C, D CD146.+ pericytes and BECs were left untreated or stimulated with TNFα or IL-1β in the presence or absence of the NFκB inhibitor, BAY 11–7082 (Bay) for 24 h. Western blot analysis shows that the increase in the levels of CD146 upon TNFα or IL-1β stimulation is abolished by treatment with NFκB inhibitor. The equal loading was assessed using anti-GAPDH antibody for CD146 and anti-p65 for phospho-p65. The densitometric analysis and unpaired Student’s t-test were performed as reported above. * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001

Since both TNFα and IL-1β are known to activate NFκB [27], we investigated the activation of the NFκB pathway in pericytes and BECs. To this end, we analyzed the phosphorylation level of p65, one of the main NFκB subunits whose phosphorylation is required for full transcriptional activity of NFκB in the nucleus [28]. Both pericytes and BECs showed a significant phosphorylation/activation of p65 already after 12 h following stimulation with TNFα or IL-1β, which further increased after 24 h (Fig. 6A, B). IL-6 stimulation did not induce any phosphorylation/activation of p65/NFκB in both pericytes and BECs (Fig. 6A, B), as expected [27], thus strengthening the hypothesis that the regulation of CD146 proceeds through NFκB activation. To confirm the involvement of NFκB in the TNFα/IL-1β-mediated CD146 up-regulation, we stimulated the cells with the cytokines in the presence or absence of the specific NFκB inhibitor, BAY 11–7082 (Bay). The presence of the inhibitor significantly decreased the levels of CD146 induced by TNFα or IL-1β stimulation in both pericytes and BECs, which appeared to be more evident in pericytes (Fig. 6C, D). The down-regulation of the phosphorylation levels of p65 after Bay treatment, confirms the efficiency of the inhibitor (Fig. 6C, D). Thus, our results demonstrated that CD146 expression is up-regulated in both pericytes and BECs after stimulation with the pro- inflammatory cytokines TNFα and IL-1β through the activation of NFκB.

Engineering of 3D pre-vascularized hydrogel with HDMECs and CD146+ pericytes

We sought to reproduce a 3D model of vascularized dermis in vitro, in which the microvascular network resembled the human dermis microvasculature as closely as possible. To this end, we improved our well-established system of 3D pre-vascularized collagen type I hydrogels [21, 29], including also CD146+ pericytes. Whole-mount staining for CD31 and Prox1 in 3D hydrogels showed a developed vascular network, in which we could reproduce both blood capillaries and lymphatic capillaries (Fig. 7A). The specificity of CD146 for blood endothelial cells was conserved in our 3D hydrogels; in fact, CD146 expression was detected only on blood capillaries (CD31+/Lyve1−), while lymphatic capillaries (CD31+/Lyve1+) were CD146− (Fig. 7B). NG2 staining displayed positive cells surrounding blood capillaries, mimicked pericyte investment of the vessel wall (Fig. 7B, C). As expected, lymphatic capillaries were missing these pericytes (Fig. 7B). Interestingly, in our 3D pre-vascularized hydrogels, we could observe the presence of NG2-single positive cells (Fig. 7B, C), as well as pericytes double-positive for CD146 and NG2 (Fig. 7C, panel II, arrows). The presence of single-positive cells for NG2 could be due to the loss of CD146 expression from some of these cells or to the differentiation of fibroblasts into pericytes. In fact, Goss and colleagues showed that mouse skin papillary and reticular fibroblasts can give rise to NG2+ pericytes in the upper and lower dermis respectively [30]. The whole-mount staining also showed the presence of CD31 + blood capillaries surrounded by CD146 + αSMA + (Fig. 7D, arrows), and CD146 + CD90 + (Supplementary Fig. S8) perivascular cells. Fibroblasts have been identified as single positive cells for CD90 (Supplementary Fig. S8).

Fig. 7

CD146 expression on 3D pre-vascularized hydrogel. Representative confocal images of HDMECs co-cultured with CD146+ pericytes and fibroblasts for three weeks in 3D collagen type I hydrogel. (n = 3 independent donors). A CD31 (red) and Prox1 (green) staining shows the developed vascular network composed of both blood and lymphatic capillaries. Scale bar: 100 μm. B Quadruple staining for CD146 (red), Lyve1 (blue), NG2 (green), and CD31 (white), shows that CD146 is expressed only on blood capillaries (CD31+Lyve1−), while lymphatic capillaries (CD31+Lyve1+) lack CD146 expression. Pericyte-like cells positive for NG2 are detected only around blood capillaries (CD31+Lyve1−) mimicking pericyte investment of the vessel and are not present on lymphatic capillaries. Scale bar: 100 μm. C, D Representative confocal images of CD146 and CD31 staining with NG2 (C) or αSMA (D) show the presence of pericytes-like cells double-positive for CD146+NG2+ (C, inset II, arrows) or CD146+αSMA+ (D, inset II, arrows), which surround blood capillaries. The presence of NG2-single positive cells is also observed around capillaries (C, inset I). Scale bar: 100 μm, inset 50 μm

In vivo transplantation of pre-vascularized dermo-epidermal skin substitutes containing CD146+ pericytes

