JAGGED1 signaling regulates hemangioma stem cell-to-pericyte/vascular smooth muscle cell differentiation

Abstract

OBJECTIVE

The aim of our study is to determine the cellular and molecular origin for the pericytes in infantile hemangioma (IH) and their functional role in the formation of pathological blood vessels.

METHODS AND RESULTS

Here we show that IH-derived stem cells (HemSCs) form pericyte-like cells. With in vitro studies, we demonstrate that HemSC-to-pericyte differentiation depends on direct contact with endothelial cells. JAGGED1 expressed ectopically in fibroblasts was sufficient to induce HemSCs to acquire a pericyte-like phenotype, indicating a critical role for JAGGED1. In vivo, we blocked pericyte differentiation with recombinant JAGGED1, and observed reduced formation of blood vessels, with an evident lack of pericytes. Silencing JAGGED1 in the endothelial cells reduced blood vessel formation and resulted in a paucity of pericytes.

CONCLUSIONS

Our data show that endothelial JAGGED1 controls HemSC-to-pericyte differentiation in a murine model of IH and suggests that pericytes have a fundamental role in formation of blood vessels in IH.

Infantile hemangioma (IH), the most common tumor of infancy, is an important model for the study of pathological disruptions of human vascular development. IH is a lesion of disorganized vasculature that appears around the second week of life, grows over the course of 6–9 months and involutes during childhood ^1–4. We previously identified and isolated stem cells (HemSCs) from proliferating IH surgical specimens ⁵. HemSCs are stem cells based on: 1- clonal expansion from a single cell in vitro, 2- multi-lineage differentiation ability, and 3- formation of functional hemangioma-like (GLUT1+) ⁶ blood vessels in primary and secondary recipient mice ⁷. In addition, the central role of HemSCs is supported by our study showing that corticosteroids, the standard treatment for IH, target the HemSCs by down-regulating VEGF-A expression ⁸.

IH blood vessels are surrounded by pericytes ^9,10 but the cellular origin of these cells has not been investigated. Here we show that clonally expanded HemSCs give rise to the pericyte/vascular smooth muscle cells, which for simplicity we will refer to as pericytes. The acquisition of the pericytic phenotype is dependent on cell contact with endothelial cells. Because of this, we investigated NOTCH ligands in pericytic differentiation. NOTCH signaling occurs through juxtacrine interactions and plays a crucial role in vascular and tumor development ^11–13. The NOTCH family comprises 4 receptors (NOTCH 1–4) and the ligands, JAGGED (JAG1, JAG2) and Delta-like (DLL1, DLL3 and DLL4) ^{14, 15}. We recently reported the expression of NOTCH receptors and ligands in tissue and cells from IH ¹⁶. JAGGED1 is upregulated in IH-derived endothelial cells (HemECs) ¹⁶, and IH endothelium ¹⁷; whereas only modest expression of the vascular tip-cell regulator Dll4^16,18 was reported in HemECs ¹⁶. However JAGGED1 signaling mechanisms in the pathogenesis of IH are still unknown.

Mutations in JAGGED1 cause the human disease Alagille syndrome (AGS) ¹⁹. AGS patients develop intracranial vessel abnormalities as well as congenital cardiac disease ^{20, 21}, cerebral aneurysms and coarctations ²⁰. Jagged1 knockout mice die between E10.5 and E11.5 with defects in the yolk sac and embryonic vasculature ²². Endothelial-specific deletion of Jagged1 results in deficit in smooth muscle differentiation, a phenotype similar to Jagged1 total knockout and Jagged1 mutants ^{18, 23}.

We report here the direct involvement of JAGGED1 in the HemSC-to-pericyte differentiation, suggesting an important role for pericytes in the formation of IH blood vessels. These results provide novel insight into mechanisms of human post-natal vasculogenesis.

