Enhancer-associated aortic valve stenosis risk locus 1p21.2 alters NFATC2 binding site and promotes fibrogenesis.

Genome-wide association studies for calcific aortic valve stenosis (CAVS) previously reported strong signal for noncoding variants at 1p21.2. Previous study using Mendelian randomization suggested that the locus controls the expression of PALMD encoding Palmdelphin (PALMD). However, the molecular regulation at the locus and the impact of PALMD on the biology of the aortic valve is presently unknown. 3D genetic mapping and CRISPR activation identified rs6702619 as being located in a distant-acting enhancer, which controls the expression of PALMD. DNA-binding assay showed that the risk variant modified the DNA shape, which prevented the recruitment of NFATC2 and lowered the expression of PALMD. In co-expression network analysis, a module encompassing PALMD was enriched in actin-based process. Mass spectrometry and functional assessment showed that PALMD is a regulator of actin polymerization. In turn, lower level of PALMD promoted the activation of myocardin-related transcription factor and fibrosis, a key pathobiological process underpinning CAVS.

Introduction

Calcific aortic valve stenosis (CAVS) is a highly prevalent heart valve disorder characterized by a fibrocalcific process (Lindman et al., 2016). There is no approved pharmacologic treatment for CAVS. For symptomatic patients in end-stage disease, the only therapeutic option is to replace the aortic valve (AV) by either surgical or transcatheter approaches, which are associated with significant mortality/morbidity and elevated cost (Lindman et al., 2016). The identification of a molecular phenome (Koonin and Wolf, 2010) (i.e., gene regulation, expression, function) is thus a priority as it could lead to the development of novel noninvasive therapies to prevent or treat CAVS. The development of CAVS is a slow process, which relies on a progressive remodeling of the extracellular matrix (ECM) (Chen and Simmons, 2011). Mesenchymally derived valve interstitial cells (VICs) are typified by a high plasticity (Schlotter et al., 2018). VICs transition from quiescent into activated cells during tissue repair (Taylor et al., 2000). The acquisition of a secretory phenotype by activated VICs is one of the earliest feature involved in the pathogenesis of CAVS. Key underpinning molecular circuits that promote the activation of VICs leading to fibrosis and remodeling of the AV are largely unknown.

During the last several years, genome-wide association (GWA) studies have identified thousands of loci associated with diverse complex trait disorders (Welter et al., 2014). GWA and Mendelian randomization studies have contributed to identify causal drivers involved in different pathological conditions and in certain cases have fueled the development of novel therapies (Dewey et al., 2017). Hence, GWA-based approaches hold promise for the identification of molecular drivers involved in the pathogenesis of several complex trait disorders. However, as the vast majority of disease-associated variants reside in the noncoding genome, the identification of causal variants and their target genes remains a challenging task. Data have shown that noncoding variants with cis-regulatory activity are enriched in distant-acting enhancers. Recently, two independent GWA studies identified rs7543130 and rs6702619, which are both located on chromosome 1p21.2 and in perfect linkage disequilibrium (LD) (r2 = 1), as being associated with CAVS at a genome-wide significance level (Helgadottir et al., 2018; Thériault et al., 2018). Cis-expression quantitative trait loci (eQTL) in AVs showed that the index variant, which is intergenic, is associated with the expression of PALMD encoding Palmdelphin (PALMD) (Thériault et al., 2018). Mendelian randomization showed that lower expression of PALMD in AVs was causally associated with CAVS (Thériault et al., 2018). In the present work, mapping and functional characterization identified the 1p21.2 risk locus as an enhancer and showed that rs6702619 alters DNA shape and disrupts a transcription factor binding site (TFBS) for nuclear factor of activated T cells 2 (NFATC2/NFAT1). We identified PALMD as a regulator of actin polymerization. Lower expression of PALMD promoted the polymerization of actin, the activation of myocardin-related TF, the expression of smooth muscle α-ACTIN/ACTA2, and the development of a fibrogenic program, a key process in the development of CAVS.

