ADAM10 controls the differentiation of the coronary arterial endothelium

Abstract

The coronary vasculature is crucial for normal heart function, yet much remains to be learned about its development, especially the maturation of coronary arterial endothelium. Here, we show that endothelial inactivation of ADAM10, a key regulator of Notch signaling, leads to defects in coronary arterial differentiation, as evidenced by dysregulated genes related to Notch signaling and arterial identity. Moreover, transcriptome analysis indicated reduced EGFR signaling in A10ΔEC coronary endothelium. Further analysis revealed that A10ΔEC mice have enlarged dysfunctional hearts with abnormal myocardial compaction, and increased expression of venous and immature endothelium markers. These findings provide the first evidence for a potential role for endothelial ADAM10 in cardioprotective homeostatic EGFR signaling and implicate ADAM10/Notch signaling in coronary arterial cell specification, which is vital for normal heart development and function. The ADAM10/Notch signaling pathway thus emerges as a potential therapeutic target for improving the regenerative capacity and maturation of the coronary vasculature.

Introduction

Coronary circulation is crucial for supplying oxygen and nutrients to the heart and is therefore essential for the proper functioning of this vital organ. Since coronary artery disease is a major cause of mortality and morbidity, accounting for one in seven deaths in the United States [1], this provides a strong incentive to learn more about the development and homeostatic maintenance of the coronary vasculature. A better understanding of these processes has the long-term potential to help design improved approaches for repair or regeneration of diseased coronary vessels. The first steps of coronary vascular development are initiated by de novo vasculogenesis by endothelial progenitor populations originating in the endocardium and the sinus venosus [2–7]. This gives rise to vascular plexuses that develop into specific regions of the coronary vasculature in the endocardial or epicardial aspects of the heart [2,8,3]. As part of this process, the immature coronary endothelial cells differentiate into coronary veins, arteries and capillaries. A recent single-cell analysis of coronary artery development identified vein-like arterial precursor cells and delineated several steps along a gradual transition to arterial differentiation, yet much remains to be learned about the signaling mechanisms that promote the final steps in the development of mature coronary arterial endothelial cells [9].

The cell surface metalloprotease ADAM10 (a disintegrin and metalloprotease 10) is an essential regulator of physiological ligand-induced Notch signaling [10] and deletion of ADAM10 in endothelial cells results in defects in organ-specific vascular beds in mice, including the coronary vasculature, the developing retinal vasculature, the liver sinusoidal vessels, the glomeruli of the kidney, and the bone vasculature [11]. These vascular defects can be recapitulated by inactivating Notch1 in endothelial cells and Notch 4 systemically (N1ΔEC/N4−/−), corroborating that the primary signaling pathway regulated by endothelial ADAM10 is the Notch pathway [12]. Endothelial cells have been shown to mainly express Notch1 and Notch4 as well as the ligands Dll4 and Jagged1 [13–15]. Proper heart development requires both Dll4- and Jag1-mediated Notch signaling [16,7], and deletion of components of the Notch pathway result in heart valve malformations, aberrant coronary artery formation and embryonic lethality [11,12,16,17,7]. Since A10ΔEC mice resemble N1ΔEC/N4−/− mice [11,12], A10ΔEC animals provide a unique opportunity to examine the consequences of inactivating input from Notch signaling into coronary endothelial cells, regardless of the Notch ligand(s) or cell type activating this signaling.

Notch signaling is known to mediate arterial differentiation in the embryo [18] and in the yolk sac [19], raising questions about the role of this signaling pathway in arterial differentiation of the coronary vasculature. Therefore, the main goal of the current study was to examine the coronary vasculature in A10ΔEC mice to learn more about the contribution of ADAM10 to the development of coronary vessels and to determine whether it could be involved in arterial maturation. First, we performed histopathological, functional and immunofluorescence analyses to better understand of how the inactivation of ADAM10 affects the differentiation and function of the heart. Transcriptome-wide gene expression analysis with RNA-seq on enriched coronary endothelial cells from A10ΔEC mice revealed the contribution of ADAM10 to the coronary angiogenesis expression program in these cells. We contextualized these results by comparing them with previously reported studies on mice with mutations in other components of the Notch signaling pathway (e.g. Jag1 or 2, Dll4, mindbomb [16,20,21]). Since ADAM10 also cleaves other membrane proteins beside the Notch receptors, we asked whether the lack of processing of other known substrates of ADAM10 might become apparent through pathway analysis of the differentially expressed genes. Our results provide the first evidence, to our knowledge, for a homeostatic role of endothelial ADAM10 in protective EGFR signaling in the heart and they suggest that ADAM10 controls the final steps of arterial differentiation in mouse coronary vessels.