METHODS

Cell Isolation and Culture

Specimens of IH were obtained under a protocol approved by the Committee on Clinical Investigation, Children’s Hospital Boston. The clinical diagnosis was confirmed in the Department of Pathology, Children’s Hospital Boston. Informed consent was obtained, according to the Declaration of Helsinki. Single cell suspensions were prepared from proliferating phase IH specimens and HemSCs were selected and expanded as described ⁷. Briefly: HemSCs were selected using anti-CD133–coated magnetic beads (Miltenyi Biotec). CD133-selected cells were cultured on fibronectin-coated (1 µg/cm2) plates with EBM-2 (CC-3156; Cambrex). EBM-2 was supplemented with 20% FBS, endothelial growth media-2 SingleQuot (CC-4176; Cambrex), and 1× PSF (100 U/ml penicillin,100 µg/ml streptomycin, 0.25 µg/ml amphotericin; Invitrogen). Hereafter this supplemented medium is called EBM-2/20% FBS. Hemangioma-derived endothelial cells (HemECs) and human umbilical cbEPCs were isolated as described ²⁴. cbEPCs were shown to form the endothelial lining of blood vessels when injected in combination with smooth muscle cells or bmMPC, in Matrigel, subcutaneously in mice ^{25, 26}. Human Dermal Fibroblasts (HDFs) were purchased from ATCC.

JAGGED1 expressing HDFs were produced by transfecting (Fugene HD, Roche) the cells with JAGGED1-vector (Origene); selection was done with G418 (Sigma-Aldrich). Short hairpin RNA (shRNA) lentiviral particles for JAGGED1 targeting (NM_00214) and non-targeting shRNA (control) were used to infect cbEPCs (Sigma-Mission®). cbEPCs with stable expression of JAGGED1 shRNA were subjected to puromycin selection, according to the manufacturer's instructions. The pool used for in vivo experiments was selected after testing five different shRNA sequences for silencing efficacy.

Co-culture differentiation protocol

The in vitro pericytic differentiation was performed by seeding HemSCs together with cbEPCs at a ratio 1:1 and total density of 10⁴ cells/cm² on fibronectin-coated plates in EBM-2/20% FBS. To inhibit the HemSC pericytic differentiation, we used recombinant human Fc-JAGGED1 and Fc-DLL4, and mouse anti-JAGGED1 (R&D Systems). Inhibitors were added 4 hours after cell plating and medium/inhibitors was replaced every other day.

(For detailed methods: see Supplemental Material)

RESULTS

IH blood vessels contain pericytes

Blood vessels that emerge during the proliferating phase of IH are encircled by perivascular cells that are defined in the literature as pericytes expressing α-smooth muscle actin (αSMA)+ ^9,10. Our results show that the majority of these perivascular cells express neural glial antigen-2 (NG2), Platelet Derived Growth Factor Receptor-(PDGFR)-β, Calponin I, αSMA and smooth muscle Myosin Heavy Chain (smMHC) (Fig.1a). To study the perivascular contribution in blood vessels, we quantified mRNA expression of pericyte markers and normalized to the endothelial content (CD31) of the IH tissue. We expected a major role for the pericytes in the maturation (i.e. involuting phase) of IH blood vessels. Instead, pericyte/smooth muscle markers were expressed in both proliferating and involuting phases, and we did not detect statistically significant differences between the 2 groups (Fig.1b and Supplementary Fig.I a). The localization of these markers indicates that perivascular cells surrounding IH vessels have features of pericytes and smooth muscle cells. For simplicity, and because the perivascular cells in IH are typically a single layer, we refer to them as pericytes.

Figure 1. Human and experimental IH blood vessels are surrounded by pericytes.

(a) Human proliferating IH tissue stained for CD31 (red) and NG2, PDGFR-β (green) or Calponin (green); VE-cadherin (green) and αSMA or smMHC (red). Nuclei are counterstained in blue (DAPI). (b) Quantitative real time PCR analysis of NG2, PDGFRβ, Calponin, αSMA and smMHC expression in IH specimens at different ages, white bars=proliferating phase IH, grey bars= involuting IH. (Relative p values: NG2 p=0.2996, PDGFRβ p=0.0735, Calponin p=0.6837, αSMA p=0.7019, smMHC p=0.1214). (c) HemSCs suspended in Matrigel and implanted in mice for 7 days; CD31 (red), Calponin (green) and αSMA (red) immunostaining of Matrigel explants. (d) GFP-HemSCs suspended in Matrigel and implanted in mice for 7 days; GFP (green) and αSMA (red) immunostaining of Matrigel explants. In c and d, asterisks mark blood vessel lumens. (e) Clonal population of GFP-labeled HemSCs, implanted in vivo in Matrigel stained for αSMA (red) and GFP (green). White arrow indicates colocalization of αSMA and GFP. Phase contrast image shows blood vessel location (black arrow). Scale bar=30µM.