Results

Prioritization of noncoding gene variant at the PALMD risk locus

We previously performed Bayesian colocalization analysis by using a GWA meta-analysis for CAVS in QUEBEC-CAVS and UK Biobank (n = 2,359 cases, n = 350,060 controls) (Li et al., 2020) and eQTL data from 233 AVs. At the PALMD locus, the analysis revealed a strong posterior probability of shared signal between genetic association for CAVS and AV eQTL (PP4 = 0.997) (Li et al., 2020). Rs6702619, which is intergenic, was identified as the variant with the strongest associations with the GWA and valve eQTL. In VICs, an assay for transposase-accessible chromatin and sequencing (ATAC-seq) showed that rs6702619 and variants in strong LD were located in open chromatin (Figure 1). We next performed chromatin immunoprecipitation (ChIP) and DNA massively parallel sequencing (ChIP-seq) with an antibody directed against H3K4me1, an enhancer-associated histone mark. At the genome-wide scale, we detected 127,329 H3K4me1 peaks in VICs. Annotation of H3K4me1 peaks with GREAT (McLean et al., 2010) showed an enrichment in Gene Ontology (GO) for focal adhesions (FAs), cell-substrate adherens junction, actomyosin, and stress fibers (Table S1). At the 1p21.2 locus, which includes variants in strong LD (spanning ∼10 kbp), the region encompasses enhancer-associated H3K4me1 peaks (Figure 1). In VICs, we performed a ChIP-seq for H3K27ac, which generated 82,662 peaks and showed that 1p21.2 has significant extended peaks for this mark and is thus compatible with the presence of active enhancers (Figure 1). Distant-acting enhancers are enriched in regulatory chromatin loops with promoters. In human primary VICs, we generated a genome-wide chromatin interaction map of enhancers and promoters by performing a H3K27ac HiChIP (Mumbach et al., 2016). Figure 2A shows the H3K27ac HiChIP interaction map of chromosome 1 and a close-up view with a resolution up to 5 kb. We implemented HOMER to identify high-confidence 3D interactions using a stringent false discovery rate (FDR) < 1 × 10−6. We found that the region encompassing rs6702619 had significant chromatin interaction with the promoter region of PALMD distant from ∼65 kb (Figure 2B). Data were next interrogated by using a virtual 4C analysis with a viewpoint as an anchor and visualized in 2D (Mumbach et al., 2016). Virtual 4C with an anchor at rs6702619 provided a visualization of the interaction between the distant-acting enhancer and the promoter of PALMD (Figure 2B). These data were next confirmed by using a chromosome conformation capture assay (3C) using specific primers designed to ensure similar efficacy among the different primer sets covering the region (Table S2). A negative random ligation control with bacterial artificial chromosome (BAC) spanning the region (GRCh37/hg19 chr1: 99966146-100154958) showed that the primers did not detect significant signal (Table S3). Figure 2C shows 3C data in VICs anchored at rs6702619 and the significant interactions with the promoter region of PALMD. As another negative control, 3C and virtual 4C (using H3K27ac HiChIP) at the beta globin locus, which is not active in VICs, showed no significant interaction between the locus control region (LCR) (Krivega et al., 2015) and the globin genes (Figures S1A and S1B and Table S4). Hence, chromatin-associated marks and chromatin contact mapping suggest that the risk locus comprises an enhancer, which is spatially coordinated with the promoter of PALMD. Regulatory loci are often conserved in the genome. Among the different variants in LD, the only polymorphism showing a high degree of conservation is rs6702619 with PhastCons and PhyloP scores of 1 and 3.39, respectively (Table S5). To further prioritize noncoding gene variants in LD (r2 > 0.6) with regulatory activity at the risk locus we performed an integrative weighted (IW)-scoring analysis (Wang et al., 2018), which combines functional annotations of 11 different scoring methods. IW-scoring identified rs6702619 as the most relevant functional variant (p = 0.009) (Table S6).

