Endothelial deletion of ADAM17 in mice results in defective remodeling of the semilunar valves and cardiac dysfunction in adults

Abstract

Global inactivation of the metalloproteinase ADAM17 during mouse development results in perinatal lethality and abnormalities of the heart, including late embryonic cardiomegaly and thickened semilunar and atrioventricular valves. These defects have been attributed in part to a lack of ADAM17-mediated processing of HB-EGF, as absence of soluble HB-EGF results in similar phenotypes. Because valvular mesenchymal cells are largely derived from cardiac endothelial cells, we generated mice with a floxed Adam17 allele and crossed these animals with Tie2-Cre transgenics to focus on the role of endothelial ADAM17 in valvulogenesis. We find that although hearts from late-stage embryos with ablation of endothelial ADAM17 appear normal, an increase in valve size and cell number is evident, but only in the semilunar cusps. Unlike Hbegf−/− valves, ADAM17-null semilunar valves do not differ from controls in acute cell proliferation at embryonic day 14.5 (E14.5), suggesting compensatory processing of HB-EGF. However, levels of the proteoglycan versican are significantly reduced in mutant hearts early in valve remodeling (E12.5). After birth, aortic valve cusps from mutants are not only hyperplastic but also show expansion of the glycosaminoglycan-rich component, with the majority of adults exhibiting aberrant compartmentalization of versican and increased deposition of collagen. The inability of mutant outflow valve precursors to transition into fully mature cusps is associated with decreased postnatal viability, progressive cardiomegaly, and systolic dysfunction. Together, our data indicate that ADAM17 is required in valvular endothelial cells for regulating cell content as well as extracellular matrix composition and organization in semilunar valve remodeling and homeostasis.

1. Introduction

Formation and remodeling of the cardiac valves is a complex process that involves many regulatory pathways (Butcher and Markwald, 2007; Combs and Yutzey, 2009a; DeLaughter et al., 2011; Hinton and Yutzey, 2011; Snarr et al., 2008). Development of the valve precursors from the endocardial cushions has been extensively studied, with >100 factors identified as mediating cushion formation (Person et al., 2005). Endocardial cushions are initially acellular and consist of a cardiac jelly, rich in the glycosaminoglycan hyaluronan and proteoglycans such as versican, covered by a layer of endothelial cells (Lockhart et al., 2011). Upon signaling by specific myocardial-derived factors, including bone morphogenetic proteins (BMPs) and members of the transforming growth factor-β family (TGFβs), the endocardial cells transdifferentiate into mesenchymal cells, which invade the cardiac jelly and undergo proliferation and differentiation (Lockhart et al., 2011; Schroeder et al., 2003). Other cell types also contribute to the cushions and assist in formation of valvuloseptal structures (Combs and Yutzey, 2009a; Hinton and Yutzey, 2011). Starting about day 12.5 in mouse embryogenesis (E12.5), primordial valves in the outflow tract (aortic and pulmonary) and atrioventricular canal (mitral and tricuspid) evolve from these cellularized cushions through elongation, apoptosis, and highly orchestrated changes in extracellular matrix (ECM) composition and organization. In the outflow tract, this remodeling process requires efficient recruitment and differentiation of cardiac neural crest cells (Jain et al., 2011; Zhang et al., 2010b). Appropriate deposition and ordering of the ECM, as well as differentiation of the mesenchymal cells, are essential for efficient valve functionality. Despite the importance of these remodeling events, the molecular mechanisms that govern valve maturation are only poorly understood. The BMP and epidermal growth factor receptor (EGFR) pathways have been implicated in valve remodeling, as disruption of EGFR signaling or sustained BMP signaling lead to enlargement of the semilunar valves, with variable effects on the atrioventricular valves, typically through hyperplasia (Chen et al., 2000; Galvin et al., 2000; Iwamoto et al., 2010; Iwamoto et al., 2003; Jackson et al., 2003; Nanba et al., 2006; Tadano et al., 2005; Yamazaki et al., 2003; Yu et al., 2011).

Mutational analysis in mice has shown that of all the ligands for the EGFR, only heparin-binding epidermal growth factor (HB-EGF) appears to be critical in valvulogenesis (reviewed in (Iwamoto and Mekada, 2006)). In the developing valves, HB-EGF is expressed by the valvular endocardium (cardiac endothelium) (Iwamoto et al., 2003; Jackson et al., 2003), and, like all EGFR ligands, is synthesized as a transmembrane precursor that requires proteolytic processing to a soluble, highly bioactive form. Global and endothelial deletion of HB-EGF or mutation of its cleavage site result in thickened semilunar and atrioventricular valves (Iwamoto et al., 2003; Jackson et al., 2003; Nanba et al., 2006; Yamazaki et al., 2003), with enlargement primarily due to an increase in cell number through acute proliferation of valvular mesenchymal cells from E14.5 to E16.5 (Iwamoto et al., 2010; Jackson et al., 2003). Because valves in Hbegf−/− embryos show persistent staining for phosphorylated forms of the BMP effectors Smad1/5/8, Jackson et al. put forth the attractive notion that soluble HB-EGF signaling through the EGFR is required for dampening the BMP pathway and suppressing proliferation of the interstitial cells of the valves (Jackson et al., 2003).

The enzymes that play a major role in proteolytic release or “shedding” of EGFR ligands and other transmembrane proteins from the cell surface, particularly during embryonic development, belong to the ADAM (A disintegrin and metalloproteinase) subfamily of metalloproteinases. ADAM17, also known as TACE (TNF-α converting enzyme) (Black et al., 1997), cleaves most EGFR ligands, including HB-EGF (Sahin et al., 2004), and is expressed throughout the mouse embryonic heart from mid to late gestation (Jackson et al., 2003; Shi et al., 2003). Global inactivation of ADAM17 gives rise to expansion of both ventricles in neonatal hearts (Shi et al., 2003), ventricular septal defects (Horiuchi et al., 2005), and enlargement of the atrioventricular valve leaflets and semilunar valve cusps (Horiuchi et al., 2007; Horiuchi et al., 2005; Jackson et al., 2003; Sahin et al., 2004). These heart defects, in addition to abnormalities in lung development, likely contribute to the perinatal mortality observed in mice lacking ADAM17 (Peschon et al., 1998; Zhao et al., 2001). Global deletion of another ADAM proteinase, ADAM19, also leads to death of most homozygous null neonates and reproduces many of the defects seen in HB-EGF- and ADAM17-null hearts (Kurohara et al., 2004; Zhou et al., 2004). Although HB-EGF was not initially thought to be a target for ADAM19 (Sahin et al., 2004), further investigation revealed that this ADAM cleaves HB-EGF specifically and constitutively in the absence of ADAM17 (Horiuchi et al., 2005). In addition, embryos lacking both ADAM17 and ADAM19 exhibit earlier embryonic mortality (E13.5 to E14.5) with more severe heart defects than either knockout alone (Horiuchi et al., 2005), highlighting the potential for mutual compensation by these two ADAMs in at least some aspects of development. Ablation of two other ADAMs, ADAM9 and ADAM12, that have been reported to shed HB-EGF produces no valvular phenotype when individually deleted (Kurisaki et al., 2003; Weskamp et al., 2002). ADAM10, another candidate HB-EGF sheddase, is required early in endocardial cushion development for the endothelial-to-mesenchymal transition (EMT), likely through cleavage of Notch1 (Zhang et al., 2010a).

Thus, the primary responsibility for cleavage of HB-EGF in heart valve development has traditionally been ascribed to ADAM17, with ADAM19 likely assuming the role in the absence of ADAM17. However, the importance of ADAM17 in the specific cell types that contribute to valvulogenesis has not been established. In this study, we sought to further address the role of endothelial ADAM17 in valve development and remodeling. We generated mice with a floxed allele of Adam17 for both global and cell-specific deletion. Using our floxed Adam17 mice in conjunction with Tie2-Cre transgenics, we show that absence of endothelial ADAM17 reproduces the valve enlargement observed in the global knockout, but only in the semilunar valves. Our results also reveal an unexpected function for ADAM17 in modulation of the valvular ECM. Our findings support the idea that ADAM17 shedding activity in endocardial cells and their derivatives is essential for normal remodeling and maturation of the semilunar valve cusps, but they also suggest that the enzyme targets additional substrates beyond HB-EGF in this process.