HemSCs differentiate into pericytes when injected into mice

To assess whether pericytes are present in the HemSC-derived blood vessels in our murine model of IH, HemSCs were suspended in Matrigel and injected into nude mice ⁷. After 7 days, Matrigel explants showed Calponin+ cells surrounding human CD31+ blood vessels. (Fig.1c, left). Pericytes, in the murine model of IH are characterized by coexpression of αSMA and Calponin (Fig 1c, right). To further test whether the αSMA+ pericytes arose from HemSCs, we injected GFP-labeled HemSCs (GFP-HemSCs). The presence of αSMA+/GFP+ pericytes confirmed their HemSC origin (Fig.1d). We previously reported that HemSCs can differentiate into ECs ⁷. To verify that pericytes and endothelial cells are derived from the same progenitor cell, we prepared 3 single cell-derived clonal populations of GFP-HemSCs derived from 2 different IH samples (Supplementary Fig.I b), and analyzed cellular fate(s) in vivo. All 3 GFP-HemSC clonal populations formed pericytes (Fig.1e and Supplementary Fig.I c) and endothelial cells (Supplementary Fig.I d).

HemSCs differentiate into pericytes when combined with cbEPCs

In previous studies, we described a model to bio-engineer human blood vessels using bone marrow-derived mesenchymal progenitor cells (bmMPCs) and cord blood endothelial progenitor cells (cbEPCs). In this model bmMPCs differentiate into αSMA+ pericytes, while cbEPCs form the endothelium ²⁵. To augment the HemSC-to-pericyte differentiation, we implanted HemSCs together with cbEPCs into mice. This resulted in a blood vessel density 4–5 fold higher than HemSCs implanted alone ⁸. GFP-HemSC+cbEPC produced vessels surrounded by GFP+/αSMA+ cells (Fig.2a), providing direct evidence that GFP-HemSCs became pericytes. Furthermore, 3/3 GFP-HemSC clonal populations differentiated into αSMA+ pericytes when injected with cbEPCs (Supplementary Fig.II a). Quantitative analysis showed that 82% (82±4) of blood vessels had αSMA+ pericytes and 33% (33±2.1) had HemSC-derived pericytes (GFP+/ αSMA+). Of the total GFP+ blood vessels, 73% (33/45) co-expressed αSMA (Fig.2a, bottom). We hypothesize that the remaining 17% of the GFP+ vessels are composed of GFP-HemSCs that underwent endothelial differentiation or remained undifferentiated. To confirm that cbEPCs contributed exclusively to the endothelial population, eGFP-cbEPCs were co-implanted with unlabeled-HemSCs. No colocalization of eGFP with αSMA was seen around blood vessels, confirming that cbEPCs do not differentiate into pericytes (Fig.2b).

Figure 2. GFP-HemSCs form αSMA+ pericytes when co-injected with cbEPCs.

Matrigel explants stained for αSMA (red) and GFP (green) (asterisks mark blood vessel lumens); nuclei counterstained in blue (DAPI). In (a) GFP-HemSCs co-injected with unlabeled cbEPCs. Perivascular GFP-HemSCs (green) expressing αSMA (red) showed yellow fluorescence (arrows). (b) HemSCs co-injected with GFP-cbEPCs, GFP staining shows no overlap with αSMA (arrows). (c) Unlabeled HemSC+cbEPCs co-injected into GFP-SCID mouse. (d) Unlabeled HemSCs-alone injected into GFP-SCID mouse. In c and d co-localization (yellow) between GFP and αSMA indicate murine cells (GFP cells) (arrowhead) contributing to pericytic coverage; cells that express αSMA only are of HemSC origin. In the images, mosaic vessels of murine and HemSC-origins are seen (arrows). Bottom (a,b) quantification of αSMA+ (red bars), GFP+ (green bars) vessels and GFP+/ αSMA+ (yellow bars) vessels in percentage relative to total number of vessels. Bottom, (c,d) quantification of GFP+ (green bars) and αSMA+ (red bars) vessels and quantification of GFP+/αSMA+ (yellow bars) (pericytes of mouse origin), and no overlap vessels (grey bars) (perivascular cells of human origin) in GFP-SCID mice. (* p=0.031, n.s. p=0.677) (n=10 fields/section, n=5 animals/group) Scale bar=30µM. Data expressed as means ± s.e.m.

To examine the contribution of murine cells to the pericytic population, we implanted unlabeled-HemSCs+cbEPCs into GFP-SCID mice ²⁷ (Fig.2c), as well HemSCs-alone (Fig.2d). In both models, we observed a mosaic of vessels with murine- and HemSC-derived αSMA+ pericytes. When HemSC+cbEPCs were injected into GFP-SCID mice, αSMA+ pericytes were derived mostly from the HemSCs (Fig 2c, bottom). When HemSCs-alone were injected into GFP-SCID mice, no statistical difference was seen between host (murine) and HemSC (human) contribution to the αSMA+ pericytes (Fig 2d, bottom). This shows that, when in contact with cbEPCs, HemSCs prefer to take on the pericytic phenotype.

Pericytes in the murine model of IH also stained for Calponin, most of erythrocytes-filled-blood vessels were surrounded by Calponin+ cells, and Calponin staining signal overlapped with αSMA (Supplementary Fig.II b).

HemSCs co-cultured with cbEPCs differentiate into pericytes

Data in Fig.2 showed that HemSCs combined with cbEPCs resulted in HemSC-to-(αSMA+)pericyte differentiation. This prompted us to analyze the role of cbEPC when co-cultured with HemSCs. After 10 days, anti-αSMA and -smMHC identified pericyte-like cells (Fig.3a, left). HemSCs and cbEPCs cultured alone did not express pericyte/smooth muscle cell markers (Supplementary Fig.III a). To confirm that HemSC differentiate into pericyte-like cells, we co-cultured cbEPC with GFP-HemSCs and showed that all of the αSMA+ and smMHC+ cells were derived from the GFP-HemSCs (Fig.3a, right); and GFP-HemSCs cultured alone did not express αSMA or smMHC (Supplementary Fig.III b). Calponin-1 and sm22α are markers of pericyte-like cells ^28,29. HemSC+cbEPC co-cultures showed significant induction of Calponin-1, sm22α, α-SMA protein (Fig.3b), and Calponin-1 and smMHC mRNA (p<0.001) (Fig.3c) compared to HemSCs alone. The expression of Calponin-1, smMHC and αSMA gradually increased from day 3 to day 10 (Fig.3c). In addition, hemangioma-derived endothelial cells (HemECs) also caused HemSC-to-pericyte differentiation (Supplementary Fig.III c).

Figure 3. HemSCs differentiate into pericyte when co-cultured with cbEPCs.

(a) HemSC+cbEPC co-cultures at day 10 stained for vWF (red) and αSMA or smMHC (green) (left). GFP-HemSC+cbEPC co-cultures at day 10 stained for GFP (green) and αSMA or smMHC (red) (right). Scale bar=100µM. Nuclei counterstained in blue (DAPI). (b) Immunoblots for Calponin-1, α-SMA and sm22α in HemSCs-alone or in HemSCs+cbEPCs. (Tubulin=loading control). (c) Quantitative real time PCR of Calponin-1 (left) and smMHC (right) in GFP-HemSCs+cbEPCs (grey bars) compared to GFP-HemSCs alone (black bars). (* p<0.001) (Results are representative of 3 independent experiments repeated with HemSCs from 3 different IH samples). (d) HemSC+cbEPC cocultured for 5 days and then subjected to immunoseparation based on CD31 expression (schematic), the CD31− fraction was analyzed by real time PCR for relative expression of NG2, PDGFRβ and CD31, and results compared to HemSC cultured alone. (e) GFP-HemSC alone (top panels) and CD31− cells immunoseparated from GFP-HemSC+cbEPC coculture (bottom panels) cultured in silicon based sustrata. Phase, fluorescence, merge. On the right quantification of wrinkle forming cells. Data expressed as means ± s.d.m.

To further confirm that the HemSCs acquired a pericytic phenotype, we cocultured HemSC+cbEPC for 5 days and, after trypsinization, selected the cells based on the expression of CD31 (CD31+ = endothelial cells), or lack of CD31 (CD31− =HemSCs and HemSC-derived pericytes). Real time analysis showed that CD31− cells from coculture significantly upregulated PDGFRβ and NG2 (p=0.004) and CD31 expression was not detected (Fig.3d).

