Myocardial differentiation is dependent upon endocardial signaling during early cardiogenesis in vitro

Congenital heart defects (CHDs) are the most common form of major congenital malformations, occurring in nearly 1% of live births and 10% of aborted fetuses (Triedman and Newburger, 2016; van der Linde et al., 2011). Despite the increase in life expectancy of patients with CHDs as a result of effective surgeries and advanced medical therapies, it remains the leading cause of mortality due to birth defects. Thus, it is imperative to continue studying normal cardiac development and cell biology as well as the underlying pathological molecular changes that give rise to CHDs.

The heart is the first functional organ to develop during mammalian embryogenesis and supplies nutrients necessary for continued growth and development throughout gestation (Srivastava, 2006). Mammalian heart development commences during gastrulation when cardiac precursors ingress through the primitive streak cranially and bilaterally to populate an area of the anterior mesoderm known as the cardiac crescent (Auda-Boucher et al., 2000; Buckingham et al., 2005; Abu-Issa and Kirby, 2007). This occurs prior to heart tube formation at which point the endocardium and myocardium can be distinguished as unique cell populations in close juxtaposition. Interactions between the endocardium and myocardium are known to be crucial for later stages of cardiac development, such as during valvulogenesis (Barnett and Desgrosellier, 2003; Lin et al., 2012) and trabeculation (Gassmann et al., 1995; Lee et al., 1995; Zhang et al., 2013). However, the role of cell-to-cell interactions between these two populations during the initial stages of cardiogenesis has not been extensively explored. Our laboratory and others have shown that myocardial and endocardial cells arise from a population of common multipotent progenitors, which could explain the juxtaposition of the differentiating cell types often observed both in vivo and in vitro (Motoike et al., 2003; Masino et al., 2004; Kattman et al., 2006; Bu et al., 2009; Misfeldt et al., 2009; Pasquier et al., 2017). It is possible that the close proximity of endocardial and myocardial cells in cardiac mesoderm is necessary to promote differentiation and maturation via paracrine signaling. However, the limited availability of developmental models that permit the study of myocardial differentiation in the absence of endocardial cells has impeded the identification of such an interaction during the early stages of heart development. We have begun to elucidate early endocardial-myocardial interactions required for normal cardiac differentiation and maturation by taking advantage of an in vitro model of cardiogenesis using mESCs and the endocardial-specific expression of NFATc1.

The mESC in vitro system mirrors in vivo pre- and post-gastrulation events to such a high degree of fidelity that it has become a well-recognized model in which to study cellular heterogeneity and interactions, and spatiotemporal molecular processes of early mouse embryonic development (Misfeldt et al., 2009; Schulz et al., 2009; Van Vliet et al., 2012; DeLaughter et al., 2016). We generated a mESC bacterial artificial chromosome (BAC) transgenic line in which the regulatory elements of the Nfatc1 genomic locus drive the expression of the human heparin-binding EGF-like growth factor [HBEGF, also known as the diphtheria toxin receptor (DTR)]. Exposure of cells expressing the transgene to diphtheria toxin (DT) results in endocardial-specific cell death. Thus, the NFATc1-DTR mESC line is a genetically pliable tool that can be used to understand the dependence of early myocardial differentiation on endocardial cells and identify additional interactions necessary for cardiomyocyte maturation. We observed a significant attenuation in the percentage of contracting cardiomyocytes and a decrease in early myocyte differentiation and contractile markers within NFATc1-DTR embryoid bodies (EBs) during endocardial ablation, which was partially rescued by exogenous administration of Bmp2. Collectively, our results show that cardiomyocyte differentiation and maturation in vitro depends on early communication with endocardial cells that involves the Bmp signaling pathway. This work provides supporting data for the growing body of work delineating the importance of Bmp in endocardial-myocardial interactions and introduces a novel interaction that occurs prior to the stages previously studied in vitro.

Crosstalk between endocardial and myocardial cells is known to occur during the mid-to-late stages of heart development; however, it is unclear if there is an earlier interaction between the two cell types required for the differentiation and proliferation of cardiomyocytes. In an effort to address this, we developed an experimental system to study differentiation of cardiomyocytes in the absence of endocardial cells. We utilized BAC recombineering to insert both the diphtheria toxin receptor and a GFP reporter into the Nfatc1 genomic locus (Fig. 1A), which allows induction of endocardial cell death upon DT treatment. The NFATc1-DTR BAC was electroporated into low passage G4 hybrid mESCs (Tompers and Labosky, 2004; George et al., 2007). Selection and expansion of NFATc1-DTR ESCs yielded an in vitro transgenic model that permits the identification and selective ablation of endocardial cells throughout differentiation.

Expression of the GFP reporter was first detected on day (D) 6 of differentiation in NFATc1-DTR EBs. Immunofluorescent (IF) staining demonstrated colocalization of GFP (Fig. 1B) with NFATc1⁺ (Fig. 1C) cells. Additional staining for CD31 (Pecam1)⁺ endothelium verified that GFP expression is restricted to a subpopulation of the endothelium that represents NFATc1⁺ endocardial cells (Fig. 1D) (Misfeldt et al., 2009). To determine whether expression of the GFP reporter could be detected during later stages of differentiation, we analyzed D10 EBs using immunofluorescence. As with D6 EBs, the nuclear localization of GFP (Fig. 1E) coincided with endogenous NFATc1 expression (Fig. 1F) and CD31 expression (Fig. 1G). Thus, expression of the NFATc1-DTR transgene faithfully demarcates the endocardium as a separate endothelial population within EBs as early as D6 and up to D10 of differentiation.