The 3D pre-vascularized hydrogels composed of endothelial cells, CD146+ pericytes, and fibroblasts were then covered with human keratinocytes (KC) to generate pre-vascularized dermo-epidermal skin substitutes (DESS) and transplanted onto the back of immuno-incompetent rats to investigate whether they might be able to form a stable and functional vascular network in vivo. Histological analysis of the skin substitutes collected after one week showed the presence of a stratified epidermis composed of stratum basale and several suprabasal layers until the stratum corneum, as well as a dermal compartment containing fibroblasts (Fig. 8A). The corresponding immunofluorescence analysis showed the expression of the differentiation marker cytokeratin 10 (CK10) throughout the suprabasal layers, while the basal layer displayed no expression of CK10. The staining for Laminin332 showed the presence of a basement membrane between the epidermis and the human dermis, which has been visualized with the antibody against human CD90 (Fig. 8B). Evaluation of epidermis homeostasis has been performed with the analysis of cytokeratin 19 (CK19) expression, which is usually confined to the stratum basale of the epidermis in engineered skin substitutes after three weeks of transplantation [20]. Our results showed that after one week from skin graft transplantation, CK19 was expressed not only in the cells of the basal layer, but also in the suprabasal layers of the epidermis, indicating that epidermis homeostasis had not yet reached (Fig. 8C).

Fig. 8

In vivo characterization of pre-vascularized DESS containing CD146+ pericytes. A The hematoxylin–eosin (H/E) staining of pre-vascularized DESS after 1 week shows a stratified epidermis and dermis containing fibroblasts. Scale bar: 100 μm. B The immunofluorescence staining for CK10 displays a positive signal of this keratinocyte differentiation marker from the suprabasal layer. Laminin332 and CD90 staining indicates the presence of basement membrane and human dermal fibroblasts, respectively. Scale bar: 100 μm. C The immunofluorescence staining for CK19 shows the presence of CK19-positive keratinocytes in both the basal and suprabasal layers. The staining for Laminin332 indicate the presence of the basement membrane. Scale bar: 100 μm. D, E Human capillaries positive for the specific human CD31 antibodies are distributed throughout the human dermis, which is marked with human CD90. Inosculation between human and rat capillaries is visualized by co-localization of humanCD31 and ratCD31 (D, E, insets). The ingrowth of rat capillaries is also observed in the human neo-dermis (arrows). White dashed lines indicate the dermo-epidermal junction. Scale bar: 100 μm, inset 50 μm. F The green autofluorescence of rat erythrocytes (ratBCs) inside the lumen of human capillaries indicates the presence of perfused capillaries. Scale bar: 100 μm, inset 50 μm. G Immunofluorescence staining for CD146 and humanCD31 shows the co-localization between these two markers and the presence of human CD146+ pericytes around capillaries. Scale bar: 100 μm, inset 50 μm. H Lymphatic capillaries stained for Lyve1 are CD146 negative. I Human pericytes double-positive for CD146 and NG2 are observed around capillaries. CD146 negative capillaries, attributable to lymphatic capillaries, show no presence of pericytes (asterisk). G, H, I (n = 3 independent donors). White dashed lines indicate the epidermal-dermal junction. Scale bar: 100 μm, inset 50 μm

To investigate the presence of the human vascular network in the skin substitutes and its ability to anastomose with the host vascular plexus, we performed immunofluorescence analysis with antibodies specifically discriminating between CD31 of human and rat origin (Fig. 8D, E). CD90 staining was performed to distinguish between the human dermis and the rat tissue (Fig. 8D). An extensive network of human blood capillaries was distributed throughout the human dermis (Fig. 8D, E). Here, the presence of ratCD31-single positive capillaries indicated the ingrowth of the host blood vessels into the human skin transplant to support blood perfusion (Fig. 8D, E, arrows). The colocalization between human capillaries and rat capillaries, particularly observed at the interface between human and rat tissues, indicated inosculation between the two microvascular networks (Fig. 8D, E insets). In addition, rat red blood cells (ratBCs) were found in the lumen of several capillaries lined by human endothelial cells (Fig. 8F, inset). All these data provided evidence for the functionality of the transplanted human microvascular network and its efficiency in blood circulation.

Further, we assessed the expression of CD146 in blood capillaries and pericytes in the pre-vascularized transplants. We found the presence of human capillaries double-positive for humanCD31 and CD146, surrounded by perivascular cells single-positive for CD146 (Fig. 8G, Supplementary Fig. S9A). Moreover, lymphatic capillaries stained with Lyve1 or Prox1 displayed negative expression of CD146 (Fig. 8H, Supplementary Fig. S9B), confirming the exclusive expression of CD146 on blood capillaries. CD146/NG2 double-positive human pericytes were also found investing CD146-CD31 positive capillaries (Fig. 8I, Supplementary Fig. S9C), while CD31 single-positive capillaries attributable to lymphatic capillaries showed no presence of pericytes around (Fig. 8I, asterisk), confirming the role of pericytes in stabilizing only the blood capillaries.

The immunofluorescence staining for CD146, humanCD31 and ratCD31 showed a co-localization of CD146 only with capillaries from human origin (Supplementary Fig. S9D), confirming the specificity of the CD146 antibody for the human tissue and excluding the possibility that the the observed CD146-positive pericyte could have been of rat origin.