We next wanted to test the HemSC-derived pericytes for contractility to assess the specific acquisition of pericyte/smooth muscle cell function. GFP-HemSCs were cultured with cbEPCs to obtain differentiation and after 10 days the cbEPCs were removed with CD31 magnetic beads. The CD31− cells were assessed for contractility on a silicon-based substrata. CD31− cells obtained from the coculture showed almost a 3-fold increase in the capacity to form wrinkles in the silicon, compared to CD31− cells from GFP-HemSCs cultured alone (Fig.3e). This indicates that the HemSC-derived pericytes acquired expression of pericyte/smooth muscle markers that functionally correlate with contractile force.

Next, to assess that each single HemSC has the ability to differentiate into a pericyte, we used single cell-derived HemSC clones ⁷. 6/6 HemSC clones tested, co-cultured with cbEPCs, showed induction of αSMA and smMHC, (Supplementary Fig.III d) and Calponin-1 (Supplementary Fig.III e). These results exclude the possibility that mixed populations of unipotential cells were responsible for the pericyte-like phenotype.

Direct and continuous HemSC-cbEPC contact is required for pericyte differentiation

The endothelial component in the coculture was crucial in the HemSC-to-pericyte differentiation. We next tested if HemSC-derived pericytes could maintain their phenotype if ECs were removed. We used GFP-HemSCs because of their puromycin resistance induced by lenti-viral insertion of a vector containing GFP and the puromycin-resistance genes. GFP-HemSCs+cbEPCs were co-cultured for 10 days to obtain αSMA/smMHC expressing cells; after re-plating, the cells were treated with puromycin (Fig.4a). As expected, cbEPCs died after 2 days, confirmed by the absence of vWF staining. The surviving cells (puromycin resistant GFP-HemSCs) were cultured for up to 20 days. smMHC expression became undetectable after 15 days. αSMA expression was retained until day 20, but cellular morphology, beginning at day 15, reverted to a spindle shape (Fig.4b). Results were confirmed by gradual decrease of Calponin-1 (Fig.4c) and Calponin1/Tubulin (Fig.4d) from day 5 to day 20. αSMA expression i n control GFP-HemSC+cbEPC (normal and puromycin resistant) co-cultures with and without addition of puromycin is shown in Supplementary Figure IV a and IV b. GFP-HemSC cultured alone did not show αSMA expression, with or without puromycin treatment (Supplementary Fig.IV c). This experiment indicates that continuous physical contact with endothelial cells is needed for the GFP-HemSCs to maintain the pericyte phenotype.

Figure 4. HemSC-derived pericytes de-differentiate after cbEPC removal.

(a) Schematic of experimental procedure. (b) Puromycin resistant GFP-HemSCs co-cultured with cbEPCs and treated with puromycin (2 µg/ml) were stained for vWF (red) and: smMHC (top) or αSMA (bottom) (green). (c) Immunoblots for Calponin-1 and Tubulin; (d) densitometric quantification of Caponin-1/Tubulin band intensity. (Representative of n=3 independent experiments). Scale bar=100µM.

To test whether HemSC differentiation could occur without direct cellular contact with cbEPCs, we cultured HemSCs alone on the bottom well and cbEPCs or HemSCs+cbEPCs on the top well of a 0.4 µm pore Transwell™. HemSCs did not express smMHC under these conditions (Supplementary Fig.IV d). We tested specific factors reported to induce smooth muscle differentiation (TGF-β1, PDGF-BB, type IV collagen-coated wells) ^30–32. After 10 days, smMHC was not detected and only rare αSMA+ cells were seen (Supplementary Fig.IV e). Without endothelial cell contact, the HemSCs were unable to differentiate toward the pericyte/smooth muscle phenotype under the culture conditions tested.

JAGGED1 is required for HemSC-to-pericyte differentiation and blood vessel formation

We and others recently reported high JAGGED1 expression in IH endothelium ^16,17. We further confirmed these findings here, as shown by the high JAGGED1 mRNA and protein expression in HemECs (Supplementary Fig.V a) and JAGGED1 in endothelium of a proliferating IH specimen (Supplementary Fig.V b). We used different approaches to explore the role of the endothelial JAGGED1 in the HemSC-to-pericyte differentiation. The first was to inhibit membrane-bound JAGGED1 binding to its ligand/s by adding recombinant human (rh) Fc-JAGGED1 (Fc-JAGGED1) to the HemSC+cbEPC co-cultures. Fc-JAGGED1 (0.1 µg/ml) limited HemSC-to-pericyte differentiation as indicated by reduced Calponin1 and smMHC expression, compared to human recombinant Fc-domain (Fig.5a).

Figure 5. JAGGED1 activity is required for HemSC-to-pericyte differentiation.