Results

A10ΔEC mice display abnormal myocardial compaction

Previous studies identified defects in the coronary vasculature of adult A10ΔEC mice [11], raising questions about how this affects the myocardium and whether any defects are also present earlier, at birth. Analysis of the H&E stained sections from 6-week-old animals revealed an abnormal myocardial compaction with dilated capillaries in the outer parts of the myocardium of A10ΔEC animals compared to controls (Fig. 1a–d). Structural defects in the myocardium were also present in several newborn animals (P0), which displayed abnormal myocardial compaction (see Fig. 1e compared to 1f, representative for 2 out of 4 mutant animals) as well as enlarged dilated capillaries in the myocardium that were not evident in controls (Fig. 1g, h, CD31 stained sections shown in Figure 1i, j, arrows indicate enlarged vessels in panels h and j). However, newborn animals did not appear to have the enlarged subepicardial vessels that are present in adult animals [11] (Figure 1f). A10ΔEC hearts occasionally display regions of focal necrosis and calcification in adults (Suppl. Fig. 1a–c, 1 out of 2 mutant animals) and focal necrosis in newborn animals (Suppl. Fig. 1d, 2 out of 4 mutant animals). Previous studies have implicated Notch signaling in heart valve morphogenesis and myocardial compaction^12–15. However, the aortic, pulmonary and mitral valves of adult or newborn A10ΔEC mice had a comparable appearance to controls (representative adult aortic and mitral valves shown in Suppl. Fig. 2a–d, representative newborn aortic and pulmonary valves shown in Suppl. Fig. 2e–h).

Fig. 1. A10ΔEC mice display abnormal myocardial compaction.

H&E-stained sections of six-week-old control (a, c) and mutant hearts (b, d) show abnormal myocardial compaction in the parts of the outer (epicardial) aspects of the myocardium in mutant A10ΔEC hearts (b, d) (parts of the left ventricular myocardium are shown in a – d). H&E-stained sections of newborn (P0) A10ΔEC hearts (f, h) also display abnormal myocardial compaction and contain enlarged myocardial capillaries (pointed by yellow arrows) when compared to control hearts (e, g). PECAM-stained sections show normal staining of myocardial capillaries in a control heart (i), but a less organized staining with some abnormally enlarged capillaries in a mutant A10ΔEC heart (pointed by black arrows (j). Images are representative of sections of hearts from at least 3 separate animals per age and genotype. Scale bars a, b; 200 microns. c, d, g - j; 20 microns. e, f; 100 microns.

A10ΔEC adult hearts display insufficient systolic contraction and are enlarged relative to body weight

Echocardiographic analysis of the hearts of A10ΔEC mice uncovered a 20% decrease in fractional shortening due to insufficient systolic ventricular contraction (Fig. 2a, b). These results provide the first direct evidence for a functional manifestation of the coronary endothelial abnormalities in A10ΔEC mice. In addition, the heart/body weight ratio in A10ΔEC mice was significantly increased (16.6%, SEM 4.5%) compared to control littermates (Fig. 2c, the mouse body and heart weights are shown separately in Suppl. Fig. 3a and b). Nevertheless, the echocardiographic measurements showed that the interventricular septal thickness, posterior wall thickness, and left ventricular mass were comparable between mutant and control mice (Suppl. Fig. 3c–e). When we examined cardiomyocytes using fluorescent cell surface staining with wheat germ agglutinin (WGA) to identify possible hypo- or hypertrophy of these cells, we found that the cross-sectional area of cardiomyocytes in A10ΔEC animals was comparable to controls, arguing against cardiomyocyte hypertrophy as a cause for the enlarged hearts in these animals (Fig. 2d, e, quantification in f, see materials and methods for details).

Fig. 2. A10ΔEC adult hearts are enlarged and display insufficient systolic contraction.

A10ΔEC animals have decreased fractional shortening (a) and abnormal systolic contraction (b) relative to control animals (n=7 for each genotype, p = 0.012). The ratio of heart mass relative to body mass in A10ΔEC mice is larger than in control animals (c) (n=18 for each genotype, p = 0.0008). Representative images of adult cardiomyocytes stained with the lectin Wheat Germ Agglutinin (WGA) from control (d) and A10ΔEC animals (e). The images in (d) and (e) are representative of sections of hearts from 3 animals per genotype. The average mean cross-sectional area covered by individual myocardial cells, as well as the 75^th percentile and 90^th percentile area of these measurements for WGA-stained control and A10ΔEC cardiomyocytes was comparable (f, see methods for details). Scale bars 100 microns.