2. Materials and Methods

2.1. Conditional deletion of Adam17-floxed alleles in mice

A detailed description of the targeting construct for generation of Adam17-floxed mice using C57BL/6 ES cells is provided in the Supplementary Materials. To generate mice with an ADAM17-null allele (Adam17^ΔEx5) by cre-mediated excision of exon 5, Adam17^Fl/Fl animals were mated with the B6.FVB-Tg(EIIa-cre)C5379Lmgd/J strain (Jackson Laboratory stock number 003724) and progeny were screened by PCR of tail biopsies (genotyping primers are listed in Table S1). For endothelial cell deletion of ADAM17, transgenic mice expressing cre recombinase from the Tek (Tie2) promoter were used (Koni et al., 2001) (strain B6.Cg-Tg(Tek-cre)12Flv/J, Jackson Laboratory stock number 004128). Adam17^Fl/Fl females were crossed with Tie2-Cre/+;Adam17^+/ΔEx5 males to generate progeny for analysis. Characterization of Adam17^ΔEx5/ΔEx5 embryos and further information about endothelial deletion of ADAM17 are outlined in the Supplementary Materials. Use of mice with the Adam17^ΔEx5 allele was previously reported (Tang et al., 2011).

Timed matings were done to generate embryos of specific gestational ages, with appearance of a copulation plug designated as day 0.5. All procedures involving mice were approved by the University of Washington Institutional Animal Care and Use Committee.

2.2. Histology and immunohistochemistry

Embryos and hearts from late-stage embryos (E18.5), juvenile (4 weeks old), young adult (7–9 weeks old) and adult (> 9 weeks old) mice were fixed in 4% paraformaldehyde or Bouin’s fluid and embedded in paraffin for sectioning and standard histological stains. Images were captured using a SPOT Insight color camera and software (version 3.5.9 for Windows) to quantify aortic valve cell numbers with NIH ImageJ and to measure cusp thickness as described (Iwamoto et al., 2010). For cell proliferation studies using bromodeoxyruridine (BrdU), pregnant dams were injected intraperitoneally with 1.2 mg Cell Proliferation Labelling Reagent (GE Healthcare) and sacrificed 2 hours later for collection of embryos. Paraffin sections of embryos fixed in 4% paraformaldehyde were heated in Antigen Unmasking Solution (Vector Laboratories) before deparaffinizing and rehydrating. DNA was denatured by incubation in 2N HCl for 20 min at 37°C. After neutralizing with boric acid/borate buffer and blocking with 5% rabbit serum, sections were incubated with a monoclonal antibody against BrdU (Dako; clone Bu20a), diluted 1:200 in PBS containing 5% rabbit serum, for 2 hours. Sections were developed using a biotinylated rabbit anti-mouse antibody (Zymed) and the Vectastain Elite ABC kit (Vector Laboratories), with DAB as the chromagen and Mayer’s hematoxylin or dilute eosin (1:8 in 70% ethanol) as the counterstain. In some experiments, NiCl₂ was added to the DAB solution to yield a dark brown/black color. For determination of the apoptotic index in semilunar valve cusps, the TACS^® 2 TdT-Blue Label In Situ Apoptosis Detection Kit (Trevigen) was used according to the manufacturer’s instructions to visualize cells with DNA fragmentation. A minimum of 6 sections was counted from 3 embryos of each genotype, for a total of 7566 cells from Tie2-Cre/+;Adam17^Fl/+ and 6977 cells from Tie2-Cre/+;Adam17^Fl/ΔEx5. In both the proliferation and apoptosis studies, sections were photographed for counting total and positively stained cells using NIH ImageJ. For immunolocalization of versican, sections were incubated with an antibody against the GAGβ domain of mouse versican (Millipore) at 6 μg/ml. Sections were developed using the Vectastain Elite ABC kit (Vector Laboratories), Nova Red (Vector Laboratories) as the chromagen and Mayer’s hematoxylin as the counterstain.

2.3. Western blotting

For detection of versican, embryos at day 12.5 in gestation were collected and tail biopsies were taken to determine genotype. Hearts were dissected above the atria and frozen individually until genotype was identified. Hearts of the same genotype were pooled (3–4 hearts per pool) and protein homogenates prepared as described previously (Schonherr et al., 1997). Tissues were extracted with GuHCl buffer (4 M guanidine-HCl, 100 mM sodium sulfate, 100 mM Tris base, 2.5 mM Na₂EDTA, 0.5% Triton X-100, pH 7.0) containing protease inhibitors (5 mM benzamidine-HCl, 100 mM 6-aminohexanoic acid, 1 mM phenylmethyl sulfonyl fluoride) overnight at 4°C. Tissue extracts were then dialyzed against urea buffer (8 M urea, 2 mM EDTA, 50 mM Tris base, 0.5% Triton X-100, pH 7.5) to remove the guanidine. To assess levels of intact versican variants containing the GAGβ domain, equal amounts of protein, as determined using the Coomassie Protein Assay kit (Thermo Scientific), from each pool were subjected to DEAE ion exchange chromatography. Eluates were ethanol precipitated, digested with chondroitin ABC lyase, and analyzed by SDS-PAGE (4–12% polyacrylamide separating gel with 3.5% polyacrylamide stacking gel). Proteins were transferred to nitrocellulose and probed with the anti-GAGβ antibody at 0.25 μg/ml. To detect the 70-kDa cleavage fragment of versican, equal amounts of protein were ethanol precipitated and subjected to SDS-PAGE (8% polyacrylamide separating gel with 3.5% polyacrylamide stacking gel) followed by transfer to nitrocellulose. Blots were probed with an anti-DPEAAE antibody (Affinity Bioreagents) at 1 μg/ml. The blot was stripped and re-probed with anti-actin antibody (Sigma) to control for loading. Bands were visualized using enhanced chemiluminescence (Thermo Scientific).

2.4. Quantification of hyaluronan

Frozen tissue samples were digested with proteinase K (250 μg/ml) for 18 hours at 60°C, followed by heat inactivation at 100°C for 20 minutes. To quantify levels of hyaluronan, we used a modification (Wilkinson et al., 2004) of a previously described competitive enzyme-linked sorbent assay (Underhill et al., 1993). Samples and standards were first mixed with biotinylated hyaluronan binding protein (bHABP), the amino-terminal hyaluronan-binding region of aggrecan, and then added to hyaluronan-coated microtiter plates that had been blocked in PBS plus 10% calf serum. After washing, plates were incubated with streptavidin-peroxidase followed by 2,2′-azinobis-3-ethyl-benzthiozoline in 0.1 M sodium citrate, pH 4.2. Absorbance was measured at 405/570 nm. In this competitive assay, the absorbance is inversely proportional to the concentration of hyaluronan added to the bHABP. To normalize the data and to allow correlation of hyaluronan levels with cell number in the tissue samples, DNA content was measured using a mouse Tfrc TaqMan® Copy Number Reference Assay kit (Life Technologies). Briefly, 1.5 μl of each extract was amplified in quadruplicate using 1X Taqman Gene Expression Master Mix and an ABI Prism 7900 HT. Relative copy number was calculated based on a standard curve.

2.5. Transthoracic echocardiography

For the evaluation of heart function and detection of heart valve defects, echocardiographic experiments were performed to measure left ventricular (LV) wall thickness, LV end-diastolic dimension (LVEDD), LV ejection fraction (EF) and velocity of blood flow using a Visual Sonics VEVO 770 system equipped with a 707B scan head, as previously described (Liao et al., 2002; Liu et al., 2010; Yamamoto et al., 2003). Male and female mice 24–28 weeks of age were lightly anesthetized with 1% isoflurane. Chamber dimensions were measured in M mode from the short axis. LVEF (in %) was calculated as follows: LVEF = [(LV Vol; d) − (LV Vol; s)]/(LV Vol; d) × 100, where (LV Vol; d) and (LV Vol; s) are LV volume at diastole and systole, respectively. LV Vol (μl) was calculated as: [7.0/(2.4+LVID)] × (LVID)³ × 1000. LVID is LV internal diameters and it was measured at both diastole and systole. The velocity of flow was measured by pulse-wave Doppler scanning in the ascending aorta distal to the aortic valve.