Figure 1

Variant rs6702619 is located in an enhancer controlling PALMD expression

The upper panel shows the genetic associations with CAVS surrounding rs6702619. The linkage disequilibrium (LD; r2 values) for all SNPs with rs6702619 is indicated by colors. The bottom panel shows ATAC and ChIP-seq in human VICs.

Figure 2

Variant rs6702619 connects with the promoter region of PALMD

(A) H3K27ac HiChIP contact matrix for VICs at 500-, 25-, and 5-kb resolutions.

(B) Virtual 4C analysis revealing contact between rs6702619 and PALMD promoter region; below are H3K27ac HiChIP 1D track and the arc representing significant normalized interaction identified by HOMER.

(C) NspI digestion pattern and chromosome conformation capture (3C) showing association between rs6702619 and PALMD promoter region.

CRISPR activation at the rs6702619 promotes the expression of PALMD

Functional annotations and chromatin looping pattern suggest that rs6702619 has regulatory activity at the 1p21.2 risk locus. Figure S2A shows the topologically associated domain (TAD) in mesenchymal cells and centered on rs6702619. TADs, which are largely conserved between cell types, are compartments that promote and restrict chromatin looping within a submegabase domain (Dixon et al., 2012). This TAD includes LPPR5, LPPR4, and FRRS1, and the nearest gene to rs6702619 is PALMD (∼65kb) (Figure S2A). To confirm that rs6702619 is within a regulatory region that exerts a control over the expression of PALMD, we used the clustered regularly interspersed short palindromic repeats (CRISPR)-mediated gene activation (CRISPRa) system (Figure 3A). Nuclease-deficient Cas9 (dCas9) was used to target the catalytic core of p300, a histone 3 lysine 27 acetyltransferase (H3K27ac) and positive regulator of enhancer/promoter activity, to the rs6702619 locus. Single guide RNA (sgRNA) targeting rs6702619 were cloned in a vector containing dCas9 fused with the catalytic core (amino acids 1048–1664) of p300. The control consisted in a vector containing identical target-specific sgRNA and a dCas9 fused with a catalytically inactive mutant p300 D1399Y. In VICs, the active vector dCas9-p300 increased the level of H3K27ac at the rs6702619 locus as shown by quantitative chromatin immunoprecipitation (Figure S2B). Compared with control dCas9-p300-D1399Y vector, the transfection of dCas9-p300 targeting rs6702619 in VICs increased the expression of PALMD by 1.3-fold (Figure 3B). Genes within 1 Mb from PALMD and sharing the same TAD were not modulated to a significant extent by CRISPRa at rs6702619 (Figure 3B). These data thus indicate that rs6702619 is functionally coordinated to PALMD and exerts a significant control over its expression in human VICs.

Figure 3

Impact of gene variant on NFATC2 binding

(A) Scheme showing dCas9 fused with the acetyltransferase p300 for epigenome editing.

(B) Expression of PALMD, FRRS1, LPPR4, and LPPR5 in VICs in response to epigenome editing (n = 5); p300-D1399Y is the inactive mutant p300.

(C) Position-weighted matrix of NFATC2 binding site sequence. Arrow indicates rs6702619 base position where reference allele is T.

(D) Representation of NFATC2/DNA complex from crystallography data (MG: major groove, mG: minor groove). Right panel shows a close-up view of the interaction between NFATC2 amino acids and DNA bases.

(E) Minor groove width (MGW) and helical twist (HelT) parameters in the presence of T or G allele.

(F) Electrostatic potential of the minor groove in the presence of T or G allele. Red color intensity indicates negative electrostatic potential.

(G–L) (G and H) Luciferase reporter assay of enhancer constructions containing T or G allele (G) without and (H) with overexpression of NFATC2 (n = 5) normalized by the minimal promoter. (I) NFATC2 mRNA (n = 5), (J) NFATC2 protein (n = 3), and (K) PALMD mRNA expression levels following NFATC2 siRNA knockdown (n = 5). (L) DNA-protein ELISA assay for NFATC2 interaction with 30-mer double-stranded oligonucleotides centered on rs6702619 and containing either the T or G allele (n = 5). Values are mean ± SEM, ∗p < 0.05. Statistical analyses, normal distribution testing: Shapiro-Wilk, (J) Wilcoxon-Mann-Whitney, (B, G–I, and K) Student's t test.