A10ΔEC mutant coronary vessels show higher endomucin expression and are Vegfr3 positive

To further characterize the coronary vasculature in A10ΔEC mice, we retro-orbitally injected a fluorescently tagged lectin to determine vascular perfusion (Fig. 3a, b). This demonstrated a comparable perfusion of the smaller myocardial vessels in A10ΔEC mice and controls. However, the abnormal sub-epicardial vessels that are found in A10ΔEC mice did not appear as well perfused as normal vessels (Fig. 3b, white arrows). Staining with anti-VEGFR2 antibodies showed a similar distribution of VEGFR2 positive vessels in mutant and control mice (Fig. 3c, d). Visualization of hypoxia in A10ΔEC hearts with hypoxyprobe revealed regions of hypoxia that were concentrated in the areas of abnormal compaction (Suppl. Fig. 4). Staining with alpha smooth muscle actin (α-SMA, Suppl. Fig. 5a) showed that most enlarged subepicardial vessels in A10ΔEC mice have a thin layer of mural cells, which is reminiscent of α-SMA staining of veins (white arrows in Suppl. Fig. 5a). Moreover, the ensheathment of larger arteries with α-SMA appeared comparable in A10ΔEC mice and controls (Suppl. Fig. 5b). Interestingly, the A10ΔEC coronary endothelium displayed stronger staining for endomucin, a marker of vein-like endothelium [22], compared to controls (Fig. 3e, f). The enlarged subepicardial vessels in A10ΔEC mice stained with anti-VEGFR3, but not with the lymphatic marker LYVE (Fig. 3g–l) indicating that they are not lymphatic vessels. This is of particular note since VEGFR3 is a marker of immature coronary endothelium[7] and VEGFR3 expression in control hearts overlaps partially with lymphatic vessels [23,24] (Fig. 3k, l).

Fig. 3. Evaluation of perfusion and endothelial cell identity markers in A10ΔEC hearts.

(a, b) The lectin perfusion of myocardial vessels in A10ΔEC animals appears similar to that of control animals. However, the abnormally enlarged sub-epicardial vessels (white arrows) did not appear to be as well perfused as normal vessels. (c, d) Vegfr2 staining is comparable in an A10ΔEC heart relative to a control. (e, f) A10ΔEC hearts have higher and more uniform levels of endomucin staining throughout the epicardium and myocardium. (g, h) The abnormal and enlarged subepicardial vessels in A10ΔEC mice stain with VEGFR3, a marker for immature arterial endothelial cells, but not LYVE, a marker for lymphatic vessels (i, j), merged images in (k, l). Each image is representative of sections from at least 3 separate animals per genotype. Scale bars: 100 microns.

RNA-seq of enriched A10ΔEC coronary endothelial cells shows immature arterial identity, evidence for cardiomyopathy and altered Notch signaling

To learn more about potential changes in the transcriptional signature of coronary vessels from A10ΔEC mice, we performed RNA-seq on poly A-captured mRNA from coronary endothelial cells enriched from six-week-old mutant animals and littermate controls. This uncovered 701 differentially expressed genes (428 upregulated and 273 downregulated, FDR < 0.05, Fig. 4a), including several genes that are involved in Notch signaling (Hey1, Smad6, Gja4 and Notch4). Hey1 and Smad6 were downregulated in A10ΔEC samples, corroborating the synergistic expression of Smad6 and Notch signaling¹⁸. Interestingly, Notch4 was upregulated in A10ΔEC samples, perhaps to compensate for the lack of ADAM10/Notch1 signaling. In addition, we observed lower expression of VEGFR1 as well as VEGFb and VEGFc, the latter of which is known to have a role in coronary artery and capillary growth^1,19. Moreover, additional genes that are considered to be markers for arterial endothelium were downregulated (Gja4, Igfbp3, and Hey1). We also found that the Apelin receptor (Aplnr) was upregulated in A10ΔEC samples, as previously observed in glomerular endothelial cells isolated from these animals²⁰. In addition, we noted a differential regulation of tetraspanin14 (Tspan14), an ADAM10-interacting protein that could potentially be involved in regulating ADAM10-dependent processing of Notch receptors²². Finally, components of several disease- and signaling pathways were significantly enriched for in differentially expressed genes (Fig. 4b). The most prominent pathways were those seen in dilated cardiomyopathy, hypertrophic cardiomyopathy, cGMP/PKG signaling (involved in Calcium signaling), and vascular smooth muscle contraction [25].

Fig. 4. A10ΔEC heart endothelial cells differentially express genes related to Notch signaling and cardiomyopathies.

(a) Scatterplot of 701 differentially expressed genes. Select genes related to the Notch signaling pathway are highlighted. (b) Top 10 pathways found to be dysregulated in A10ΔEC coronary endothelial cells compared to controls by Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. (c) Top 10 enriched gene sets found when the A10ΔEC data set was compared to the Library of Integrated Network-Based Cellular Signatures (LINCS l1000) database.

