HAND2-mediated epithelial maintenance and integrity in cardiac outflow tract morphogenesis

During embryogenesis, epithelial construction establishes the foundation for organogenesis, in particular, for establishing the tubular structure during the early stage of organ development (Kalluri and Weinberg, 2009; Nieto, 2013). The cardiac outflow tract (OFT) is a transient tubular structure that connects the main vasculature to cardiac ventricles. Morphogenesis of OFT is important for future septation of OFT into the trunks of aorta and pulmonary artery to establish the systemic and pulmonary circulation, which is required for the maintenance of physiology and survival (Harvey, 2002; Bruneau, 2008). In mice, the anterior second heart field (aSHF) progenitors from the splanchnic mesoderm (SpM) in the dorsal pericardial wall (DPW) migrate to the arterial poles of heart tube to form OFT and continue to facilitate its elongation and expansion for OFT morphogenesis (Black, 2007; Kelly et al., 2014). Perturbation of the OFT morphogenic process results in a spectrum of congenital heart defects affecting OFT septation and alignment, which severely impairs cardiovascular function and compromises embryonic development and viability (Srivastava and Olson, 2000; Srivastava, 2006; Fahed et al., 2013).

Recently, several groups have recognized the epithelial properties of aSHF cardiac progenitors in the anterior DPW (aDPW) and adjacent transition zone (TZ) connecting the proximal OFT (TZ as the most distal OFT, Fig. 1A,B) (Cortes et al., 2018). For instance, it was reported that the Wnt5a-Disheveled (Dvl) planar cell polarity (PCP) signaling regulates epithelial conversion of the SHF progenitors in the SpM for cell deployment (Sinha et al., 2012; Li et al., 2016). This research also suggested that the core PCP gene Vangl2 is involved in this process. A detailed and more refined analysis has since confirmed this possibility, demonstrating that Vangl2-regulated polarization is essential for epithelial establishment and cardiomyocyte differentiation in OFT morphogenesis; this also highlighted the relationship between cell polarity, the epithelial features of the distal OFT and differentiation of SHF progenitors to cardiomyocytes (Ramsbottom et al., 2014). Additionally, TBX1, a cardiac transcription factor required for OFT development, has been identified as a regulator of the epithelial properties of aSHF progenitors (Ramsbottom et al., 2014). These pioneering studies have defined a process of epithelial formation in OFT morphogenesis, whereby the cardiac progenitors with mesenchymal features in the aDPW constitute atypical apicobasally polarized epithelium that further develops into genuine epithelium in the adjacent, most distal, OFT, termed ‘TZ’, facilitating OFT morphogenesis.

Cell adhesive interactions, including adherens junctions (between cells) and cell-extracellular matrix (ECM) adhesion, are important elements of epithelial cells that help sustain tissue integrity in morphogenesis (Gumbiner, 2005; Takeichi, 2014). Adherens junctions are specialized structures that join together adjacent cells in the epithelium (Perez-Moreno et al., 2003). Two major adhesion molecules, E-cadherin and N-cadherin, are transmembrane proteins that bind epithelial cells together and, through their intracellular tails, connect to cytosolic α-, β- and γ-catenins (Thiery, 2003; Gumbiner, 2005). Furthermore, these catenins interact with the cytoskeleton of actin filaments and/or actomyosin bundles to achieve strong cell-cell adhesion (Parsons et al., 2010). By contrast, the epithelial cells also make contact with the ECM through membrane-localized integrins binding to matrix molecules, such as laminin and fibronectin (Hynes, 2002; Berrier and Yamada, 2007). Integrins further associate with the cytoskeleton of actin filaments and/or actomyosin bundles to result in stronger adhesion and/or help transmit signals from the extracellular environment to the cell. Therefore, actin filaments and/or actomyosin bundles have a crucial role in organizing cell adhesive interactions, the foundation for epithelial tissue formation (Parsons et al., 2010). Although these primary adhesive molecules have been characterized and described in the epithelium of aSHF progenitors in the DPW and TZ (Gumbiner, 2005; Takeichi, 2014), the regulators of these adhesion proteins remain unknown.

HAND2 is a key cardiac transcription factor essential for heart development, in particular for right ventricular (RV) formation (Srivastava et al., 1997; Yamagishi et al., 2000; Schindler et al., 2014; Vandusen et al., 2014). Although the roles of HAND2 in heart development and cardiomyocyte trans-differentiation have been extensively explored, the regulatory mechanisms exerted by HAND2 at molecular and cell biological levels in heart development remain less well understood. In zebrafish, Hand2 was found to be required for myocardial polarity in the early cardiac epithelium through negative modulation of fibronectin level, suggesting a role in tissue morphogenesis (Le and Stainier, 2004).

In this study, we investigated the regulation of OFT morphogenesis with an emphasis on the epithelial properties of the progenitors. We identified a gradient of HAND2 protein specifically in cardiac progenitors from aDPW and TZ epithelium. Deletion of Hand2 in these progenitors caused abnormal cell arrest and accumulation at the TZ, hindering OFT morphogenesis. Although cell polarity was unaffected, adherens junction and cell-matrix adhesion was aberrant in the differentiating progenitors in the TZ of Hand2 mutant mice. RNA-seq analysis also revealed disrupted regulation of the contractile fiber and actin cytoskeleton in Hand2 mutant mice. Furthermore, we showed that Stars (Abra) is transcriptionally controlled by HAND2. STARS facilitates actin polymerization that is essential for cell adhesion formation. Thus, here we report a new role for HAND2 in mediating epithelial maintenance and integrity through regulation of actomyosin bundle assembly to construct cell adhesive interactions, which are crucial for OFT morphogenesis.

OFT morphogenesis starts from embryonic day (E) 8.5-9.0 and then actively occurs between E9.5 and E10.5. During this period, OFT elongation and expansion occur through progressive addition of aSHF progenitors from the SpM to the distal OFT. To understand the regulation of OFT morphogenesis, we dissected distal OFT together with the aDPW, part of the SpM, from mouse embryos at E9.5 and E10.5 for gene expression profiling by RNA-seq analysis.

This analysis revealed the high level expression of several cardiac transcription factors, including Nkx2-5, Hand2, Gata4, Hand1, Srf and others (Fig. 1C). These results are consistent with previous studies from whole-mount in situ hybridization analyses (Prall et al., 2007; Ma et al., 2008; Vincentz et al., 2011). To better investigate the potential role of these genes in OFT morphogenesis, we generated specific monoclonal antibodies (in collaboration with Abcam), resulting in a HAND2-specific antibody worked successfully with immunofluorescence (IF) staining. Therefore, we performed a systematic study of HAND2 protein expression in early embryonic cardiovascular tissues.

At E8.0, HAND2 protein expression was found in both SpM and somatic mesoderm (SoM), where it largely overlapped with expression of Isl1 (Fig. 1D-K). At E8.5, high levels of HAND2 protein expression were observed in the aDPW, OFT and endocardium (Fig. 1L-O). At E9.5, high HAND2 protein expression was identified specifically in the aDPW, with more concentrated and intensive expression in the adjacent TZ, showing a gradient expression pattern (Fig. 1P-Q′). HAND2 accumulation peaked in the TZ at E9.5, after which its level declined, being then only found in a small number of cells in the RV. These results provide new information regarding the distribution of HAND2 in the cardiovascular tissues and suggest that HAND2 has an important role in aSHF progenitors in the aDPW and TZ for OFT morphogenesis.

We also observed HAND2 expression in cardiovascular tissues from E10.5 to E15.5 (Fig. S1).

To test the possible involvement of HAND2 in OFT morphogenesis, we specifically deleted Hand2 in aSHF progenitors using Mef2c-AHF-Cre-mediated ablation (Verzi et al., 2005). IF staining for HAND2 showed successful deletion of Hand2 in aSHF progenitors (Fig. 2A-D). Hand2 mutant mice (Hand2^f/^f; Mef2c-Cre) displayed substantially smaller OFT and RV compared with controls (Fig. 2E-F′). Given that this RV phenotype has been intensively investigated elsewhere (Tsuchihashi et al., 2011), we focused our study on OFT morphogenesis.