2.6. Quantitative reverse transcription-PCR

Pregnant dams at day 11.5 and 12.5 in gestation were killed and perfused with PBS followed by RNAlater (Invitrogen), which was also used for heart dissection. Yolk sacs stored at −20°C were used for genotyping. Total RNA was isolated from each heart individually using TRIzol Reagent (Invitrogen) by homogenizing tissue with a 1-ml syringe fitted with a 20-G needle. Glycogen (10 μg) was added at the alcohol precipitation step to enhance nucleic acid recovery. cDNA was synthesized from 1.5 μg RNA using the Superscript III kit (Invitrogen) and a combination of oligo dT primers and random hexamers. cDNAs were diluted to 12.5 ng/μl and 2 μl (25 ng) of each were amplified in triplicate using SYBR® Green PCR Master Mix (Applied Biosystems) and intron-spanning primer pairs (see Table S2 for sequences). Amplification was performed using an ABI Prism 7900 HT. For each sample, the change in expression level of each gene was calculated against a single calibrator sample using the 2^ΔCt method to obtain the relative quantity Q. Q values for each gene of interest were then normalized using the geometric mean of the Q values for the two most stable reference genes identified for this tissue, Hprt1 and Ywhaz (determined using the geNorm VBA applet for Microsoft Excel, version 3.5), as described (Vandesompele et al., 2002). Amplification efficiency was determined for each run and primer pair using a cDNA dilution series.

2.7. Transmission electron microscopy

Four week-old male mice were killed by isoflurane overdose and perfused with PBS, followed by half-strength Karnovsky’s (0.1 M cacodylate, pH 7.4, 3.4 mM CaCl₂, 0.01% paraformaldehyde, 0.025% glutaraldehyde). Hearts were harvested and stored in half-strength Karnovsky’s on ice during dissection of the aortic valves. Valves were incubated 3 h at room temperature in half-strength Karnovsky’s containing 0.2% ruthenium red for visualization of proteoglycans (Wight and Ross, 1975), then stored at 4°C. Following a wash (0.1 M cacodylate, pH 7.4, 3.4 mM CaCl₂, 0.1 M sucrose, 0.1% ruthenium red), valves were post fixed for 1 h at room temperature in 1% OsO₄ containing 0.1 M cacodylate, pH 7.4, 3.4 mM CaCl₂, 0.05% ruthenium red. Valves were washed three times in distilled water, then incubated in 2% uranyl acetate for 1 h at room temperature. After washing in water, tissues were dehydrated through a graded series of ethanol, and embedded in epoxy resin. Ultrathin sections were imaged using a JEOL JEM 1200 EXII electron microscope and images were captured with an Olympus Morada camera.

2.8. Statistical analysis

All parametric data are expressed as mean ± SEM. Significance was determined using the unpaired two-tailed Student’s t test with Welch correction (GraphPad Instat®, version 3.0b). For nonparametric data, the Mann-Whitney test was used to calculate significant differences between the medians (GraphPad Instat®, version 3.0b). Single-factor ANOVA was used for statistical comparison of multiple groups (Microsoft Excel version 11.6.6). For all statistical tests, a p value of less than 0.05 was considered significant.

3. Results

3.1. Reduced postnatal survival of progeny with developmental ablation of ADAM17 in endothelial cells

For conditional deletion of ADAM17, we developed a line of mutant mice on the C57BL/6 background with floxed exon 5 in the Adam17 allele (denoted Adam17^Fl). Cre recombinase-mediated excision of exon 5 produces the null allele, Adam17^ΔEx5 (Fig. S1). To generate mice heterozygous for the null allele (Adam17^+/ΔEx5), Adam17^Fl/Fl animals were mated with EIIa-Cre transgenics as described in the Materials and Methods. Both Adam17^Fl/Fl and Adam17^+/ΔEx5 mice are viable and fertile and show no overt defects in development. For selective elimination of ADAM17 in endothelial cells, we first crossed a Tie2-Cre transgene (Koni et al., 2001) into the Adam17^+/ΔEx5 background. Males with the genotype Tie2-Cre/+;Adam17^+/ΔEx5 were then bred to Adam17^Fl/Fl females to produce offspring for study. This breeding scheme was used to avoid recombination that occasionally occurs in the female germline of this transgenic (de Lange et al., 2008) and has been employed by other investigators (Beppu et al., 2009; Shen et al., 2005; Sridurongrit et al., 2008; Yu et al., 2011). With this strategy, an equivalent number of progeny with the four genotypes listed in Table 1 was expected. However, mutants with Tie2-Cre mediated deletion of ADAM17 (Tie2-Cre/+;Adam17^Fl/ΔEx5) were under-represented by 75% at weaning (Table 1). This was an early postnatal effect on survival, as the majority of Tie2-Cre/+;Adam17^Fl/ΔEx5 embryos examined at E18.5 were found alive and were obtained at a frequency close to the expected percentage (25%; Table 1). By contrast, progeny globally homozygous for the Adam17^ΔEx5 null allele, produced through breeding of Adam17^+/ΔEx5 mice, died late in embryogenesis or right after birth (Table S3), in accordance with the original Adam17 inactivating mutation (deletion of the Zn-binding domain in catalytic site, denoted as Adam17^ΔZn) (Peschon et al., 1998). Thus, although endothelial-selective deletion of ADAM17 using this Tie2-Cre transgenic strain improved postnatal survival over that observed with the global knockout (Adam17^ΔEx5/ΔEx5), the relative paucity of live Tie2-Cre/+;Adam17^Fl/ΔEx5 mutants suggests that endothelial ADAM17 has a critical role in development.

TABLE 1.

Distribution of genotypes from crosses of Adam17^Fl/Fl females and Tie2-Cre/+;Adam17^+/ΔEx5 males

Agea:	E18.5		P21–P28
Genotype	#b	%	#	%
Adam17Fl/+	15	19	232	33
Adam17Fl/ΔEx5	18	23	211	30
Cre/+;Adam17Fl/+	20	25	223	31
Cre/+;Adam17Fl/ΔEx5	26 (2)	33	45	6
Total	79		711

^aE, embryo; P, postnatal

^bNumber in parentheses indicates nonviable embryos in that group

3.2. Valve enlargement appears restricted to the semilunar valves and persists postnatally with deletion of endothelial ADAM17

Like its substrate HB-EGF, ADAM17 is broadly expressed throughout the mouse embryonic heart, including the endocardium overlaying the valves (Jackson et al., 2003; Shi et al., 2003). Because lineage tracing studies indicate that leaflets of the atrioventricular valves and cusps of the semilunar valves are largely or in part derived from the endocardium (de Lange et al., 2004; Kisanuki et al., 2001; Snarr et al., 2008), we evaluated these structures in Tie2-Cre/+;Adam17^Fl/ΔEx5 embryos in comparison to Adam17^+/+ and Adam17^ΔEx5/ΔEx5 at stage E18.5. As expected, in hearts from the global knockout (Adam17^ΔEx5/ΔEx5), there was pronounced enlargement of both the semilunar and atrioventricular valves (Fig. 1), consistent with previous observations for Adam17^ΔZn/ΔZn (Horiuchi et al., 2005; Jackson et al., 2003; Sahin et al., 2004). However, while the semilunar valves were similarly enlarged in the Tie2-Cre/+;Adam17^Fl/ΔEx5 endothelial mutants, the atrioventricular valve leaflets were not overtly dysmorphic (Fig. 1). This result was surprising, especially as the inflow valves are almost completely populated by cells emanating from the endothelium (de Lange et al., 2004; Kisanuki et al., 2001; Snarr et al., 2008) and would be expected to be virtually devoid of ADAM17 activity with endothelial ablation of its expression. Because the mutant semilunar valve cusps appeared hyperplastic, we quantified both the total number of cells and the thickness of cusps of the aortic valve at this stage (Fig. 1M). In mutant cusps, the median number of cells was increased by 43% and cusp thickness by 57% over that in controls (p = 0.014 and 0.0037, respectively, two-tailed Mann-Whitney test).

Fig. 1.

