Procollagen IIA mediates positive feedback control of the mouse cardiogenic transcriptional network

Significance

We found that during formation of the mammalian heart tube, the activity of cardiac-enriched transcription factors including Nkx2-5 is self-enhanced via a positive feedback mechanism involving the extracellular matrix (ECM) protein procollagen IIA, which modulates the signaling activity of bone morphogenetic proteins (BMP) and transforming growth factors-beta (TGFβ) proteins. BMP/TGFβ-SMAD4 signaling is diminished in mouse embryos lacking procollagen IIA, which impacts the specification of cardiac progenitors resulting in heart defects. This study highlights the role of the ECM, especially procollagen IIA as a mediator of transcriptional activity in the formation of the heart.

Abstract

Cardiogenesis relies on the integrated interplay between cardiac transcription factors and signaling pathways. Here, we uncover a role for type IIA procollagen (IIA), an extracellular matrix (ECM) protein encoded by an alternatively spliced Col2a1 transcript, encoding a N-terminal cysteine-rich domain, as a critical regulator in a cardiac gene regulatory feedback loop. The cysteine-rich domain of IIA protein was previously reported to interact with bone morphogenetic proteins (BMPs) and transforming growth factors-beta (TGFβ) in in vitro binding assays and acts as a BMP antagonist in amphibian embryo assays. We show that the Col2a1 gene in mice is activated in the developing heart by core cardiogenic factors (NKX2-5, GATA4, MEF2, and SRF) via cis-regulatory enhancer elements. IIA loss (ΔIIA) in mice results in depletion of Isl1- and Nkx2-5-expressing progenitors, causing outflow tract defects resembling disrupted BMP/TGFβ-SMAD signaling, alongside reduced nuclear pSMAD1/5/8 in cardiac tissues. Compound +/ΔIIA; Smad4+/− mutants exhibit aggravated malformations. IIA enhances BMP-responsive reporter activity in cells in transactivation assays. We propose that IIA supports a positive functional role on SMAD4-dependent signaling, fine-tuning BMP/TGFβ signaling, thereby regulating GATA4 and NKX2-5 activity during second heart field progenitor specification. These findings position IIA procollagen as a key ECM component that integrates BMP/TGFβ signaling with cardiac transcription factors such as NKX2-5, revealing a feedback loop essential for cardiogenesis. Given its role in cardiac development, IIA emerges as a potential congenital heart disease risk factor.

The mammalian heart is constituted within the cardiogenic mesoderm from the first heart field (FHF) and the pharyngeal and splanchnic mesoderm of the second heart field (SHF) (1). The cooperative action of cardiac transcription factors such as NK2 Homeobox 5 (NKX2-5), GATA binding protein 4 (GATA4), serum response factor (SRF), Myocyte Enhancer Factor 2C (MEF2C), and ISL LIM homeobox 1 (ISL1) is required to drive cardiogenesis and morphogenesis of the heart and the outflow tract (2–6). These transcription factors interact with cardiac specific enhancers of their target genes and act in an autoregulatory feedback network to regulate and stabilize the cardiac program (7–14). During mouse cardiogenesis, bone morphogenetic proteins (BMP) signaling acts in concert with SMAD Family member 1 (SMAD1), GATA4, and chromodomain helicase DNA binding protein 7 (CHD7) to activate the transcription of Nkx2-5 (7, 12, 15). NKX2-5 in turn downregulates the transcription of Bmp2 and SMAD1-mediated BMP signaling, creating a negative feedback loop essential to fine-tune cardiomyocyte proliferation and outflow tract development (16). Parallel to the direct regulation of the transcription of signaling factors, the connectivity of BMP signaling with the transcriptional network may be modulated by the availability of BMP to elicit the signaling cascade. Proteins containing evolutionarily conserved cysteine-rich (CR) domains display context- and protein partner-dependent modes of BMP regulation that can either promote or repress BMP signaling (17–19). Null mutants of the long form of latent transforming growth factors-beta (TGFβ) binding proteins (LTBP), which contain conserved cysteine-rich domains, display structural defects of outflow tract and valve (20, 21). In this case, the LTBP putatively acts as a sink to control the bioavailability of TGFβ proteins in the extracellular matrix (ECM). There may also be other feedback mechanisms controlling the bioavailability of BMP/TGFβ ligands that act upstream in regulating the activity of cardiac-specific transcription factors during murine cardiogenesis.

The Col2a1 gene encodes two procollagen II isoforms, IIA and IIB, as a result of alternative splicing of exon 2 (22). The isoform IIA is transiently expressed during early human and mouse embryogenesis in nonchondrogenic tissues including the heart (23–25) (Fig. 1A). Loss of IIA function in anterior mesendoderm in mice is associated with defective closure of the cephalic neural tube and holoprosencephaly (26). Like other CR-domain containing proteins that modulate the availability of BMP/TGFβ factors (18, 19, 27), IIA procollagen contains a conserved CR domain in its amino-terminal propeptide (28, 29). The IIA isoform has been implicated in regulating negatively the activity of BMP ligands in in vitro binding assays and Xenopus embryo studies (28). However, it remains unclear whether IIA, like other CR-domain proteins, could potentiate BMP signaling in vivo. These findings, and the early expression of IIA during the specification phase of heart development, raises the question about whether IIA may contribute to the cardiac transcription factor gene regulatory network involving NKX2-5 and other BMP-dependent transcription factors.

Fig. 1.