(a) Real Time PCR for Calponin-1, smMHC, HES1 and HEYL in Fc-JAGGED1 (0.1 mg/ml) -treated cultures (grey bars) compared to matching concentration of Fc-domain treated cultures (black bars). (b) HemSC+cbEPC co-cultures treated with 2 µg/ml Fc-JAGGED1 for 5 days; HemSCs were separated from cbEPCs using anti-human CD31-magnetic beads and analyzed by immunoblot for Calponin1 and sm22α after treatment with Fc-JAGGED1 or control Fc-domain. (c) Immunofluorescence for vWF (red) and α-SMA (green) in co-cultures treated with Fc-domain or Fc-JAGGED1 with quantification of α-SMA+ cells shown in the bar graph. (Data are expressed as means, normalized to control, of 3 independent experiments with n=3, * p=0.006) (d) H&E-stained sections of Matrigel explants containing HemSC+cbEPCs treated with 1 mg/ml Fc-domain (top) and Fc-JAGGED1 (bottom) and quantification of MVD and human CD31+ vessels (right). Scale bar, 100µM. (e) Immunoblot for JAGGED1 in Human Dermal Fibroblasts (HDF) after transfection with control (empty) and JAGGED1-vector. (f) Real time PCR showing Calponin1, smMHC, NOTCH3, HES1, HEYL, NG2, PDGFRβ, αSMA and sm22α in HemSCs co-cultured with JAGGED1-expressing-HDFs (grey bars) compared to HDFs expressing empty-vector (black bars). (n=3 independent experiments). Data are expressed as means ± s.d. (a,c, e) or as means ± s.e.m. (d).

Upon NOTCH receptor signaling, the intracellular domain of NOTCH (NICD) is cleaved by γ-secretases and is translocated to the nucleus where it associates with the RBP-Jk proteins to induce transcription of the bHLH genes HES and HEY, which are the most prominent NOTCH effector molecules ³³. We assessed NOTCH ligand-induced signaling by measuring HES and HEY. Treatment with Fc-JAGGED1 significantly decreased (p<0.05) HES1 and HEYL mRNA expression (Fig.5a). Fc-JAGGED1 also prevented Calponin1 and sm22α protein upregulation in HemSCs purified from the co-cultures (Fig.5b). Additionally, the number of αSMA+ cells in co-cultures treated with Fc-JAGGED1 showed an 80% reduction compared to the Fcdomain control, (Fig.5c). In vivo, Fc-JAGGED1 significantly (p<0.05) reduced the total MVD and the human CD31+ MVD (Fig.5d). Fc-JAGGED1 affected vessel integrity, suggesting a disruption in the perivascular coverage (Fig.5d)

The NOTCH ligand DDL4 was reported to have a role in the pericyte/smooth muscle differentiation of bone marrow cells and in the vessel maturation ^34,35. Therefore, we investigated the effect of DDL4 blockade in the HemSC-to-pericyte differentiation. HemSC+cbEPC and HemSC+HemEC cocultures were treated with rhJAGGED1, rhDLL4 or combination of both. Results showed that DLL4 blockade did not inhibit Calponin, αSMA and sm22α expression, and we did not detect any additional reduction in Calponin, αSMA and sm22α when rhDLL4 was added with rhJAGGED1 (Supplementary Fig.V c). These results are on line with our previous report on modest levels of DLL4 in HemEC compared to HDMEC, and suggest that, as DLL4 expression is low in HemEC, DLL4 blockade may not effect HemEC-induced HemSC-to-pericyte differentiation.

To confirm that blockade of JAGGED1 signaling inhibits HemSC-to-pericyte differentiation, we treated cbEPC+HemSC co-cultures with anti-JAGGED1 antibody. Results showed significant decreases (p<0.05) in Calponin1, smMHC and HES1 and HEYL mRNA compared to cocultures treated with control mouse IgG (Supplementary Fig.V d).

Fc-JAGGED1 blocked rather than induced pericytic differentiation, as reported in the literature ^{36, 37}, therefore, we tested whether cell-bound JAGGED1 was sufficient to stimulate differentiation. We transfected Human Dermal Fibroblasts (HDFs) with an expression plasmid for JAGGED1 (Fig.5e, left), and co-cultured them with HemSCs. After 5 days, HemSC+JAGGED1-HDF cocultures exhibited increased (p<0.05) mRNA expression of Calponin1, smMHC, NOTCH3, HES1, HEYL, NG2, PDGFRβ, αSMA and sm22acompared to HemSC + empty vector-HDF co-cultures (Fig.5e, right). Therefore, JAGGED1 on a non-endothelial membrane was sufficient to stimulate HemSCs to differentiate toward a pericyte-like phenotype in vitro.