Risk variant rs6702619 disrupts a NFATC2 binding site

Cis-acting variants may impact gene expression by altering TF occupancy. Variants in strong LD (r2 > 0.9) at the 1p21.2 risk locus were thus analyzed for their potential to alter TFBS by using an algorithm based on position weight matrix score (Kumar et al., 2017). Of the five variants in strong LD (r2 > 0.9) only rs6702619 is included in a TF motif. The risk allele G at rs6702619 is within a core TFBS [5′-TT(T/G)TCCA-3′] for NFATC2, a TF involved in heart valve morphogenesis (Chang et al., 2004) (Figure 3C). Structural crystallographic data show that monomeric NFATC2 interacts with the major and minor grooves of DNA (PDB: 1OWR) (Figures 3D and S3). In the major groove, Q571 and Y424, which are in a loop of the Rel Homology Region (RHR) of the N-terminal portion of NFATC2, interact with AT base pair of the protective T allele at rs6702619 (Figures 3D and S3). Distribution of hydrogen bonds, which confer base-amino acid specificity, shows that glutamine (Q) has a preference for AT base pairs and thus substitution of T by G may affect the interaction with NFATC2 in the major groove (“base readout”) (Luscombe et al., 2001). We next evaluated the effect of the risk variant on the predicted DNA shape (three-dimensional structure), an important feature for DNA-protein interaction, by using an experimentally validated algorithm (Rohs et al., 2009). Substitution of T by G at rs6702619 is predicted to modulate local DNA three-dimensional structure with lower degree in helical twist and increased minor groove width (Figures 3E and S4A). Side chain of a positively charged arginine residue (R537) in NFATC2 is projected in a narrow DNA minor groove of the consensus motif and provides electrostatic interaction (Figure 3D). Modeling and non-linear Poisson-Boltzmann calculations indicate that an increment in minor groove width, which is associated with the risk allele G, diminishes electronegativity and thus affects arginine-dependent electrostatic interaction and may alter the binding of NFATC2 (“shape readout”) (Siggers and Gordân, 2014) (Figure 3F). Hence, the atomic resolution structure of NFATC2 with DNA indicates that the risk allele G alters base and shape readouts. This may significantly change the affinity and biological effect of NFATC2 (Yang et al., 2017). To substantiate the modeling derived from atomic-level data, we cloned a region of 500 bp centered on rs6702619 with a minimal promoter coupled to luciferase (see Figure S4B for the cloning details). Reporter vectors were transfected in human VICs. Compared with the minimal promoter, the reporter activity was increased by 3.5-fold in the construct containing the reference allele T (Figure 3G). In contrast, the vector containing the risk allele G decreased the reporter activity by 68% compared with the protective allele (Figure 3G). In addition, after normalization for the minimal promoter, the co-transfection of the enhancer reporter and NFATC2-encoding vectors showed that the activity of the protective T allele was increased by 5.9-fold compared with the risk allele (Figure 3H). These data suggested that NFATC2 is an important regulator of this locus and that risk allele G significantly reduces enhancer activity. To this effect, short interfering RNA (siRNA) for NFATC2 in human VICs (Figures 3I and 3J) lowered the expression of PALMD by 37% (Figure 3K). Next, we performed a DNA-binding ELISA assay. Biotin-labeled 30-mer double-stranded oligonucleotides centered on rs6702619 were synthesized and linked to streptavidin-coated plates to assess the binding of NFATC2. This experiment showed that replacement of T by G at rs6702619 significantly lowered the binding affinity for NFATC2 between 60% and 89% (Figure 3L). Hence, the risk allele G at rs6702619 modifies a TFBS and negatively regulates the affinity for NFATC2, which leads to lower expression of PALMD.