Expression of IIA procollagen and transcription factors Isl1 and Nkx2-5. (A) Overlapping domain of expression of IIA, Isl1, and Nkx2-5 in the cardiac crescent of 1 to 2 somite stage embryo (section-i to vi: sections of the foregut portal). Isl1 and Nkx2-5 expressing cardiac progenitors are localized to medial position (arrows) in the SHF while Nkx2-5 expressing progenitors are localized laterally in the FHF progenitors. (B) A 5 to 6 somite stage (E8.25) embryo showing localization of IIA protein in the dorsal mesocardium (bold arrows) juxtaposing the Isl1- and Nkx2-5-expressing mesenchyme (arrows). (C) In E9.0 embryo, IIA protein is localized in Isl1-expressing tissues of the outflow tract (arrows) and Nkx2-5-positive tissue in the heart tube. Bar = 0.1 mm.

To elucidate the role of IIA in the feedback control of the cardiogenic program, we first conducted a genome-wide bioinformatic analysis of cardiac-specific cis-regulatory element signatures in gene loci across the mouse genome and identified Col2a1 as a gene targeted by cardiac transcription factors during heart development (30). By transgenic analysis, we confirmed the presence of an evolutionarily conserved upstream cis-regulatory element in mouse Col2a1 encompassing a putative enhancer bound by NKX2-5 and GATA4. Phenotypic analysis of the ΔIIA/ΔIIA mouse mutant revealed that loss of IIA function is associated with heart defects accompanied by reduced Mothers against decapentaplegic (SMAD) activity and down-regulation of Nkx2-5 expression during early cardiogenesis. We propose that IIA procollagen, itself regulated by Nkx2-5 and GATA4 activity, is key to mediating positive feedback control of Nkx2-5 expression via enhancement of BMP-SMAD4 activity during cardiac cell differentiation and heart morphogenesis.

Results

Expression of IIA Transcript and Protein in the First and Second Heart Fields (FHF and SHF).

The alternatively spliced transcript Col2a1 IIA has been shown to be expressed in the embryonic heart of humans (24) and mice (23, 25). In the cardiac crescent of E8.0 (1 to 2 somite stage) mouse embryos, IIA expression overlapped with that of Nkx2-5 and Islet1 in cells of the FHF and SHF (16, 31) (Fig. 1 A and B). IIA messenger RNA (mRNA) was coexpressed with Nkx2-5 in FHF myocardium and with Nkx2-5 and Isl1 in the SHF mesoderm. In E8.5 (5 to 6 somite stage) embryos, IIA protein was detected in the ECM of the heart tube, the dorsal mesocardium and the dorsal pericardial (splanchnic) mesoderm, in which Isl1 and Nkx2-5 were expressed (Fig. 1C).

Identification of the Cardiac Enhancer in Col2a1.

Applying a cis-regulatory element-directed bioinformatic pipeline that searches for cardiac-specific enhancers and the upstream transcription factors that control gene expression in the embryonic heart (30), we identified in the region spanning ~chr15:97,838,600 to 97,840,200 (mm9) of the Col2a1 locus (Fig. 2A), a cardiac-specific cis-regulatory element that is enriched for histone marks H3K4me1 and H3K27ac in the heart, however, these were absent in brain and liver (Fig. 2 C–E). This heart-specific region (termed Block C) is located -5 kb to -3.5 kb region upstream of Col2a1 transcription start site (Fig. 2B). RNA-seq data of heart tissues showed Col2a1 expression (>20 fragments per kilobase of transcript per million mapped reads, confirming the significance of Col2a1 expression in heart tissues (SI Appendix, Fig. S1). This Block C element also has enrichments for NKX2-5 binding in E9.5 embryos (Fig. 2F) (32) and a core 176 bp element is highly conserved among vertebrates (Figs. 2H and 3 A and B). Within a broader 5′ flanking region of the Col2a1 gene, we identified additional evolutionarily conserved blocks of sequences upstream of Block C (-9 kb and -15.5 kb) (Fig. 3C). Block C, which overlaps with a H3K4me1- and H3K27ac-marked predicted enhancer region (Fig. 2C), is anticipated to be a cardiac-specific enhancer due to harboring putative binding sites for NKX2-5, GATA, MEF2, and SRF (Fig. 3 B and C). Notably, the NKX2-5 site is positioned adjacent to one of two predicted GATA factor binding sites and a MEF2 binding site (position -4205 to -4199 bp, Fig. 3B).

Fig. 2.

Regulatory landscape in the Col2a1 locus. (A) Part of the Col2a1 locus, including the 5′ upstream genomic region, its 5′ untranslated region, the first exon and the first intron, is oriented from 3′ to 5′ (left to right). (B) The predicted heart-specific enhancer identified from our pipeline (30) is highlighted in blue (upper track), along with the location of the Block C highlighted in purple. (C–E) Tracks displaying H3K4me1, H3K27ac, and H3K4me3 chromatin marks across four tissues (heart, brain, liver, and limb) (33). The peaks for H3K4me1 and H3K27ac are located slightly upstream of the Block C element. This may relate to the mapping resolution of the sequencing data, the stage of analysis in heart and other tissues and tissue complexity. (F) NKX2-5 ChIP-seq duplicates (GSM3711204 and GSM3711025) from Wang and others (32). (G) assay for transposase-accessible chromatin using sequencing peaks from four tissues (heart, forebrain, liver, and limb) at E14.5. (H) Vertebrate conservation track generated by PhastCons.

Fig. 3.

Conserved noncoding sequences and a potential cardiac enhancer at the Col2a1 locus. (A) Sequence traversal panel and similarity plots showing alignment of mouse Col2a1 sequence with that of the rat, human, dog, cow, and opossum by VISTA (34). The X-axis represents the nucleotide position in the mouse sequence, and the Y-axis represents the degree of similarity. A window length of 100 bp is set for calculation of percent identity with conservation level at 75%. Only ≥50% conservation level is shown in the plot. (B) Potential transcription factor binding sites for SRF, GATA4, MEF2, and NKX2-5 are present in Block C which is conserved across rat, human, dog, and opossum (A). (C) Col2a1–LacZ BAC reporter constructs (Left column) and their expression patterns (middle column) in E9.5 transgenic mouse embryos, with magnified images of the boxed regions lined with dashes in i to x (Right column). A LacZ gene with a polyadenylation signal (pA) was inserted within the exon 1 sequence of Col2a1. Abbreviations: N, NKX2-5; G, GATA4; M, MEF2; S, SRF; h, heart; nd, notochord; om, periotic mesenchyme. Bar: 0.5 mm.

A bacterial artificial chromosome (BAC) vector containing the Block C cardiac enhancer element and a LacZ reporter inserted into exon 1 of Col2a1 (bCol2) was used to generate a reporter construct for visualizing tissue-specific Col2a1 expression in transgenic embryos. The reporter recapitulated with high fidelity and efficiency endogenous Col2a1 expression in the periotic mesenchyme, notochord, and heart at E9.5, indicating that this BAC vector harbors the essential enhancer elements driving tissue-specific Col2a1 transcription (Fig. 3 C, i).

Attributes of NKX2-5, GATA4, SRF, and MEF2 Binding Sites in the Cardiac Enhancer for Col2a1 Expression in the Heart.

Disruption of the Block C enhancer element, achieved by inserting a neomycin resistance cassette (bCol2-neo) into the NKX2-5 binding site, resulted in the complete loss of LacZ expression in the heart possibly because of a general disruption of the cis-regulatory architecture (Fig. 3 C, ii and SI Appendix, Fig. S2). To evaluate the relative roles of NKX2-5, GATA, MEF2, and SRF binding sites, individual and combined mutations of each of these sites in the bCol2 vector were tested in transgenic embryos. Mutations of each site down-regulated the transgene expression to different degrees in the heart, but none completely abolished expression (Fig. 3 C, iii–vii and SI Appendix, Fig. S2). Combinatorial mutations of NKX2-5, GATA (site 1) and MEF2 binding motifs (bcol2-NGM) significantly reduced the level and frequency of transgene expression (Fig. 3 C, viii and ix and SI Appendix, Fig. S2). Additional mutation of the SRF site in the triple mutant context (bcol2-NGMS) abolished reporter expression in the heart, while expression at other sites was not affected (Fig. 3 C, x and SI Appendix, Fig. S2). Collectively, NKX2-5, GATA, SRF, and MEF2 drive cardiac expression of Col2a1, consistent with the activation of cardiac-specific genes by these factors (10).

NKX2-5 and GATA4 Regulate Col2a1 Expression.

To explore functionally whether NKX2-5 regulates Col2a1 expression, microarray, and quantitative polymerase chain reaction (qPCR) analyses were conducted on cells of Nkx2-5^GFP/GFP (=Nkx2-5-null) purified by fluorescence-activated cell sorting (FACS) for green fluorescent protein (GFP) expression (6). Col2a1 expression was down-regulated by 20 to 50% in Nkx2-5^GFP/GFP cells of 6 to 12 somite stage embryos, and further to 80 to 90% in those of 16 to 21 somite stage embryos (Fig. 4A). In situ hybridization showed that Col2a1 expression was reduced in Nkx2-5^GFP/GFP hearts of E9.5 embryo (n = 3) (Fig. 4B), while expression in other tissues was unaffected. Chromatin immunoprecipitation-PCR (ChIP-PCR) assays revealed that NKX2-5 binds preferentially to Block C (Fig. 4C), supporting that NKX2-5 can directly regulate Col2a1.

Fig. 4.

NKX2-5 and GATA4 directly regulate heart tube expression of Col2a1. (A) Microarray/qPCR analysis of heart tissues of 6 to 21 somite stage Nkx2-5^GFP/GFP embryos showing reduced Col2a1 expression. Level 1 = normalized level of Col2a1 expression in Nkx2-5^+/GFP embryos. (B) Complete lack of Col2a1 expression in E9.5 Nkx2-5^GFP/GFP heart which displayed a truncated outflow tract. (C) NKX2-5 binds to region C but not regions A and B of the Col2a1 upstream sequence in E11.5 heart. Control: γ-crystallin promoter. Significant difference between binding to Col2a1 A and C (**P = 0.0049) and between Col2a1 B and C (*P = 0.0146) by the unpaired t test. (D) Reduced Col2a1 expression in Gata4 conditional knockout; Nkx2-5 compound mutants. Microarray data confirmed by qPCR indicated a 50% down-regulation of Col2a1 transcript levels in Gata4^flox/flox; Nkx2-5^Cre/+ E9.5 hearts compared to Gata4^flox/+; Nkx2-5^+/+ hearts (**P < 0.001). An intermediate level was seen in Gata4^flox/+; Nkx2-5^Cre embryos (*P < 0.05), indicating that GATA4 regulates Col2a1 expression in a dose-dependent manner. (E) GATA4 ChIP. Fold enrichment was determined for the indicated Col2a1 genomic regions against the γ-Crystallin promoter sequence. **P < 0.01. Bar = 0.5 mm.

NKX2-5 and GATA4 are known to regulate expression of cardiac genes cooperatively (11, 14, 35). One of the two GATA factor-binding sites in Block C sits adjacent to an NKX2-5 site (-4217 to -4212). Col2a1 expression was affected by the loss of Gata4 in conditional mutant embryos, generated by Nkx2-5-Cre-mediated deletion of Gata4 in cardiac tissue (36). Microarray data confirmed by qRT-PCR indicated a 50% down-regulation of Col2a1 transcript levels in Gata4^flox/flox; Nkx2-5^Cre/+ hearts at E9.5 compared to Gata4^flox/+; Nkx2-5^+/+ hearts (P < 0.001) (Fig. 4D). An intermediate level was seen in Gata4^flox/+; Nkx2-5^Cre/+ embryos (P < 0.05), indicating that GATA4 may regulate Col2a1 expression in a dose-dependent manner. ChIP-PCR assays revealed that GATA4 preferentially bound to the Block C sequence (Fig. 4E), demonstrating that GATA4 may directly regulate Col2a1.

Loss of IIA Results in Developmental Cardiac Defects.

To investigate the role of IIA procollagen in heart development, we examined the cardiac phenotypes of ΔIIA/ΔIIA mouse mutant with knockout of the exon 2 sequence of Col2a1 gene. Perinatal ΔIIA/ΔIIA mice displayed features of severe congenital heart diseases (CHD) including atrioventricular septal defects and abnormal atrioventricular valves (30%), heavily trabeculated ventricular myocardium (80%), double-outlet right ventricle with subpulmonic ventricular septal defect (Taussig-Bing Syndrome) (45%), complete transposition of the great arteries (35%), and/or persistent truncus arteriosus (15%) (Fig. 5 A–D, n = 20, SI Appendix, Table S1), which are also reported in mouse hypomorphs or null mutants for Nkx2-5, Mef2c, and Gata4 (16, 36–41). The spectrum of cardiac phenotypes observed are consistent with malformations which arise as a consequence of SHF defects (42).

Fig. 5.

Newborn and embryonic heart defects of ΔIIA/ΔIIA mutants. (A) Wild-type heart showing the atrial septum (arrow) and the interventricular septum (ivs) between the left (lv) and right (rv) ventricles. (B) Transposition of the great arteries with the aorta (a) emerging from the right ventricle (arrow) and the patent ductus arteriosus (pda) connecting to the left descending aorta (da). (C) Ventricular septal defect (arrow) beneath the pulmonary artery (p) opening to both ventricles. (D) Atrial septal defect, common atrioventricular canal, and endocardial cushion defect (double arrowheads). (E and F) Right side (RHS) and frontal views of Col2a1^+/+ and ΔIIA/ΔIIA E9.5 embryos. A solid arrow points to the fluid-filled pericardial cavity. Curved arrows indicate the direction of blood flow from the ventricle to the outflow tract. (G and H) Absence of IIA protein and incomplete heart tube looping in the E10.5 ΔIIA/ΔIIA heart. (I and J) In situ hybridization of Isl1 in E8.0 and E8.5 wild-type (+/+) and ΔIIA/ΔIIA embryos. (K and L) Nkx2-5 expression in the embryonic heart at (K) E8.0 and (L) E9.5. Arrows and asterisks indicate the SHF and FHF, respectively. (M and N) Col2a1 whole mount in situ hybridization in E8.5 and E9.5 wild-type (+/+) and IIA null (ΔIIA/ΔIIA) embryos. Bars: 100 μm (A–D, I–L, and N) or 50 μm (G, H, and M).

At E9.5, pericardial effusion and incomplete cardiac looping were found in 40.5% (n = 37) of ΔIIA/ΔIIA embryos (Fig. 5 E and F). By E10.5, IIA protein expression was undetectable in the incompletely looped ΔIIA/ΔIIA heart tube (Fig. 5 G and H). Outflow tract-related phenotypes and incomplete cardiac looping were accompanied by reduced expression of Isl1, a SHF marker (31), in the cardiac crescent and the looping heart at E8.0 and E8.5 (83.3%, n = 12; Fig. 5 I and J). Expression of Nkx2-5 was significantly reduced in the cardiac progenitors of the forming heart tube and in the SHF progenitors within the dorsal mesocardium at E8.0 (n = 2; Fig. 5K). At E9.5, Nkx2-5 expression was also dramatically reduced in the SHF derived outflow tract of ΔIIA/ΔIIA embryo (80%, n = 5; Fig. 5L). These findings suggest that IIA, whose transcription is regulated by NKX2-5 activity, controls Nkx2-5 expression. Overall Col2a1 transcription was down-regulated in ΔIIA/ΔIIA heart (E8.0-9.5, 57.1%, n = 14; Fig. 5 M and N) but appears relatively unaffected in cartilages in E13.5 embryos where NKX2-5 is not expressed (SI Appendix, Fig. S3). Altogether, this is consistent with a potential positive feedback loop by IIA on early Nkx2-5 expression, and downstream transcription of cardiac specific target genes.

Reduced SMAD Signaling Activity in ΔIIA/ΔIIA Embryo.

In line with BMP/SMAD activity being required for Nkx2-5 expression and the specification of cardiac progenitors (7, 12), we investigated the expression of phosphorylated (p)SMAD1/5/8 as a readout for BMP signaling activity in the heart tissue. Loss of IIA was associated with reduced immunoreactive pSMAD1/5/8 in the progenitors in the SHF and the outflow tract of E8.5 to 9.5 ΔIIA/ΔIIA embryos (33.3%, n = 12; Fig. 6 A and B). pSMAD1/5/8 staining was completely absent in the interventricular septum and right ventricle of E13.5 ΔIIA/ΔIIA heart (Fig. 6 C and D). In the embryonic E9.5 heart, BMP2 and BMP4 are colocalized with IIA protein (Fig. 6E). However, BMP2 expression appeared to be unaffected in ΔIIA/ΔIIA embryos (Fig. 6F). A reduced level of immunoreactive pSMAD1/5/8 in ΔIIA/ΔIIA mutant embryo implicates that IIA may act to enhance or maintain BMP signaling during heart development. To test this hypothesis, we studied the synergistic impact of reduced co-SMAD activity and partial loss of IIA on heart development. Among the offspring from crosses of IIA and Smad4 mutant mice, compound heterozygotes (+/ΔIIA; Smad4^+/−) were present at a reduced Mendelian ratio at weaning (SI Appendix, Table S3). In the +/ΔIIA; Smad4^+/− E9.0 embryos, Isl1 expression was reduced in the splanchnic mesoderm and the outflow tract (n = 3; Fig. 6G), and these embryos presented abnormal to arrested looping of the heart tube and pericardial edema (Fig. 6H) more frequently than the wild-type controls (SI Appendix, Table S3). The phenotypic outcome is consistent with the notion that the disruption of IIA/SMAD4 activity in +/ΔIIA; Smad4^+/− compound heterozygotes leads to defective differentiation of the SHF cardiac progenitors during the specification and chamber/outflow tract patterning phases of heart tube development. To test further the activating effect of IIA on SMAD-mediated BMP signaling, we cotransfected mouse embryonic stem (ES) cells with the cDNAs encoding full-length IIA, full-length IIB, or the N-propeptide of IIA together with 3GC-Lux, a BMP-responsive reporter (43). We found that reporter activity was promoted specifically in the presence of full-length IIA procollagen cDNA but not with full-length IIB or the N-propeptide (Fig. 6I). As IIA procollagen is a trimer of pro-alpha chains that requires the C-terminal propeptide for assembly, this result suggests that bioactivity is specifically attributed to trimeric procollagen IIA. The absence of activity in the mixed IIA and IIB assay suggests that BMP reporter activity is elicited specifically by homotrimeric IIA and that heterotrimers are not functionally active.

Fig. 6.

Loss of IIA function impacts on SMAD activity in the mouse heart. (A and B) pSMAD1/5/8 expression visualized by immunofluorescence in the foregut epithelium (fg) and outflow tract of (A) E8.5 and (B) E9.5 ΔIIA/ΔIIA embryos. (C and D) pSMAD1/5/8 expression is absent in the intraventricular septum (IVS) (C) and right ventricle (RV) (D) of E13.5 ΔIIA/ΔIIA embryos. (E) Localization of IIA, BMP2, and BMP4 proteins in the heart tissue. In the E8.5 heart tube (HT), IIA protein, BMP2 and BMP4 are colocalized in the ECM of the endocardium (arrowheads). In the E9.5 outflow tract, these proteins are also colocalized in the ECM near the myocardium (arrowheads). (F) BMP2 protein immunostaining in E8.5 wild-type and ΔIIA/ΔIIA hearts. (G) Isl-1 expression in the outflow tract of wild-type and +/ΔIIA; Smad4^+/− embryos visualized by in situ hybridization. (H) Frontal view of E9.0 wild-type, +/ΔIIA, Smad4^+/−, and +/ΔIIA; Smad4^+/− embryos, illustrating the looping pattern of the heart tube from the right ventricle to the outflow tract (curved arrows). Note the distended pericardial cavity and incomplete rightward looping of the +/ΔIIA; Smad4^+/− heart tube (asterisk). (I) In vitro assay demonstrating the effect of IIA procollagen (but not IIB procollagen or IIA N-propeptide) on activating BMP signaling in mouse ES cells. (J) A proposed positive feedback loop mediated by Col2a1 and IIA procollagen. We hypothesize that the cysteine-rich domain of IIA protein binds to BMP, enhancing BMP’s bioavailability to BMP receptors. This interaction maintains or increases BMP signaling transduction and the downstream SMAD-dependent activities that promote Nkx2-5 expression in the cardiac mesoderm. NKX2-5 collaborates with GATA4, MEF2C, and SRF to regulate Col2a1 expression through direct interaction with the cardiac-specific enhancer. SM: splanchnic mesoderm; TSS: transcription start site. ***p value < 0.0001. Bars: 50 μm (A–G); 100 μm (H).

Discussion

Cardiac transcription factors are known to work cooperatively in a gene regulatory network (2) to control various aspects of cardiogenesis including progenitor specification and outflow tract morphogenesis. Among these transcription factors, NKX2-5 plays a key role in the induction and maintenance of the SHF cardiac mesoderm, and its differentiation into cardiomyocytes (8, 16). NKX2-5 activity is regulated by both positive and negative feedback mechanism. NKX2-5 autoregulates its own expression by directly binding to an upstream enhancer in the avian or indirectly via MEF2C binding to the same enhancer in mammals (8). NKX2-5 also negatively regulates its own expression via repression of Bmp2 transcription and SMAD1 activity (16). Disruption of these feedback loops causes CHD including malformations of the outflow tract and arterial pole of the heart (44, 45).

Our study identified a cardiac cis-regulatory element located at -5 kb to -3.5 kb upstream of the mouse Col2a1 gene, termed Block C enhancer element. This enhancer element was independently identified via genome wide interrogation of H3K4me1 and H3K27ac marks in the mouse (30). This element, which is conserved in the human genome, is predicted in silico as an enhancer in cardiac tissues (46). GATA4 and NKX2-5 coactivate the promoters of cardiac genes including Actc1 (cardiac alpha-actin), Nppa (natriuretic peptide type A), and Nppb (natriuretic peptide type B) (14, 35, 47). Kinnunen et al. (48) additionally identified conserved residues in mouse GATA4 protein that are responsible for NKX2-5 interaction and synergism with NKX2-5 to promote activation of cardiac gene promoters. Consistent with activation of Col2a1 expression by the actions of the NKX2-5 and GATA4 and other cardiac transcription factors, our genetic and biochemical assays revealed the direct binding of these two cardiac transcription factors to the cardiac-specific enhancer element.