After verifying the endocardial expression of the NFATc1-DTR reporter, we tested the specificity and function of the diphtheria toxin receptor. To ablate endocardial cells, NFATc1-DTR EBs were treated with DT from D0 to D6 of differentiation. Although GFP could be detected in non-treated NFATc1⁺/CD31⁺ endocardial cells at D6 (Fig. 2A), few or no GFP⁺ cells were detected in DT-treated D6 NFATc1-DTR EBs (Fig. 2B). The attenuated number of NFATc1⁺ cells with the absence of GFP expression suggests that DT treatment resulted in endocardial ablation within D6 NFATc1-DTR EBs. Quantification of immunofluorescent staining revealed a decrease in NFATc1⁺ endocardial cells as a result of DT-mediated ablation but a similar number of CD31⁺ endothelial cells within both untreated and DT-treated NFATc1-DTR EBs (Fig. 2C). Therefore, deletion of endocardial cells did not significantly affect the overall presence of the endothelium at D6. To determine if endocardial cells could be ablated during a time frame that would most likely encompass the specification, differentiation and proliferation of myocardial cells, NFATc1-DTR EBs were treated with DT from D0 to D10. This extended ablation allowed us to interrogate more deeply the consequences of removing all endocardial signaling required by arising cardiomyocytes during myocardiogenesis. DT treatment of EBs for 10 days resulted in a more dramatic reduction of NFATc1⁺/GFP⁺/CD31⁺ endocardial cells relative to untreated EBs (Fig. 2D,E) as demonstrated by IF. Quantification of cells expressing the GFP reporter and endogenous NFATc1 confirmed a significant decrease in endocardial cells whereas the remaining endothelial population (CD31⁺) appeared unperturbed in D10 EBs (Fig. 2F). Significant reductions in NFATc1and GFP message were also confirmed by qPCR (data not shown). From these results we can conclude that GFP expression is restricted to endocardial cells, faithfully recapitulating endogenous Nfatc1 expression. We have also shown that treatment with DT successfully and specifically ablates endocardial cells expressing DTR, without noticeably affecting other endothelial populations. In addition, DT treatment did not appear to overtly affect cell lineages that arose from other germ layers (Fig S1A,B), further supporting the endocardial specificity of this model.

To elucidate the effects of ablating endocardial cells on myocardial differentiation, we first looked for differences in functional myocytes by assessing the number of contracting EBs that developed during DT-treated and untreated conditions. During these experiments, both NFATc1-DTR and wild-type (WT) G4 ESCs were treated with DT from D0 to D10. G4 ESCs were used to generate the transgenic line and served as a WT control. DT treatment of WT EBs resulted in no significant decrease in the percentage of contracting foci compared with untreated WT EBs (Fig. 3A). However, there was a significant decrease in the number of beating myocytes within DT-treated NFATc1-DTR EBs compared with untreated NFATc1-DTR EBs on D8 persisting through D10 of differentiation (Fig. 3B). Thus, we conclude that although untreated NFATc1-DTR EBs behave like WT EBs, the presence of DTR allowed for specific ablation of NFATc1⁺ endocardial cells with DT, which interrupted necessary endocardial-myocardial interactions and resulted in a significant reduction of contracting EBs.

To ascertain how endocardial ablation via DT treatment influenced myocyte gene expression in D10 NFATc1-DTR EBs, we analyzed the expression of transcripts indicative of early cardiac differentiation. The expression of Nkx2.5 (Nkx2-5), one of the earliest heart development markers (Fig. 3C), Mef2c (Fig. 3D) and Isl1 (Fig. 3E), a marker of the secondary heart field, were all decreased in DT-treated NFATc1-DTR EBs (Tanaka et al., 1999; Cai et al., 2003). Additional markers of early myocardial differentiation, including Gata4, Gata6, Hand1 and Hand2, were also reduced in expression (data not shown) (Lescroart et al., 2014). bMHC (Myh7), a marker of late myocyte maturation (Fig. 3F) (Boheler et al., 2002) was decreased in D8 and D10 DT-treated NFATc1-DTR EBs concomitant with decreased myocardial function (Fig. 3B), as were other transcripts indicative of later myocyte differentiation including cTnT (Tnnt2), aMHC (Myh6), Mlc2a (Myl7) and Mlc2v (Myl2) (Fig. 3G) (Lyons et al., 1990; O'Brien et al., 1993; Kubalak et al., 1994). The reduced expression of genes involved in early cardiac differentiation along with decreased levels of those indicative of myocyte maturation suggest that the decreased contraction in DT-treated NFATc1-DTR EBs is due to fewer differentiating myocytes maturing into functional contractile myocardium. Of note, DT treatment resulted in specific ablation of the endocardium without perturbing differentiation of remaining endothelial populations as the expression of Er71 (Etv2) and Cgnl1, two markers of endothelial precursors, was similar in DT-treated and untreated NFATc1-DTR EBs (data not shown). To substantiate the specificity of DT treatment on endocardial ablation, we analyzed the temporal expression of mesodermal markers brachyury (T) and eomesodermin (Eomes), which did not differ between untreated and DT-treated NFATc1-DTR EBs (Fig. S1C,D). Most importantly, we did not observe a difference in the temporal expression of the early cardiac mesodermal specification marker Mesp1 between the two groups of EBs (Fig. S1E). Thus, ablation mediated by the NFATc1-DTR transgene did not hinder the differentiation of other mesodermal lineages but was restricted to endocardial cells during in vitro differentiation.