LINCS analysis provides evidence for dysregulated EGFR signaling in A10ΔEC mice

ADAM10 has several other substrate proteins besides Notch [26,27], including the EGFR ligands BTC and EGF [28,29] and the amyloid precursor protein [30]. We were therefore interested in determining whether other signaling pathways, which do not directly depend on the Notch pathway, might be affected in A10ΔEC endothelial cells. When we entered the 701 differentially regulated genes into the LINCS database (http://www.lincsproject.org/), split into upregulated and downregulated genes, we found a highly significant overlap with gene sets that are dysregulated in tumor cells treated with ligands of the EGF-receptor (EGFR), such as betacellulin (BTC), TGFα (TGFA), EGF and epiregulin (EPR) (Fig. 4c). Since all of these ligands can activate the EGFR (ErbB1), these results provide evidence for a possible dysregulation of the EGFR signaling pathway in A10ΔEC endothelial cells. Moreover, these findings are consistent with the known role of ADAM10 as the primary sheddase of BTC and EGF [28,31]. Interestingly, two of the three the other signaling protein-dependent pathways implicated in this comparison (Gas6, and HRG) rely on receptors that can be shed by ADAM10 (the Gas6 receptor Axl [32] and the HRG receptor c-Met [33]).

Comparison of gene sets dysregulated in A10ΔEC to other Notch pathway mutants

ADAM10 is a critical regulator of physiological ligand-induced Notch signaling (reviewed in [10]), and the vascular phenotype of A10ΔEC resembles that of Notch1ΔEC/Notch4−/− mice [12,11]. We therefore compared the A10ΔEC coronary endothelial gene set gene expression measurements from mice lacking other components of the Notch signaling pathway in endothelial cells or in other cell types in the heart^{13 24}, as well as with models for heart development and disease (Fig. 5a)²⁵. Importantly, we took the direction of change of the differentially expressed genes into account when assessing the similarity with other gene sets (see materials and methods for details). The highest level of similarity with the A10ΔEC coronary endothelial gene set was with developing heart endothelial cells, consistent with a role for ADAM10/Notch signaling in promoting the maturation of coronary endothelial cells. In addition, there was a significant overlap of the A10ΔEC coronary endothelial gene signature with that of L-type relative to H-type endothelial cells in bone [21]. Since ablation of endothelial Notch signaling results in more L-type endothelial cells in bone at the expense of H-type cells^17,24,26, this suggests that these different Notch-dependent endothelial cell fate decisions share at least some common underlying features (Figure 5a).

Fig. 5. A10ΔEC gene set enrichment analysis compared to other mouse models.

(a) A10ΔEC differentially expressed genes compared to other published gene sets derived from mutations that affect different components of the Notch signaling pathway, * p<0.05. (b) Venn-diagram comparison of A10ΔEC to Dll4-Tie2Cre and Jag1Jag2cTnT data sets [16]. (c) Venn diagram comparison of data from A10ΔEC to Dll4-Tie2Cre and to genes enriched in L-type vs. H-type bone endothelium [21]. (d) Summary of overlapping genes from (b) and (c).

When we compared the A10ΔEC RNA-seq data set with results from other studies on Notch signaling in the heart, we found significant similarity with data from mice in which the Notch ligand Dll4 was inactivated by a pan-endothelial deletion (Tie2-Cre) or endocardial deletion (Nfatc-Cre) et E9.5 (Fig. 5a) [16]. The differentially expressed genes observed upon deletion of Dll4 with Tie2-Cre are more similar with the A10ΔEC gene set than with data from endocardial cells lacking Dll4. Moreover, when compared to the A10ΔEC data set, the genes affected by Notch1 deletion in the endocardium are not as similar as those affected by endocardial deletion of Dll4, consistent with our previous finding that endothelial deletion of Notch1 and deletion of Notch4 is required to recapitulate the full range of vascular phenotypes seen in A10ΔEC animals⁹. Knockout of manic fringe (Mfng) and mindbomb (Mibc) also resulted in similar differential gene expression when compared to the A10ΔEC gene set. Interestingly, the similarity to the data set with the cumulative effects of deleting the Notch ligands Jag1 and Jag2 in cardiomyocytes (cTNT-Cre) was stronger (16 genes similarly affected) than the comparison to data from individually deleting Jag1 or 2 (only 3 genes similarly affected when combined), suggesting a partial functional redundancy of Jag1 and 2 with respect to activating ADAM10-dependent Notch signaling in coronary endothelial cells. Finally, the overlap with differentially expressed genes in isolated endothelial cells from hearts following a myocardial infarct was not statistically significant, consistent with the notion that the A10ΔEC phenotype is mostly likely developmental or chronic in nature, but not acute.

