Increased Proteoglycanases in Pulmonary Valves after Birth Correlate with Extracellular Matrix Maturation and Valve Sculpting

Increased mechanical forces on developing cardiac valves drive formation of the highly organized extracellular matrix (ECM) providing tissue integrity and promoting cell behavior and signaling. However, the ability to investigate the response of cardiac valve cells to increased mechanical forces is challenging and remains poorly understood. The developmental window from birth (P0) to postnatal day 7 (P7) when biomechanical forces on the pulmonary valve (PV) are altered due to the initiation of blood flow to the lungs was evaluated in this study. Grossly enlarged PV, in mice deficient in the proteoglycan protease ADAMTS5, exhibited a transient phenotypic rescue from postnatal day 0 (P0) to P7; the Adamts5−/− aortic valves (AV) did not exhibit a phenotypic correction. We hypothesized that blood flow, initiated to the lungs at birth, alters mechanical load on the PV and promotes ECM maturation. In the Adamts5−/− PV, there was an increase in localization of the proteoglycan proteases ADAMTS1, MMP2, and MMP9 that correlated with reduced Versican (VCAN). At birth, Decorin (DCN), a Collagen I binding, small leucine-rich proteoglycan, exhibited complementary stratified localization to VCAN in the wild type at P0 but colocalized with VCAN in Adamts5−/− PV; concomitant with the phenotypic rescue at P7, the PVs in Adamts5−/− mice exhibited stratification of VCAN and DCN similar to wild type. This study indicates that increased mechanical forces on the PV at birth may activate ECM proteases to organize specialized ECM layers during cardiac valve maturation.


Introduction
Adult cardiac valves comprise a specialized, stratified extracellular matrix (ECM) that gives structural support to ensure unidirectional blood flow. In addition to imparting the mechanical properties of tissues, the ECM instructs cell behavior and signaling [1,2]. The reciprocal interactions of cells and ECM are driven by biomechanical forces. However, mechanisms that integrate cell-matrix responses with mechanical load are not well understood.
The provisional ECM within endocardial cushions, the precursors of the cardiac valves, is rich in hyaluronic acid and the aggregating proteoglycan Versican (VCAN). This early ECM provides an environment conducive to cell migration and proliferation. As the embryo grows and mechanical forces are increased, the provisional ECM is remodeled by synthesis of fibrillar collagens, assembly of elastin, and proteolytic cleavage and clearance of VCAN [3]. The mature ECM components become stratified within the developing cusp (also referred to as leaflet) while the cusp morphology changes from block shaped to sculpted. Previous studies have revealed that the valvular endocardium senses changes in mechanical forces and propagates the signals to underlying valvular interstitial cells (VICs, specialized fibroblasts) through multiple mechanotransductive pathways [4,5]. Although cardiac valve

Valve Cusp and Hinge Quantification
To quantify the cusps and hinge regions of the PV as well as the AV, histological sections over a depth of 30 µM were quantified for each heart; i.e., the widest portion of the anterior, right, and left cusps of the PV, and the right coronary, left coronary, and non coronary cusps of the AV were measured. The narrowest part of the hinge regions was also measured and quantified. Measurements were averaged then used for graphs and statistical analysis. Of note, hearts that exhibited a bicuspid phenotype or had other cardiac anomalies such as a ventricular septal defect were not utilized in this study. Amira™ 3D 2021.1 (Visage Imaging, Andover, MA, USA) was used to generate three dimensional (3D) reconstructions from 5 µm-thick paraffin sections. For each P0 or P7 reconstruction, approximately 100-140 sections per valve were used.

