Nppa and Nppb act redundantly during zebrafish cardiac development to confine AVC marker expression and reduce cardiac jelly volume

At early stages of development [24-30 h post fertilisation (hpf) in the zebrafish], the heart is a double-layered linear tube, with an inner layer of endothelium (called the endocardium) and an outer muscular layer called the myocardium. During cardiac morphogenesis in both zebrafish and mice, the prospective atrial and ventricular chambers of the heart ‘balloon’ out of the heart tube, becoming morphologically and genetically distinct from the atrioventricular canal (AVC). The AVC is a specialised region of the developing heart that partitions the two chambers. It is characterised by Bmp2/4 (Keyes et al., 2003; Ma et al., 2005; Walsh and Stainier, 2001) and Tbx2 (Chi et al., 2008; Christoffels et al., 2004; Harrelson et al., 2004) gene expression in the myocardium and Has2 gene expression in the endocardium (Tien and Spicer, 2005). Has2 encodes a synthetic enzyme of the extracellular matrix component, hyaluronic acid (HA), and, congruent with its expression, localised swellings of extracellular matrix (ECM; known as ‘cardiac jelly’) are observed at the AVC. These swellings are the endocardial cushions: primitive structures that will later remodel to form cardiac valve leaflets.

The hierarchical relation between Bmp2/4 and Tbx2 has been extensively described, both in fish and mammals. Bmp2/4 and Tbx2 depletion share the phenotype of impaired AV boundary establishment (Chi et al., 2008; Harrelson et al., 2004; Ma et al., 2005) and reduced Bmp signalling results in undetectable expression of Tbx2 in the AVC myocardium, placing Bmp signalling upstream of Tbx2 (Ma et al., 2005; Shirai et al., 2009; Verhoeven et al., 2011). The zebrafish jekyll mutant, which harbours a mutation in the UDP-glucose 6-dehydrogenase gene (another enzyme of the HA biosynthetic pathway), and Has2-null mice show a key function for the cardiac jelly in determining both correct patterning and early localised growth of the endocardial cushions (Camenisch et al., 2000; Walsh and Stainier, 2001).

Although Bmp2/4 and Tbx2 are restricted to the AVC, Nppa (also known as ANF, atrial natriuretic factor) and Nppb (also known as BNP, brain natriuretic peptide) represent their chamber myocardium counterpart: Nppa and Nppb expression are enriched in the ballooning surfaces of the cardiac tube, and Nppa is actively excluded from the AVC by the repressive function of Tbx2 (Becker et al., 2014; Christoffels et al., 2004; Habets et al., 2002). Despite their compelling expression patterns, little has been reported about the function of Nppa and Nppb during heart development. Nppa- or Nppb-null mice lack any obvious developmental phenotype (John et al., 1995; Tamura et al., 2000). Nppa and Nppb lie in a conserved gene cluster (Inoue et al., 2003; Tamura et al., 1996) with common regulatory regions that become activated during development (Sergeeva et al., 2016). This co-regulation, their homology and the absence of developmental phenotypes observed in response to single gene loss of function has led to the hypothesis that these two genes are functionally redundant.

Here, we describe the generation of single and double nppa and nppb zebrafish mutants, which is now possible with genome-editing methods. Consistent with previous reports, single mutants exhibit no overt developmental phenotype. In contrast, double mutants exhibit pericardial oedema, expanded AVC markers and thickening of the cardiac jelly in the atrium of the early embryonic heart. These data support the notion that nppa and nppb function redundantly, and demonstrate that nppa/nppb double mutants play a previously unappreciated role in cardiac development by restricting AVC gene expression and repressing ECM accumulation in the atrial chamber.