Identification of common transcriptional changes in selected data sets

Jagged and Delta-like ligands have been implicated in different stages of Notch signaling in retinal endothelial cells, with Dll4 implicated in tip-stalk cell signaling, and Jagged1 in modulating the interaction between Dll4 and Notch [34,35]. Since inactivation of ADAM10 in coronary endothelial cells should block input from these two structurally and functionally different Notch-ligands [36,37], we compared the differentially expressed genes from all three analyses to identify dysregulated genes that they have in common. Such genes could serve as markers of for ADAM10/Notch-dependent endothelial cell fate decisions. Out of the 4 genes uncovered in this analysis (Fig. 5b), Adora2a, Kcne3, Glul and Gramd1b, the first two are consistent with defects in Notch signaling. Specifically, Adora2a is involved in pathological neovascularization²⁷, whereas Kcne3 is a marker for tip cells [38]. In addition, we found that Gmfg was upregulated in the A10ΔEC, Dll4-Tie2Cre, Bone Vasculature and Developing Heart data sets (Fig. 5c, 5d, Suppl. Fig. 6A). Gmfg is linked to tip cells²⁸ and blood islands found in developing yolk sacs²⁹. Gja4, a known arterial marker and Notch target gene [39], is one of the nine genes present in the L-type to H-type, Dll4-Tie2Cre, and A10ΔEC data sets (Fig. 5d).

Finally, we wanted to determine whether the gene expression profile after inactivation of Dll4 in the endocardium (Dll4-NfatcCre) is similar to the gene expression profile of the coronary endothelium in A10ΔEC mice (A10-Tie2Cre) and Dll4-Tie2Cre mice. Using the A10ΔEC data set as a point of comparison, we analyzed the genes that are only found in the Dll4-Tie2Cre and A10ΔEC data sets, but not in the Dll4-NfatcCre data set, based on the assumption that the latter is more representative of endocardially-derived endothelial cells. We found 44 differentially expressed genes (12 downregulated and32 upregulated), including Gmfg and AplnR receptor, which can be considered potential markers for ADAM10/Notch signaling in the epicardial coronary vasculature. (Suppl. Fig. 6B). Finally, 23 differentially expressed genes (7 downregulated and 16 upregulated) were found in all three gene sets, including the known Notch signaling targets, Adora2a, Kcne3 and Gja4.

Discussion

The premise for this study was our previous observation that inactivation of ADAM10 in endothelial cells results in the development of abnormal coronary vessels [11]. To learn more about the nature of these defects and their underlying causes, we performed a histopathological and functional analysis of the hearts of A10ΔEC mice and a transcriptional analysis of coronary endothelium from these animals. The functional studies uncovered a reduction in fractional shortening of A10ΔEC hearts, providing the first evidence that their coronary vascular abnormalities have functional consequences. The general perfusion of the heart, as assessed by lectin staining, seemed normal, although the coronary vasculature in CD31-stained sections had a different appearance in the mutant animals. The presence of areas of hypoxia and, in some cases, focal necrosis in the mutant hearts could provide an explanation, at least in part, for the reduced fractional shortening, which could be further exacerbated by the lack of myocardial compaction in the subepicardial aspect in A10ΔEC hearts.

Our gene expression analysis revealed that the A10ΔEC coronary endothelial cells displayed a transcriptional profile resembling that of endothelial cells in developing hearts. This suggests that ADAM10/Notch signaling is important for the maturation of the coronary endothelium. Recent studies have shown that coronary endothelial cells are initially vein-like, and then differentiate through a series of steps into arterial and venous vessels⁷. The transcription factor COUP-TFII is crucial for the first step of differentiation of veins from the coronary vascular progenitors, but the final steps towards arterial differentiation have not yet been fully defined. Our observations that arterial markers (e.g. Gja4, Hey1, Igfbp3) are downregulated in A10ΔEC coronary endothelia relative to controls support the notion that ADAM10-Notch-dependent signaling promotes the acquisition of the mature coronary arterial cell fate (see Fig. 6). Additionally, the presence of VEGFR3-positive enlarged vessels in the subepicardial portion of mature A10ΔEC hearts further serves as an indicator that these vessels contain immature coronary arterial cells, and potentially coronary angiogenic precursors [7].

Fig. 6. Schematic of artery-vein specification of coronary endothelium.

Coronary progenitor cells originating from the sinus venosus and endocardium first undergo a cell fate decision, regulated by COUP-TFII, to become veins or pre-artery cells. To fully differentiate into mature arterial endothelium, pre-artery cells must then undergo ADAM10-dependent canonical Notch signaling.