To test if JAGGED1 is necessary in the HemSC-to-pericyte differentiation, we downregulated JAGGED1 in cbEPCs using short-hairpin- (sh-)RNA-mediated silencing. Two cbEPCs pools with downregulated JAGGED1 (shJAGGED1-pool1 and shJAGGED1-pool2) (Fig.6a), were combined with HemSCs and co-injected into mice. The shJAGGED1 cbEPCs showed decreased MVD and human-derived CD31+ MVD compared to the control shNon-targeting (Fig.6b and Supplementary Fig.VI a) that correlated with the level of decreased JAGGED1. To assess pericytic coverage of the vessels, sections were immunostained for human CD31 in combination with anti-αSMA antibody. Specificity of anti-human CD31 is shown in Supplementary Fig.VI b. Quantification of the human CD31+ vessels with αSMA+ pericytes versus murine vessels (human CD31−) with αSMA+ pericytes revealed that JAGGED1 silencing affected only the pericyte coverage of the human-CD31+ vessels (Fig.6c). These results confirmed that JAGGED1 in cbEPCs is responsible for the HemSC-derived vasculogenesis. In addition, we tested Matrigel explants for presence of GFP-HemSC-derived αSMA + pericytes. Explants of GFP-HemSC+shJAGGED1cbEPCs exhibited a significant reduction (p≤0.004) of GFP+/αSMA+ cells (compared to control, non transfected cells, and shNon targeting treated cells) (Representative pictures in Fig.6d, quantification in Fig.6e).

Figure 6. Downregulation of JAGGED1 in cbEPCs prevents pericytic differentiation and blood vessel formation.

(a) Immunoblot for JAGGED1 in cbEPCs infected with lentivirus encoding non-targeting-shRNA (shNon targeting), and JAGGED1-targeting-shRNA (shJagged1), pool 1 and pool 2. (b) shRNA-expressing-cbEPCs were co-implanted with HemSCs in nude mice; at day 7 MVD and CD31+ vessels were quantified. (c) Sections double-labeled with anti-α-SMA and anti-human CD31 to quantify αSMA+ pericytes in human (black bars) versus murine (gray bars) endothelium. An example of human CD31 (brown)/α-SMA (pink) double-staining; murine vessels are negative for human CD31 but are detected by presence of erythrocytes. (d) GFP (green) and αSMA (red) (top) staining of Matrigel explants from experiment in d. Phase contrast images show blood vessel location (arrows, bottom row). Scale bar=100 µM. (e) Quantification of vessels containing GFP-HemSC derived αSMA positive pericytes (* p=0.004 vs Control and p=0.0006 vs shNon targeting). (n=10 fields/section, 5 animals/group). (Data are expressed as means, normalized to control, of 2 independent experiments with n=6 and n=8). Data expressed as means ± s.e.m.

DISCUSSION

In this study we show that HemSCs can differentiate into pericytes in vitro and in vivo, suggesting pericytes in IH originate from HemSCs. Furthermore, we present the first evidence of a functional role for JAGGED1 in the pericytic differentiation of HemSCs, suggesting JAGGED1 may be involved in the pathogenesis of IH. Since we showed previously that HemSCs can differentiate into endothelial cells, the HemSC-to-pericyte differentiative capability provides, for the first time to our knowledge, evidence for a human post-natal vascular progenitor cell.

For many years, IH has been thought to originate from aberrant endothelium and excessive angiogenesis ^{38, 39}. However, attempts to create an IH model by injecting HemECs into immune-deficient mice have failed ⁷. Our HemSC-IH model represents an important advance to previously published models ^{40, 41} because the patient-derived HemSCs form hemangioma-like GLUT1+ ⁶ vessels within 7 days and adipocytes by 8 weeks ^{4, 7}, thereby recapitulating important events in the IH life-cycle. HemSCs have also been shown to be the cellular target of corticosteroids ⁸, supporting the critical role of HemSCs during the growth of IH. We propose that IH arises by HemSC-driven vasculogenesis, defined as the de novo formation of blood vessels from progenitor cells.