To investigate further the cause behind the reduced number of contracting cardiomyocyte foci and decreased expression of key myocyte differentiation markers observed in DT-treated NFATc1-DTR EBs, we analyzed the effects of DT treatment on cardiomyocyte proliferation in D6 NFATc1-DTR EBs. Dual IF staining for the proliferation marker phospho-histoneH3 (pHH3) and cardiac myosin heavy chain (MHC) was performed on untreated NFATc1-DTR EBs (Fig. 4A) and on NFATc1-DTR EBs treated with DT from D0 to D6 of differentiation (Fig. 4B). Quantification of the IF analyses revealed significantly fewer MHC⁺ cardiomyocytes per field in DT-treated NFATc1-DTR EBs (Fig. 4C). However, there was no detectable difference in the number of MHC⁺ cardiomyocytes also positive for pHH3 between untreated and DT-treated NFATc1-DTR EBs (Fig. 4C). Thus, the decrease in cardiomyocytes as a result of endocardial depletion does not appear to be the consequence of fewer proliferating cardiomyocytes.

To determine if the reduction in cardiomyocytes detected in D6 DT-treated NFATc1-DTR EBs was due to cell death, we next utilized the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) fluorescence assay. Although there were fewer MHC⁺ cardiomyocytes detected in NFATc1-DTR EBs treated with DT from D0 to D6 of differentiation, TUNEL staining revealed no overt difference in cardiomyocytes undergoing cell death between the untreated and DT-treated NFATc1-DTR samples (Fig. 4D,E). Assessment of MHC⁺/TUNEL⁺ cells in control and treated D6 NFATc1-DTR EBs demonstrated that apoptosis was not an underlying cause of the decrease in functioning myocardial cells or of the misexpression of key cardiac differentiation markers (Fig. 4F).

Next, we attempted to define a window during which endocardial ablation initially perturbed the differentiation of myocardial cells. DT treatment of differentiating NFATc1-DTR ESCs during sequential time intervals of in vitro development revealed a similar decrease in myocardial function as measured by the percentage of contracting NFATc1-DTR EBs with treatment from D4 to D8 (Fig. 5A,B) as was seen with treatment from D0 to D10 (Fig. 3B). This significant diminution in the number of beating NFATc1-DTR EBs treated with DT from D4 to D8 was detected as early as D8 and continued to D10. Similar decreases in transcripts corresponding to bMHC and additional myocyte maturation markers (cTnT, aMHC, Mlc2a and Mlc2v) were detected in D10 NFATc1-DTR EBs treated from D4 to D8 of differentiation (Fig. 5C). Thus, DT treatment of NFATc1-DTR EBs from D4 to D8 yields fewer differentiated and functional cardiomyocytes than observed with untreated EBs. Interestingly, this restricted time frame, which is highly equivalent to the cardiac developmental window of embryonic day (E) 7.5 to E11.5 in vivo, coincides with the initial expression of Nfatc1 as well as both early and late markers of myocyte differentiation (Misfeldt et al., 2009; Li et al., 2015).

After ascertaining the time frame in which the in vitro differentiation of cardiomyocytes is dependent on the presence of endocardial cells, we sought to determine if this reliance was due to paracrine signaling between the two cell populations. To identify candidate genes, we compared transcriptomes of differentiating endocardial cells versus pan-endothelial cells from in vitro differentiations using NFATc1-mCherry ESCs generated with BAC recombineering of the Nfatc1 locus (Fig. 6A) (Misfeldt et al., 2009). An E9.5 heart from a mouse generated using NFATc1-mCherry ESCs demonstrated restricted, colocalized expression of a nuclear cyan reporter and a glycosylphosphatidylinositol (GPI)-anchored mCherry reporter within endocardial cells lining the atrioventricular canal (AVC) (Fig. 6B,C). Thus, expression of the NFATc1-mCherry transgene faithfully recapitulates endogenous endocardial Nfatc1 expression in vivo. ESCs were differentiated as described above and immunostaining of D8 NFATc1-mCherry EBs confirmed endocardial-specific expression of the mCherry reporter within a subset of the endothelium (CD31⁺ cells) (Fig. 6D). D8 differentiated NFATc1-mCherry ESCs were dissociated and isolated by fluorescence-activated cell sorting (FACS), which revealed that 608,771 endocardial cells (3.5% of viable cells) were mCherry⁺/CD31⁺ and 953,344 endothelial cells (7.5% of viable cells) were mCherry⁻/CD31⁺ (Fig. 6E).

We performed RNA-seq on isolated endocardial cells compared with endothelial cells, using four independent replicates of each sample. We obtained 166 million read pairs, with no major variations in coverage between libraries (Fig. S2A). Direct transcript quantification revealed a compatible fragment ratio of over 99% for all libraries, demonstrating high library quality. We confirmed the identity of the samples and verified the expected expression profile by examining Nfatc1 expression pattern, which was used to isolate the cells (Fig. S2B). Therefore, our selection strategy identified the endocardial cell population that expressed Nfatc1. We further confirmed that the gene expression pattern observed was consistent with adult endocardial gene expression (Fig. S2E,F), by comparison with the RNA-seq dataset generated by Lother et al. (2018).

Differential expression analysis between endocardial (EC) and endothelial (ET) cells revealed 811 differentially expressed genes, of which 292 were upregulated and 519 were downregulated in the endocardium (Fig. 6F,G, Table S1). These results demonstrate a clear transcriptional identity of endocardium distinct from other endothelium. Gene Ontology (GO) enrichment analysis identified terms highly relevant to endocardiogenesis (Fig. 6H, Tables S2-S4), including ‘cardiovascular system development’, ‘regulation of vasculature development’ and ‘embryonic heart tube development’. Although unexpected, we also observed an endocardial enrichment in expression of genes important for general vascular development, angiogenesis and vasculogenesis over that detected in other endothelial cell populations, suggesting that the endocardium is one of the most active sites of vascular development during this period of cardiac development. This is consistent with the observation that the most enriched gene in the endocardium compared with the endothelium was VE-cadherin (Cdh5), which is known to be essential for angiogenic remodeling (Paatero et al., 2018).