Weighted gene co-expression network analysis

Gene variants and altered gene expression profile during CAVS are likely part of a larger network organization, which modifies biological responses. We performed an unbiased weighted gene co-expression network analysis (WGCNA) in 233 surgically explanted mineralized AVs from which transcriptome-wide expression data were generated (Thériault et al., 2018). A clustering dendrogram shows the 20 different modules of co-expression profile (Figure S4C). The red module includes 283 genes and encompasses PALMD (Table S7). As expected, several gene co-expression modules were significantly related to hemodynamic indices of CAVS severity (Figure 4A). Weighted co-expression of the PALMD module (red) was negatively and positively related to the peak transaortic gradient (p = 1 × 10−7) and the AV area (p = 3 × 10−4), respectively (Figure 4A). These data are consistent with the inverse relationship between the expression level of PALMD and CAVS risk. The PALMD (red) module has strong connection (adjacency) with the salmon module (Figure S4C), which encompasses 94 genes (Table S8). GO analysis showed that developmental pathways such as heart morphogenesis (P = 3.5 × 10−4) and actin filament-based process (P = 9.7 × 10−5) were enriched in the PALMD module (red) and were shared with the connected (salmon) module (Figure 4B and Table S9). These data thus suggest that PALMD is associated in diseased AVs with the reactivation of genes involved in developmental pathways, regulation of cell fate, and actin-dependent processes.

Figure 4

Weighted gene co-expression network and relationships with biological and clinical data

(A) Module-trait relationships for WGCNA in 233 calcified aortic valves.

(B) GO analysis for the red (MEred) (includes PALMD) and salmon (MEsalmon) modules.

PALMD regulates cell shape and actin-based dynamics

To document the function of PALMD in VICs, we first evaluated its cellular distribution. Immunofluorescence in Triton-treated VICs showed that a significant amount of PALMD was co-distributed with F-actin (Figure S5A). Knockdown of PALMD with small interfering RNA (siRNA), which lowered the mRNA and protein levels (Figure 5A), revealed a striking modification in cell shape. siRNA-treated VICs had higher cell area and lower length/width ratio, indicating a major function for PALMD to control cell shape (Figures 5B and 5C) (see Figure S5B for a wide field view). Phalloidin staining of F-actin showed that VICs with a knockdown of PALMD had strong stress fibers (Figure 5D). Quantification showed a significant increase of F-actin area per cell after a knockdown of PALMD (Figure 5E). FAs are dynamic and active structures, which are co-regulated with actin stress fibers. We thus quantified FAs in response to PALMD modulation. In VICs, siRNA-mediated knockdown of PALMD increased the number of FAs (as shown by immunofluorescence for vinculin) by 3-fold (Figures 5F and 5G). Taken together, these data suggest that PALMD controls in VICs the dynamic of actin and FAs, which could thus modify signaling and gene expression.

Figure 5

PALMD levels affect VICs morphology

(A–G) siRNA-mediated knockdown of PALMD on (A) mRNA (n = 8) and protein levels (n = 5), (B) cell area, (C) length/width ratio, (D and E) F-actin polymerization (n = 4), and (F and G) focal adhesion formation (vinculin) (n = 3). Scale bars, 10 μm. Values are mean ± SEM. ∗p < 0.05. Statistical analyses, normal distribution testing: Shapiro-Wilk, (B, C, E, and G) Wilcoxon-Mann-Whitney, (A) Student's t test.