Late-stage embryos deficient in endothelial ADAM17 exhibit hyperplastic semilunar valves. Coronal sections of E18.5 hearts were stained with hematoxylin and eosin (H&E) to show the semilunar valve cusps and atrioventricular valve leaflets (denoted by the asterisks) in Adam17^+/+ (A, D, G, and J), global knockout Adam17^ΔEx5/ΔEx5 (B, E, H, and K), and endothelial mutant Tie2-Cre/+;Adam17^Fl/ΔEx5 (C, F, I, and L). Images shown of mutants are representative of 7 Adam17^ΔEx5/ΔEx5 and 8 Tie2-Cre/+;Adam17^Fl/ΔEx5 hearts analyzed. Scale bar, 25 μm. (M) Total cells (including mesenchymal and endothelial) in and thickness of aortic valve cusps were determined using sections of control (Adam17^Fl/+ and Tie2-Cre/+;Adam17^Fl/+) and endothelial mutant (Tie2-Cre/+;Adam17^Fl/ΔEx5) embryo hearts at E18.5. For the cell counts, each point corresponds to the average number of cells obtained from 3 sections representing the largest area of each cusp, using 7 cusps from 4 control hearts and 8 cusps from 5 mutant hearts. The thick lines denote the median values.

From E18.5 into adulthood, mouse cardiac valves continue to mature and undergo considerable elongation and thinning (Hinton et al., 2008; Hinton et al., 2006). Therefore, we assessed valves in adult Tie2-Cre/+;Adam17^Fl/ΔEx5 mice to determine if remodeling in these structures is merely delayed in the outflow tract. In contrast to control Tie2-Cre/+;Adam17^Fl/+ hearts, cusps of the aortic and pulmonary valves in Tie2-Cre/+;Adam17^Fl/ΔEx5 specimens showed persistent thickening (Fig. 2), indicating that the remodeling defect is irreversible. As in the late-stage embryos, the atrioventricular valve leaflets in Tie2-Cre/+;Adam17^Fl/ΔEx5 adult hearts appeared normal (Fig. 2E and data not shown). Our results emphasize that similar to other models of semilunar valve thickening characterized by hyperplasia, including the HB-EGF knockout (Chen et al., 2000; Iwamoto et al., 2003; Jackson et al., 2003; Tadano et al., 2005; Yamazaki et al., 2003; Yu et al., 2011), adults with endothelial deletion of ADAM17 retain the valve enlargement phenotype.

Fig. 2.

Semilunar valve enlargement persists in adult Tie2-Cre/+;Adam17^Fl/Ex5 mice. Shown are coronal sections of hearts from Tie2-Cre/+;Adam17^Fl/+ control (A and C) and Tie2-Cre/+;Adam17^Fl/ΔEx5 mutant (B, D, and E) mice. Sections were stained with hematoxylin and eosin (H&E). The low magnification image in E illustrates the degree of aortic valve (Ao) enlargement in comparison with the normal leaflets of the mitral valve (Mi) in a Tie2-Cre/+;Adam17^Fl/Ex5 mutant adult. Cusps and leaflets are denoted by the asterisks. Scale bar, 50 μm.

3.3. Tie2-Cre/+;Adam17^Fl/ΔEx5 mice develop cardiomegaly postnatally

In addition to thickening of the outflow valves, hearts from surviving Tie2-Cre/+;Adam17^Fl/ΔEx5 mice developed a marked enlargement relative to body mass (Fig. 3A). This cardiomegaly occurred postnatally, as there were no differences in heart-to-body mass at E18.5 between Tie2-Cre/+;Adam17^Fl/ΔEx5 and any of the other control genotypes (Adam17^Fl/+, Adam17^Fl/ΔEx5, and Tie2-Cre/+;Adam17^Fl/+) (Fig. 3A). By contrast, Adam17^ΔEx5/ΔEx5 embryo hearts were already enlarged at this stage in development (Fig. 3A), in agreement with reported findings for Adam17^ΔZn/ΔZn (Shi et al., 2003). Lack of overall heart enlargement in the Tie2-Cre/+;Adam17^Fl/ΔEx5 embryos confirmed that ADAM17 deletion is restricted in these mutants as compared to the global knockout. The increase in heart size was primarily due to expansion of the left ventricle (Fig. 3, B and C), likely a compensatory response to compromised outflow valve function and poor cardiac output in the context of the increase in pressure that occurs after birth. These findings emphasize crucial differences between global and endothelial deletion of ADAM17, in that ADAM17-null mice have perinatal enlargement of both left and right ventricles (Shi et al., 2003).

Fig. 3.

Postnatal compensatory heart enlargement in mice lacking endothelial ADAM17. (A) Hearts were dissected from embryos and adults and weighed for comparison to body mass. Controls include the following genotypes: Adam17^Fl/+, Adam17^Fl/ΔEx5, and Cre/+;Adam17^Fl/+. The number of mice used for the analysis is indicated in each column. The y-axis scale starts with 6 in this panel. Hearts from Tie2-Cre/+;Adam17^Fl/+ control (B) and Tie2-Cre/+;Adam17^Fl/ΔEx5 mutant (C) mice were processed for hematoxylin and eosin staining. Shown are transverse sections at 2.5x magnification. Note that the images in B and C were assembled from photographs of 2 and 4 overlapping fields, respectively. LV, left ventricle. RV, right ventricle.

3.4. Defects in Tie2-Cre/+;Adam17^Fl/ΔEx5 outflow valves are not accompanied by alterations in proliferation or apoptosis

To understand the mechanisms underlying semilunar valve enlargement in mice with endothelial deletion of ADAM17, we assessed development and remodeling of the valves at stages earlier than E18.5 during embryogenesis. We focused particularly on E14.5, since increases in valve size and proliferation were first noted at this time point for the HB-EGF knockout (Iwamoto et al., 2010; Jackson et al., 2003) and the anticipation was that these changes would be phenocopied in ADAM17-null valves. We first used comparative RT-PCR of whole hearts at E11.5 and E12.5 to confirm a partial reduction in Adam17 mRNA levels in Tie2-Cre/+;Adam17^Fl/ΔEx5, consistent with restricted expression of cre recombinase from the Tie2 promoter, versus complete ablation in Adam17^ΔEx5/ΔEx5 (Fig. S2).

Histological analysis showed that outflow valve cusps in Tie2-Cre/+;Adam17^Fl/ΔEx5 and control hearts are virtually indistinguishable by morphology at E14.5 (Fig. 4, A and B) and E16.5 (Fig. S3). There are also no discernible differences in cellularization of the outflow tract cushions at E12.5 (data not shown). To assess acute cell proliferation in the Tie2-Cre/+;Adam17^Fl/ΔEx5 hearts, we measured BrdU incorporation in semilunar valve cusps at E12.5 and E14.5. As previously reported (Hinton et al., 2006), cell proliferation is high at E12.5 and declines as development proceeds (Fig. 4C). The number of BrdU-positive endocardial and mesenchymal cells in the outflow endocardial cushions was similar between mutants and controls at E12.5 (Fig. 4C). However, we also observed no differences in the proliferation index at E14.5 (Fig. 4C), supported visually by the lack of overt changes in the frequency of BrdU-positive cells (Fig. 4, D and E). Because an elevation in cell number was detected at E18.5 in Tie2-Cre/+;Adam17^Fl/ΔEx5 aortic valves (Fig. 1M), we also examined proliferation at this stage in development. As expected for this time point (Hinton et al., 2006), the proliferation index was low (<5 %) and was similar between the genotypes (data not shown). Although this analysis does not rule out a discrete window of increased proliferation occurring from E15.5 through E17.5, lack of an effect at E14.5, the stage at which two groups independently showed enhanced BrdU incorporation in Hbegf−/− valves (Iwamoto et al., 2010; Jackson et al., 2003), suggests key differences between deletion of HB-EGF and ADAM17.

Fig. 4.