In zebrafish, the nppa and nppb loci reside in a gene cluster on chromosome 8, separated by approximately 2.5 kb (Fig. 1A). This proximity precludes a classical strategy of generating double homozygous mutants by homologous recombination. To overcome this, we undertook sequential mutagenesis of the two loci, first generating an nppa mutant line using TALEN technology, and then targeting the nppb locus in the nppa mutant background using the CRISPR/Cas9 system. Simultaneously, nppb single mutants were created. For nppa, exon 1 was targeted, producing a 4 bp deletion at 55 bp of the coding region (nppa^uq19ks). This results in normal protein sequence until amino acid 18, followed by frameshifted sequence to amino acid 23 and then translational termination (Fig. 1A,B), preventing generation of the 139 amino acid wild-type protein. For nppb, exon 2 was targeted and two alleles were generated. The mutant locus nppb^uq20ks, which is linked in cis to the nppa^uq19ks allele, carries a 39 bp insertion containing a stop codon, resulting in a shorter protein of 80 amino acids (Fig. 1A,B; full-length wild-type protein contains 129 amino acids). The single nppb^uq21ks mutant allele harbours a 10 bp frameshift insertion at amino acid 80, encoding 15 out-of-frame amino acids until termination at 95 amino acids in length (Fig. 1A,B). In all three instances, the mutated forms of Nppa and Nppb lack the C-terminal region of the protein, which contains the mature peptides produced following processing (Fig. 1B) (Bloch et al., 1985). Given that targeting the propeptide region of Nppa has previously been shown to result in undetectable amounts of Nppa protein (John et al., 1995), we surmise that the nppa^uq19ks, nppb^uq20ks and nppb^uq21ks alleles are functional nulls.

The incross of nppa^+/uq19ks and nppb^+/uq21ks single carriers generated embryos that, at the developmental stages analysed [48, 72 and 96 h post fertilisation (hpf)], appeared wild type, thus in line with the absence of phenotype in Nppa^−/− and Nppb^−/− mouse embryos as well as zebrafish single morphants (Becker et al., 2014; John et al., 1995; Tamura et al., 2000). In contrast, double homozygous nppa^{uq19ks/uq19ks}/nppb^{uq20ks/uq20ks} embryos displayed pericardial oedema starting at 48 hpf (Fig. 1D-H). The penetrance of the phenotype was incomplete and the expressivity variable. Approximately 20% of double mutants presented a wild-type phenotype and, in the remaining 80% of the genotypically mutant embryos, pericardial oedema was accompanied by additional cardiac abnormalities, such as looping defects or blood regurgitation at the inflow tract, or both (Fig. 1H; Movie 1). One day later, the penetrance increased (Fig. 1H). These data confirm that Nppa and Nppb do indeed function redundantly and play a role during vertebrate cardiac development.

To establish whether compensation at the level of gene expression exists between nppa and nppb, we performed in situ hybridisation expression analysis on both single and double mutant embryos (Fig. 1C, Fig. S1). The intensity of both nppa or nppb staining was unaltered at all stages and genotypes analysed, with the exception of nppb (Fig. 1C and Fig. S1). To quantify the expression of nppa and nppb levels in order to address the issue of compensation, we performed Q-PCR analysis on single nppa and nppb mutant embryos compared with wild-type sibling controls. No significant difference was observed for any groups tested (Fig. S1). This demonstrates that there is no compensation occurring for these genes and is consistent with reports that nppa and nppb are co-regulated (Becker et al., 2012; Sergeeva et al., 2016, 2014).

Previous studies inhibiting nppa and nppb expression via double morpholino knockdown have described increased cardiomyocyte numbers in double morphant embryos (Becker et al., 2014). To examine whether we observe a difference in cardiomyocyte number in double mutant versus sibling embryos, we crossed double heterozygous carriers onto the Tg(myl7:dsRed-nls) background (labelling cardiomyocyte nuclei) and quantified cell number from confocal z-stack images. In contrast to double morpholino-treated embryos, we observed no significant difference in cardiomyocyte number in double mutant embryos compared with siblings (Fig. 2A,A′).

Natriuretic peptides are well known for counteracting cardiac hypertrophy (Lerman et al., 1993; Vanderheyden et al., 2004) and, conversely, mice null for Nppa or the receptor shared by Nppa and Nppb, Npr1, display cardiac enlargement and augmented cardiomyocyte diameter (Ellmers et al., 2007; Holtwick et al., 2003; John et al., 1995). Quantifying cardiomyocyte size in double mutant hearts compared with siblings demonstrated no difference between double mutants and siblings (Fig. 2B,B′). Although this result is in contrast with that observed in Nppa^−/− and Npr1^−/− mice, these previous studies examine effects after months of Nppa deficiency (Ellmers et al., 2007; John et al., 1995). It is perhaps this chronic loss of Nppa signalling that accounts for the discrepancy between our work and previous reports.