Immunohistochemistry
Standard histological procedures were used [31]. Decorin (DCN) antibody (AF1060) was purchased from R & D Systems Inc. Antibodies to α-smooth muscle actin (SMA) (Sigma, A 5228) were used to identify smooth muscle cells. Fluor-conjugated secondary antibodies were purchased from Jackson ImmunoResearch (West Grove, PA, USA). Antibodies were used in murine tissues fixed in Amsterdam (Amst: 5% methanol, 35% acetone, 5% acetic acid) [14] or 4% paraformaldehyde (para). All para-fixed tissue was treated with either citric acid antigen unmasking (H-3300, Vector laboratories, Burlingame, CA, USA) for ECM antibodies to murine VCAN (gift from Dr. S. Hoffman) and DCN or proteinase K treatment for ADAMTS1 (SC5468), MMP2 (AB19167) and MMP9 (AB38898) localization. Post-fixation involved 80% ethanol for 5 min, followed by 50% ethanol for 5 min, and then two rinses in phosphate-buffered saline. Imaging was performed on the Leica TCS SP5 or SP8 AOBS Confocal Microscope System (Leica Microsystems Inc., Exton, PA, USA). Images in panels are an example of a minimum of three different experimental replicas of Adamts5 −/− and littermate controls (Adamts5 +/+ ; wild type-WT). Digital images of Adamts5 −/− and WT heart sections were acquired at identical confocal settings using the Leica TCS SP5 or SP8 AOBS Confocal Microscope System.

Statistics
For quantification of the valve cusp and hinge width, data were analyzed using Graph-Pad Prism 9.1 (GraphPad Software Inc.). A one-way analysis of variance (Anova) or the nonparametric Kruskal-Wallis test was used to determine the differences between the groups wild type (+/+) and Adamts5 −/− (−/−), at timepoints, P0, P7 and 1 mo. When Anova assumptions of normality and equal variances could not be met, a Kruskal-Wallis test was used to compare groups. Adjusted p-values are indicated for each group comparison, with α = 0.05. Anova and Kruskal-Wallis tests revealed significant differences between groups. GraphPad was utilized to generate graphs. In graphs, each symbol represents data from a single mouse (n = 1) for evaluation of cardiac valve cusp and hinge regions. The colored bar height represents the mean of the measurements, with small bars above and below indicating the standard deviation. The animal numbers were assigned randomly prior to genotyping, and this served to blind the investigators until the grouping for statistical analysis. To ensure the fidelity of the Adamts5 −/− homozygous phenotype Adamts5 −/+ het X Adamts5 −/+ het matings were performed; this approach also generated Adamts5 −/− and Adamts5 +/+ littermate controls. The number of mice utilized in this study represents a minimum of 10 different litters per timepoint to obtain an appropriate n for this study for each developmental stage analyzed.

The Pulmonary Valves of Adamts5 −/− Mice Exhibited a Transient Phenotypic Rescue at Postnatal Day 7
We previously published that mice deficient in the ADAMTS5 protease have larger PV, AV, and MVs at embryonic day 17.5 (E17.5) and in adult mice (>6 mo) [14]. In this study, the PV and AV of Adamts5 −/− mice at P0 (birth), P7, and 1 mo were analyzed. Histological sections and 3D reconstructions of the WT at P0 (n = 6) and P7 (n = 3) as well as Adamts5 −/− PV at P0 (n = 7) and P7 (n = 4) were generated and analyzed ( Figure 1). Histological comparisons revealed significant morphological correction of the Adamts5 −/− enlarged valve cusps from P0 to P7 (Figure 1). Comparisons of PV 3D reconstructions from P0 ( Figure 1D), and P7 Adamts5 −/− mice ( Figure 1H) highlighted the thinner and more symmetric valve cusps that were similar to WT at P7. The morphological changes in valve shape and size indicated an apparent rescue of the enlarged valve phenotype found in prenatal and P0 Adamts5 −/− mice to the sculpted morphology observed in WT.