IH vessels are surrounded by closely associated perivascular cells that express markers of pericytes (NG2, PDGFR-β and αSMA) and smooth muscle cells (αSMA, calponin and smMHC) (Fig.1a,b). In contrast, tumor vessels have perivascular cells that are often loosely associated or appear detached ^42–45. Tumor blood vessels expand through an angiogenic program where VEGF-A negatively regulates perivascular coverage preventing tumor EC maturation ⁴⁶. In contrast, we show here that pericytes in IH are critical because inhibition of HemSC-to-pericyte differentiation severely compromised blood vessel assembly in our IH murine model (Fig.5d, Fig.6).

Embryonic vascular progenitor cells have been shown to differentiate into endothelial and pericytic phenotypes in response to VEGF-A and TGFβ or PDGF-BB ^30,32,47. In contrast, cellular contact with an endothelial cell, cbEPC or HemEC, was needed for HemSC-to-pericyte differentiation (Fig.3 and Fig.4). In addition, JAGGED1-expressing fibroblasts were sufficient to induce the HemSCs to acquire a pericyte-like phenotype (Fig.5e–f). Furthermore, soluble recombinant Fc-JAGGED1 thwarted differentiation of the HemSCs (Fig.5a–c) and inhibited blood vessel formation in the IH murine model (Fig.5d). This was perhaps due to disruption of juxtacrine interactions between endothelial JAGGED1 and a NOTCH receptor on HemSCs ¹⁶.

Silencing of JAGGED1 in the cbEPCs resulted in a dose-dependent decrease in the vascular density and pericytes in the IH murine model (Fig.6). Endothelial JAGGED1 is crucial for vascular smooth muscle development ²³,⁴⁸. Furthermore, angiogenic sprouting in the retina is inhibited in the endothelial Jag1 loss of function model ¹⁸. In this model, the authors reported a lack of αSMA+ smooth muscle cells on developing arteries but the pericyte coating seemed unaffected. Our results may differ because IH blood vessels form through a vasculogenic process, with angiogenic sprouting as secondary event.

We also tested inhibitors of receptor-ligand pathways that have been implicated in vascular development, including γ-secretase inhibitors to prevent NOTCH juxtacrine signaling. Recombinant Tie2, an ERK inhibitor (U0126) and a PDGFRβ inhibitor (AG1295) did not prevent HemSC-to-pericyte differentiation in the co-culture assay (Supplementary Fig.VII), but two different γ-secretase blockers (DAPT and Compound E) were inhibitory (Supplementary Fig.VIII). DAPT was tested in the IH murine model and resulted in reduced microvessel density and reduced α-SMA+ cells in proximity to blood vessels. Importantly, DAPT had little effect on proliferation or viability of HemSCs and cbEPCs in vitro. This supplemental data supports the idea that JAGGED1 is exerting an effect on NOTCH signaling.

In summary, our study demonstrates that HemSCs differentiate into pericytes as well as endothelial cells, and are thus akin to a vascular progenitor cell. We speculate that the first event of IH vasculogenesis is the differentiation of the HemSCs into HemECs. HemECs in turn induce neighboring HemSCs to become pericytes, thereby initiating a feed-forward mechanism for vascular assembly and further growth. Recruitment of circulating ECs, through VEGF-A and matrix metalloproteinase (MMP)-9, into the hemangioma could further contribute to the vasculogenesis ^{49, 50} and would diminish the need for HemSC-to-endothelial differentiation enabling them to differentiate into pericytes.

VEGF-A is highly expressed in HemSCs, and is required for the HemSC-derived blood vessel formation in the IH murine model ⁸. VEGF-A plays a positive role in angiogenesis in concert with NOTCH signaling ^51,52, furthermore authors reported VEGF-A as a negative regulator of pericyte and vascular smooth muscle cells recruitment ⁴⁶. In contrast, pericytes are prevalent in the proliferating phase of IH ⁸ despite high levels of VEGF-A (Fig.1a,b), Furthermore, preliminary data showed that when HemSCs and cbEPCs were cocultured, VEGF-A was upregulated, and DAPT inhibited this upregulation. We thereby speculate that in our model VEGF-A expression can also be regulated downstream of NOTCH signaling.

Non-standard abbreviations

IH

infantile hemangioma

SC

stem cells

cbEPCs

cord blood EPCs

MVD

microvessel density

vWF

von Willebrand factor