PALMD regulates fibrogenesis

As modification of actin dynamics is linked to cell activation, we measured the level of ACTA2, a marker of VIC activation with a secretory phenotype and considered as an early marker in the development of CAVS (Rutkovskiy et al., 2017). In VICs, after a knockdown of PALMD, the mRNA and protein levels of ACTA2/α-ACTIN increased by 1.5- and 1.3-fold respectively (Figures 6A and 6B). In line with this finding, we documented that siRNA-mediated knockdown of PALMD also increased the mRNA levels for TGFB2, COL1A2, MMP2, and MMP9 (Figures 6C–6F). These findings suggested that PALMD regulates tissue remodeling and fibrogenesis, a process involved in the development of CAVS. To buttress these findings we measured the protein level of collagen type I, a predominant collagen in the AV and highly expressed in surgically explanted mineralized AVs (Hutson et al., 2016). In VICs, the knockdown of PALMD for 72 h increased the synthesis of collagen type 1 by 1.3-fold (Figure 6G). The expression of COL1A2, MMP9, and TGFB2 is under the regulation of serum response factor (SRF) and myocardin-related transcription factors (MRTFs) (Gilles et al., 2009; Haak et al., 2014; Kuzniewska et al., 2013; Li et al., 2012). MRTFs contain an N-terminal RPEL domain by which they interact with G actin. During the polymerization of actin, reduced G actin pool contributes to release MRTFs, which translocate to the nucleus and act as cofactors for SRF. Thus, we reasoned that a lower expression of PALMD may activate the nuclear translocation of MRTF-A (also known as MKL1 or MAL). In isolated VICs, 24 h after the transfection with an siRNA-targeting PALMD, we observed a significant rise of MRTF-A in nuclear extracts (Figure 6H). In a reporter assay for SRF response elements conducted in COS-7, which do not express appreciable level of PALMD (Figure S5C), we evaluated the response to a vector expressing PALMD. Consistently, this assay showed that the overexpression of PALMD in COS-7 reduced by 38% the SRF reporter activity (Figure 6I). These data thus underlined that PALMD regulates the MRTF-SRF pathway.

Figure 6

Lowering PALMD induces fibrogenesis

(A–H) siRNA-mediated knockdown of PALMD on (A) ACTA2/α-ACTIN mRNA and (B) protein levels; (C) TGFB2, (D) COL1A2, (E) MMP2, and (F) MMP9 mRNA levels (n = 5); (G) COL1A1 protein expression (n = 6); and (H) MRTF-A nuclear localization (n = 3).

(I) SRF response element luciferase reporter assay (n = 4). Values are mean ± SEM, ∗p < 0.05. Statistical analyses, normal distribution testing: Shapiro-Wilk, (A and C–F) Wilcoxon-Mann-Whitney, (B and G–I) Student's t test.

PALMD is a regulator of actin polymerization

PALMD is a member of the paralemmin family and evolved in vertebrates from gene duplication events (Hultqvist et al., 2012). Analysis of amino acid sequence in different vertebrate species (Dereeper et al., 2008) indicates that PALMD is highly conserved, especially among the mammalian clades (Figure S5D). Hence, considering its localization in the cytosol and its evolutionary conservation, we hypothesized that PALMD has important regulatory functions, which may be shaped by its interactome. We examined the protein interactome of PALMD by using liquid chromatography-tandem mass spectrometry (LC-MS/MS). A vector encoding for hemagglutinin (HA)-PALMD was transfected in HEK293T and immunoprecipitated for subsequent LC-MS/MS assay. This analysis showed that PALMD may interact with up to 595 proteins (Tables S10 and S11). The protein interactome was enriched in GO for cell adhesion, structural molecule activity, and actin-dependent ATPase activity (Figure 7A and Table S12). Among the different proteins, actin was among the strongest hits. This interaction was verified in co-immunoprecipitation experiments. In VICs, the pull down of native PALMD recovered α-ACTIN confirming data obtained in mass spectrometry (Figure 7B). Conversely, the immunoprecipitation of native α-ACTIN in VICs also recovered PALMD (Figure 7C). In situ proximity ligation assay was also performed to spatially locate the intracellular interactions between PALMD and α-ACTIN. In isolated VICs, we found that PALMD interacted with α-ACTIN in the cytosol (Figure 7D). Next, we used a blot overlay assay to evaluate the interaction of PALMD with actin. F- and G-actin were immobilized on nitrocellulose membranes, which were incubated with cell extracts transfected with a vector encoding for HA-PALMD. The interactions were detected with an anti-HA antibody. Both G and F actin interacted with PALMD (Figure 7E). Having shown that PALMD is a protein that interacts with actin, we reasoned that it could regulate the polymerization process. We thus tested recombinant PALMD in actin polymerization assay with fluorescent pyrene. In this assay, we found that PALMD dose-dependently and negatively impacted the polymerization of actin (Figure 7F). PALMD decreased the area under the curve of the polymerization assay by up to 37% (Figure 7G). These data thus indicate that PALMD is an actin binding partner and negative regulator of the polymerization process.