ADAM10 is crucial for activating Notch signaling in cis, i.e. in the same cell that is expressing a given ligand-activated Notch receptor [40,41]. Consequently, inactivating ADAM10 in endothelial cells should block the input from ligand-induced canonical Notch signaling into these cells, regardless of the ligand or cell type providing the input. This notion is supported by the significant overlap with differentially expressed genes from mice lacking other components of the Notch signaling pathway in coronary endothelial cells or in other cell types in the heart. Specifically, inactivation of the Notch-ligand Dll4 in endothelial cells affected several of the same genes as inactivation of ADAM10. In addition, targeting both Notch ligands Jagged1 and 2 in cardiomyocytes also resulted in transcriptional changes that significantly overlapped with those observed in A10ΔEC mice, but had almost no overlap in Dll4 mutant mice. Thus, different ligands can apparently elicit distinct ADAM10-dependent Notch signals. Presumably, this can be explained by differences in the spatiotemporal expression of different Notch ligands as well as by the context provided by the differentiation states of the signal-receiving cells. On the other hand, four genes were similarly affected in A10ΔEC as in mice carrying mutations in Dll4 or Jag1/2, two of which are known to have prominent roles in Notch signaling and tip/stalk cell fate decisions (Kcne3 and Adora2a) [38,42]. These results suggest that that inactivation of ADAM10 in endothelial cells also affects a limited number of common genes that respond to ligand-induced Notch signaling, regardless of the ligand and Notch receptor. Thus, inactivation of ADAM10 in endothelial cells apparently represents an efficient approach to inactivate canonical Notch signaling in these cells.

ADAM10 is known to have numerous other substrates besides the Notch receptors, including the amyloid precursor protein and the EGFR-ligands BTC and EGF [26–30]. Nevertheless, Notch is considered to be a principal substrate of ADAM10, at least during mouse development, since almost all of the developmental phenotypes caused by inactivation of ADAM10 signaling can be explained by the resulting lack of Notch signaling (reviewed in [10]). It was therefore interesting to find that A10ΔEC endothelial cells showed gene expression signatures consistent with defects in EGFR signaling. Ligands of the EGFR, such as BTC, EGF or TGFα, are made as membrane-anchored precursors that must be processed, typically by ADAM10 or the related ADAM17, to activate the EGFR (reviewed in [31]). Most phenotypes of mice lacking ADAM17 can be attributed to a defect in EGFR signaling [43–47], which is consistent with the essential role of ADAM17 in processing several EGFR ligands, including TGFα, HBEGF and Amphiregulin [45,44,43]. However, as noted above, two out of the seven known ligands of the EGFR, BTC and EGF, are preferentially processed by ADAM10 [28,29], which is consistent with a possible role for ADAM10 in cardiac EGFR signaling. Interestingly, EGFR inhibitors used to treat cancer can cause cardiomyopathy [48–50], providing further support to the notion that EGFR signaling is cardioprotective. These results are thus the first to suggest a possible role of ADAM10 in protective EGFR signaling in the heart, most likely through the release of EGF or BTC, which have been reported to be expressed in endothelial cells (Bend3 [51]). However, we cannot rule out possible indirect effects of ADAM10 on EGFR signaling, e.g. through a Notch-dependent process that affects other components of the EGFR signaling pathway.

ADAM10/Notch signaling-dependent endothelial cell fate decisions are known to be crucial for the development of organ-specific vascular beds, such as in the glomerular and bone endothelium^17,20. The mature glomerular and coronary vasculature is thought to derive from progenitor cells that arise from outside the main vascular tree and must undergo vasculogenesis in order to generate these specialized vascular structures (reviewed in [10]). This is conceptually similar to the vasculogenesis events that represent the first steps of vascular development in the yolk sac or in the embryo, which require Notch signaling [18,10]. Interestingly, the expression of Gmfg, a marker of the blood islands of the developing yolk sac²⁹, is increased in both the heart and bone endothelium in A10ΔEC mice, further supporting the notion that there is an impediment in vascular maturation in these specialized vascular niches. On the other hand, there are also key differences in the dysregulated genes in different vascular beds of A10ΔEC mice, presumably because of the specific requirements of the final differentiation state of a given organ. An example is provided by glomerular endothelial cells [52], where ADAM10/Notch signaling regulates glomerular fenestrae formation and the vascular morphology of glomerular tufts²⁰. Notch signaling also mediates the transition from L-type to H-type bone endothelium, which is crucial to long bone growth²⁴. In the case of the coronary endothelium, the results presented here support a model in which ADAM10/Notch signaling controls the final maturation of arterial endothelial cells (Figure 6). The inability of the coronary arteries to mature properly in A10ΔEC mice correlates with abnormal myocardial compaction, enlarged heart mass relative to body mass as well as heart dysfunction, illustrating the importance of the differentiation of arterial coronary endothelium for normal heart development and function.

Providing a better understanding of the cellular signaling pathways involved in the maturation of the coronary vascular network is crucial to understanding cardiac development and to help uncover new approaches for harnessing the regenerative potential of the coronary endothelium. In addition, studies on the differentiation of coronary vessels have the potential to elucidate the homeostatic signaling provided by fully differentiated coronary endothelial cells in normal and pathologic states [53]. This study indicates that the modulation of the endothelial ADAM10/Notch signaling pathways could be a potential tool for future therapeutic approaches to promote arterial differentiation in coronary disease and organ regeneration. Furthermore, our data suggests that ADAM10 plays a crucial role in the acquisition of organ-specific endothelial identity in several distinct vascular structures, warranting future investigations into the common themes that unite these unique vascular beds.