In this study, IIA procollagen mRNA and protein expression has been detected in the Nkx2-5⁺Isl1⁺ cardiac mesoderm. In IIA loss-of-function mutants, molecular and morphological defects were evident at the heart crescent stage (E8.0) and early somite stage (E8.5), respectively, thus placing IIA procollagen as an ECM protein that functions at early stages of mammalian cardiogenesis. We have shown here, in the cardiac mesoderm, that NKX2-5 and other transcription factors including GATA4 and MEF2C cooperatively regulate IIA expression via direct interaction with conserved cardiac-specific enhancers within the Block C cis-regulatory element of Col2a1 and that IIA protein colocalizes with BMP2/4 proteins in the ECM of the murine heart tube. We hypothesize that IIA protein in the ECM modulates BMP signaling and transduction via SMADs, which in turn drive Nkx2-5 and Isl1 expression in the SHF. These findings revealed an IIA-dependent positive feedback loop that connects an ECM protein to the transduction of BMP signaling, linking it to the NKX2-5-GATA4 transcription network responsible for cardiac differentiation and morphogenesis (Fig. 6J). The genetic interaction between IIA and Smad4, as reflected by the exacerbated cardiac phenotype in compound ΔIIA/Smad4 mutants, further supports this possibility.

Overall, the results of mouse genetics and cell-based studies show a role of IIA in enabling BMP signaling in vivo. IIA procollagen, in modulating BMP activity in the cardiac mesoderm via its cysteine rich domain, may facilitate the bioavailability of the ligand to the receptor complex to elicit BMP signaling activity and downstream SMAD transduction activity (17, 29), which in turn feeds forward to the gene regulatory network to modulate the activity of the cardiac-specific transcription factors. Extracellular titration of dosage of signaling factors such as BMPs are essential for maintaining threshold activity for cardiac induction and differentiation (49). IIA procollagen, expected to assemble into fibrils of triple helices containing the intact N-propeptide (29, 50), may contribute, in concert with other ECM proteins, to establishing or maintaining the threshold and bioavailability of BMP signaling for orchestrating cardiogenesis.

Previous in vitro and frog model studies have demonstrated a potential role of IIA in constraining the bioavailability of BMP ligands (28, 29). Ectopic overexpression of IIA procollagen in Xenopus embryos induced a secondary axis, suggesting an antagonistic function of IIA procollagen on BMP signaling (28). In contrast, our study of the mouse embryo reveals a link between IIA procollagen loss and impaired BMP signaling in cardiac development, suggesting a positive role in regulating BMP signaling in vivo. These contrasting results may reflect species differences as well as context and concentration-dependent effects of overexpression assays. Many BMP binding proteins such as Glypicans and Collagen IV and CR-containing ECM proteins such as CRIM1 and CV2 (BMPER) can either promote or inhibit BMP signal in a context-dependent manner (51). IIA could engage in the following modes of action that have been exploited by other BMP-binding proteins to regulate BMP signaling: a) regulating the processing of BMP precursors prior to their activation and secretion from cells; b) regulating the diffusion or sequestration of preprocessed or active BMPs in the ECM; c) regulating the formation of complex with or dissociation of BMPs from other BMP-binding proteins, and d) as a coreceptor regulating the presentation of the BMPs to their receptors on the cell surface (51, 52). In the last context, the integrity of trimeric procollagen IIA was important to IIA’s BMP regulatory function. Results of our cell reporter assay showed that only full-length IIA procollagen promoted BMP signaling with exogenous mature BMP ligands. The CR domains of IIA procollagen are retained and localized on the surface of type II collagen fibrils (29). These CR domains protruding from the collagen fibril surface may interact with free BMPs in the extracellular space and act as a sink for BMP ligands, potentially increasing the local concentration of BMP ligands to attain the threshold for receptor activation. It is possible that IIA can physically act at a level between BMP ligands and their receptors at cell surface (28). Type II collagen interacts with α1β1 and α2β1 integrins. The interaction of IIA procollagen with these integrins may enhance the response of cultured cells or mouse cardiac mesoderm to BMP signaling, analogous to that previously observed for type I collagen in enhancing BMP signaling in osteoblast via integrins (53). Potential cooperative binding at the cell surface mediated by multiple receptor types including BMP receptors and integrins may exist between the collagen fibrils containing IIA N-propeptides and the cells receiving BMP signaling. Furthermore, in our assays, mouse IIA procollagen in homotrimeric form, but not IIB procollagen, IIA/IIB heterotrimer, or the IIA N-propeptide alone, activated BMP signaling. This result underscores that the bioactivity of IIA depends on its intact trimeric procollagen form and is specific to IIA procollagen. Questions to be addressed in future studies are whether, in vivo, variables such as the context of the tissues involved and the processing enzymes that act on IIA [e.g., MMPs (54)] or the procollagen-integrin and BMP ligand receptor complexes, are modifying determinants in the mechanisms of IIA-dependent control of BMP signaling. For example, does IIA by binding BMP form a ternary complex with BMP receptor and facilitate or influence the establishment of a BMP gradient?

In Drosophila, TSG (twisted gastrulation) can displace Sog (short gastrulation)/Chordin from collagen (51). Identifying other ECM regulators that synergize with, enhance, or antagonize IIA bioactivity is also important. Recent genetic screens in Drosophila embryos have identified a number of ECM proteins as players in cardiogenesis, including a number of collagen orthologs such as Multiplexin (Collagen XV/XVIII), Viking (Collagen IV), and Pericardin (Collagen IV) (55). Multiplexin is known to interact with the Slit/Robo signaling pathway but the roles of Multiplexin and other orthologs in controlling the invertebrate cardiac transcriptional program are unclear. In mammals, collagens including types I, III, V, XI, XIV, and XV were found to be expressed in the developing and adult mouse heart (56–58). That IIA and not IIB as the major isoform expressed early in heart development (23–25), is consistent with IIA as a regulator at key stages of cardiogenesis. Studies of various single and compound mouse mutants pointed to diverse roles played by collagens later during development or cardiac repair. Specifically, collagens played a role in the maintenance of the structure and organization of ECM fibril network in cardiac tissues such as valves and ventricular myocardium, the hemodynamic requirements of the cardiac cycle, the resistance of the heart to cardiac failure under stress conditions or the proper reconstitution of cardiac tissues (57–60). Other studies have reported functional roles of noncollagen ECM proteins, such as the proteoglycans HSPG2 (or Perlecan) and VCAN (or Versican), participating in signal transduction, tissue morphogenesis, and the control of migration of specific cardiac cell types (61, 62). Unlike IIA procollagen, these ECM proteins act late in the development of the cardiac mesoderm. For instance, VCAN removal was found to suppress phosphorylation of SMAD2 during the remodeling of endocardial cushions (62). Mouse embryos deficient in HSPG2 had abnormal accumulation of mesenchyme in the outflow tract, potentially due to a premature and uncontrolled migration of neural crest cells into the looped heart tube (61). Interestingly, disruption of HAS2, encoding the synthase for the glycosaminoglycan hyaluronic acid, rich in IIA procollagen containing cartilage tissues, resulted in incomplete heart looping and trabeculation defects, as well as defective anterior/head development (63), reminiscent of the phenotype of ΔIIA/ΔIIA mutants.

ΔIIA/ΔIIA mutants displayed incomplete penetrance of cardiac phenotypes. This could be due to multiple factors: a) genetic background—C57BL/6 mice, the breed used in this study, are more proned to severe cardiac defects (64) and sensitive to changes in BMP signaling (65); b) fibrillar collagens like COL1A1 and COL3A1 also contain IIA-like N-propeptides (22); hence, the presence of these other fibrillar collagens during outflow tract development (66) might partially compensate for loss of IIA procollagen; and c) partial penetrance may be associated with transcription factors functioning close to critical thresholds. The myriad of cardiac defects spanning AV canal, cardiac chambers and outflow tract seen in ΔIIA/ΔIIA mutants are also consistent with cardiac transcription factors functioning in aspects of heart development below or close to critical thresholds (67). As in the case for IIA mutants, incomplete penetrance has been reported for genetic mutants of BMP antagonists, chordin and noggin (68, 69).

In sum, we have identified Col2a1 as both a transcriptional target and a regulator of the cardiogenic molecular program, and a candidate causative factor of congenital heart disease. Procollagen IIA, present in the ECM of cardiac tissue, acts to maintain cardiac progenitor programs and modulate signaling responses during cardiac differentiation and morphogenesis via positive feedback to the cardiac gene regulatory network. Of interest, apart from Block C, strong NKX2-5 enrichments could be identified at E9.5 in the coding and intronic sequences of Col2a1 [Fig. 2F, (32)]. Also, in the reference map of cardiac transcription factor chromatin occupancy, multiple overlapping binding sites could be identified around the Col2a1 locus at E12.5 for key cardiac transcription factors GATA4, MEF2C, SRF, TBX5, and NKX2-5 (70) suggestive of the presence of multiple cardiac regulatory elements driving Col2a1 expression. Additionally, human COL2A1 expression is dysregulated in cardiac progenitors in heart diseases associated with pathogenic variants of TBX5 and GATA4 (71, 72). Taken together with the fact that ΔIIA/ΔIIA mouse mutants display many features of severe congenital heart defects, our study provides insights into the role of the ECM in modulating signaling pathways in cardiogenesis and highlights that loss of IIA procollagen function may be a risk factor of CHD.

Materials and Methods

Mouse Mutants.

The mutant Col2a1 allele (Col2a1^+/ΔIIA-neo) was generated in the Col2a1 locus by homologous recombination in R1 ES cells (26). Two ES cell clones were used to generate heterozygous Col2a1^+/ΔIIA-neo mice on a mixed 129S/SvEv; C57BL/6 background, both exhibiting similar phenotypes. These heterozygous mice were crossed with the β-actin Cre recombinase deleter to excise the neo^r cassette, resulting in +/ΔIIA mice, which were then back crossed to C57BL/6 background. The heterozygote offspring were intercrossed to generate homozygous ΔIIA/ΔIIA mutants. Notably, significantly fewer ΔIIA/ΔIIA mutants were observed at birth in the heterozygous crosses (SI Appendix, Table S2). Genomic primers for ΔIIA allele are listed in SI Appendix, Table S4. To investigate the genetic interaction between IIA and Smad4 we performed crossings of +/ΔIIA mice (backcrossed to C57BL/6 background) with Smad4^+/− (73), as well as crossings of viable +/ΔIIA; Smad4^+/− mice with wild-type C57BL/6 mice (SI Appendix, Table S3). Offspring from these crosses were collected from E9.0 to newborn for gene expression and phenotypic analyses.

Immunohistochemistry, Immunofluorescence, In Situ Hybridization, and X-Gal Staining.

All embryos were fixed and processed using standard procedures. Immunohistochemistry was carried out on 4% paraformaldehyde fixed sections, using the Envision™+ System (DAKO) and Peroxidase (DAKO EnVision™+System, horse radish peroxidase) colorimetric detection and hyaluronidase for antigen retrieval. Staining signals were visualized using VECTASTAIN Elite ABC Kit (Goat IgG) (Vector Laboratories) and counterstained with hematoxylin. Immunofluorescence was carried on 4% paraformaldehyde fixed cryosectioned embryos, blocked with 10% goat serum and 1% Triton in phosphate-buffered solution. Images were scanned with a Bio-Rad confocal microscope. Anti-pSMAD1/5/8 (a gift of Peter ten Dijke, Leiden University, Leiden, Netherlands, 1:200 dilution), and anti-BMP2/4 (Santa Cruz, sc-6895 and sc-6896, 1:400 dilution) were used. Affinity purified procollagen IIA antibody directed against peptide PICPADLATASGRKL was generated by Covalab (Lyon, France) and used at 1:400 dilution. X-Gal staining was performed as described (74). Whole mount and section in situ hybridizations were carried out by following standard procedures using digoxigenin-labeled riboprobes specific for IIA(pSOK10) (25), IIA+IIB (pNJ61) (75), Nkx2-5 (16), and Isl1 (31). For in situ hybridization for Col2a1 on Nkx2-5 mutants, the probe was made from the NIH clone H3026G09 (https://www.ncbi.nlm.nih.gov/nuccore/BG078341.2). Anti-dig-alkaline phosphatase antibody and BM Purple (Roche) or NBT/BCIP colorimetric signal detection were used for whole mount and nonamplified sections, respectively. Expression analyses were performed on at least three embryos of the same genotypes.

RNA Analyses, Expression Microarrays, and Real-Time PCR.

Total RNA was isolated from E8.5 whole embryos and subjected to total RNA extraction by the RNeasy mini kit (Qiagen). 1 µg of each RNA sample was reverse transcribed using SuperScript™ First strand synthesis system (Invitrogen) and cDNAs amplified with forward and reverse primer pairs spanning exons 1 to 8 of the Col2a1 gene. Microarray expression data obtained from Nkx2-5-GFP+ cells collected by flow cytometry from embryos (5 to 12 somite stage) and dissected heart tubes (15 to 21 somite stage) from mutant mice carrying the Nkx2-5-GFP allele were analyzed as described in ref. 16. Quantitative reverse transcription polymerase chain reaction (RT-PCR) for Col2a1 was performed using the Rotorgene and Roche SYBR green PCR mix according to the manufacturer’s instructions (Rotor-Gene3000, Corbett Technologies) using Col2a1 primers. RNA was isolated from four groups of five E9.5 from Gata4^flox/+; Nkx2-5^+/+, Gata4^flox/+; Nkx2-5^Cre/+, and Gata4^flox/flox; Nkx2-5^Cre/+ littermate embryos. Amount of Col2a1 mRNA was measured by RT-PCR and by Affymetrix expression arrays as described (36). Sequences of all PCR primers used are listed in SI Appendix, Table S4.

ChIP.

ChIP was performed with anti-NKX2-5 (sc-8697, sc-4033) and anti-GATA4 (sc-25310, Santa Cruz Biotechnology) antibodies. Chromatin was prepared from mouse E11.5 hearts. Target fragments from the Col2a1 genomic regions A, B, and C and a region from the γ-crystallin promoter were tested for enrichment after immunoprecipitation with the corresponding antibodies compared to mock immunoprecipitation (no antibody control) by qPCR using SYBR green master mix (SI Appendix, Table S4). Data were obtained from at least three independent experiments, and the results are presented as the mean ±SD of three determinations (raw Ct (cycle threshold) readings and calculations included in SI Appendix, Table S5). The fold enrichment was calculated relative to the negative control (promoter region of γ-crystallin) using the delta Ct method. Primer sequences are provided in SI Appendix, Table S4.

BAC Transgenic Studies.

The Col2a1-LacZ reporter BAC (bCol2) was generated by inserting the β galactosidase gene and frt flanked kanamycin into exon 1 of Col2a1 by recombineering (76) using as template the RP23-61F08 BAC clone containing the Col2a1 gene with 74 kb 5′ and 33 kb 3′ flanking DNA. BAC recombineering was used to insert a frt flanked neomycin resistant gene into the NKX2-5 site (generating bCol2-neo) and to mutate the NKX2-5, MEF, GATA, and SRF sites in the cardiac enhancer (Block C) in bCol2 to generate the different vectors for transgenesis. Transgenic mice were produced by pronuclear injection into F1 CBA/C57BL6 oocytes which were transferred into pseudopregnant mice for complete development.

Bioinformatics Analyses.

Organ-specific promoters and enhancers were obtained via localization of H3K4me3 and H3K4me1/H3K27ac marks from ChIP-sequencing data (H3K4me1, H3K4me3, H3K27ac), obtained from the heart, liver, limb, and whole brain in mouse E14.5 embryos, downloaded from ENCODE (33). The data were processed as previously described (30), to produce heart-specific H3K4me1 and H3K4me3 cis-regulatory elements. Cis-regulatory elements were assigned to genes using GREAT (77) version 3.0.0, species assembly: mm9, association rule: basal+extension with default parameters (5,000 bp upstream, 1,000 bp downstream, 1,000,000 bp max extension). The GREAT-assigned gene lists (heart-specific H3K4me3 and H3K4me1) were overlapped to create a Venn diagram. Tracks were visualized using the UCSC Genome Browser (78).

Luciferase Reporter Assay.

IIA and IIB cDNAs including the endogenous 5′-UTR, signal peptide, and full-length coding sequences of IIA (NM_031163.4) and IIB (NM_001113515.3), were amplified from E12.5 mouse (CBA x C57BL/6) heart total RNA using RT-PCR. These cDNAs were subsequently subcloned into the HindIII and XbaI sites of the pRc/CMV expression vector to generate the CMV-IIA and CMV-IIB plasmids. 3GC-Lux, a BMP-responsive luciferase reporter driven by a GC-rich sequence in the Smad6 promoter (43), Renilla-luciferase plasmid (Promega #E2231), with or without pRc/CMV-IIA, pRc/CMV-IIB, or pRc/CMV-NpIIA expression vectors, were cotransfected using FuGene HD transfection reagent into L4 mouse ES cells (derived from 129 mouse blastocysts). ES cells were chosen for the assay because they do not express Col2a1 but possess the necessary molecular machinery for BMP signal transduction (79). Total amount of DNAs transfected per well were kept constant at 200 ng, except for the group 1× (IIA+IIB), 298 ng total DNAs were transfected. Transfected cells were lysed 48 h posttransfection. Luciferase activity was determined using a luminometer and normalized against Renilla luciferase readings.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.