Interestingly, our pathway analysis identified ‘cellular response to Bmp stimulus’ as enriched in the endocardium (Fig. 6H). Therefore, it was no surprise that Bmp2 was listed among four other growth factors (Tgfb1, Fgf3, Igf1 and Pdgfb) that were significantly upregulated in endocardial cells (Fig. 6I). To confirm these growth factors as potential candidates for a rescue experiment, we analyzed their expression in NFATc1-DTR EBs treated with DT from D4 to D8 of differentiation relative to the untreated control (Fig. 7A). Given the decrease in expression of Bmp2, Fgf3, Igf1, Pdgfb and Tgfb1 in DT-treated NFATc1-DTR EBs and their importance during heart development (Engelmann et al., 1992; Van den Akker et al., 2008; Li et al., 2011; Urness et al., 2011; Papoutsi et al., 2018), we attempted to rescue the differentiation and maturation of myocardial cells by singly introducing these prospective endocardial candidates during DT treatment from D4 to D8. Of the six factors tested, Bmp2 showed the most potential, as it partially rescued the number of contracting D10 NFATc1-DTR EBs treated with DT relative to untreated EBs (Fig. 7B). Additionally, Bmp2 treatment rescued the differentiation of myocardial cells as indicated by increased expression of Nkx2.5, Isl1, Gata4, Gata6 and Hand2 in EBs treated with DT from D4 to D8 compared with EBs only treated with DT (Fig. 7C). Furthermore, in agreement with increased beating in response to Bmp2 treatment during endocardial ablation, there was also an increase in the expression of the late myocyte maturation markers cTnT, aMHC, Mlc2a, Mlc2v and Hcn4 (Fig. 7D). Taken together, our data show that, in the absence of endocardium, exposure to Bmp2 during a key developmental window partially rescues the effects of endocardial ablation on cardiomyocyte differentiation and maturation by augmenting normal myocardial gene expression pathways. Of note, addition of Bmp2 to NFATc1-DTR EBs in the absence of DT did not increase the percentage of contracting EBs (data not shown).

To ascertain whether the partial rescue of cardiomyocyte gene expression and function observed as a result of Bmp2 treatment was due to an increase in cardiomyocytes during endocardial ablation, we investigated changes in proliferation at D6 of differentiation. The addition of Bmp2 during endocardial ablation increased the number of MHC⁺ cells present within D6 NFATc1-DTR EBs compared with those only treated with DT (Fig. 8A,B). However, quantification of MHC⁺/pHH3⁺ cells revealed an insubstantial increase in proliferating cardiomyocytes as a result of Bmp2 treatment (Fig. 8C). Thus, although Bmp2 treatment increased the number of myocardial cells detected in DT-treated NFATc1-DTR EBs, the results suggest that the increase is not due to a considerable change in proliferation. Therefore, we considered the likelihood that Bmp2 treatment reduced the number of dying cardiomyocytes. To test this possibility, we used a TUNEL assay to examine myocardial cell death in NFATc1-DTR EBs treated with DT from D4 to D6. Results from the assay revealed similar low levels of apoptosis among cardiac muscle cells between NFATc1-DTR EBs treated with DT (Fig. 8D) and NFATc1-DTR EBs simultaneously treated with DT and Bmp2 (Fig. 8E). This observation was further supported by quantification of TUNEL⁺ cardiomyocytes identified in D6 EBs following endocardial ablation and Bmp2 rescue (Fig. 8F). Taken together, the data show that Bmp2 treatment during endocardial ablation increases the number of cardiomyocytes present in NFATc1-DTR EBs at D6 of differentiation, but not by increasing cardiomyocyte proliferation or survival.

The above results showing decreased myocardial cells and cardiogenic gene expression due to endocardial ablation provide evidence that differentiation and maturation of cardiomyocytes require an interaction with endocardium that involves Bmp2 signaling. However, another possible explanation for our results could be the presence of multipotent cardiovascular progenitors that normally give rise to both the myocardium and endocardium, which express NFATc1 and thus DTR. If this is the case, then DT treatment would annihilate the progenitor population and effectively prevent the cells from branching off into separate populations of cardiomyocyte-fated and endocardial-fated cells. To exclude that possibility, we aimed to validate the NFATc1-DTR transgene specificity to endocardial tissue within the heart and emulation of endogenous Nfatc1 expression utilizing a mouse line derived from NFATc1-DTR ESCs (Fig. 9A) (Poueymirou et al., 2007). Expression of the transgene reporter, GFP, was detected within endocardial cells that lined the outflow tract (OFT; Fig. 9B) and AVC (Fig. 9C) of whole-mount E9.5 NFATc1-DTR embryos (Zhou et al., 2005; Misfeldt et al., 2009). This observation was supported by IF staining of an E9.5 NFATc1-DTR heart that revealed colocalization of endogenous NFATc1 and the GFP reporter within endocardial cells of the AVC (Fig. 9D-F). Overall, these results reveal that the NFATc1-DTR transgene faithfully recapitulates the expression pattern of NFATc1 during early heart development in vivo.