Proteoglycan Proteases ADAMTS1, MMP2, and MMP9 Were Present in Postnatal Cusp and Hinge Regions of the Pulmonary Valve
We reasoned that the transient phenotypic correction of the Adamts5 −/− PV may be due to proteolytic cleavage and clearance of VCAN from proteoglycanases other than ADAMTS5 that were upregulated at birth in the Adamts5 −/− mice.
1. ADAMTS1 is the closest family member to ADAMTS5, exhibits catalytic activity to VCAN, and may compensate for the loss of ADAMTS5 from P0 to P7.
ADAMTS1 was observed in PVs at late stages of valve maturation. Images are representative of n=3 for each genotype and at each stage. At E17.5, there was prominent staining of ADAMTS1 in the endocardium of WT PV cusps with localization to the arterial side ( Figure 3; white arrows). At E17.5, the enlarged cusps in the Adamts5 −/− PV exhibited reduced ADAMTS1 localization compared to WT ( Figure 3B; white outline arrow). However, shortly after birth there was considerably more ADAMTS1 in the Adamts5 −/− PV cusp VICs than late-stage gestation ( Figure 3D, white asterisks). At P0, the WT cusps exhibited a stratified expression of ADAMTS1, with predominant localization on the arterial side

Proteoglycan Proteases ADAMTS1, MMP2, and MMP9 Were Present in Postnatal Cusp and Hinge Regions of the Pulmonary Valve
We reasoned that the transient phenotypic correction of the Adamts5 −/− PV may be due to proteolytic cleavage and clearance of VCAN from proteoglycanases other than ADAMTS5 that were upregulated at birth in the Adamts5 −/− mice.
3.2.1. ADAMTS1 Is the Closest Family Member to ADAMTS5, Exhibits Catalytic Activity to VCAN, and May Compensate for the Loss of ADAMTS5 from P0 to P7 ADAMTS1 was observed in PVs at late stages of valve maturation. Images are representative of n = 3 for each genotype and at each stage. At E17.5, there was prominent staining of ADAMTS1 in the endocardium of WT PV cusps with localization to the arterial side ( Figure 3; white arrows). At E17.5, the enlarged cusps in the Adamts5 −/− PV exhibited reduced ADAMTS1 localization compared to WT ( Figure 3B; white outline arrow). However, shortly after birth there was considerably more ADAMTS1 in the Adamts5 −/− PV cusp VICs than late-stage gestation ( Figure 3D, white asterisks). At P0, the WT cusps exhibited a stratified expression of ADAMTS1, with predominant localization on the arterial side and undetectable expression on the ventrialis side ( Figure 3C,D; white arrows, yellow asterisk). In WT cusps at P7, ADAMTS1 was expressed in the cusps but not in the hinge regions ( Figure 3E, yellow asterisks), while the Adamts5 −/− P7 cusps exhibited ADAMTS1 localization in both the cusp and hinge regions ( Figure 3F). At 1 mo, ADAMTS1 was localized to the distal regions of the cusps in the WT but observed throughout the Adamts5 −/− cusps ( Figure 3G,H). and undetectable expression on the ventrialis side ( Figure 3C,D; white arrows, yellow asterisk). In WT cusps at P7, ADAMTS1 was expressed in the cusps but not in the hinge regions ( Figure 3E, yellow asterisks), while the Adamts5 −/− P7 cusps exhibited ADAMTS1 localization in both the cusp and hinge regions ( Figure 3F). At 1mo, ADAMTS1 was localized to the distal regions of the cusps in the WT but observed throughout the Adamts5 −/− cusps ( Figure 3G,H).   Figure 4A,B). MMP2 expression was stratified on the arterial side of the cusps shortly after birth in WT mice but was not observed in the ventricularis region ( Figure 4C; white arrows and yellow asterisk). Adamts5 −/− cusps exhibited a more homogenous staining pattern of MMP2 in the cusp than WT (Figure 4D). At P7, the WT and the Adamts5 −/− cusps exhibited similar MMP2 localization in the endocardium and VICs.
2. MMP2 was evident in the pulmonary valve and may play a role in late-staged valve maturation.
MMP2 was localized within the endocardium and valvular interstitial cells of the cusps in late gestation ( Figure 4A,B). MMP2 expression was stratified on the arterial side of the cusps shortly after birth in WT mice but was not observed in the ventricularis region ( Figure 4C; white arrows and yellow asterisk). Adamts5 −/− cusps exhibited a more homogenous staining pattern of MMP2 in the cusp than WT (Figure 4D). At P7, the WT and the Adamts5 −/− cusps exhibited similar MMP2 localization in the endocardium and VICs. 3. MMP9 was localized in the maturing valve cusps in regions where the mesenchymal cells were highly compacted.
At E17.5, MMP9 was localized in the PV hinge and cusp regions of the WT ( Figure  5A, arrows) but MMP9 was not detected in the distal cusp regions of the Adamts5 −/− (Figure 5B, white asterisk) where the VICs are not as condensed as the WT [14]. At P0, MMP9 exhibited a stratified expression pattern in the WT cusps ( Figure 5C, yellow asterisk-no staining, arrows-staining); and was localized in the hinge regions ( Figure 5C, h); the Adamts5 −/− cusps exhibited homogeneous expression throughout the cusps and localization in the hinge regions ( Figure 5D). At P7, the localization of MMP9 was similar in the hinge and cusp regions in WT and Adamts5 −/− PV ( Figure 5E,F).  Figure 5B, white asterisk) where the VICs are not as condensed as the WT [14]. At P0, MMP9 exhibited a stratified expression pattern in the WT cusps ( Figure 5C, yellow asterisk-no staining, arrows-staining); and was localized in the hinge regions ( Figure 5C, h); the Adamts5 −/− cusps exhibited homogeneous expression throughout the cusps and localization in the hinge regions ( Figure 5D). At P7, the localization of MMP9 was similar in the hinge and cusp regions in WT and Adamts5 −/− PV ( Figure 5E,F). Localization of The Small Leucine-Rich Proteoglycan Decorin was overlapping with versican at P0 in the Adamts5 −/− cusps, but at P7 DCN and VCAN were restricted to separate layers consistent with 'Phenotypic Rescue'.
The Small Leucine Rich Proteoglycan (SLRP) Decorin (DCN), named because it decorates Collagen I, was evident in the venticularis of the cusps, as well as the anchor and hinge regions by P0 in both the WT and Adamts5 −/− PV ( Figure 6A-F). The organization of DCN at P0 in the PVs appeared more prominent in the WT ( Figure 6A,B) than Adamts5 −/− ( Figure 6D,E). In the WT PV at P0, the immunolocalization of DCN ( Figure 6B, green, yellow asterisk) and versican (VCAN) ( Figure 6C, blue, yellow asterisk) exhibited a complementing pattern while in the Adamts5 −/− PV at P0, DCN and VCAN were overlapping ( Figure 6E,F, yellow line). However, at P7, the organization of DCN in the Adamts5 −/− PV was similar in the WT ( Figure 6G-L); i.e., there was a distinction in the ECM layers of DCN and VCAN expression in the WT ( Figure 6G,H, yellow asterisks) as well as the Adamts5 −/− PV at P7 ( Figure 6J,K, yellow asterisks).