Materials and Methods

Immunofluorescence Reagents.

The antibodies against VEGFR3 (goat anti-mouse, R&D Systems (Minneapolis, MN) #AF743) and VEGFR2 (goat anti-mouse, R&D Systems #AF644) were used at a 1:50 dilution; the anti-LYVE (rabbit anti-mouse Abcam,Cambridge, MA) #ab14917) was used at 1:250; the anti-αSMA (mouse anti-mouse Dako,Santa Clara, CA) #M0851) was used at 1:100, and anti-PECAM/CD31 antibody (rat monoclonal, catalog #DIA-310; Dianova, Hamburg, Germany) was used at 1:250. Secondary antibodies were Donkey anti-rabbit 594 (#21207 ThermoFisher, Grand Island, NY); Donkey anti-goat 488 (Abcam #ab150129); Donkey anti-goat 546 (ThermoFisher #A11056); and Donkey anti-mouse 488 (Abcam #ab150105) were all used at a dilution of 1:250. Sections from a total of 3 pairs of A10ΔEC mice and littermate controls were analyzed per antibody, with at least one pair of littermates of each gender.

Mice.

The A10ΔEC mouse strain used here has been previously described [52]. Adult female mice of mixed genetic background (129Sv/C57Bl6) carrying two floxed alleles of ADAM10 were mated with male mice carrying two floxed alleles of ADAM10 and one allele of the Tie2-Cre transgene [54]. Male or female offspring from such a mating containing two alleles of floxed ADAM10 and one allele of the Tie2-Cre transgene were referred to as A10ΔEC mice, whereas littermates with two floxed alleles of ADAM10, but no Tie2-Cre, were used as controls. All experiments were performed with gender-matched littermate male or female mice, as indicated. All procedures were approved by the Animal Care and Use Committee of Weill Cornell Medicine.

Histopathologic examination.

Hearts were fixed by immersion in 10% neutral buffered formalin, processed routinely in alcohol and xylene, embedded in paraffin, sectioned at 5-micron thickness, and stained with hematoxylin and eosin. Sections were examined and images prepared by a board-certified veterinary pathologist (SM). Sections from a total of 2 pairs of A10ΔEC mice and littermate controls were analyzed for 6-week-old mice, and 4 pairs for newborn (P0) mice.

Chromogenic immunohistochemistry.

Chromogenic immunohistochemistry for PECAM-1 (CD31) on formalin fixed paraffin embedded sections was performed on a Leica Bond RX automated staining platform (Leica Biosystems). Following heat-induced epitope retrieval at pH 9.0, the primary antibody was applied at a concentration of 1:250 and was followed by application of a rabbit anti-rat secondary antibody, mouse adsorbed (Vector Laboratories BA-4001), and a polymer detection system (Novocastra Bond Polymer Refine Detection, Leica Biosystems). The chromogen was 3,3 diaminobenzidine tetrachloride (DAB), and sections were counterstained with haematoxylin. Sections were examined and images prepared by a board-certified veterinary pathologist (SM). Sections from a total of 4 pairs of newborn (P0) A10ΔEC mice and littermate controls were analyzed.

Echocardiogram.

Echocardiography was performed on 6-week-old adult mice using the Vevo 770 Imaging system (VisualSonics, Toronto, Canada). The mice were anesthetized using isoflurane, and measurements were collected for 3 to 6 consecutive cycles, as previously described [55]. In short, left ventricle contraction (diastolic (LVd) and systolic dimensions (LVs)) was measured using data collected in M-mode traces. Fractional shortening was calculated by the following equation: ([LVd-LVs]/LVd × 100). A total of 14 mice were analyzed (7 control and 7 A10ΔEC), each A10ΔEC mutant animal was matched to a littermate.

Retro-orbital injection.

Mice were anesthetized with isoflurane and then injected retro-orbitally with 100μl of a lectin solution (for 1ml; 900μl PBS and 100μl Lycopersicon esculentum tomato lectin (Vector Labs #DL-1178)), left in a cage for 10 minutes and then sacrificed so that the tissues could be prepared for further analysis.

Immunofluorescence.

The hearts were fixed in 4% paraformaldehyde overnight at 4°C, then dehydrated in 15% sucrose followed by 30% sucrose overnight. The tissue was then frozen and sectioned at 12 microns. Slides with frozen heart sections were incubated in blocking solution (1% donkey serum (Jackson ImmunoResearch, West Grove, PA) in 0.5% Triton X-100 and 0.1% Saponin in PBS (TSP)) for 1 hour at 37°C. Slides were then treated with primary antibodies in blocking solution for 1 hour at 37°C. Next, slides were washed in TSP three times for five minutes at room temperature, then treated with secondary antibodies in blocking solution for 1 hour at 37°C. After the staining with secondary antibodies, the slides were washed 3× in TSP for five minutes each at room temperature and mounted with Prolong Diamond solution (ThermoFisher #P36965) and a coverslip. The images were collected on a Nikon Ni-E microscope with an Andor Zyla camera. All images were analyzed using either Nikon NIS Elements or ImageJ software.