Next, to investigate heart function of double mutants, high-speed imaging of siblings and double mutants was performed. By observing movies of siblings versus double mutant hearts, blood flow appeared slower when examining the inflow tract of beating hearts, which would account for the pericardial oedema observed in double mutant embryos (Movie 1). However, quantitative measurements of stroke volume and ejection fraction showed no significant difference between sibling and mutant embryos (Fig. S2). These measurements rely on comparable morphology between different groups being measured. Considering the distention of the hearts observed from pericardial oedema, this is likely to have impaired measurements and may account for the lack of functional difference detected between siblings and mutants.

Given the repressive role Tbx2 plays on Nppa expression (Christoffels et al., 2004; Habets et al., 2002), we sought to investigate whether a reciprocal repressive function for Nppa on AVC-restricted gene expression takes place. To examine this, we performed in situ hybridisation analysis for a range of AVC markers in sibling and double mutant embryos. Interestingly, expression of all markers tested, namely bmp4, tbx2b, has2 and versican, was expanded in double mutant embryos at 72 hpf (Fig. 2C,D). To investigate whether this resulted in a concomitant increase in the number of cells transitioning to endocardial cushion (EC) identity, immunostaining for Alcama on the Tg(fli1a:nEGFP)^y7 transgenic line was performed and EC cell number quantified. Surprisingly, no significant difference was observed in the number of EC cells between sibling and double mutant embryos (Fig. S3). Instead, the expansion of AVC marker gene expression appears to affect thickening of the jelly between the layers of the cardiac wall. When high-speed movies from Fig. S2 and Movie 1 were examined closely, an increased endocardial-to-myocardial thickness was consistently observed in double mutant embryos. To quantify this, the area from the outer surface of the atrial myocardium to the luminal surface of the heart was measured and a significant difference was observed between sibling and double mutant embryos (Fig. 2E,E′).

We recently reported a genetically encoded biosensor to detect HA in the developing embryo (De Angelis et al., 2017), using a similar approach used in cultured cells (Zhang et al., 2004). This biosensor uses the HA-binding domain of the protein, Neurocan, fused to EGFP with a synthetic secretion sequence to export it extracellularly and localise it to HA. Given that HA is a major component of the cardiac jelly, we applied this tool to examine double mutant embryos. Although transient mRNA injection could be used for analysing early stages of zebrafish development, we found poor reproducibility by 48 hpf in the cardiac jelly. We therefore generated a stable transgenic line, Tg(ubi:ssNcan-EGFP)^uq25bh, using the ubiquitin (ubi) promoter and observed robust and reproducible localisation of EGFP in the cardiac jelly (Fig. 3 and Movie 2). To validate that ssNcan-EGPF was indeed indicative of HA localisation and not simply an accumulation of secreted protein, we generated a ubiquitously expressing secreted GFP transgenic line [Tg(ubi:ssEGFP)^uq7ks] and found that, not only did it have different GFP localisation from Tg(ubi:ssNcan-EGFP), but almost no GFP was detected in the cardiac jelly (Fig. 3 and Movie 2), substantiating the use of the Tg(ubi:ssNcan-EGFP) line as a biosensor for HA and the cardiac jelly.

To measure the cardiac jelly phenotype observed in Fig. 2E, we crossed double nppa/nppb carriers onto the Tg(ubi:ssNcan-EGFP) background and took confocal z-stacks of the hearts of anaesthetised live zebrafish. An obvious expansion of the fluorescent area was observed in double mutant embryos. Quantification showed a significant increase in ssNcan-EGFP area in the atrium but not ventricle of double mutants compared with siblings (Fig. 4A,B). Interestingly, this correlated with the region of AVC marker gene expansion: into the atrium but not the ventricle. Of note, this included the ECM-synthetic genes has2 and vesican. In vivo and in vitro models have previously demonstrated that Nppa and Nppb exert anti-fibrotic function by inhibiting TGF-β-mediated fibroblast transformation and the production of ECM proteins (Ellmers et al., 2007; Horio et al., 2003; Li et al., 2008; Tamura et al., 2000; Tsuruda et al., 2002; Wang et al., 2003). We show here, that Nppa and Nppa exert a similar effect on ECM production during cardiac development.