Both the in vitro and in vivo data substantiate the specificity of the NFATc1-DTR transgene within the endocardium; however, the lineage of NFATc1-DTR⁺ cells was unknown. To ensure that only endocardial cells were ablated during DT treatment and to rule out the possibility that regulatory elements of Nfatc1 promoted the expression of DTR in additional cardiac progenitors, which would result in the ablation of both cardiomyocytes and endocardial cells, we used an NFATc1-Cre BAC transgenic mouse line in which Cre recombinase is specifically expressed in endocardial cells as described above. Whole-mount X-gal staining of E10.0 embryos from a mating between NFATc1-Cre and Rosa26^lacZ reporter mice revealed the presence of β-galactosidase (β-gal)-positive cells predominantly in the endocardium of the heart (Fig. 9G,H). Endocardial-specific expression of NFATc1-Cre was more apparent following Eosin staining of sectioned E10.0 NFATc1-Cre; Rosa26^lacZ embryos, which revealed expression of β-gal within the endocardial lining of the OFT (c′) and the AVC (c″) (Fig. 9I). There was also β-gal expression demarcating mesenchymal cells in the endocardial cushion separating the left atrium (a) and the left ventricular lumen (v). This is unsurprising, as studies have reported that the majority of these cells are derived from endocardial cells, which, upon epithelial-to-mesenchymal transition (EMT) during the initiation of valvulogenesis, first suspend Nfatc1 expression and then migrate into the cardiac jelly (Wu et al., 2011). Thus, the signal within the cushions represent derivatives of NFATc1⁺ endocardial cells that contribute to the cardiac cushions (reviewed by DeLaughter et al., 2011). Importantly, little to no β-gal expression was detected in the myocardium, suggesting there was no Cre expression in myocardial progenitors during earlier stages of cardiac development. Detection of β-gal⁺ cells within the liver (li) (Fig. 9I) is also in agreement with published studies regarding the expression of Nfatc1 in other cell populations and further confirms the fidelity of the NFATc1-Cre (Zetterqvist et al., 2013). Comprehensively, these results suggest that the NFATc1-Cre transgene is specifically expressed in a subset of cells emulating the transcription of endogenous Nfatc1. The transcription of lacZ as a result of Cre-mediated recombination in NFATc1-Cre; Rosa26^lacZ mice led to the labeling of all NFATc1-Cre⁺ cells and their progeny. In an attempt to investigate expression of the NFATc1-Cre transgene at the onset of Nfatc1 transcription, we analyzed the earliest time point NFATc1⁺ cells have been detected in the developing heart (Misfeldt et al., 2009). The frontal view of a whole-mount X-gal-stained E7.75 NFATc1-Cre; Rosa26^lacZ embryo revealed endocardial-restricted β-gal expression within the cardiac crescent (Fig. 9J,K). Moreover, we observed β-gal activity in endocardial cells distinctly positioned alongside myocardial progenitors within the cardiac primordium of a sectioned E7.75 NFATc1-Cre; Rosa26^lacZ embryo (Fig. 9L). Overall, the lineage tracing of NFATc1-Cre expression demonstrates that the Nfatc1 promoter is not active within progenitor cells that contribute to cardiomyocytes; rather, the expression pattern within the developing heart is specific to the endocardium. As the same promoter elements were utilized in the NFATc1-Cre construct as in the NFATc1-DTR construct, these data substantiate the endocardial-specific ablation strategy during in vitro cardiac differentiation. Therefore, cardiomyocytes are not among the cells ablated during DT treatment as only endocardial cells express DTR and respond to DT treatment at the onset of differentiation. Thus, the decrease in myocardial cell number and disrupted cardiomyocyte gene expression patterns are directly related to an attenuation of the endocardial cell population.

Though there have been many studies highlighting crucial interactions that occur between endocardial and myocardial cells during later stages of heart development, it is unclear whether such interactions occur at the onset of cardiogenesis. We are addressing this gap in the literature with the use of a new ESC line that permits the diphtheria toxin receptor-mediated ablation of endocardial cells at any time during ESC differentiation in vitro. In addition, we utilized the enhancer-promoter regulation of the Nfatc1 locus to ensure that the expression of DTR emulated endogenous expression of Nfatc1 specifically in endocardial cells. As a result, we have evidence of an interaction between endocardial and myocardial cells that occurs during the early stages of cardiac myogenesis. Furthermore, to our knowledge, the data presented in this study are the first direct demonstration of a crucial role for endocardium on subsequent myocardial differentiation, maturation and function.

During in vitro differentiation, the critical period in which myocardial cells are reliant on the presence of endocardial cells encompasses the initial expression of early myocyte markers Nkx2.5, Mef2c and Isl1. The temporal expression of the markers observed in this study is in agreement with previous in vitro studies of differentiated mESCs (Bondue et al., 2008; Misfeldt et al., 2009; Tanwar et al., 2014; Prados et al., 2018). Interestingly, null mutations of these genes impaired but did not entirely inhibit the development of a primitive heart in murine embryos (Lyons et al., 1995; Lin et al., 1997; Cai et al., 2003; Prall et al., 2007; Li et al., 2016). This suggests that, though these genes are not required for initiation of cardiomyogenesis, they are essential for morphogenesis and the continuation of heart development. This early period of crucial endocardial-myocardial interactions also includes the emergent expression of the late-stage cardiac sarcomere marker bMHC, suggesting the importance of endocardium for both specification and determination of at least a subset of cardiomyocyte progenitors throughout the continuum of myocyte differentiation.

Interestingly, expression of myocardial differentiation markers was still detected following ablation, albeit at significantly reduced levels. This could be explained by the fact that ablation of the endocardium was not complete in this model, resulting in persistence of a residual population of endocardial cells that supported general myocyte differentiation. However, a recent study using single-cell analysis of early cardiomyocyte progenitor cell differentiation demonstrated that cardiomyocytes that differentiate as distinct subpopulations after E6.75 are committed to different lineages and contribute to distinct regions of the mature heart (Lescroart et al., 2018). Thus, it is equally plausible that some of these myocyte cell populations are more sensitive to endocardial-derived factors than others. We favor this later explanation for two reasons: (1) we did not see either an increase in myocardial cell death or a decrease in myocyte proliferation following DT treatment and (2) treatment with Bmp2 was able to partially rescue the phenotype observed. This is consistent with a subset of myocardial cells that are dependent on endocardial signaling and subsequently undergo expansion when components of endocardial signaling are restored by exogenous growth factor administration. It is also possible that signaling from the endocardium is required for mesodermal induction of myocyte differentiation. However, most Mesp1⁺ cardiac progenitors differentiate into either cardiomyocytes or endothelial cells early during gastrulation (Lescroart et al., 2014). The fact that no myocyte population was identified in our NFATc1-Cre lineage analysis suggests that the primary effect of DT administration occurred after mesodermal specification of endocardium and myocardium. Additionally, further analysis of the single-cell datasets from Lescroart et al. (2018) revealed the absence of cells that co-expressed Nfatc1 (and thus the NFATc1-DTR transgene) and early myocardial markers in E6.75 murine hearts (Fig. S3A-C). Thus, our data would suggest that endocardial ablation did not alter mesodermal specification of myocardial cells from a common cardiac progenitor. We speculate that once myocardial specification has occurred, further myocyte differentiation and maturation is dependent on paracrine signaling from the endocardial cell population. This resident subpopulation of mesodermal cells that are specified to become myocardial cells can be induced to continue maturation by exogenous BMP2 in the absence of endocardium.

Results from the current study suggest that interactions between myocardial and endocardial cells occur during the early differentiation and maturation phase of cardiomyogenesis in vitro, when cardiomyocyte precursors are committed to specific cell fates. The reduced number of contracting NFATc1-DTR EBs treated with DT, and the decrease in myocyte maturation markers supports this conclusion. Indeed, our in vitro experiments define a critical period of maximal effect (Fig. 5) between days 4 and 6 of culture, which is roughly equivalent to E7.5-9.5 in vivo. This correlation is supported by an in vivo study that clearly delineated the juxtaposition of endocardial and myocardial cells in the cardiac crescent at E7.5 (Misfeldt et al., 2009). Future experiments that may help elucidate the interactions between endocardial and myocardial cells during this early stage of differentiation and maturation include performing transcriptional analysis of cardiomyocytes cultured with and without endocardial cells.

Data from this study also imply that interactions between endocardial and myocardial cells during early cardiomyogenesis involve Bmp signaling. However, our results differ from those of Palencia-Desai et al. who demonstrated that chemical and genetic ablation of Bmp signaling in the tailbud-stage zebrafish embryo resulted in severe inhibition of endocardial differentiation with little effect on myocardial development (Palencia-Desai et al., 2015). We cannot explain this discrepancy other than to suggest this might reflect species variability in dependence on Bmp signaling during early cardiogenesis as chicken (Schlange et al., 2000), mice and humans (Kattman et al., 2011) depend on Bmp signaling for early myocardial differentiation.

Though addition of Bmp2 partially rescued the effects of endocardial ablation on cardiomyocyte number and the percentage of beating NFATc1-DTR EBs on D10 of differentiation, the source of this growth factor under normal conditions is unclear. Multiple tissues secrete Bmp2 in a spatiotemporal manner to promote the initiation and morphogenetic events of cardiogenesis. Myocardial cells serve as a source of Bmp2 during early heart development (Palencia-Desai et al., 2005) and during later stages to promote endocardial EMT and patterning of the AVC during valvulogenesis. Inactivation of the growth factor specifically in the atrioventricular myocardium or deletion of the Bmp type 1A receptor, Bmpr1a, in endocardial cells resulted in a lack of cushion formation (Ma et al., 2005). In addition, ectopic Bmp2 expression by ventricular myocardium can drive ventricular endocardium to differentiate into valve endocardium (Papoutsi et al., 2018). Recently, studies have focused on Bmp2 expression in endocardial cells. Saxon et al. demonstrated a role for Bmp2 in the elongation, remodeling and maturation of the atrioventricular (AV) endocardial cushion into AV valves and the ventricular septum following analysis of endocardial-specific conditional knockouts of Bmp2 in E13.5 and E16.5 embryos (Saxon et al., 2017). Therefore, given the different sources of Bmp2, there are at least two possible explanations for the partial rescue of cardiomyocyte differentiation during ablation. One possibility is that endocardial cells signal to myocardial cells to secrete Bmp2, which acts in an autoregulatory loop to promote specification of myocardial precursors. A second possibility is that Bmp2 is released directly from endocardial cells to signal through Bmp receptors of nascent cardiomyocytes to promote further differentiation and maturation. These two possibilities are not mutually exclusive and further investigation will be required to define these crucial cell-to-cell interactions.

In summary, these studies demonstrate a crucial interaction between endocardial and myocardial cells in vitro during early cardiogenesis. Whether the key signaling events observed are the result of cell-to-cell contact or paracrine signaling, or both, cannot be determined with these experiments. However, these experiments do suggest that alterations in very early endocardial-myocardial interactions may play a role in the development of CHDs as endocardial ablation resulted in attenuated expression of the highly conserved transcription factors Gata4 and Nkx2.5, which are associated with a range of CHDs (Li et al., 1997; Schott et al., 1998; Garg et al., 2003; Davidson and Erwin, 2006; Zaidi and Brueckner, 2017; Russell et al., 2018). Additionally, these studies suggest that optimal in vitro myocyte differentiation of progenitor cells for tissue regeneration may require further delineation of the essential components of endocardial-myocardial interactions that drive normal myocyte differentiation. It is known that co-culture of ESCs with endothelial cells promotes enhanced in vitro maturation of cardiomyocytes (Pasquier et al., 2017). We demonstrated that embryonic endocardium is a transcriptionally unique endothelial subpopulation. Thus, it is possible that co-culture with endocardial cells, rather than ‘generic’ endothelium, will further enhance the utility of differentiated progenitor cells for cell therapies (Zhang et al., 2018). Altogether, insights gathered from these studies should assist the development of novel stem cell-based replacement therapies with the potential to promote the differentiation of myocardial cells and generate functional myocardium.