Intriguingly, histological analysis revealed multiple layers of cells in the distal OFT of Hand2 mutant mice, whereas, in the OFT of control mice, there were only one or two layers of cells (Fig. 2G-G′,I-I′). Detailed analysis showed up to four or five layers of cells in the distal OFT in Hand2 mutant embryos compared with one or two layers of cells in control embryos (Fig. 2H,J). This change in cell layers occurred in the distal OFT but not in the aDPW, although Hand2 was also deleted in the aDPW (Fig. S2A,B). Rosa26-mTmG reporter mice have cell membrane-localized green fluorescent protein (GFP) and red fluorescent protein (RFP) that can precisely outline the shape of cells. Therefore, we used these mice to study OFT wall thickness, with horizontal sections of the distal OFT further confirming the multiple layers of cells in Hand2 mutant mice (Fig. 2K,L). Quantification of the distal OFT wall thickness also demonstrated substantially increased OFT thickness in Hand2 mutant mice (Fig. 2M). We also used the Rosa26-mTmG reporter line to reveal differences in the morphology of OFT cells in the Hand2 mutant versus control mice (Fig. 2N-Q). Cell circularity analysis further revealed the profoundly reduced roundness of OFT cells in Hand2 mutant mice compared with controls (Fig. 2R).

Taken together, these results demonstrated defective OFT morphogenesis and cell accumulation in the distal OFT of Hand2 mutant mice.

To understand the cause of cell accumulation in the distal OFT of Hand2 mutant mice, we examined cell proliferation and survival. Previous work showed that enhanced thymoma viral proto-oncogene 1 (Akt) signaling in aSHF progenitors promotes cell proliferation in the caudal SpM, the increased progenitor pool of which can be visualized directly by using a GFP reporter mouse line at E9.5 (Luo et al., 2015). Thus, we used this strategy to examine the progenitor pool in Hand2 mutant and control mice at E9.5. The results showed little difference between the two groups at both E9.5 (Fig. S3A,B) and E10.5 (Fig. S3C,D). We then performed IF staining to study KI67⁺ (Mki67⁺) and Isl1⁺ cells at E9.5, and showed that the number of Isl1⁺ cells and the percentage of KI67⁺/Isl1⁺ cells in aDPW were comparable between Hand2 mutant and control mice (data not shown). Further IF staining for phosphohistone H3 (pHH3)⁺ at E9.5 obtained similar results (Fig. S3E-G). IF staining for pHH3⁺ cells in the OFT failed to detect a difference between the two groups (Fig. S3H-J). In addition, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining to examine cell apoptosis indicated no difference between Hand2 mutant and control mice (Fig. S3K-M).

Collectively, these studies demonstrated unchanged cell proliferation and survival in aSHF progenitors of Hand2 mutant mice, indicating that cell proliferation and apoptosis do not account for the defective OFT morphogenesis in these mice.

OFT morphogenesis relies on the addition (i.e. movement and/or migration) of aSHF progenitors to the distal OFT during E9.5-10.5 (Kelly et al., 2014). Therefore, we examined the movement and/or migration of aSHF progenitor cells into OFT. The primary characteristics of aSHF progenitors are robust cell proliferation but delayed cell differentiation, which contribute to the elongation and expansion of the OFT (Vincent and Buckingham, 2010). By IF staining for Isl1⁺ cells, we first examined the number of aSHF progenitors deployed in the OFT and showed that, whereas the total cell number was comparable between Hand2 mutant and control mice at both E9.5 and E10.5, the distribution pattern differed between the two groups. In the control mice, Isl1⁺ cells lined the whole OFT; by contrast, they cumulated in the TZ in Hand2 mutant mice (Fig. 3A-F).

We also performed 5-ethynyl-2’-deoxyuridine (EdU) pulse-chase assays to trace the movement and/or migration of aSHF progenitors towards the distal OFT. A large dose of EdU was administrated to embryos at E9.25, followed by thymidine administration 2 h later to compete against EdU incorporation (Fig. 3G). Continuous and similar numbers of EdU-labeled progenitors in the aDPW were found between the two groups at E9.25 (Fig. 3H-J). However, the assays at E10.0 revealed a markedly reduced distance of cell movement and/or migration in Hand2 mutant compared with control mice (Fig. 3K-M).

Taken together, these results indicate that the cell arrest of aSHF progenitors at the TZ is the primary cause of defective OFT morphogenesis in Hand2 mutant mice.

Recent studies have recognized the epithelial properties of aDPW and adjacent TZ, which primarily comprise aSHF cardiac progenitors (Francou et al., 2014; Ramsbottom et al., 2014). To understand the causes of cell arrest of aSHF progenitors at the distal OFT at the cell biology level, we investigated the epithelial features in both aDPW and TZ.

One epithelial property is cell polarity. IF staining for ZO-1 (tight junction protein 1; Tjp1) and atypical protein kinase C, zeta [aPKCζ (Prkcz)] showed comparable patterns in the aDPW and TZ between Hand2 mutant and control mice at E9.5, indicating normal cell polarity (Fig. 4A-D′). By contrast, when another epithelial feature, adherens junctions, was examined through IF staining for the two essential adhesive molecules of N-cadherin and E-cadherin, it demonstrated a severely disrupted pattern in the TZ of Hand2 mutant compared with control mice (Fig. 4E-H). Interestingly, the protein levels of both N-cadherin and E-cadherin were unaltered in OFT tissues via western blot analysis in Hand2 mutant mice compared with controls (Fig. S4A). We also studied α- and β-catenin by IF staining and failed to detect an obvious difference between the two groups (Fig. S4B-E″).

Next, we investigated another type of cell adhesion, cell-matrix adhesion, in these areas. The ECM protein fibronectin was normally distributed in Hand2 mutants (Fig. 5A-D). However, another ECM protein, laminin, showed both apical and basal distribution in TZ of Hand2 mutant embryos compared with a predominantly basal location in controls (Fig. 5E,F). Other important molecules involved in cell-matrix adhesion, namely integrin β1 (ITGB1), vinculiln and ILK, all displayed severely attenuated intensity, particularly in the lateral interfaces between cells of the TZ in Hand2 mutant mice compared with controls (Fig. 5G-L″).

In addition, the cytoskeleton of actin filaments and actomyosin bundles that associate with adhesive molecules in both adherens junction and cell-matrix adhesion, showed remarkably altered patterns in Hand2 mutant mice. Phalloidin staining for F-actin demonstrated obvious attenuated actin filaments in the tissue of TZ in Hand2 mutant mice (Fig. 6A-B″). Compared with controls, the signal in the basal side of the TZ was weaker and was almost completely lost in the lateral side in mutant mice. We also detected an increase in monomeric G-actin levels in the distal OFT of Hand2 mutant mice (Fig. 6C-G), suggesting weakened actin polymerization or excessive depolymerization. Consistent with the change in actin filaments, investigation of the patterns of associated nonmuscle myosin II (NMMIIB) together with its regulator, RMLC (active form of p-RMLC, p-Ser19), revealed nearly identical changes to those of actin filaments (Fig. 6H-K; Fig. S5A-F). Staining of Phalloidin and p-RMLC in the aDPW showed comparable patterns between control and Hand2 mutant mice (Fig. S5G-J).

Collectively, these data reveal severely disrupted cell adhesion in the TZ in Hand2 mutant mice: although HAND2 was not needed for cell polarity establishment, it was required for the formation of cell adhesions.

To understand the molecular regulation of cell adhesion in the progenitors of TZ, we collected tissue from aDPW and OFT at ∼E9.0-E9.5 for RNA-seq analysis. Heatmaps showed that, compared with controls, most genes were downregulated in Hand2 mutant mice, suggesting a role for HAND2 in activating gene transcription (Fig. 7A). Gene Ontology (GO) analysis revealed that the genes with significantly reduced expression level were within the contractile fiber and actin cytoskeleton catalogs (Fig. 7B). RT-quantitative (q)PCR analysis confirmed the changes of several genes (Fig. S6A). The RNA-seq data (Fig. 7A) revealed a substantial reduction in the expression of Stars (Akiko et al., 2002; Koichiro et al., 2005), a conserved actin-binding protein that facilitates actin polymerization. We verified this change by RT-qPCR (Fig. 7C). IF staining demonstrated a near-complete loss of STARS throughout the OFT of Hand2 mutant mice (Fig. 7D,E). Western blotting analysis also confirmed this change (Fig. S6B).

We also examined the levels of differentiation markers of cardiomyocytes in control and Hand2 mutant mice at E9.5. The expression levels of Nkx2.5, Mef2c and Gata4, together with those of their proteins, were comparable between the two groups (Fig. S6C-I). Next, we performed IF staining for cTNT (troponin T2, cardiac; Tnnt2) and α-actinin (key components of the sarcomere apparatus) to check cardiomyocyte maturation in the TZ. The results revealed profoundly reduced protein levels together with aberrant localization patterns in cardiomyocytes of Hand2 mutant mice (Fig. S6J-M) indicating cardiomyocyte differentiation defects. The expression levels of Tnnt2 and Actn2 (encoding α-actinin) were also significantly reduced in Hand2 mutant mice (Fig. S6A).

The above data suggested that HAND2 directly activates the transcription of Stars. Therefore, we explored the relationship between HAND2 and Stars. HAND2 is a member of the Twist family of basic helix-loop-helix (bHLH) transcription factors and binds the consensus elements termed E-boxes (CANNTG) or degenerate sequences of D-boxes (CGNNTG) as either homodimers or heterodimers, to activate the transcription of downstream genes (Firulli et al., 2003). Through bioinformatics analysis, three well-conserved E-boxes were identified located upstream of the mouse Stars transcription start site (TSS) (Fig. 7F). To test whether exogenous HAND2 impacts the Stars promoter, we cloned the 1608-bp and 400-bp fragments upstream of the Stars TSS to PGL3-basic plasmids for luciferase reporter assays. The results showed that the 400-bp fragment could be activated by HAND2 (Fig. 7G). Next, we generated mutation constructs to abolish the E-box sequences of E2 and E3, with both mutants showing strongly reduced reporter activity in response to HAND2, suggesting that these two E-boxes are responsible for transcriptional regulation by HAND2 (Fig. 7H,I). The combined double mutations of E2 and E3 showed a similar effect as the single mutation of E2, suggesting the more important role of E2 in this regulation (Fig. 7I).

Furthermore, we investigated the regulatory relationship between endogenous HAND2 and STARS in H9C2 rat cardiomyocyte-like cells. Knockdown of Hand2 significantly reduced STARS levels, as revealed by western blot analysis (Fig. 7J,K), and also profoundly reduced luciferase reporter activity (Fig. 7L). Finally, chromatin immunoprecipitation (ChIP) studies confirmed that HAND2 directly binds to Stars (Fig. 7M).

Taken together, these data demonstrated that HAND2 directly regulates Stars expression. STARS also controls an intracellular signaling cascade to stimulate the MRTF-SRF transcriptional pathway. The SRF protein level in the OFT was normal (Fig. S7A-C), and the MRTF-A pattern in the TZ was unchanged (Fig. S7D-E′) in Hand2 mutant mice.

Collectively, these results are consistent with the results of, and support the conclusions made by, Métais and colleagues in their study (Métais et al., 2018), suggesting the important role of actin cytoskeleton organization and remodeling in cardiomyocyte differentiation in the TZ.

In this study, we characterized the epithelial features of the aDPW and TZ in early mouse embryos by investigating alterations in these tissues in Hand2 mutant mice, providing additional understanding of the new role for the cardiac-specific transcription factor HAND2 in mediating epithelial maintenance and integrity for OFT morphogenesis.

The HAND2-specific antibody enabled us to observe a specific gradient expression pattern in the aDPW and TZ. Removal of Hand2 from these areas disrupted cell adhesion, including adherens junction and cell-matrix adhesions. Mechanistically, HAND2 regulates the expression of the actin and myosin cytoskeleton together with their regulators to establish cell adhesions for epithelial formation in the TZ, facilitating OFT morphogenesis.

The epithelial behavior of the aDPW and TZ has only recently been recognized through research that has provided a foundation for understanding the regulatory signaling pathways involved in the developing cardiac epithelium and the principles for establishing cell polarity. However, understanding of this paradigm of epithelial structure has been limited to the fact that Wnt5a-Dvl PCP signaling is necessary to facilitate epithelial formation, whereas PKC, TBX1 and Vangl2 are responsible for cell polarity establishment in these areas.

Our study demonstrates a new role for HAND2 in maintaining the epithelial integrity in OFT morphogenesis, and we propose a model based on this study (Fig. S8A,B). Our results also help us to understand the contribution of SHF progenitors to the OFT. For more than a decade, it has been well accepted that these progenitors move and/or migrate into the OFT, although the mechanisms involved have been unclear. Our study demonstrated that, by organizing epithelial tissue, SHF progenitors contribute to the OFT and that HAND2 has a pivotal role in maintaining the epithelial integrity. Consequently, the OFT morphogenic process can be understood from a new perspective, whereby the epithelium is continuously generated in the TZ, which pushes the caudal extension of the coherent epithelial sheet of primary OFT (Fig. S8A), a process similar to the involution of mesoderm and endoderm movement during in Xenopus gastrulation (Wolpert et al., 2006).

Our results also identified a new role of HAND2 in regulating cell adhesions and shed light on the underlying mechanisms. By transcriptionally regulating the expression of actomyosin cytoskeleton genes and their regulators, HAND2 is crucially involved in the construction of the cell adhesions in the cardiac epithelium and also maintains the epithelial integrity. Our results also suggest that the HAND2-regulatory actin cytoskeleton is required for correct cellular distribution and stability of N- and E-cadherin. Recently, Chien and colleagues reported N-cadherin-mediated cell proliferation and differentiation in the second heart field through systemic investigation of SHF-specific N-cadherin-deficient mice (Boon-Seng et al., 2014). Their results pinpoint the importance of N-cadherin in regulating SHF development and OFT morphogenesis.

Analysis of the RNA-seq data identified Stars as one of the most downregulated genes in Hand2 mutant mice. Previous studies revealed STARS to be actin-binding protein facilitating actin polymerization (Koichiro et al., 2005). Therefore, we investigated whether Stars is a direct transcriptional target of HAND2 through a series of biochemical analysis and in vivo immune-staining studies, which demonstrated Stars to be a downstream effector of HAND2. A recent study reported defective heart development in zebrafish with knocked down stars, indicating a crucial function of STARS in heart formation (Chong et al., 2012). Although Stars mutant mice are viable, this does not contradict our mechanistic conclusions. A panel of genes encoding distinct types of actin, such as Actc1 and Acta2, showed significant reduction in expression levels in Hand2 mutant mice. Actin molecules are materials for actin filament assembly to establish the actin cytoskeleton and STARS promotes this process. In Hand2 mutant mice, both actin materials and their assembly regulator, STARS, were reduced significantly. Therefore, the actin cytoskeleton defects are a combinatory effect of reduced actin and STARS.

The HAND2-regulatory actomyosin filament formation and STARS-mediated actin polymerization are also involved in the myofibril assembly in the cardiomyocytes. We observed a disrupted sarcomere structure in the RV of Hand2 mutant mice, which might account for the heart failure that leads to embryonic lethality by E13.5. How HAND2 couples the regulation of actin cytoskeleton together with that of actin-based sarcomere assembly is well understood. However, the regulatory mechanisms of actin-sarcomere assembly are more complex and require further detailed investigations.

Our study also defines a role of mammalian HAND2 that is distinct from its zebrafish counterpart in establishing cardiac epithelial organization. Zebrafish HAND2 is uniquely required for cell polarity of the early myocardial epithelium, whereas we identified a new function of mammalian HAND2 in cell adhesions.

Srivastava was one of the first to generate Hand2 (dHAND) global null mice for delineating HAND2 function in heart development, leading to the seminal discovery of HAND2 involvement in early heart formation (Srivastava et al., 1997). Recent research by his group using single cell transcriptome analyses revealed HAND2 to regulate OFT myocardial specification (De Soysa et al., 2018 preprint). This is consistent with our results and provides further insights into SHF progenitor cell fate and differentiation.

A recent report suggests the involvement of a Wnt5 signaling cascade in regulating actin polymerization (Li et al., 2016). In combination with our study, it is possible that HAND2 functions downstream of Wnt5a signaling to control actin filament and/or bundle formation.

In summary, we have revealed an important role for HAND2 in mediating cell adhesion by maintaining epithelial integrity, which is crucial for OFT morphogenesis. We believe that this study will help researchers to understand further the pathogenesis of certain congenital heart defects.

The following mouse strains were used in this study: Hand2 floxed mice (Holler et al., 2010), Mef2c-AHF-Cre mice (Verzi et al., 2005) and Rosa26-mTmG mice (from the Jackson Laboratory, Stock No. 023035). All these mouse lines were previously reported. The experimental animal facility has been accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC); the Institutional Animal Care and Use Committee (IACUC) of Model Animal Research Center of Nanjing University, China, approved all the animal protocols used in this study.

HAND2 antibodies were generated as a collaborative project with Abcam (R&D Center, Hangzhou, China). We first tested the specificity of the antibody product using cell transfection and Hand2-knockout tissues. We obtained a good batch of rabbit monoclonal antibody that is suitable for both IF staining and western blotting analysis. Abcam provided a portion of this product without charge, but it is now on the market with the catalog number of Ab200040.

Mouse embryos were fixed in 4% PFA for 2 h, dehydrated, embedded in paraffin or in optimal cutting temperature compound (OCT), and then sectioned. Histological sections were stained with Hematoxylin and Eosin (H&E) or used for TUNEL, IF and immunohistochemical (IHC) analysis. IF and IHC were carried out using standard procedures (Xiao et al., 2017). Embryos were fixed in 4% PFA, then rinsed in PBS and incubated in 30% sucrose at 4°C until sinking. The embryos were then embedded in OCT medium and snap-frozen in liquid nitrogen and stored at −80°C. Sections were cut at a thickness of 8 μm. Detailed information on the primary antibodies and dilutions is provided in Table S1. Fluorescent secondary antibodies were obtained from Jackson ImmunoResearch. Phalloidin-543 was obtained from Sigma (P1951) and Phalloidin-647 from Abcam (ab176759). Images were captured using a Leica SP5 laser confocal microscope.

An EdU pulse-chase assay was performed as described previously (Hu et al., 2018). Pregnant female mice were injected with 500 μg EdU (Invitrogen) at E9.25-9.5 and embryos were harvested 2 h later to detect the labeled cells proliferating during this period. To chase the EdU⁺ cells, female mice were injected with 5 mg thymidine (Sigma) after the 2-h EdU pulsing period to compete against the incorporation of EdU. Embryos were collected 19 h later at E10.0-10.5. All the embryos were processed for serial paraffin sections, and then stained for EdU according to the manufacturer's recommendation (Invitrogen).

aDPW and OFT tissues were dissected from control and Hand2 mutant embryos at E9.0-E9.5 and pooled. Total RNA was extracted from these tissues using Trizol. Three biological replicates for each group (control and Hand2 deletion, 21 tissues per group) were processed and assessed in the Illumina HiSeq™2500/Miseq sequencing platform (Novogene). The differentially expressed genes were analyzed with DESeq (Novogene). GO enrichment was analyzed via GOseq (Novogene). For RT-qPCR, tissue collection and RNA isolation were performed as described above. Primers used for qPCR analysis are listed in Table S2.

OFT tissues microdissected from control and Hand2 mutant embryos or H9C2 cells collected from normal control (NC) and Hand2 small interfering RNA (siHAND2) groups were used for protein extraction with RIPA (Beyotime Biotechnology, P0013B). Tissue or cell lysates were prepared with RIPA [20 mM Tris, 150 mM NaCl, 10% glycerol, 20 mM glycerophosphate, 1% NP40, 5 mM EDTA, 0.5 mM EGTA, 1 mM Na₃VO₄, 0.5 mM PMSF, 1 mM benzamidine, 1 mM DTT, 50 mM NaF, 4 M leupeptin (pH 8.0)]. Anti-HAND2 (generated in collaboration with Abcam, b200040; 1:1000), anti-E-cadherin (Bioworld, BS1098; 1:1000), anti-N-cadherin (Abcam, ab98952; 1:1000), anti-STARS (Proteintech, 22673-1-AP; 1:1000) and anti-SRF (Santa Cruz, sc-335; 1:1000) antibodies were used, with anti-β-actin (Bioworld, BS6007M; 1:5000), anti-α-tubulin (Bioworld, BS1699; 1:5000) or anti-GAPDH (Bioworld, AP0063; 1:5000) antibodies as internal controls. Samples were resolved by SDS-PAGE and transferred to PVDF membranes (Millipore). Membranes were blocked with 5% non-fat milk in TBST [50 mM Tris, 150 mM NaCl, 0.5 mM Tween-20 (pH 7.5)] and then incubated with primary antibodies overnight at 4 °C. Protein levels were quantified using ImageJ software. Each experiment was repeated at least twice.

The full-length fragment of the mouse Hand2 coding sequence (CDS) was cloned into the pCMV7.1-3×FLAG vector. Two ∼1.6 kb and ∼0.4 kb fragments, corresponding to the Stars 5′ flanking sequence, were amplified via PCR from mouse genomic DNA and cloned into the pGL3-Basic Vector (Promega) to generate a Stars-luc vector. The Stars mut-luc vector, which contained a 3-bp mutation in the mouse Stars E-box (A/T to G), was generated by mutation PCR. The primers used are listed in Table S2. Luciferase reporter assays in HEK293T and H9C2 cells were performed using the Dual Luciferase Assay System (Promega), following a standard protocol. Each assay was repeated three times.

For RNAi analysis of Hand2 in H9C2 cells, a 21-nucleotide small interfering RNA (siRNA) duplex was synthesized as follows: sense: 5′-UCAAGGCGGAGAUCAAGAATT-3′; antisense: 5′-UUCUUGAUCUCCGCCUUGATT-3′. siRNA with sense: 5′-UUCUUCGAACGUGUCACGUTT-3′; antisense: 5′-ACGUGACACGUUCGGAGAATT-3′ was used as negative control. Transfections were performed using Lipo3000 (Invitrogen) according to the manufacturer's instructions. Cells were used for the luciferase reporter assays or harvested for western blot analysis.

ChIP assays were performed according to a protocol from Dr Aibin He's lab (Peking University). In brief, H9C2 cells with exogenous mouse HAND2 overexpression were collected and crosslinked with 1% formaldehyde for 10 min at room temperature and subsequently quenched with glycine to a final concentration of 0.125M for another 10 min. Chromatin was sonicated with a Bioruptor (Diagenode), cleared by centrifugation (10 min, 4°C, 15,000 g), and incubated overnight at 4°C with 5 µg anti-Flag antibody (Sigma F1804). Immunocomplexes were immobilized with 100 μl protein-G Dynal magnetic beads (Invitrogen) for 4 h at 4°C, followed by stringent washes and elution. Eluates were subject to reversal of crosslinks overnight at 65°C and deproteinated. DNA was extracted with phenol chloroform, followed by ethanol precipitation. ChIP-qPCR analyses were performed using ABI StepOne Plus (ABI). The primers used are shown in Table S2.

All data are expressed as mean±s.e.m. Statistical analyses were performed using paired or unpaired t-tests as appropriate. Significance of differences was calculated with GraphPad Prism6. P-values <0.05 were considered significant.

We thank Dr Marthe J. Howard (University of Toledo Health Sciences Center, Ohio) for providing the Hand2 floxed mice.