Given the observed increase in bmp4 expression in nppa/nppb double mutants and the knowledge that has2 expression is downstream of Bmp signalling (Shirai et al., 2009), we investigated whether inhibition of Bmp signalling is sufficient to rescue the has2 expansion in double mutants. We injected previously published bmp4 morpholinos (Chocron et al., 2007) into a clutch of incrossed double heterozygous carriers, grew them until 72 hpf, performed has2 in situ hybridisation, imaged and categorised embryos then genotyped them. We observed a robust expansion of has2 into the atrial chamber of double mutant embryos (as observed in Fig. 2). However, bmp4 knockdown rescued the has2 expression pattern to that of uninjected sibling controls (Fig. 4C,D).

To investigate whether the expanded cardiac jelly in double mutants was also downstream of Bmp signalling, we injected bmp4 morpholinos into nppa/nppb double mutants on the Tg(ubi:ssNcan-EGFP) background and measured cardiac jelly area. As observed for has2 gene expression, the increased area of cardiac jelly in the atrium was restored to that of siblings following bmp4 knockdown. We repeated this analysis using DMH1, a chemical inhibitor of Bmp signalling and, again, observed the cardiac jelly area restored to that of siblings (Fig. S4). Together, these data suggest that nppa/nppb disruption results in ectopic bmp4 transcription into the atrial chamber, and in a resultant increase in HA synthesis and production of cardiac jelly.

These data demonstrate that Nppa and Nppb are required to repress the AVC genetic program within the atrial chamber, providing complementarity to the activity of Tbx2b in the AVC, which represses the chamber genetic program, including nppa and nppb. This mutually exclusive repressive activity must assist in establishing a sharp boundary between these two tissue compartments. Presumably, these ligands perform this function via receptor-mediated signalling, although this remains to be tested. What is still unknown is how receptor-mediated signalling acts to repress bmp4. We show that Nppa/Nppb signalling functions to inhibit bmp4 transcription in the chambers, repressing the Bmp>Tbx2>Has2 cascade. The most noteworthy outcome of nppa/nppb deficiency is increased transcription of the extracellular matrix components versican and overproduction of HA.

In summary, these data reveal a previously unappreciated role for Nppa and Nppb in cardiac development. Zebrafish nppa/nppb double mutants develop pericardial oedema as early as 48 hpf. The functional consequence of losing Nppa/Nppb is unrestricted AVC boundary establishment, resulting in ectopic production of cardiac jelly in the atrial chamber. This role for Nppa/Nppb in restricting the Bmp-Tbx2-Has2 pathway from the chamber to the AVC, places Nppa/Nppb in a regulatory loop with Tbx2, upstream of Bmp.

All zebrafish strains were maintained and animal work performed in accordance with the guidelines of the animal ethics committee at The University of Queensland, Australia. The previously published transgenic lines used are Tg(myl7:dsRed-nls) (Mably et al., 2003); Tg(kdrl:mCherry) (Hogan et al., 2009) and Tg(fli1a:nEGFP)^y7 (Lawson and Weinstein, 2002). Generation of the Tg(ubi:ssEGFP) and Tg(ubi:ssNcan-EGFP) lines was performed by injecting 1 nl of 25 ng/μl purified plasmid DNA (constructs described by De Angelis et al., 2017) into single-cell stage embryos together with tol2 transposase mRNA (25 ng/μl). Embryos expressing EGFP mosaically were selected and raised to adulthood, and screened for germline transmission to generate stable transgenic lines.

TALEN monomers (Table S2) for nppa were generated using the golden gate kit (Cermak et al., 2011), cloned into the pCS2TAL3RR and pCS2TAL3DD backbones, and transcribed as previously described (Dahlem et al., 2012). CRISPR gDNA (Table S2) and co-injection with cas9 mRNA for targeting nppb was performed as described (Capon et al., 2017). Fish were screened by high-resolution melt analysis (HRMA) using a Viia7 Real-Time PCR System (Applied Biosystems) and F1 embryos carrying a 4 bp deletion and 10 bp insertion for nppa and nppb, respectively, were outcrossed to the wild-type strain to generate an F2 population. To generate double mutants, nppa homozygous fish were incrossed and nppb CRISPR/Cas9 RNA injected into embryos. F1 embryos identified with a 39 bp insertion for nppb were outcrossed to the wild-type strain to generate an F2 population. Primers for HRMA and sequencing of individual mutations are listed in Table S2.