Figure 7

PALMD interacts with actin and regulates its polymerization

(A) Gene ontology (GO) terms enriched for PALMD protein interactome.

(B and C) PALMD and α-ACTIN co-immunoprecipitation (n = 3).

(D) Proximity ligation assay for PALMD and α-ACTIN (n = 2).

(E) Blot overlay assay of F- and G-actin interaction with PALMD (n = 4) by using various quantities of recombinant actin.

(F and G) Actin polymerization in the presence of PALMD 1 μM (n = 4) or 2 μM (n = 3) and area under curve of actin polymerization profiles.

(H) The G allele at rs6702619 hinders NFATC2 binding to the enhancer, resulting in decreased PALMD transcription. Reduced PALMD level in VICs leads to increased actin polymerization, enhancing MRTF-A nuclear translocation. MRTF-A/SRF binding to DNA response element initiates VICs activation toward myofibroblast-like cells, which promotes CAVS development. Scale bars, 10 μm. Values are mean ± SEM, ∗p < 0.05. Statistical analyses, normal distribution testing: Shapiro-Wilk, (E and G) Wilcoxon-Mann-Whitney.

Discussion

In this work, mapping and fine-grained molecular analyses of the 1p21.2 CAVS risk locus revealed that the risk variant rs6702619 is located in an enhancer region within a core TFBS for NFATC2. Atomic resolution data indicate that the risk allele G modifies target recognition and DNA shape, which impair the binding of NFATC2, a TF with key regulatory function in the AV (Chang et al., 2004). Functional assessment showed that PALMD is a regulator of actin polymerization. In turn, genetically determined lower expression of PALMD promotes MRFT/SRF-mediated gene expression and activation of VICs toward myofibroblast-like cells, an early process involved in the development of CAVS (Figure 7H).

Recently, GWA studies identified 1p21.2 as a risk locus for CAVS (Helgadottir et al., 2018; Thériault et al., 2018). Herein, by using fine mapping and functional assessment we provide evidence that rs6702619 is causally associated with the expression of PALMD. The index variant rs6702619 and SNPs in LD reside in an enhancer and may thus play a critical role in the cell identity and fate of VICs. By using CRISPRa, we underscored that the risk locus is mainly involved in the regulation of PALMD, whereas other genes sharing the same TAD are not. These data are in line with chromatin contact mapping, which showed spatial coordination between the risk locus and PALMD. Growing evidence suggests that noncoding variants may affect protein-DNA interactions by altering local atomic structure (Rohs et al., 2009). Analysis of atomic-level data revealed that the risk variant alters significantly both base-readout and DNA shape, which could affect the binding of NFATC2. Our data showed in DNA-binding ELISA assay that the risk variant significantly decreased the occupancy of NFATC2, suggesting a functional implication for this TF. To this effect, in VICs, the knockdown of NFATC2 significantly lowered the expression of PALMD. NFATC2 is a key regulator of heart valve biology and morphogenesis (Chang et al., 2004); Nfatc2 deficiency in mice results in lethal defects due to developmental abnormalities of cardiac valves and septa (de la Pompa et al., 1998; Ranger et al., 1998). Hence, disruption of TFBS and lower NFATC2 occupancy at 1p21.2 is a key process in lowering the expression of PALMD.