Localization of the Small Leucine-Rich Proteoglycan Decorin Was
Overlapping with Versican at P0 in the Adamts5 −/− Cusps, but at P7 DCN and VCAN Were Restricted to Separate Layers Consistent with 'Phenotypic Rescue' The Small Leucine Rich Proteoglycan (SLRP) Decorin (DCN), named because it decorates Collagen I, was evident in the venticularis of the cusps, as well as the anchor and hinge regions by P0 in both the WT and Adamts5 −/− PV ( Figure 6A-F). The organization of DCN at P0 in the PVs appeared more prominent in the WT ( Figure 6A,B) than Adamts5 −/− ( Figure 6D,E). In the WT PV at P0, the immunolocalization of DCN ( Figure 6B, green, yellow asterisk) and versican (VCAN) ( Figure 6C, blue, yellow asterisk) exhibited a complementing pattern while in the Adamts5 −/− PV at P0, DCN and VCAN were overlapping ( Figure 6E,F, yellow line). However, at P7, the organization of DCN in the Adamts5 −/− PV was similar in the WT ( Figure 6G-L); i.e., there was a distinction in the ECM layers of DCN and VCAN expression in the WT ( Figure 6G,H, yellow asterisks) as well as the Adamts5 −/− PV at P7 ( Figure 6J,K, yellow asterisks). The immunolocalization of DCN and VCAN indicated that the mature fibrosa and spongiosa layers, respectively, were distinct at P7 in both the WT and Adamts5 −/− PV which correlated with the P7 phenotypic rescue of the enlarged Adamts5 −/− PV.