Hypoxyprobe.

Three pairs of 6-week-old adult mice (2 male pairs and 1 female pair) were injected with 60mg/kg of Hypoxyprobe solution via intraperitoneal injection. After 1 hour, the mice were sacrificed, their hearts isolated and then fixed with 4% PFA overnight. The hearts were dehydrated with 15% and 30% sucrose overnight, then frozen and sectioned with a thickness of 12 μm. The resulting sections were blocked and then stained with Hypoxyprobe-Red549 at a dilution of 1:200 (Hypoxyprobe #HP7–100kit, Hypoxyprobe, Burlington, MA).

Isolation of Coronary Endothelial Cells.

Mice were sacrificed with CO₂ following the guidelines of the American Veterinary Association. Mouse hearts were collected and washed in PBS to remove excess blood. Samples were then minced into small pieces and moved to a 50 ml Falcon tube, digested in 3 ml solution consisting of 1mg/ml Collagenase Type II (Sigma #C6885, St. Louis, MO), 1ml of Dispase (Stemcell #07923, Cambridge, MA), and 1ml PBS. Samples were digested for an hour at 37°C and resuspended and mixed with a pipette after 30 and 60 minutes. The dissociated tissue was then filtered through a 70 μm mesh sieve (Corning #352350, Corning, NY) and washed three times with PBS. The suspension was spun down in a centrifuge for 8 minutes at 200 × g, then the supernatant removed, and the pellet resuspended in PBS, followed by a second identical washing step. The final pellet was resuspended in 1ml wash buffer (PBS with 0.5% BSA and 2 mM EDTA) and moved to a 5 ml round bottom Falcon tube. Magnetic Dynabeads (ThermoFisher #11035) with coupled Pecam antibodies (BD Biosciences #553370, Sparks, MD) were added to the resuspended pellet, and the suspension left on a rotating incubator for 40 minutes at 4°C. The beads with bound cells were isolated using a Dynal magnet and washed 5 times to remove loosely bound cells while the tube was still attached to the magnet. The RNA from the isolated cells was purified using an RNeasy isolation kit (Qiagen #74104, Germantown, MD)

RNA Sequencing.

Library preparation and RNA sequencing was performed on 6 samples from 3 control and 3 A10ΔEC animals (a total of 3 pairs of gender matched littermates from different litters, 2 male pairs, and 1 female pair) by the Weill Cornell Genomic Core Facility. The libraries were prepared using TruSeq Stranded mRNA Library Prep (poly-A selection; Illumina #20020594, San Diego, CA). The sequencing was performed on an Illumina HiSeq4000 with 50 base pair single-end reads.

RNA Sequencing Data Analysis.

Raw sequencing reads were aligned to the GRCm38 reference genome with STAR [56]. Gene expression was quantified with RSEM [57]. Differential expression was assessed with limma using voom with observations weights [58]. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway [59] enrichment was determined with the Fisher’s Exact Test. Permutation tests were performed to test for statistically significant overlaps between two differentially expressed gene sets. In particular, for i = 1,2 consider differentially expressed gene set S_i consisting of n_u,i upregulated genes and n_d,i downregulated genes. The aim is to test if the total number of commonly regulated genes between S₁and S₂, i.e. upregulated in both S₁ and S₂ or downregulated in both S₁ and S₂, is higher than expected by chance. For n = 10⁶ permutations, we sampled n_u,2 genes and n_d,2 genes and counted the total number overlapping genes by comparing to those upregulated and downregulated in gene set S₁, respectively.

Quantification of Cardiomyocyte Area.

Frozen sections of adult heart (12 μm) were stained with fluorescently tagged wheat germ agglutinin (Sigma #L4895). The stained tissues were then imaged, with 3 images per animal. Each image was of a section from a different region of the heart (ranging from the apex to base). Regions were isolated from the images of the left ventricle and subjected to the morphological segmentation plugin in MorphoLibJ (ImageJ) to record the area of the segmented cardiomyocytes using the measure particle feature. Only the smallest cross-sectional areas of the measured cardiomyocytes were used in statistical analysis. These values were screened from the obtained data set by solely analyzing cardiomyocytes whose longest dimension was only twice the shortest dimension. A total of 3 six-week-old male littermate pairs were analyzed, each consisting of one A10ΔEC and one control animal.

Statistical analysis.

A 2-tailed t-test performed with GraphPad Prism 7 was used to measure statistical significance of the microscopy or physiological and functional experiments. P-values of 0.05 or less were regarded as statistically significant.