Previous work conducted in myocytes underlined that PALMD regulates cell shape (Nie et al., 2017). However, the molecular process whereby PALMD affects cell shape and dynamics was largely unknown until now. In isolated human primary VICs, we found that lower level of PALMD led to increased cell area. Moreover, we found that this cell modification was coordinated with higher level of FAs and expression of ACTA2/α-ACTIN, a marker of myofibroblasts. Data have consistently shown that transition of VICs into myofibroblast-like cells was an early process in the development of CAVS (Latif et al., 2015; Schlotter et al., 2018; Taylor et al., 2000). It is worth pointing out that myofibroblastic transformation of VICs following a knockdown of PALMD was accompanied by higher expression of genes involved in matrix remodeling such as TGFB2, MMP9, MMP2, and COL1A2. Also, lower level of PALMD in VICs was conducive to higher production of collagen type I, a major component of the ECM in AV and contributor to the remodeling process during CAVS. These data are in agreement with previous work showing that increased synthesis of collagen and production of ECM are among the key processes leading to valve thickening during CAVS (Hutson et al., 2016).

In the present work, LC-MS/MS data and immunoprecipitation in human VICs have identified PALMD as a protein that interacts with actin. Actin is highly conserved and interacts with a large number of proteins (Dominguez and Holmes, 2011). Numerous proteins thus fine-tune the complex process of actin polymerization as it is involved in a myriad of cellular functions (Davidson and Wood, 2016). Noteworthy, the present findings indicate that PALMD is a direct regulator of actin polymerization. In isolated cells, proximity ligation assay showed cytosolic interaction of PALMD with actin. In vitro, PALMD is a potent negative regulator of actin polymerization. In isolated cells, the knockdown of PALMD promoted the formation of F-actin. Hence, lower expression of PALMD, which dynamically and negatively regulates the polymerization of actin, promotes cell activation and a fibrotic gene program that relies on SRF-MRTF-A. To this effect, we observed an increased and significant nuclear translocation of MRTF-A in cells with a knockdown of PALMD. Consistently, SRF-dependent genes such as TGFB2, MMP9, and COL1A2 were also overexpressed in cells with a knockdown of PALMD (Gilles et al., 2009; Haak et al., 2014; Kuzniewska et al., 2013; Li et al., 2012).

This study has several important clinical implications as previous work suggested, by using Mendelian randomization, a causal role for a lower expression of PALMD in CAVS (Thériault et al., 2018). The identification of PALMD as a new regulator of actin dynamics and polymerization in VICs provides substantive evidence for a major function of PALMD in the biology of the AV. In addition to CAVS, a GWA study also found at a genome-wide significance level an association between rs6702619 and aortic root diameter (Wild et al., 2017). This association has a directional effect that is concordant with CAVS risk. Hence, the aortic root and the AV, which share embryologic and developmental pathways, are under the regulation of this locus. Of note, the 1p21.2 locus is not, however, associated with coronary artery disease risk (van der Harst and Verweij, 2018). Clinically, the present findings may foster the development of therapies, which could be based on the biology of PALMD and its signaling pathway.

Noncoding risk variant at 1p21.2 modifies a TFBS for NFATC2. In turn, the activity of a distant-acting regulatory element residing in an enhancer is substantially impacted and it lowers the expression of PALMD. Genetically determined lower level of PALMD promotes the activation of VICs and fibrogenesis, an early pathogenic process in the development of CAVS. Hence, comprehensive assessment of the 1p21.2 risk locus provides a detailed map for the functional and molecular genetics in CAVS. These findings pave the way for further research, which may lead to therapies.

Limitations of the study

The impact of CRISPRa on the expression of PALMD was limited in human VICs as these primary cells were largely resistant to selection exhibiting a high level of mortality. The expression was determined on bulk non-selected transfected cells. It is thus expected that the effect of CRISPRa on the expression would be much higher if cells could have been selected. The process whereby PALMD regulates the polymerization of actin and activates an MRTF-SRF pathway was determined in isolated cells. Further work in vivo should be performed to investigate the contribution of PALMD to the fibrogenesis of the AV.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Patrick Mathieu (patrick.mathieu@fmed.ulaval.ca).

Materials availability

Plasmids generated in this study are available on request.

Data and code availability

Raw data for ChIP-Seq, ATAC-seq, and HiChIP are available in the NCBI's Gene Expression Omnibus (GEO) repository (GSE154513).

Methods

All methods can be found in the accompanying Transparent methods supplemental file.