Discussion
In this study, we investigated ECM protease-substrate localization in the early postnatal timeframe where biomechanical forces are altered due to the initiation of blood circulation to the lungs. The fully penetrant enlarged PV cusp and hinge phenotypes in the Adamts5 −/− mice were resolved to a sculpted appearance indistinguishable from WT at P7. This apparent phenotypic rescue of the Adamts5 −/− PV may be the result of increased ECM proteoglycanase activity in the VICs and endocardium after birth; of note, these proteoglycanases were reduced in the Adamts5 −/− PV compared to WT prior to birth. ADAMTS1, MMP2, and MMP9, that also cleave VCAN, may compensate for the loss of ADAMTS5 by decreasing VCAN levels thereby restoring normal PV morphology. The apparent increase in proteoglycanase activity at birth in the PV may be the result of increased blood flow to the lungs. It is likely that ADAMTS-VCAN-substrate interactions are activated by mechanotransducive pathways that drive cardiac valve maturation (Figure 7). The immunolocalization of DCN and VCAN indicated that the mature fibrosa and spongiosa layers, respectively, were distinct at P7 in both the WT and Adamts5 −/− PV which correlated with the P7 phenotypic rescue of the enlarged Adamts5 −/− PV.

Discussion
In this study, we investigated ECM protease-substrate localization in the early postnatal timeframe where biomechanical forces are altered due to the initiation of blood circulation to the lungs. The fully penetrant enlarged PV cusp and hinge phenotypes in the Adamts5 −/− mice were resolved to a sculpted appearance indistinguishable from WT at P7. This apparent phenotypic rescue of the Adamts5 −/− PV may be the result of increased ECM proteoglycanase activity in the VICs and endocardium after birth; of note, these proteoglycanases were reduced in the Adamts5 −/− PV compared to WT prior to birth. ADAMTS1, MMP2, and MMP9, that also cleave VCAN, may compensate for the loss of ADAMTS5 by decreasing VCAN levels thereby restoring normal PV morphology. The apparent increase in proteoglycanase activity at birth in the PV may be the result of in-creased blood flow to the lungs. It is likely that ADAMTS-VCAN-substrate interactions are activated by mechanotransducive pathways that drive cardiac valve maturation (Figure 7). Mechanosensors of shear force such as Krüppel-like factor 2 (Klf2/4) and endothelial nitrous oxide synthetase are expressed by endocardial cells of developing valves and their genetic perturbation in mice disrupts cardiac valve development [32][33][34][35][36]. Proteoglycanases ADAMTS5 [14], ADAMTS1 [37], and MMP2 are localized to the valvular endocardium, where shear and compressive forces are generated. Moreover, ADAMTS-cleaved VCAN fragments are found in the endocardial ECM as well as the adjacent VICs [14]. ADAMTS5 interacts with Klf2/4 as part of the endothelial cell cerebral cavernous malformation pathway which, when disrupted, leads to over-expression of Adamts5 and premature loss of provisional cardiac valve ECM (cardiac jelly) with reduced proliferation of adjacent myocardial cells. Loss of endocardial expression of Adamts19 also perturbs shear stress signaling and leads to increased mesenchymal cells and proteoglycan deposition in cardiac valves [26]. Klf2 together with Adamts5 are upregulated in the Adamts19 −/− valvular endocardium [26]. It is not clear if Adamts5 and Adamts19 are direct targets of the mechano-induced Klf2, but they are integrated within mechanosenstive pathways localized to the endocardium.
During valve development, mechanical signals are transmitted from the endocardium to the underlying VICs, in part through incorporation of nitrous oxide. Mechanosensitive ion channels also regulate endothelial response to shear flow [38,39]. By midgestation, a subset of VICs, adjacent to the endocardium, is differentiated from others by their condensed cell behavior, as well as localization of cytoskeletal proteins, αSMA and Filamin A [16]. This subset of VICs is devoid of intact VCAN, but resides in an ECM of cleaved VCAN (referred to as DPEAAE or Versikine) [14]. Deletion of ADAMTS5 with loss of cleaved VCAN results not only in the disruption of VIC condensation, but also loss of αSMA, and Filamin A localization [16]. Since the actin cytoskeleton is highly responsive to changes in mechanical forces [40,41], this further implicates the ADAMTS5-VCAN, protease-substrate pair as an essential component of mechanically responsive pathways during cardiac valve development. The mechanisms by which cleaved VCAN facilitates a cytoskeletal response, and/or increased VCAN inhibits the cytoskeletal organization, will require further investigation. Mechanosensors of shear force such as Krüppel-like factor 2 (Klf2/4) and endothelial nitrous oxide synthetase are expressed by endocardial cells of developing valves and their genetic perturbation in mice disrupts cardiac valve development [32][33][34][35][36]. Proteoglycanases ADAMTS5 [14], ADAMTS1 [37], and MMP2 are localized to the valvular endocardium, where shear and compressive forces are generated. Moreover, ADAMTS-cleaved VCAN fragments are found in the endocardial ECM as well as the adjacent VICs [14]. ADAMTS5 interacts with Klf2/4 as part of the endothelial cell cerebral cavernous malformation pathway which, when disrupted, leads to over-expression of Adamts5 and premature loss of provisional cardiac valve ECM (cardiac jelly) with reduced proliferation of adjacent myocardial cells. Loss of endocardial expression of Adamts19 also perturbs shear stress signaling and leads to increased mesenchymal cells and proteoglycan deposition in cardiac valves [26]. Klf2 together with Adamts5 are upregulated in the Adamts19 −/− valvular endocardium [26]. It is not clear if Adamts5 and Adamts19 are direct targets of the mechano-induced Klf2, but they are integrated within mechanosenstive pathways localized to the endocardium.
During valve development, mechanical signals are transmitted from the endocardium to the underlying VICs, in part through incorporation of nitrous oxide. Mechanosensitive ion channels also regulate endothelial response to shear flow [38,39]. By mid-gestation, a subset of VICs, adjacent to the endocardium, is differentiated from others by their condensed cell behavior, as well as localization of cytoskeletal proteins, αSMA and Filamin A [16]. This subset of VICs is devoid of intact VCAN, but resides in an ECM of cleaved VCAN (referred to as DPEAAE or Versikine) [14]. Deletion of ADAMTS5 with loss of cleaved VCAN results not only in the disruption of VIC condensation, but also loss of αSMA, and Filamin A localization [16]. Since the actin cytoskeleton is highly responsive to changes in mechanical forces [40,41], this further implicates the ADAMTS5-VCAN, protease-substrate pair as an essential component of mechanically responsive pathways during cardiac valve development. The mechanisms by which cleaved VCAN facilitates a cytoskeletal response, and/or increased VCAN inhibits the cytoskeletal organization, will require further investigation.
The limitations of this study include the assumption that mechanical load changes on the PV when blood circulation to the lungs is initiated at birth. Proteoglycans are also regulated by transcription, not evaluated here. ADAMTS and MMP activity is inhibited by TIMP3 and not investigated in this study. While the known repertoire of substrates for the ADAMTS proteoglycanases is limited, our understanding of their role in development and disease will be significantly advanced as additional in vivo substrates are identified [7,42,43].

Conclusions
This study indicates that proteolytic cleavage of proteoglycans may be induced by altered mechanical load to promote ECM maturation during cardiac valve development. In the murine model of ADAMTS5 deficiency, loss of this protease results in enlarged cardiac valves that contain excess VCAN; the increased localization of other proteoglycanases in the maturing valve cusps corrects the morphological defect by P7 and correlates with ECM organization, specifically VCAN in the spongiosa and DCN in the Collagen I-rich fibrosa layer. This study indicates that proteoglycanase activity may be an essential downstream mechanism of mechanotransduction that is important for normal cardiac valve maturation. Since increases in proteoglycans are evident in diseased cardiac valve tissues, understanding mechanisms that reduce proteoglycan content is essential for developing effective therapeutic strategies to treat cardiac valve diseases.