The NFATc1-DTR BAC transgene was generated using a multicistronic vector comprising a nuclear-localized GFP reporter (H2B-eGFP) linked by the self-cleaving 2A peptide to DTR, the human heparin binding EGF-like growth factor (HBEGF), which acts as a receptor to diphtheria toxin (Mitamura et al., 1995; Donnelly et al., 2001; Stewart et al., 2009). The transgene was inserted between the 500 bp 5′ translational start site and the 400 bp 3′ splice site of exon 1 corresponding to homologous arms of the murine Nfatc1 gene. The bacterial homologous recombination of the transgene with a BAC consisting of the entire Nfatc1 genomic locus was achieved utilizing a previously described protocol (Lee et al., 2001). The NFATc1-mCherry ESC line was generated in a similar manner, except the multicistronic vector introduced into the BAC containing the Nfatc1 genomic locus consisted of a nuclear localized cyan reporter (H2B-CFP) and an mCherry reporter tagged with the GPI signal sequence. The BAC transgene was then purified with a cesium chloride gradient and dialyzed in microinjection buffer (10 mM Tris, 15 µM EDTA pH 7.4) prior to electroporation. ESCs that successfully incorporated the purified NFATc1-DTR BAC or NFATc1-mCherry BAC were expanded on mouse embryonic fibroblasts (MEFs) with the neomycin-resistance cassette (Millipore PMEF-NL) and cultured in selection media with 500 µg/ml of G418 Sulfate (Corning).

To generate NFATc1-DTR and NFATc1-mCherry BAC transgenic mouse lines, eight NFATc1-DTR ESCs were introduced into the space between the zona pellucida and blastomeres of 2.5 days post-coitum (dpc), 8-cell ICR (Mus musculus) embryos via laser-assisted microinjections (Poueymirou et al., 2007). The embryos were then incubated at 37°C for 24 h to permit progression of the embryo to compaction and the blastocyst stage at 3.5 dpc. The embryos were then transplanted into the uterine horns of pseudo-pregnant ICR dams for further development. The NFATc1-Cre BAC transgenic mouse line was generated by Cyagen Biosciences (Guangzhou, China) using the PiggyBac transgenic method. F0 chimeric males identified by 100% coat color and genotyped by PCR were mated with C57BL/6 dams and exhibited a high capability of germline transmission. The F1 generation of litters resulting from the mating was maintained in accordance with protocols approved by the Vanderbilt University Animal Care and Use Committee (IACUC). All the mice were housed in mating pairs under standard housing conditions (12 h/12 h light/dark cycle at 25°C).

NFATc1-DTR ESCs were cultured on irradiated MEFs in media consisting of Dulbecco's Modified Eagle Medium (DMEM, Invitrogen) supplemented with 4.5 g/l of D-glucose and 2 mM of L-glutamine. The media also contained 15% fetal bovine serum (FBS; Atlanta Biologicals), 100 µM non-essential amino acids, 1 mM sodium pyruvate, 100 U/ml Pen/Strep, 100 µM β-mercaptoethanol and 1000 U/ml recombinant leukemia inhibitory factor (ESGRO mLIF, Millipore). Hanging drops of 2.5×10⁻⁴ NFATc1-DTR ESCs were cultured in differentiation media (DM) for 48 h to form EBs. The DM consisted of Iscove's Modified Dulbecco's Medium (IMDM) supplemented with 2 mM L-glutamine and 25 mM HEPES, as well as 15% FBS (Atlanta Biologicals), 100 U/ml Pen/Strep, 0.5 mM L-ascorbic acid (Sigma), 15 µg/ml transferrin (Roche) and 4.5×10⁻⁴ M 1-thioglycerol. The EBs were suspended in DM on non-treated suspension culture dishes for another 48 h before being transferred onto gelatinized culture dishes where they were maintained from D4 and thereafter. The DM was changed every 48 h. Cells were treated with diphtheria toxin (Sigma, 0.1 µg/ml), rhBmp2 (R&D Systems, 100 ng/ml and 200 ng/ml; because no difference was observed with use of the higher concentration, all experiments reported are with a dose of 100 ng/ml), Nrg1b1 (100 ng/ml), Pdgfb (10 ng/ml), Tgfβ1 (5 ng/ml), Igf1 (100 ng/ml) or Fgf3 (10 ng/ml).

Total RNA was extracted from NFATc1-DTR EBs according to the TRIzol protocol (Invitrogen). RNA concentrations were measured using a Nanodrop spectrophotometer prior to treatment of 1 µg RNA with DNase I (Invitrogen). RNA was reverse-transcribed into cDNA using SuperScript III first-strand synthesis system (Invitrogen). qPCR was performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) with gene-specific primers. The cycling conditions were as follows: 94° for 5 min, 94° for 1 min, and 35 cycles of amplification (94° denaturation for 30 s, 55° annealing for 30 s, 72° elongation for 60 s) using the C1000 Touch Thermal Cycler (Bio-Rad). Experiments were performed in triplicate, and relative expression was calculated as 2^−ΔΔCT with 18 s as an internal control.

Whole EBs were cultured in 8-well chamber slides (Nunc, Lab-Tek) from D4 of differentiation to the day of analyses. For immunofluorescence, EBs were fixed with 2% paraformaldehyde (PFA) and blocked in 0.1% PBS/0.1% Tween 20 (Fisher Biotech) with 5% bovine serum albumin (Sigma) and 10% normal goat serum prior to incubation with primary antibodies overnight at 4°C. Cryosections (10 µm) of NFATc1-DTR and NFATc1-mCherry embryos were fixed in 4% PFA and incubated in a similar blocking solution before incubation with primary antibodies overnight at 4°C. Primary antibodies, GFP (Invitrogen, 1:200, rabbit, A-11122), cTnT [Developmental Studies Hybridoma Bank (DSHB), 1 µg/ml, mouse CT3], MHC (DSHB, 1 µg/ml, mouse, MF20), CD31 (BD Pharmingen, 1:150, rat MEC13.3, 550274) and pHH3 (Cell Signaling Technology, 1:200, rabbit, 9701), were used according to the manufacturer's instructions. Sections that were stained with the NFATc1 antibody (BD Pharmingen, 1:50, mouse 7A6, 556602) were processed according to the protocol associated with the mouse on mouse (M.O.M.) immunodetection kit (Vector Laboratories). Cell death was detected in NFATc1-DTR EBs using the ApopTag Red In Situ Apoptosis Detection kit and the TUNEL method (Millipore). Samples were incubated in Alexa Fluor-conjugated secondary antibodies (ThermoFisher Scientific, 1:500; goat anti-rabbit 488, A-11034; goat anti-rat 488, A-11006; goat anti-mouse 555, A-21424; goat anti-rat 647, A-21247) for 1 h in the dark at room temperature before they were mounted with the ProLong Gold Antifade Mountant with DAPI (Life Technologies). Images were obtained using a Nikon Eclipse E800 epifluorescence microscope or an Olympus FV-1000 inverted confocal microscope. Adobe Photoshop CS6 was used to examine the stained samples and for quantitative analysis by averaging the total number of proliferating and dying cells between three frames per sample within control and treated groups of three.

For whole-mount β-gal staining, E10.0 NFATc1-Cre; Rosa26^LacZ embryos were dissected in cold PBS and fixed in 4% PFA for 4 h at 4°C. The embryos were then incubated in X-gal staining buffer consisting of 1 mg/ml X-gal, 5 mM potassium ferro/ferricyanate, 0.02% NP40, 2 mM magnesium chloride and 0.01% sodium deoxycholate in PBS overnight at 37°C with constant, gentle agitation. The following day, the embryos were washed and dehydrated using a series of ethanol and xylene washes before being processed for embedding in paraffin. Sections of the embryos (6 µm) were incubated in xylene, rehydrated, and stained with Eosin before they were imaged with a Nikon AZ 100 M microscope.

D8 EBs were first dissociated using cell dissociation buffer (Gibco). The resultant single cells were then re-suspended in a solution consisting of Hanks' Buffered Salt Solution (HBSS), 25 mM HEPES and 2% FBS. The cells were incubated with CD16/CD32 (mouse, Ebiosciences, 1 µg/ml) to prevent nonspecific binding before incubation with an Allophycocyanin-conjugated (APC) antibody against CD31 (rat anti-mouse MEC 13.3, BD Pharmingen, 0.5 µg/ml). The cells were also stained with propidium iodide (Molecular Probes) in an effort to identify and exclude dead cells from collection. The FACSAria from Becton Dickinson along with FACSDiva 6.0 was then used to physically sort out endocardial cells (mCherry⁺/CD31⁺) and endothelial cells (mCherry⁻/CD31⁺) from the heterogeneous NFATc1-mCherry population of 3×10⁶ cells per 1 ml.

NFATc1-mCherry mESCs (Misfeldt et al., 2009) were differentiated using the hanging drop method. FACS was used to identify and isolate pools of endocardial cells (NFATc1⁺/CD31⁺) and endothelial cells (NFATc1⁻/CD31⁺). RNA was prepared from the frozen, sorted cells using a Qiagen RNeasy Mini Kit (Cat No: 74104; Lot No: 145038477) according to the manufacturer's instructions. RNA quality was determined using a TapeStation high sensitivity assay where a RIN value of at least 8.0 was required. Qubit quantification was also performed and a minimum of 0.1 µg of RNA was isolated from each sample. Four replicate RNA-seq libraries were prepared from independent embryoid body cultures for both endocardial and endothelial cells using an Illumina stranded mRNA LT Kit (Part Number: RS-122-2101; Lot: 401884). The libraries were poly(A) selected and directional, and the protocol from the True-seq stranded mRNA sample preparation guide was followed with the fragmentation step adjusted to 1 min to generate fragment sizes compatible with 100 bp paired-end sequencing (approximately 300 bp). The barcoded libraries were diluted 1:1000 and quantified using qPCR, normalized to 10 mM, and pooled. Cluster generation and sequencing was performed in paired-end mode with 100 to 101 cycles per read with indexing on an Illumina HiSeq 2000 instrument at the Biomedical Research Centre Genomic Core facility at King's College.

Base calling was performed using Casava (version 1.8.0) on the Biomedical Research Centre High Performance Computing Cluster at King's College. We obtained 166 million reads pairs, with no major variations in coverage between libraries. Transcript quantification was performed using Salmon (version 0.9.1) (Patro et al., 2017). Differential expression analysis was performed using R package DESeq2 (version 1.18.1) (Love et al., 2014). Principal component analysis was performed using the prcomp R function. GO analysis was performed using the package GOstats (version 2.44.0) (Falcon and Gentleman, 2007). Gene set enrichment analysis was performed using Liger R package (https://github.com/JEFworks/liger). Transcriptional signatures using the data from Lother et al. (2018) were generated using the same processing methods used for the data presented here.

Statistical data presented in this study represent mean±s.e.m. A two-tailed Student's t-test was utilized to identify statistical significance in the difference of mean values compared during quantitative analyses. Differences were designated as significant if P<0.05.

We thank Christopher B. Brown PhD for his support on experimental design and resources. Experiments/data analysis/presentations using the Nikon AZ 100M were performed in part through the use of the Vanderbilt Cell Imaging Shared Resource. We are grateful for services provided by the Biomedical Research Centre High Performance Computing Cluster, The Biomedical Research Genomic Core facility at King's College, London and the genomics core facility. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. Flow Cytometry experiments were performed in the VMC Flow Cytometry Shared Resource. The VMC Flow Cytometry Shared Resource is supported by the Vanderbilt Ingram Cancer Center (P30 CA68485) and the Vanderbilt Digestive Disease Research Center (DK058404)