Morphology of and acute cell proliferation in semilunar valves of endothelial ADAM17 mutant embryos after mid-gestation. (A and B) Transverse sections of E14.5 embryos stained with hematoxylin and eosin to show the similarities in morphology of the pulmonary valve cusps (denoted by the asterisks) in Tie2-Cre/+;Adam17^Fl/+ control (A) and Tie2-Cre/+;Adam17^Fl/ΔEx5 mutant (B) hearts. (C) Percentage of BrdU-positive cells in semilunar endocardial cushions at E12.5 and cusps at E14.5. The number of cushions/cusps analyzed is indicated in each column. Data were obtained from 3 individual hearts of each genotype at E12.5, and from 4 and 5 individual Tie2-Cre/+;Adam17^Fl/+ control and Tie2-Cre/+;Adam17^Fl/ΔEx5 mutant hearts, respectively, at E14.5. (D and E) Transverse sections of E14.5 embryos labeled with and subsequently immunostained for BrdU and counterstained with dilute eosin to compare the frequency of positivity in pulmonary valve cusps (outlined by the dotted lines) from Tie2-Cre/+;Adam17^Fl/+ (D) and Tie2-Cre/+;Adam17^Fl/ΔEx5 (E). Scale bar, 50 μm.

Because enhanced cell proliferation at E14.5 did not account for the increased number of cells in the semilunar valve cusps of Tie2-Cre/+;Adam17^Fl/ΔEx5 mutant embryos at E18.5, we next assessed if there is a deficiency in apoptosis. By E16.5, both cell proliferation and the cardiac neural crest component have decreased and apoptosis has increased as part of the remodeling process in the valve cusp precursors (Hinton et al., 2006; Jain et al., 2011). Analysis of Tie2-Cre/+;Adam17^Fl/Δ+ and Tie2-Cre/+;Adam17^Fl/ΔEx5 outflow valve cusps at this stage using a TUNEL assay showed a similar frequency of apoptotic cells (mean values of 0.39 ± 0.11% for Tie2-Cre/+;Adam17^Fl/Δ+ and 0.47 ± 0.04% for Tie2-Cre/+;Adam17^Fl/ΔEx5). These results indicate that differences in apoptosis are unlikely to explain the increase in valve cell number in hearts from embryos with endothelial deletion of ADAM17.

3.5. Altered extracellular matrix in postnatal Tie2-Cre/+;Adam17^Fl/ΔEx5 hearts

Our data indicate that enlargement of semilunar valve cusps in Tie2-Cre/+;Adam17^Fl/ΔEx5 mice involves both the accumulation of mesenchymal cells as well as an expansion of cusp thickness. To address the nature of the of the ECM comprising this increased area in the mutant cusps, sections of heart from young adult Tie2-Cre/+;Adam17^Fl/+ controls and Tie2-Cre/+;Adam17^Fl/ΔEx5 mutants were analyzed using Movat’s pentachrome staining. Representative sections of the aortic valve are shown in Fig. 5. The dominant phenotype that we observed is an increase in the valve cusp area staining positive (blue) for glycosaminoglycan-containing molecules (compare Fig. 5B to 5A), although staining was not uniform throughout the cusp. To assess the ECM at the ultrastructural level, we performed transmission electron microscopy (TEM) analysis of isolated aortic valves from juvenile mice. We focused on using juvenile mice to minimize scoring changes in the ECM that may occur in response to altered blood flow in the mutants over time. Inclusion of ruthenium red during fixation of the tissue, although not useful in detecting glycosaminoglycans such as hyaluronan, allowed visualization of proteoglycans as granules (Wight and Ross, 1975). While the glycosaminoglycan/proteoglycan-rich spongiosa layer in cusps from control mice was compact and dense with ruthenium red-positive proteoglycan granules (Fig. 5C), this layer in the cusps of a mutant mouse was enlarged and exhibited increased space between the granules (Fig. 5D). This phenotype was even more pronounced in a second mutant that we analyzed (Fig. 5E). These findings are suggestive of increased osmotic swelling and/or potential changes in the levels of proteoglycans and glycosaminoglycans in the mutant aortic cusps.

Fig. 5.

Alterations in extracellular matrix of aortic valve cusps from Tie2-Cre/+;Adam17^Fl/ΔEx5 mice. (A and B) Transverse sections of hearts from Tie2-Cre/+;Adam17^Fl/+ (A) and Tie2-Cre/+;Adam17^Fl/ΔEx5 (B) young adult mice were stained with Movat’s pentachrome to visualize matrix components, with glycosaminoglycan-containing molecules staining blue. Scale bar, 50 μm. (C–E) Transmission electron microscopy was used in conjunction with ruthenium red staining to assess the glycosaminoglycan-enriched layer (spongiosa) of aortic valve cusps from juvenile mice. (C) The spongiosa layer in a Tie2-Cre/+;Adam17^Fl/+ control aortic valve cusp is densely packed with ruthenium red-positive proteoglycan granules. (D and E) The spongiosa layer in cusps from two individual mutants (Tie2-Cre/+;Adam17^Fl/ΔEx5) is shown. In contrast to controls, this layer is less compact and contains a more widely dispersed population of proteoglycan granules, particularly in the mutant cusp in (E). Much of the spongiosa in this mutant cusp is devoid of electron-dense matrix molecules. Images are representative of 5 controls and 3 mutants analyzed. Scale bar, 2 μm.

3.6. Expression levels of versican and hyaluronan in embryonic Tie2-Cre/+;Adam17^Fl/ΔEx5 hearts

The glycosaminoglycan hyaluronan and the proteoglycan versican play key roles in the different phases of endocardial cushion formation and valve remodeling (Lockhart et al., 2011; Schroeder et al., 2003). Hyaluronan is a major component of the cardiac jelly comprising the endocardial cushions prior to EMT and cellularization, and both versican and hyaluronan are maintained in human and rodent adult valves (Angel et al., 2011; Hellstrom et al., 2006; Stephens et al., 2012) to provide compressibility to the cusps and leaflets (Hinton and Yutzey, 2011; Schoen, 2008). Because our histochemical and ultrastructural analysis of postnatal aortic valves implicated changes in the glycosaminoglycan-containing component of Tie2-Cre/+;Adam17^Fl/ΔEx5 mutant cusps, we sought to determine if levels of versican or hyaluronan are affected. Our approach was to assess levels of these molecules in whole embryo hearts at E12.5, when remodeling of the endocardial cushions begins, to establish the congenital nature of any differences detected.

To examine versican content, lysates of pooled embryo hearts were subjected to DEAE ion exchange chromatography to isolate proteoglycans and digestion with chondroitin ABC lyase for SDS-PAGE and immunoblotting. An antibody recognizing the GAGβ domain present in versican variants V0 (full length) and V1 was used for the Western analysis, since these are the predominant versican isoforms in the endocardial cushions (Lockhart et al., 2011). Compared to the controls (Adam17^Fl/+ and Tie2-Cre/+;Adam17^Fl/+), mutant embryo hearts showed dramatically reduced levels of versican (Fig. 6A). To address the possibility that the decreased levels of intact versican are due to its enhanced proteolysis in the mutant hearts, we assessed whole heart protein extracts by immunoblotting with an antibody specific for the DPEAAE epitope indicative of the 70-kDa versican cleavage fragment (Sandy et al., 2001). Hearts from Tie2-Cre/+;Adam17^Fl/ΔEx5 E12.5 embryos also had lower levels of this fragment in comparison to the controls (Fig. 6B), consistent with an overall diminishment in versican. In addition, we found no differences in the expression of the major versicanases (ADAMTS1, ADAMTS5, and ADAMTS9) that give rise to the 70-kDa cleavage fragment in the embryonic heart (Dupuis et al., 2011; Kern et al., 2006), as well as another endocardial cushion-associated metalloproteinase (MMP2) (Lange and Yutzey, 2006) with versican-degrading ability (Kern et al., 2006; Passi et al., 1999) (Fig. 6C). Note that at this stage of development, versican expression in the ventricles is low ((Henderson and Copp, 1998; Kern et al., 2006) and our unpublished observations), so that the levels seen by Western analysis of the heart primarily reflect those in the semilunar and atrioventricular endocardial cushions. Immunohistochemistry for the GAGβ domain showed abundant versican in the atrioventricular canal of both mutants and controls (data not shown).

Fig. 6.

Content of versican, both intact and cleaved, and hyaluronan in E12.5 hearts. (A and B) Hearts were dissected from embryos at day 12.5 in gestation and 3–4 hearts were pooled for protein extraction and Western analysis of versican variants containing the GAGβ domain (A) or the cleaved fragment with the carboxyl-terminal sequence DPEAAE (B). Each lane of the blots corresponds to one pool. (A) Equal amounts of protein extracts were subjected to DEAE chromatography and chondroitin ABC lyase digestion and eluates were assessed by immunoblotting with an antibody specific to the GAGβ domain of versican. Shown are three pools for Adam17^Fl/+ and Tie2-Cre/+;Adam17^Fl/ΔEx5 and two pools for Tie2-Cre/+;Adam17^Fl/+. (B) The 70-kD versican cleavage fragment was visualized using an antibody recognizing the DPEAAE epitope. Actin protein levels were used as a loading control. (C) mRNA levels of the versicanases ADAMTS1, ADAMTS5, ADAMTS9, and MMP2 in whole hearts from E12.5 mutants (Tie2-Cre/+;Adam17^Fl/ΔEx5) are shown relative to Adam17^Fl/+ controls. n = 5 per genotype. (D) Hearts from E12.5 embryos were digested with proteinase K and analyzed individually using a hyaluronan enzyme-linked sorbent assay. The concentration of hyaluronan (in ng) was normalized to DNA content (measured as described in the Materials and Methods). The number of hearts analyzed is indicated in each column. p = 0.2 by one-way ANOVA.

Because versican typically associates with hyaluronan (Wight, 2002), we next asked if levels of this glycosaminoglycan were reduced in a manner similar to versican. An enzyme-linked sorbent assay (ELSA) for hyaluronan was used to measure its concentration in proteinase K-digested embryo hearts and did not reveal any differences between mutants and controls (Fig. 6D). Yet, this apparent imbalance in the levels of hyaluronan and its aggregating proteoglycan versican in the mutant embryo hearts could potentially favor functions of the glycosaminoglycan in either its free state or bound to other molecules. Collectively, our findings indicate that the ECM differs in the hearts of Tie2-Cre/+;Adam17^Fl/ΔEx5 embryos as compared to controls at early stages in valve remodeling.

3.7. Expression and localization of ECM in adult Tie2-Cre/+;Adam17^Fl/ΔEx5 aortic valve cusps

In light of our data on versican and hyaluronan levels in mutant embryo hearts, we next assessed expression of these macromolecules in adult aortic valve cusps. We first measured hyaluronan content in isolated cusps using the quantitative ELSA. While cusps from control hearts showed a significant correlation between hyaluronan concentration and relative DNA copy number (a surrogate for cell number), levels of hyaluronan in a small set of mutants tended to be higher and to show less correlation with DNA than controls (Fig. S4). A postnatal hyaluronan-dominant matrix would be consistent with the TEM showing more electron-lucent areas and a wider dispersion of ruthenium red-proteoglycan granules in the mutant aortic valve cusps compared to controls (Fig. 5). In addition, we examined expression of intact versican in aortic valves by immunohistochemistry. Although not a quantitative approach, this analysis showed that the pattern of versican immunolocalization is heterogeneous in the mutant cusps, which have distinct versican-poor regions compared to the dense, quite uniform staining in controls (Fig. 7). In areas with reduced versican staining, we frequently observed that the mesenchymal cells are aligned parallel to each other but independently of cells in other regions of the cusp (see high magnification images in Fig. 7, D and E). Further evaluation of these areas by Masson’s trichrome histochemical staining revealed that they are typically rich in collagen. We found that 9 of 13 mutants had at least one aortic valve cusp with aberrant localization and/or excessive levels of collagen, as exemplified in Fig. 7F and G. By contrast, in controls, the bulk of the collagen detectable by Masson’s trichrome staining is concentrated in a thin, well-delineated layer (the fibrosa) on the arterial aspect (Fig. 7C). Thus, there may be additional changes that occur in the ECM of mutant aortic valve cusps postnatally and these changes likely impact the phenotype of the valvular interstitial cells as well as valve function.

Fig. 7.

Localization of versican and collagen in adult aortic valve cusps. Sections of hearts from Tie2-Cre/+;Adam17^Fl/+ (A–C) and Tie2-Cre/+;Adam17^Fl/ΔEx5 (D–G) adults (16–21 weeks old) were used for versican immunohistochemistry (A, B, D, and E) with the anti-GAGβ antibody or Masson’s trichrome staining (C, F, and G) for collagen (blue). Panels A, B, D, and E represent coronal sections of the heart. The high-power images of the boxed regions in panels D and E show the differential alignment of the mesenchymal cells (nuclei stained blue by hematoxylin). In panels C, F, and G, hearts were sectioned transversely and the orientation of the cusps relative to the lumen is indicated. The arrows in panel C denote the collagen-dense fibrosa layer. Scale bar = 50 μm in all panels except the high-power images from D and E (scale bar = 20 μm).

3.8. Tie2-Cre/+;Adam17^Fl/ΔEx5 adults exhibit aortic valve stenosis and significant systolic dysfunction

To evaluate parameters of cardiac function in Tie2-Cre/+;Adam17^Fl/ΔEx5 mice quantitatively, adult mice were subjected to non-invasive transthoracic echocardiography, using gender- and age-matched Cre/+;Adam17^Fl/+ animals for comparison (Table 2). We found that heart function was more severely affected in females than males, with mutant females significantly different from control females in all measured parameters except body mass (Table 2). Only females showed statistically significant changes in heart rate and left ventricular wall thickness and end diastolic dimension, as well as aortic valve peak velocity. However, the latter two parameters, left ventricular end diastolic dimension and aortic valve peak velocity, were close to significantly different in mutant males (p = 0.054 and 0.078, respectively). The increase in left ventricular end diastolic dimension further supports the ventricular hypertrophy and dilation observed histologically (Fig. 3). Both male and female mutants exhibited declines in percent ejection fraction and fractional shortening, indicative of systolic dysfunction (Table 2). Pulse-wave Doppler analysis did not reveal evidence of aortic regurgitation (data not shown). Taken together, our data are consistent with postnatal development of aortic valve stenosis in both male and female endothelial ADAM17 deletion mutants. Collectively, our findings suggest that developmental defects in semilunar valve cusps, potentially in conjunction with compensatory and/or flow-related alterations in ECM composition of the cusps, profoundly impact valve functionality in Tie2-Cre/+;Adam17^Fl/ΔEx5 adults.

TABLE 2.

Echocardiography parameters^a,^b in 24- to 28-week old male and female Tie2-Cre/+;Adam17^Fl mice

Genotype	Body Mass, g	Heart Rate, bpm	LV WT, mm	LVEDD, mm	EF, %	FS, %	AV Peak V, mm/s
Male
Cre/+;Adam17Fl/+	30.6 ± 1.8	511 ± 18	0.82 ± 0.04	3.49 ± 0.08	65.99 ± 1.68	35.54 ± 1.27	2163.4 ± 262.2
Cre/+;Adam17Fl/ΔEx5	29.0 ± 0.2	511 ± 12	0.83 ± 0.02	4.64 ± 0.42	49.75 ± 3.65*	25.28 ± 2.13*	3830.1 ± 707.5
Female
Cre/+;Adam17Fl/+	26.9 ± 2.2	439 ± 25	0.80 ± 0.02	3.34 ± 0.11	68.70 ± 2.52	37.67 ± 1.94	1554.6 ± 119.5
Cre/+;Adam17Fl/ΔEx5	22.3 ± 0.5	530 ± 10*	0.90 ± 0.03*	4.55 ± 0.33*	50.52 ± 4.54*	25.81 ± 2.72*	4292.1 ± 358.8*

^aValues are mean ± SEM. n = 5 per group.

Asterisks denote values significantly different from the same gender controls (p <0.05).

^bLV WT, left ventricular wall thickness; LVEDD, left ventricular end-diastolic dimension; EF, ejection fraction; FS, fractional shortening; AV Peak V, aortic valve peak velocity.

4. Discussion

4.1. Overall summary

In this study, we analyzed for the first time the basis for cardiac valve enlargement in mice lacking ADAM17 in endothelial cells during embryogenesis. By restricting deletion of the enzyme to the endothelial cells, we could assess the phenotype of the valves in a context where ADAM17 is replete in other cell types that contribute to valvulogenesis and to evaluate heart function in mutant adults, which has not been possible with the global knockout because of perinatal lethality. In comparison to the global deletion mutant, ablation of ADAM17 in endothelial cells (Tie2-Cre/+;Adam17^Fl/ΔEx5) reduced the spectrum of cardiac defects and allowed survival of a low percentage of neonates, although the mice developed a progressive cardiomegaly and showed pronounced systolic dysfunction as adults. In addition, only the semilunar valves showed the prominent enlargement previously described for Adam17−/−. Given that HB-EGF is a key modulator of valvulogenesis and a substrate for ADAM17, we expected to see a recapitulation of the Hbegf−/− valvular phenotype (Iwamoto et al., 2010; Iwamoto et al., 2003; Jackson et al., 2003) in Tie2-Cre/+;Adam17^Fl/ΔEx5 mice. Instead, we found differences between the models that indicate at least partial compensatory shedding of HB-EGF in the absence of ADAM17 and suggest that there are other potential targets for this metalloenzyme in valve remodeling. In postnatal mutant valves, which should have matured into thin elongated cusps with well-organized ECM architecture, there were significant changes in the levels and compartmentalization of glycosaminoglycan-rich and collagenous components of the ECM. That alterations in the matrix are in part developmentally regulated was demonstrated by the marked reduction in levels of the proteoglycan versican at very early stages in valve remodeling. Taken together, our findings provide further insights into the role of ADAM17 in valve maturation – a role that is not compensated for by other ADAMs in the absence of ADAM17 – and highlight a previously unrecognized relationship between ADAM17 and modulation of the ECM in semilunar valve cusps.

4.2. Possible compensatory shedding of HB-EGF

The striking similarity of the valve enlargement in HB-EGF and ADAM17 knockouts (Horiuchi et al., 2007; Horiuchi et al., 2005; Iwamoto et al., 2003; Jackson et al., 2003; Sahin et al., 2004) strongly supports a link between this substrate-enzyme pair in valvulogenesis. However, while there was an almost 50% increase in the total number of cells in the aortic valve cusps of Tie2-Cre/+;Adam17^Fl/ΔEx5 embryos at E18.5, there were no differences from controls in the number of BrdU-positive cells at either this stage or earlier (E12.5 and E14.5), in contrast to the hyperproliferation observed at E14.5 in Hbegf−/− valves (Iwamoto et al., 2010; Jackson et al., 2003). Thus, we conclude that soluble HB-EGF is likely still being generated in endothelial ADAM17 deletion mutants, although a lack of reagents precluded our confirming this hypothesis. Because ADAM19 has been shown to proteolytically release cell-surface HB-EGF specifically in the absence of ADAM17 (Horiuchi et al., 2005), we speculate that such compensation may be occurring in our model. Analysis of valves in mice with endothelial deletion of both ADAM17 and ADAM19 would help address potential redundancy between these two enzymes in HB-EGF processing. Regardless, it is likely that neither enzyme can completely compensate for lack of the other, since both global deletion mutants present with similar cardiac abnormalities, including valve malformation and ventricular septal defects (Horiuchi et al., 2007; Horiuchi et al., 2005; Jackson et al., 2003; Kurohara et al., 2004; Sahin et al., 2004; Zhou et al., 2004). These septal defects in Adam17−/− embryos point to a role for this enzyme in cardiac neural crest cells, as was elegantly demonstrated for ADAM19 (Komatsu et al., 2007). Interestingly, cardiac septal defects were not observed in Hbegf−/− mice (Jackson et al., 2003), supporting the idea that ADAM17 and ADAM19 could target substrates other than HB-EGF in valvulogenesis. In our model of endothelial ADAM17 ablation, ventricular septal defects were not evident, emphasizing that the valve phenotype is associated with functions of ADAM17 in the endocardial cells and their derivatives.

4.3. Hyperplasia of the semilunar valves

As noted in section 4.2, we were unable to detect enhanced acute cell proliferation in Tie2-Cre/+;Adam17^Fl/ΔEx5 semilunar valves over that in controls at several stages of embryogenesis. Nevertheless, mutant valves presented with a quantifiable increase in cell number by E18.5 and an elevation in cellular content appeared to be maintained in adults. Our data indicated that apoptosis in the mutant outflow valves in late gestation is not altered, making this an unlikely mechanism to account for the excessive cells. It may be that a burst of proliferation has occurred at a time point not measured in our studies. For example, if ADAM19 indeed compensates for the absence of ADAM17 by shedding HB-EGF, but only at E14.5, there could be a shift in the stage at which the increased proliferation is manifested. Alternatively, lack of ADAM17-mediated processing of one or more unidentified substrates in the outflow valve cusps may impact slow cell turnover in a manner not detectable by BrdU incorporation. Another possibility is a gradual, ongoing EMT in semilunar valves of Tie2-Cre/+;Adam17^Fl/ΔEx5 mice, especially as the receptor ALK-5 for TGFβ, a key factor mediating EMT in valve formation, has been identified as a substrate for ADAM17 (Liu et al., 2009; Mu et al., 2011). Although the expression and importance of ALK-5 on endothelial cells in vivo is somewhat controversial (Seki et al., 2006; Sridurongrit et al., 2008), in vitro studies suggest that both the endothelial-specific receptor ALK-1 as well as ALK-5 cooperate in mediating the effects of TGFβ (Goumans et al., 2003; Goumans et al., 2002; Oh et al., 2000). Cell-surface retention of ALK-5 on ADAM17-null endocardium could act to potentiate cellular responses to low levels of TGFβ.

4.4. Limitation of the valve enlargement phenotype to the outflow tract

A surprising element of our studies is that only the semilunar valves were affected by loss of endothelial ADAM17, with the atrioventricular valve leaflets appearing grossly normal in Tie2-Cre/+;Adam17^Fl/ΔEx5 mice. Endocardial cells, cardiac neural crest cells, and cells of the second heart field all contribute to the cusps of the semilunar valves, but endocardially derived cells constitute the bulk of the atrioventricular valve leaflets (Combs and Yutzey, 2009a; Hinton and Yutzey, 2011); thus, our prediction was that an endothelial cell-based phenotype would either be represented in both sets of valves or more pronounced in the atrioventricular over the semilunar. Since homozygosity of hypomorphic alleles of Egfr (wa2) and Adam17 (woe) both lead to enlargement of only the semilunar valves (Chen et al., 2000; Hassemer et al., 2010), it was proposed that these valves are more sensitive than the atrioventricular to levels and activity of the EGFR (Jackson et al., 2003; Tadano et al., 2005). This interpretation could apply to other substrates targeted by ADAM17 in valvulogenesis as well. It may be that ADAM17 present in the surrounding tissue of the atrioventricular valves is sufficient for supporting normal development and remodeling in these valves. Alternatively, there is the possibility that ablation of ADAM17 in the inflow tract is less efficient than in the outflow in our Tie2-Cre/+;Adam17^Fl/ΔEx5 model, allowing a low level of ADAM17 expression and activity in the atrioventricular valve leaflets.

4.5. ECM changes in the semilunar valves

Another explanation for the restriction of the valve enlargement phenotype to the outflow tract could be impaired interactions between ADAM17-null mesenchymal cells and ADAM17-replete cardiac neural crest cells at the remodeling stage. Such interactions have been shown to be important for ECM remodeling in the valve cusps (Jain et al., 2011), although the role of the ECM and the specific molecular mechanisms involved in sculpting of the primitive endocardial cushions into elongated valve structures are not well understood (Hinton and Yutzey, 2011; Lockhart et al., 2011). The most prominent phenotype in aortic valves from postnatal Tie2-Cre/+;Adam17^Fl/ΔEx5 mice, in addition to the hyperplasia, was an expansion of the cusp area staining positive for carbohydrate moieties in the matrix. However, ultrastructural analysis indicated that these moieties are not necessarily associated with an increase in proteoglycans (Fig. 5). In fact, in mutant embryo hearts, levels of the proteoglycan versican were decreased at early stages in valve remodeling, yet there was no obvious impairment of EMT and no visible morphological differences between mutant and control endocardial cushions. The reduced versican content at this stage is unexpected, as there have been no previous reports linking ADAM17 with ECM regulation in heart development (Lockhart et al., 2011) or tissue remodeling. However, in ADAM19-null embryos, there are changes in the appearance and distribution of fibrils containing fibronectin and latent TGFβ binding protein-1 in the atrioventricular endocardial cushions (Kurohara et al., 2004), establishing a precedent for dysregulation of the ECM associated with inactivation of an ADAM proteinase. Because the myocardium initially produces versican and hyaluronan and other cardiac jelly components (Butcher and Markwald, 2007; Combs and Yutzey, 2009a; DeLaughter et al., 2011; Hinton and Yutzey, 2011; Snarr et al., 2008), perturbations in signaling interactions between the ADAM17-null endocardial cells and the myocardium could be a mechanism accounting for the decrease in versican. At later stages in valvulogenesis, production of these molecules and other ECM proteins by the valvular mesenchymal cells is important in both developmental and homeostatic remodeling of the valves (DeLaughter et al., 2011). Absence of ADAM17 in either the endocardium or mesenchyme, or both, may alter the differentiation program of these cells and thus remodeling potential in the valves. Because valvular endocardial cells expressing the transcription factor NFATc1 do not undergo EMT (de la Pompa et al., 1998; Ranger et al., 1998; Wu et al., 2011), the Nfatc1-enhancer Cre line of transgenic mice developed by Wu et al. (Wu et al., 2011) could be used in conjunction with our floxed Adam17 line to separate effects of ADAM17 ablation on endocardial cells versus their mesenchymal derivatives. It will also be of interest in future work to assess the levels of other valve-associated proteoglycans, as well as additional matrix proteins, at early stages in remodeling of Tie2-Cre/+;Adam17^Fl/ΔEx5 semilunar valves.

As mentioned previously, analysis of mutant adult semilunar valves by histochemical staining and electron microscopy pointed to an expansion of the layer dominated by glycosaminoglycan-containing molecules (the spongiosa). Although this expansion may be partially due to increased osmotic swelling, we did find that levels of hyaluronan tended to be higher in aortic valve cusps from mutant adult mice as compared to controls (Fig. S4). However, no difference in hyaluronan content between the genotypes was noted in whole heart extracts at E12.5. A likely explanation for this discrepancy is that the ECM of adult mutant valve cusps undergoes changes in response to altered hemodynamics, whereas such forces would not be as significant at embryonic stages. Along these lines, we observed that versican immunolocalization was much less homogeneous in mutant aortic valve cusps as compared to controls. Associated with this phenotype, about 70% of the Tie2-Cre/+;Adam17^Fl/ΔEx5 adults that we evaluated showed evidence of increased collagen outside the major zone of its localization (the fibrosa) in the aortic valve cusp, suggesting a loss of appropriate ECM stratification. Whether this collagen deposition is merely a compensatory response or a consequence of phenotypic changes in the valvular interstitial cells is still an open question. To our knowledge, a compensatory increase in collagen has not been noted in other models of semilunar valve enlargement involving expansion of the proteoglycan/glycosaminoglycan component (Dupuis et al., 2011; Kern et al., 2010) or a deficiency in EGFR signaling (Chen et al., 2000).

4.6. Valve-related alterations in left ventricular responses and hemodynamics

Measurement of valve competency in 6–7 month old mice by echocardiography demonstrated significant aortic valve stenosis in Tie2-Cre/+;Adam17^Fl/ΔEx5 mutants. Our data suggest that changes in ECM composition, coupled with congenital thickening of the valve cusps, likely increase their stiffness and impede blood flow. This alteration in flow may also impact the hemodynamics of the mutant heart by increasing filling pressures, resulting in the left ventricular hypertrophy and dilation we observed in our mutant mice, as well as heart failure. While we have not performed intrauterine echocardiography on the embryos, our prediction is that there would be little to no difference in valve competency between the mutants and controls throughout most of gestation, as thickening of the semilunar valve cusps was not obvious until late in embryogenesis (Fig. 1 and section 3.4). Our findings strongly suggest that the defects we observed in the left ventricle occur postnatally and are secondary to the semilunar valve abnormalities. Although it is possible that deletion of ADAM17 in coronary endothelium or aberrant cre-mediated ablation of the enzyme in the cardiomyocytes could produce alterations in the ventricular myocardium, these possibilities are unlikely, as we might expect both ventricles to be affected. However, crossing the Adam17 floxed allele with a Cre driver that specifically targets valvular endocardium, such as the Nfatc1-Cre transgenic (Wu et al., 2011) mentioned earlier, would help to eliminate these possibilities.

Expression of ADAM17 in endothelial cells has been associated with promoting angiogenesis and cell migration, particularly under pathological conditions (Gooz et al., 2009; Kwak et al., 2009; Lin et al., 2011; Weskamp et al., 2010). However, we and others have not noted gross abnormalities in development and homeostasis of the vasculature in mice with endothelial deletion of ADAM17 or a global reduction in its levels (Chalaris et al., 2010; Lucitti et al., 2012; Weskamp et al., 2010), apart from a recently discovered decrease in cerebral collateral vessel formation in the absence of endothelial ADAM17 (Lucitti et al., 2012). Collectively, these findings contrast with a report of significant vascular defects, consisting of abnormal vessel branching and capillary network formation and resulting in hemorrhage and edema, in Adam17^ΔZn/ΔZn embryos after day 14.5 in gestation (Canault et al., 2010). The profound defects in the vascular system of global ADAM17 knockouts may be due to earlier and much more widespread deletion of the enzyme in comparison to endothelial ablation. Although we would not rule out subtle alterations in the endothelium of our mutants, lack of a generalized vascular phenotype in unchallenged Tie2-Cre/+;Adam17^Fl/ΔEx5 mutants supports highly specific roles for endothelial ADAM17 in embryogenesis.

4.7. Potential substrates for ADAM17

Based on our data and previous work by others, we conclude that while ADAM17 is likely the major HB-EGF processing enzyme in wild-type semilunar valves, there may be additional cell-surface substrates for ADAM17 in valve morphogenesis and remodeling. As discussed previously, the TGFβ receptor ALK-5 is a possible ADAM17 target (Liu et al., 2009; Mu et al., 2011), and increased cell surface levels of ALK-5 in the absence of ADAM17 could conceivably impact EMT, cell turnover, and ECM production in valve maturation and maintenance. Among other membrane-bound proteins associated with valvulogenesis, RANK (receptor activator of NF-κB) is cleaved by ADAM17 (Hakozaki et al., 2010). RANK and its ligand RANKL are important in feedback regulation of NFATc1 expression in endocardial cells (Combs and Yutzey, 2009b; Lange and Yutzey, 2006), and reduced cleavage of RANK in ADAM17-null cells could augment this signaling pathway. Although the extracellular domain of Notch1, which is clearly required for valve formation (reviewed in (High and Epstein, 2008)), is proteolytically released by ADAM17, this cleavage is ligand independent (Bozkulak and Weinmaster, 2009; Murthy et al., 2012; van Tetering et al., 2009), and ADAM10 is considered the primary physiological Notch1 sheddase in development (Esteve et al., 2011; Gibb et al., 2010; Gibb et al., 2011; Weber et al., 2011). Further studies will be required to evaluate these and other candidate substrates of ADAM17 in valvulogenesis.

4.8. Concluding remarks

In conclusion, deletion of endothelial ADAM17 unexpectedly leads to early changes in the ECM of the primordial semilunar valves, and, in adult mutants, culminates in aortic valve cusps with some features reminiscent of stenotic valve disease. Because congenital malformations may underlie the development and progression of aortic valve disease (Combs and Yutzey, 2009a; Hinton and Yutzey, 2011), further analysis of the mechanisms involved in ADAM17-associated modulation of the ECM could provide unique insight into the etiology of this disorder.

Highlights.

• Deletion of ADAM17 in mouse embryonic endothelial cells reduces postnatal viability

• Persistent valve enlargement occurs in mutants but is limited to semilunar cusps

• Mechanism of valve hyperplasia differs from mice lacking HB-EGF, an ADAM17 substrate

• Mutant hearts show decreased levels of versican early in valve remodeling

• Adult mutants develop significant heart dysfunction due to semilunar valve stenosis