In situ hybridisation analysis was carried out as previously described (Smith et al., 2011). Immunohistochemistry was performed as previously described (Smith et al., 2008). Antibody details are provided in Table S2. The splice-targeting bmp4 MO was used as previously described (Chocron et al., 2007). The standard p53 control MO from Gene Tools was used.

DNA was extracted from the trunk of 48 hpf individual embryos as previously described (Dahlem et al., 2012) and genotyped by PCR using primer sequences depicted in Table S1. Following amplification, digestion with BsrGI-HF which cuts the wild-type sequence, was performed and successful digest resolved by agarose gel electrophoresis. nppb embryos were genotyped by PCR using primer sequences from Table S1, which amplify wild-type and mutant amplicons (10 bp or 39 bp insertion) resolved by electrophoresis on a 2.5% sodium boric acid gel (Brody and Kern, 2004) at 200 V.

RNA was subsequently extracted from the remaining head/heart from a pool of four mutants or wild-type siblings in triplicate using a Direct-zol kit (Zymol Research) as per the manufacturer's guidelines. qPCR for nppa and nppb was performed on a Viia 7 system (Applied Biosystems). Efficiency corrected data was normalised to the geometric average of ef1α and rpl13 using GeNorm (Vandesompele et al., 2002) as previously described (Coxam et al., 2014).

DMH1 was dissolved in DMSO. The DMSO control and DMH1 in DMSO were diluted with E3 medium to a concentration of 1 µM. Embryos were dechorionated and incubated in either DMSO or DMH1 solutions for 24 h (from 48-72 hpf), their hearts stopped with Tricaine and live imaging performed.

Whole-mount in situ hybridisation embryos were imaged using an Olympus BX51 Upright Microscope Stand with Olympus DP70 CCD camera (10× dry objective). Confocal imaging was performed on a Zeiss Axiovert 200 Inverted Microscope Stand with LSM 710 Meta Confocal Scanner using 20× and 40× dry objectives (Zeiss Zen 2012 Black Software). High-speed movies (100 fps) were acquired using a Nikon Ti-E Inverted Stand with Hamamatsu Flash 4.0 sCMOS camera (20× dry objective - NIS Elements 4.3 Software). Fixed and live embryos for confocal imaging, as well as live embryos for movie acquisition, were mounted in low-melting point 1% agarose (Sigma-Aldrich).

Wall thickness was measured from movie acquisitions. Files were opened with the ImageJ 2.0.0 Software and individual frames analysed. The inner surface (IS) of the heart (as delimited by the endocardium) was subtracted from the myocardium-lined outer surface, in order to obtain the overall cardiac wall area. This value was normalised against the overall perimeter, calculated as the arithmetic mean of inner and outer perimeters. For each embryo, the corresponding wall thickness was obtained by averaging the measurements of three consecutive beating events. The extent of HA staining in the atrium was measured from confocal acquisitions of Tg(ubi:ssNcan-EGFP) embryos, calculating the ratio of EGFP-labelled area out of the total chamber area. For each embryo, the final value was the average of three measurements on subsequent optical sections, chosen to be at the level of the atrioventricular canal (AVC) and inflow tract. As for the quantification of HA in the ventricle (see Fig. 4A,B), relevant sections were selected to allow as clear a visualisation of the chamber inner and outer outlines as possible, preferably at the level of the AVC.

Image analysis and quantification was performed using Fiji (ImageJ 2.0.0 Software). Measurements of cell size were conducted as previously described (Auman et al., 2007). All cellular measurements were performed on three cells per condition from at least three different embryos. Movie analyses to determine ‘stroke volume’ and ‘ejection fraction’ were performed as described previously (Bagatto and Burggren, 2006). For statistical analysis, Gaussian distribution of the data and equal variance among groups was assumed although not formally tested. The unpaired, two-tailed Student's t-test or two-way ANOVA followed by Tukey's post hoc test for multiple comparisons was employed (Prism 7.0a Software), as reported in figure legends.

The bmp4 morpholino was a kind gift from Jeroen Bakkers. Confocal microscopy was performed at the Australian Cancer Research Foundation's Cancer Ultrastructure and Function Facility at The University of Queensland, Brisbane, Queensland 4072, Australia. The Alcama antibody (zn-8) was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA.