Genetic interaction between pku300 and fbn2b controls endocardial cell proliferation and valve development in zebrafish

The heart tube, at the time of its formation, has a two-layer structure consisting of an inner endothelium and an outer myocardium, connected by the extracellular matrix (Fishman and Chien, 1997). While the myocardium has been the focus of heart development during the past decades, the origin, proliferation and differentiation of the endocardium has gained less attention and so remains largely unknown. Lineage tracing studies support the suggestion that endocardial and myocardial progenitors are located near each other ventro-laterally in the zebrafish blastula (Keegan et al., 2004; Lee et al., 1994; Stainier et al., 1993). These progenitors migrate to form the cardiac field (or anterior lateral plate mesoderm) in the early somite zebrafish embryos, where the anterior Scl⁺ and Etsrp⁺ progenitors give rise to endocardial/endothelial cells and the posterior Nkx2.5⁺ mesodermal cells to myocardial progenitors (Palencia-Desai et al., 2011; Schoenebeck et al., 2007). In zebrafish, etsrp, cloche and lycat are critical for the generation of endocardial and endothelial progenitors (Palencia-Desai et al., 2011; Stainier et al., 1995; Xiong et al., 2008). Tie2 (Tek) is required for the formation of the endocardium, but dispensable for the formation of vessels in mice (Puri et al., 1999). Hedgehog signaling is required for the specification and differentiation of endocardial progenitors in zebrafish (Wong et al., 2012). Scl/tal1 is not needed for the formation of endocardial progenitors but is essential for endocardial cell migration (Bussmann et al., 2007). While VEGF-induced NFATc1 activation promotes endocardial cell proliferation for sustaining endocardial cell numbers in heart valve development (Combs and Yutzey, 2009b; Graef et al., 2001; Johnson et al., 2003; Lee et al., 2006), we have limited knowledge on how endocardial cell proliferation is regulated before and during the formation of the atrio-ventricular canal (AVC) and cardiac valves. The common cardinal vein (CCV) is on the yolk syncytial layer connecting the posterior cardinal vein (PCV) and the sinus venosus of the heart. The CCV appears ∼24 hours post-fertilization (hpf) and finishes remodeling ∼72 hpf (Isogai et al., 2001).

Endocardial differentiation is associated with remarkable morphological and molecular changes during cardiac valve development in the zebrafish, chicken, and mouse (Beis et al., 2005; Gitler et al., 2003; Hu et al., 2000; Markwald et al., 1975; Mjaatvedt et al., 1998; Person et al., 2005; Timmerman et al., 2004). In zebrafish, AVC valve development progresses through three major and continuous stages: first, AV endocardial cells extend processes and invaginate into the space between the endocardium and myocardium around three days post-fertilization (dpf) (Rutenberg et al., 2006; Scherz et al., 2008); second, a subset of AVC endocardial cells form endocardial cushions through the epithelial–mesenchymal transition (EMT) ∼6 dpf (Beis et al., 2005; Martin and Bartman, 2009); and third, endocardial cushions are subsequently remodeled into mature valve leaflets between 16 and 28 dpf (Martin and Bartman, 2009). RNA in situ analysis in mouse and zebrafish embryonic hearts reveals that several genes are preferentially expressed in the endocardial cushions/valves, including vegf, neuregulin, erbB3, NFATc1, notch1, versican, bmp2/4, tbx2b, foxn4 and pkd2 (Chi et al., 2008; de la Pompa et al., 1998; Del Monte et al., 2007; Gitler et al., 2003; Just et al., 2011; Ranger et al., 1998; Rivera-Feliciano and Tabin, 2006; Walsh and Stainier, 2001). Analyzing these signaling pathways in mouse and zebrafish mutants further supports their roles in endocardial cushion and valve formation (Armstrong and Bischoff, 2004; Chi et al., 2008; Combs and Yutzey, 2009a; Hinton and Yutzey, 2011; MacGrogan et al., 2010; Scherz et al., 2008; Smith et al., 2011; Totong et al., 2011). In the AVC of mouse embryos, myocardial Bmp2/Tgfβ2 and endocardial Notch1 activate Snail1, then suppress VE-cadherin to promote the EMT and endocardial cushion formation (for reviews. see Armstrong and Bischoff, 2004; Hinton and Yutzey, 2011; MacGrogan et al., 2010). However, the cellular and molecular events underlying this important process remain largely unknown, particularly how notch signaling is regulated in the endocardium. In this work, we investigated this important process by positional cloning of an ethylnitrosourea (ENU)-induced genetic mutant, scotch tape^te382 (sco^te382), and revealed simultaneous mutations of pku300 and fbn2b in sco^te382, and their genetic interactions when endocardial cells proliferate and differentiate into cardiac valves.

To gain insights into the molecular basis of cardiac valve development, we analyzed the zebrafish genetic mutant sco^te382, which was isolated from a large-scale ENU-induced mutagenesis of the zebrafish genome for cardiovascular mutants (Chen et al., 1996). Homozygous sco^te382 mutants were indistinguishable from their siblings before 24 hpf, but had pericardial edema (Fig. 1B), dilated heart chambers (Fig. 1D), and abnormal caudal veins in the trunk (supplementary material Fig. S3B) at 48 hpf. Histological sections revealed that sco^te382 mutants had abnormal AVC at 48 hpf (Fig. 1F), and failed to form cardiac valve leaflets at 6 dpf (Fig. 1H), compared with wild-type siblings (Fig. 1E,G). By taking high-speed movies of sco^te382 mutant hearts, we found that major blood stream followed the direction from the inflow tract to atrium, ventricle and outflow tract, but a fraction of blood cells were pushed back from ventricle to atrium, suggesting the presence of valve defects in the AVC (supplementary material Movie 1). A systematic examination revealed that ∼3% homozygous mutant embryos with milder pericardial edema and some circulation at 48 hpf survived to adulthood. Therefore, we maintained both heterozygous and homozygous adult sco^te382 mutant zebrafish for subsequent experiments.

To identify cellular and molecular defects in sco^te382 mutants, we analyzed and compared the expression patterns of endocardial and myocardial genes between sco^te382 mutants and control siblings. We found that endocardial fli1, tie1 and kdrl were almost undetectable in the atria of sco^te382 mutants (Fig. 1J,L,N) while they were normally expressed in both the atria and ventricle of control siblings at 48 hpf (Fig. 1I,K,M). The myocardial genes myl7, nkx2.5 and tbx5 were normally expressed in sco^te382 mutants (supplementary material Fig. S1). To evaluate vascular development in sco^te382 mutants, we crossed the Tg(kdrl:EGFP) transgene into wild-type and sco^te382 mutant zebrafish, and found that the overall vasculature formed in both the sco^te382 mutants and the Tg(kdrl:eGFP) siblings (supplementary material Movies 6, 7). However, the organization of endothelial cells in the caudal vein was disrupted in the sco^te382 mutants at 48 hpf (supplementary material Fig. S3B). Together, our data suggested that primary endocardial and endothelial cell defects led to abnormal cardiac valve and caudal vein development in sco^te382 mutants.

For genetic analysis, we collected ∼25% homozygous mutants with both abnormal heart and caudal vein from heterozygous sco^te382 mutant zebrafish crosses. We mapped the sco^te382 locus on linkage group 22 by bulk segregation analysis with microsatellite markers. Fine mapping determined the genetic interval between Z10673 and Z33723 by genotyping 1632 informative meioses. Further genetic and physical walking identified closer flanking genetic markers 9 (1/1632) and 6 (2/1632), as well as marker 7 (0/1632) (Fig. 2A). Therefore, the sco^te382 locus was defined between markers 9 and 6, which are covered by the contigs CR751224.20, CR394531.15, CU570881.11, CU582903.5, CU207311.10, CU929300.5 and CU076074.3 in Zv9 on the Zebrafish Genome Browser Gateway. We identified the four genes calreticulin (calr), LAG1 longevity assurance homolog 4 (trh1), 182 (ZDB-GENE-070705-222 or ENSDART00000147116) and calcium homeostasis endoplasmic reticulum protein (cherp) from this genetic interval. A recent report showed that fbn2b is mutated in the sco^te382 locus (Mellman et al., 2012), however fbn2b was located outside of this genetic interval on our genetic mapping panel (Fig. 2A).

To search for mutations in sco^te382 mutant embryos, we isolated full-length coding regions of the four genes by RT-PCR. The 182 gene is predicted as ENSDART00000147116 in the Ensembl. Isolating 182 cDNA from zebrafish embryos at 24 hpf revealed that the correct 182 mRNA transcript contained 1976 nucleotides encoding a peptide with 540 amino acids (supplementary material Fig. S7). A C946T mutation in exon 3 led to a premature stop code (TAG), generating a truncated protein with only 121 amino acids (Fig. 2B). This nonsense mutation was also confirmed by sequencing genomic DNA of sco^te382 mutants (Fig. 2C). These findings suggested that the 182 gene was responsible for the sco^te382 mutant locus. We named this 182 cDNA pku300. pku300 is located within CU570881.11 (contig 3) and fbn2b is within FP360035.6 (contig 15), where the two genes are ∼570 kb apart in Zv9 (Fig. 2A). Using InterProScan, we found that the Pku300 protein had one transmembrane domain, three extracellular immunoglobulin (IgG) domains, and a short 19-amino-acid cytoplasmic tail (Fig. 2D), suggesting that Pku300 may be a receptor or co-receptor on the cell surface. By blast searching in the human genome with the zebrafish Pku300 protein as a query, we identified several homologs of IgG-containing proteins such as CD44, the hyaluronic acid receptor. However, the sequence homology is very low between zebrafish Pku300 and human CD44, reflecting the sequence diversity in this family of IgG-domain-containing proteins across species. We found that pku300 mRNA was widely expressed, including in the heart, during zebrafish embryogenesis by in situ hybridization (Fig. 2E–G). Also, RT-PCR showed that pku300 was expressed in the heart (Fig. 2H), and was reduced in cloche^m39 mutant embryos that lack endothelial/endocardial lineages (Fig. 2I), showing that pku300 is expressed in the endocardial and endothelial cells. Endocardial expression of the fusion gene pku300-mCherry revealed a cell membrane localization (Fig. 2J), suggesting that Pku300 is a cell membrane protein.

To investigate whether both pku300 and fbn2b are responsible for the phenotypes of sco^te382 mutants, we carefully evaluated and classified the mutants into two groups by analyzing 2341 offspring from heterozygous sco^te382 mutant crosses (Fig. 3; supplementary material Table S2). Group I mutants represented 6.7% of the total offspring and had defects only in the heart, lacking most of the Tg(kdrl:EGFP) atrial endocardium (Fig. 3C; supplementary material Fig. S4C). Group II mutants represented 25.5% of the total offspring and had defects in the heart, lacking most of the Tg(kdrl:EGFP) atrial endocardium, and abnormalities in the CCV (Fig. 3B; supplementary material Fig. S4B) and the caudal vein (supplementary material Fig. S3B; supplementary material Fig. S5B). Wild-type siblings represented 67.8% of the total offspring (Fig. 3A; supplementary material Fig. S3A; supplementary material Fig. S4A; supplementary material Fig. S5A). These data suggested that more than one gene contribute to the phenotypes of sco^te382 mutants. Indeed, sequencing analyses revealed that group I mutants were all heterozygous for both genes; group II mutants were all homozygous for both genes; and wild-type siblings were 1) heterozygous for both genes, 2) heterozygous for pku300, 3) heterozygous for fbn2b, or 4) wild-type (supplementary material Table S2). Furthermore, simultaneous knockdown of fbn2b and pku300 with low-dose morpholinos led to higher percentages of morphants resembling group II mutants (supplementary material Fig. S2D), while each applied separately was less potent (supplementary material Fig. S2B,C). These data strongly suggest a genetic interaction between pku300 and fbn2b in regulating the development of the atrial endocardium.

To further investigate the roles of pku300 and fbn2b in sco^te382 mutants, we applied loss-of-function and gain-of-function analyses of the two genes in zebrafish embryos. First, we attempted to isolate pku300^+/− and fbn2b^+/− zebrafish lines from the sco^te382 mutant line by segregating the pku300 and fbn2b mutant loci. We detected either pku300^+/− or fbn2b^+/− mutant embryos from wild-type offspring of heterozygous sco^te382 crosses, suggesting that the two loci can be segregated (supplementary material Table S2). By sequencing 64 adult offspring of heterozygous sco^te382 mutant crosses, we only isolated a pair of pku300^+/− adult zebrafish, named sco^pku300. The pku300^+/− mutants lost most of the Tg(kdrl:EGFP) atrial endocardium and common cardinal vein (Fig. 3E) but had mild defects in the caudal vein (supplementary material Fig. S3E), compared with their controls (Fig. 3D; supplementary material Fig. S3D). We designed a morpholino targeting the ATG of pku300 (MO1), and a morpholino targeting the first exon splicing donor site of pku300 (MO2). Microinjections of either MO1 or MO2 into embryos phenocopied the sco^te382 mutant phenotypes in the heart (Fig. 3G). MO1 appeared to be more potent in knocking down pku300 function. Therefore we used MO1 to carry out the rest of the experiments, unless otherwise specified. pku300MO knockdown led to very few Tg(kdrl:EGFP) endocardial cells (Fig. 3G), little tie1 or fli1 expression (data not shown) in the atrial endocardium, an abnormal CCV (Fig. 3G), and mild defects in the caudal vein (supplementary material Fig. S3G). Consistent with the previous report (Mellman et al., 2012), we found that fbn2bMO morphants lost most of the Tg(kdrl:EGFP) atrial endocardium but had no defect in the common cardinal vein (Fig. 3I) and defect in the caudal vein (supplementary material Fig. S3I), compared with their controls (Fig. 3H; supplementary material Fig. S3H). These data suggest critical roles of pku300 in the development of the atrial endocardium, the CCV, and the caudal vein, as well as fbn2b in the development of the atrial endocardium and caudal vein.

We then asked whether overexpression of pku300 can rescue the mutant phenotypes of sco^pku300 or sco^te382. Microinjection of pku300 mRNA rescued the Tg(kdrl:EGFP) atrial endocardium and common cardinal vein of a pku300^−/− mutant (Fig. 4C), compared with an uninjected mutant (Fig. 3E). Nine pku300^−/− mutants out of 182 offspring from heterozygous sco^pku300 crosses had little or no rescue (Fig. 4B), showing that overexpression of pku300 mRNA had rescued most of the pku300^−/− mutants. The genotypes of pku300 mutants and siblings were confirmed (data not shown). On the other hand, we failed to rescue mutant phenotypes of sco^te382 by overexpression of pku300 mRNA (data not shown). To explore pku300 transgenic rescue, we generated Tg(kdrl:pku300) transgenic zebrafish where pku300 was overexpressed in the endocardium and endothelial cells. We found that the Tg(kdrl:pku300) transgene rescued endocardial Tg(kdrl:EGFP) expression in the atrium of 24 out of 89 (27%) Tg(kdrl:pku300);sco^te382/te382 mutant embryos (Fig. 4F,G), while there was no rescue in the sco^te382/te382 mutant (Fig. 4E). We also generated Tg(myl7:pku300) transgenic zebrafish that overexpressed pku300 in the myocardium, but none of 33 Tg(myl7:pku300);sco^te382/te382 mutant embryos showed Tg(kdrl:EGFP) expression in the atrium. They did not show additional defects (data not shown). The genotypes of sco^te382/te382 mutants and siblings as well as transgenic Tg(kdrl:pku300) and Tg(myl7:pku300) were confirmed (data not shown). These results demonstrate that pku300 is responsible for sco^te382 and plays a role in the development of the endocardium, common cardinal vein and caudal vein.

Previous studies reported that endocardial cells are specified in the anterior lateral plate mesoderm in early somite embryos, then migrate and fuse in the midline ∼18 hpf, and turn leftward at ∼20 hpf (Schoenebeck et al., 2007; Bussmann et al., 2007; Wong et al., 2012). We did not see any abnormality, including formation and patterning of the endocardium within the heart tube in sco^te382 mutants at 24 hpf. Consistently, Tg (kdrl: EGFP) endocardial cells migrated properly to the heart tube in the midline and then turned toward the left side in sco^te382 mutants (supplementary material Movie 3), as in wild-type embryos (supplementary material Movie 2). Live imaging suggested that increased Tg(kdrl:EGFP) endocardial cells came from cell proliferation and migration along the looped heart tube. Those Tg(kdrl:EGFP) endocardial cells coming from outside the heart tube mostly contributed to the sinus venosus (supplementary material Movies 2, 4). We did not detect endocardial cell apoptosis in sco^te382 mutant and wild-type embryos by TUNEL assays (data not shown). These findings suggest that the reduced numbers of atrial endocardial cells were not due to defects in endocardial cell migration or apoptosis in sco^te382 mutants. Therefore we hypothesized that the reduced numbers of atrial endocardium of 48 hpf mutant embryos result from defects of endocardial cell proliferation. By using Tg(fli1a:nuEGFP)^y7 transgenic labeling (Roman et al., 2002), we revealed that endocardial cells failed to divide in sco^te382 mutants from 29 to 36 hpf (supplementary material Movie 5), compared with an average of 12 dividing endocardial cells (13.3% proliferation rate) in the heart tube of wild-type embryos (supplementary material Movie 4). A set of representative images, from one of the three experiments, showed that there was one dividing cell between 29 hours 30 minutes (Fig. 5A) and 29 hours 40 minutes (Fig. 5B), and three dividing cells between 29 hours 40 minutes and 29 hours 50 minutes (Fig. 5C) in wild-type embryos. In contrast, we did not see any dividing cells in sco^te382 mutants at the same time points (Fig. 5D–F). The ratio of proliferating to total endocardial cells was much smaller in sco^te382 mutants than in wild-type siblings (Fig. 5G). Importantly, we found that the numbers of endocardial cells in the atrium and the whole heart were reduced, while they were slightly increased in the ventricle, in sco^te382 mutants compared with wild-type siblings at 48 hpf (Fig. 5H–J). In contrast, the numbers of myocardial cells in the ventricle and atrium were slightly increased in sco^te382 mutants compared with wild-type siblings at 48 hpf (Fig. 5K–M). By using BrdU labeling or immunostaining with anti-pH3 we also found that endocardial cells gradually failed to enter mitotic phases in sco^te382 mutant hearts from 23 to 33 hpf (supplementary material Fig. S6E,F). Representative images showed that there were more dividing endocardial cells in wild-type (supplementary material Fig. S6A,C) than in sco^te382 mutant (supplementary material Fig. S6B,D) embryos at 25 hpf and 30 hpf. Interestingly, we did not find defects in vascular endothelial cell proliferation in the trunk of sco^te382 mutants (supplementary material Movie 7), compared with wild-type embryos (supplementary material Movie 6) from 23 to 33 hpf. Together, these results suggest that pku300 and fbn2b are required for endocardial cell proliferation during heart looping and chamber formation.

The presence of cardiac valve defects in sco^te382 mutants urged us to address the underlying molecular mechanisms. A previous report showed that Notch1b and bmp4 are restricted in the respective endocardium or myocardium of the AVC during heart valve development in zebrafish (Walsh and Stainier, 2001). Endocardial notch 1b and its ligand dll4 were weakly expressed in the ventricle at 48 hpf and gradually restricted within the AVC from 48 to 72 hpf (Fig. 6A–C,M,N; data not shown), however they were ectopically expressed in the ventricle of sco^te382 mutant hearts (Fig. 6D–F,O,P). In contrast, myocardial bmp4 were normally expressed in the inflow tract, AVC and outflow tract in both wild-type (Fig. 6G–I) and sco^te382 (Fig. 6J–L) embryos from 48 to 72 hpf. To assess whether the endocardial–mesenchymal transition (EMT) is affected in sco^te382 mutants, we examined the expression of secreted phosphoprotein 1 (spp1), a previously known factor that is upregulated in the AVC endocardial cushion in zebrafish (Just et al., 2011; Peal et al., 2009). We found that spp1 was not expressed in the AVC (red arrowhead) while was reduced in the outflow tract (black arrowhead) of a sco^te382 mutant (Fig. 6R), compared with those in a wild-type embryo (Fig. 6Q), suggesting that EMT was also affected in sco^te382 mutants. Together, these data support a notion that the compromised expression of AVC endocardial and valve genes led to abnormal development of the endocardial cushion and cardiac valve in sco^te382 mutants.

We then asked whether endocardial notch signaling was accordingly changed in the AVC of sco^te382 mutants. The Notch reporter zebrafish line Tg(Tp1:mCherry) is generated in which mCherry is placed downstream to 12 repeated RBP-Jk sites (Parsons et al., 2009). To determine Notch signaling in sco^te382 mutant hearts, we crossed the Notch reporter transgene into sco^te382 mutant zebrafish to make sco^te382/te382; Tg (kdrl:eGFP); Tg (Tp1:mCherry) double transgenic mutant zebrafish. In wild type Tg (kdrl: eGFP) transgenic embryos, we found endocardial Notch signaling was initially activated in the endocardium of the AVC and ventricle at 48 hpf (Fig. 7A), and gradually concentrated in the AVC at 60 hpf and 72 hpf (Fig. 7C,E). However, Notch signaling was not activated in the endocardium of the AVC of sco^te382 mutant hearts (Fig. 7B,D,F) while was ectopically activated in the ventricles of sco^te382 mutant hearts at 60 and 72 hpf (Fig. 7D,F), which is consistent with dll4 and notch1b ectopic expression in mutant ventricles (Fig. 6D–F,O,P). In particular, Tg(Tp1:mCherry) positive cells did not overlap with the endocardial cells (Fig. 7D,F), suggesting that the Notch signaling might be ectopically activated in the myocardium in sco^te382 mutants. These data support the idea that pku300 and fbn2b are required for the endocardial Notch signaling during cardiac cushion and valve development in zebrafish.

During the first three days of development, endocardial invagination is formed as a temporary structure in the AVC to secure directional blood flow in the heart, and EMT might take place at larval stages after 6 dpf (Martin and Bartman, 2009; Scherz et al., 2008). However, it is unclear how the endocardial invagination is related to the cardiac valve during heart development, and what are the underlying cellular and molecular mechanisms engaging in cardiac valve development during the first three days. EMT is normally involved in cell adhesion, cell junctions and cell polarity (Combs and Yutzey, 2009a). From the previous studies (Beis et al., 2005; Just et al., 2011; Peal et al., 2009) and our data (Fig. 6Q), endocardial cushion and the EMT might take place during the first three days of development in zebrafish. We found endocardial cell adhesion gene cdh5 (VE-cadherin) was expressed in the whole heart at 48 hpf, downregulated and restricted in the AVC of wild type hearts from 60 to 72 hpf (Fig. 8A,C,E), while it remained ectopically expressed in the endocardium of sco^te382 mutant ventricles from 48 to 72 hpf (Fig. 8B,D,F). To address morphological changes of cell junctions, we assessed ZO1 expression during the first three days of development by immunostaining. We found that ZO1 was normally expressed around the AV endocardial cells at 48 hpf (n = 13) and was gradually reduced at 60 hpf (n = 12) and 72 hpf (n = 15); ZO1 was not expressed in the basal part of the endocardium of wild type embryos at 60 hpf and was not expressed in both basal and epical parts of the endocardium of wild type embryos at 72 hpf (Fig. 8G,I,K). However, in sco^te382 mutant embryos at 48 hpf (n = 18), ZO1 was slightly upregulated; at 60 hpf (n = 7) and 72 hpf (n = 11), ZO1 was abnormally expressed around endocardial cells in the AV canal, and was particularly upregulated in the basal side of endocardial cells in the AVC (Fig. 8J,L). These data suggest that abnormal endocardial cell adhesion and tight junction likely interfered endocardial EMT or morphological changes that are required for heart valve formation in sco^te382 mutant hearts.

Our major findings in this work were (1) positional cloning of sco^te382 as a novel gene, pku300, that encodes a putative transmembrane protein, and mutations of both pku300 and fbn2b contribute to the defects in the atrial endocardium and caudal vein in sco^te382 mutants; (2) deciphering pku300 as a novel player in endocardial cell proliferation; and 3) demonstrating the critical roles of pku300 and fbn2b in regulating endocardial Notch signaling, endocardial cell adhesion and tight junctions during valve morphogenesis. Our data support the hypothesis that endocardial cell proliferation is critical for correct lining of the endocardium during heart tube expansion and chamber formation. Also we demonstrated that dramatic morphological changes and possible endocardial EMT in the AVC take place in the first three days of heart development in zebrafish. Both pku300 and fbn2b play essential roles in this important process in the AV endocardial cells, partly by regulating the temporo-spatial patterning of endocardial Notch signaling.

By positional cloning, we clearly demonstrated that sco^te382 encoded a novel putative transmembrane protein by revealing a nonsense mutation in sco^te382 mutants, phenocopying sco^te382 mutant phenotypes by antisense pku300 morpholino knockdown, rescuing sco^te382 mutant phenotypes by overexpressing the wild-type pku300 gene in the endothelium and endocardium, and determining pku300 mRNA expression in the endocardium. As reported recently (Mellman et al., 2012), we also found the G3935T missense mutation of fbn2b in our sco^te382 mutant. Mutational analyses of fbn2b and pku300 allowed us to further classify two groups of mutants in sco^te382, where group I mutants were heterozygous for both genes and group II mutants were homozygous for both genes (supplementary material Table S2; Fig. 3A–C). But only ∼13% of embryos that were heterozygous for both genes showed group I mutant phenotypes. Together with the work by Mellman et al., our data support the conclusion that simultaneous mutations of both fbn2b and pku300, which were ∼570 kb apart in Chromosome 22, contributed to the defects in sco^te382 mutants. This very rare event, presumably induced by ENU, provides an important lesson that the possibility of more than one causative gene mutation should be explored if the mutant phenotypes from a particular locus are not identical. Furthermore, the genetic interaction between fbn2b and pku300 will guide future studies on how the two proteins interact in regulating endocardial cell proliferation.

The Pku300 protein was predicted to contain three extracellular IgG domains, a single transmembrane domain, and a short intracellular tail without any conserved kinase motifs according to InterProScan (http://www.ebi.ac.uk/interpro/scan.html), which is similar to a large class of cell adhesion molecules and receptors, including CD44, one of the hyaluronan receptors (Bechara et al., 2007). Blast searching with human CD44 protein as bait did not identify any potential homologs in the zebrafish genome. IgG-domain-containing proteins are generally sequence diversified. Our current hypothesis is that Pku300 functions as a hyaluronic acid (HA) receptor that participates in cardiac jelly formation and then modulates Notch, Fbn2b and other signaling pathways during heart valve development. However, true Pku300 protein homologs in other vertebrates remain to be identified.

Cardiac jelly is the extracellular matrix between endocardium and myocardium, and its major components are glycosaminoglycans (GAGs) and proteoglycans. GAGs are long-chain sulfated sugar polymers of three subtypes: chondroitin sulfate (CS), heparan sulfate (HS) and HA. The generation and breakdown of cardiac jelly are precisely controlled, as they are essential for proper heart development. Mice lacking hyaluronan synthase-2, the enzyme required for the production of mucopolysaccharide hyaluronan, do not develop trabeculae and have significantly reduced cardiac jelly (Camenisch et al., 2000; Mjaatvedt et al., 1998). This is mainly due to the failure of the cardiac endothelium to undergo EMT with inactivation of ERBB2/ERBB3 and RAS signaling (Camenisch et al., 2002). Hyaluronan binds to either two major receptors (CD44 and RHAMM) or other receptors at the cell surface (Toole, 2004). In zebrafish, ENU-induced mutagenesis screens have led to the identification of many valve mutants, of which two are known to have cardiac jelly defects: scotch tape and jekyll (Chen et al., 1996; Walsh and Stainier, 2001). Jekyll is a null mutation of UDP-glucose dehydrogenase, which is required for the synthesis of HS, CS, and HA (Walsh and Stainier, 2001). Jekyll mutants indeed show abnormal expression of several valve markers at 48 hpf, suggesting that cardiac jelly is involved in heart valve development at very early embryonic stages. As reported previously (Chen et al., 1996), we found cardiac jelly was reduced in sco^te382 mutant hearts, however, CS, a component of cardiac jelly, was normally expressed in these mutants (data not shown). The other two components, HS and HA, need to be examined in sco^te382 mutant hearts when the required assay reagents become available. Identifying the roles of the pku300 and fbn2b genes here will allow further study of the molecular basis of cardiac jelly defects and their relationship to heart valve development.

Our data suggest that both pku300 and fbn2b are required for endocardial cell proliferation but are dispensable for vascular endothelial cell proliferation in the trunk. To our knowledge, this is one of the very few reports on identifying regulators in endocardial cell proliferation. Previous studies have shown that myocardial Nfat2/3/4 represses Vegf expression in the AVC that allows local EMT, and that later endocardial Calcineurin-Nfatc1 promotes valve elongation and remodeling in mice (Chang et al., 2004). Early studies suggested that VEGF-Nfatc1 is essential for endocardial cell proliferation (Lee et al., 2006; ,Combs and Yutzey, 2009b). On the other hand, others reported that VEGF is dispensable for endocardial cell proliferation, cell death and Nfatc1 nuclear localization in the endocardial cells of elongating mitral valves (Stankunas et al., 2010). Therefore, the underlying signaling pathways that regulate endocardial cell proliferation remain largely unknown. Identifying pku300 and fbn2b as essential genes for endocardial cell proliferation will likely help to further delineate related molecular mechanisms in the future.

As in jekyll mutants, cardiac valve genes were abnormally expressed in the AVC of sco^te382 mutants. Previously, Hey1 and Hey2 were shown to suppress Bmp2/4 and then Tbx2 expression in the heart chambers, so Bmp2/4 and Tbx2 are restricted to the myocardium of the AVC where Hey1 and Hey2 are not expressed (Fischer et al., 2007; Kokubo et al., 2007; Rutenberg et al., 2006). On the other hand, it remains unknown how endocardial notch1 is regulated in the AVC. Here, we found that endocardial Notch signaling was not activated and dll4/notch1b were not properly expressed in the AVC of sco^te382 mutant hearts, suggesting that pku300 and fbn2b modulate Notch signaling during endocardial cell proliferation and valve development. Ectopic Notch activity, particularly in the ventricle of sco^te382, might promote ventricular myocyte proliferation, which is consistent with the finding of more myocytes in the mutant ventricle (Fig. 5K–M) and Notch activation of cell cycle entry and proliferation in mammalian cardiac myocytes (Campa et al., 2008). Previous studies suggest that the TGFβ/BMP and Notch signaling pathways are essential for endocardial EMT and valve development (Combs and Yutzey, 2009a; MacGrogan et al., 2010; Rivera-Feliciano and Tabin, 2006; Timmerman et al., 2004). In addition, overexpression of a constitutively active Notch intracellular domain in zebrafish embryos increases endocardial cell proliferation (Timmerman et al., 2004). The lack of endocardial Notch signaling and the endocardial EMT marker spp1 in the AVC substantiates the roles of fbn2b and pku300 function in valve development. Therefore, our work provides another important entry point to determine endocardial signaling in cardiac valve morphogenesis.

sco^te382 is a cardiac valve mutant isolated from an ENU-induced mutagenesis screen in Tubingen (Chen et al., 1996). We acquired this mutant fish from Mark Fishman's laboratory at Massachusetts General Hospital, Boston. Both heterozygous and homozygous sco^te382 fish were maintained. Transgenic line Tg(kdrl:EGFP) that EGFP was driven by the zebrafish kdrl promoter (Choi et al., 2007; Cross et al., 2003) was crossed with heterozygous sco^te382/+ fish. Individual fish of Tg(kdrl:EGFP)/+, sco^te382/+ and Tg(kdrl:EGFP);sco^te382/+ were identified. Heterozygous sco^pku300/+ was derived from offspring of heterozygous sco^te382/+ crosses. Tg(myl7:EGFP) and Tg(myl7:nuDsRed) lines were from Geoffrey Burns at Massachusetts General Hospital (Burns et al., 2005). Tg(fli1:nuEGFP) was provided by Feng Liu at Institute of Zoology, Chinese Academy of Sciences (Roman et al., 2002). The Notch reporter transgenic line Tg(Tp1:mCherry) was obtained from Dr Michael Parsons, Johns Hopkins University (Parsons et al., 2009). Tg(kdrl:pku300-mCherry), Tg(kdrl:pku300) and Tg(myl7:pku300) transgenic zebrafish lines were generated by using Tol2-based transgenesis as described (Kawakami et al., 2004). Individual fish of Tg(kdrl:pku300)/+;sco^te382/te382 and Tg(myl7:pku300);sco^te382/+ were isolated from crosses between sco^te382/te382 and transgenic zebrafish. Zebrafish were raised and handled in accordance with the Peking University and Massachusetts General Hospital institutional guidelines.

About 25% siblings from heterozygous sco^te382 mapping zebrafish were mutants with defects in both the heart and caudal vein, which were sorted for generating our mapping DNA panels. The sco^te382 locus was mapped between the microsatellite markers Z10673 and Z33723 on chromosome 22 using bulk segregation analysis. Using sequence length polymorphism (SLP) and single nucleotide polymorphism (SNP) markers (supplementary material Table S1), we narrowed down the scotch tape genetic interval into about 0.18 cM between markers 6 and 9. SLP markers were assayed by agarose gel electrophoresis of PCR products. SNP markers were analyzed by sequencing PCR products. Marker 6 was in the 4th intron of pku300. The sco^te382 genetic interval was covered and sequenced in the assembled zebrafish genome sequence by the Sanger Institute, UK, where four genes were determined, including calr, trh1, pku300 and cherp genes. fbn2b was ∼570 kb away from pku300.

Two antisense morpholinos were synthesized to target the zebrafish pku300 ATG site (pku300MO1): AAATCATACCTCAGCGAAGCTCACA; the first splicing donor site (pku300MO2): GTGTTGTAGCACCTGTAATTTTCA; fbn2b start site targeted morpholino (fbn2bMO): GCGACTCCTGAAGCGCCGGTAAATG (Mellman et al., 2012); and a control morpholino (CtrMO): GCAGCGGGCACTGCTGGTGGAAGT (Gene Tools, LLC). The full length of wild type and mutant pku300 cDNAs were cloned into T7TS vector and linearized, and capped mRNA was synthesized using Ambion mMESSAGE mMACHINE mRNA transcription synthesis kits. Morpholinos or mRNA were injected into 1- or 2-cell-stage wild-type or sco^te382 or sco^pku300 mutant embryos. Injected embryos were incubated at 28.5°C for examination of phenotypes or were fixed for in situ hybridization and immunostaining as described (Xiong et al., 2008). Photographs were taken using a Leica MZ16 fluorescence microscope.

In situ hybridization was done using antisense myl7, tbx5, nkx2.5, ve-cadherin, fli1, tie1, notch 1b, dll4, versican, bmp4, spp1, pku300 and tbx2b RNA probes. Histology was done by JB4 section according to a protocol by the manufacturer (Polysciences, Inc., Warrington, PA, USA).

Wild type TL, sco^te382 and cloche^m39 mutant embryos were collected. Hearts from Tg(myl7:eGFP) embryos at 48 hpf were isolated under the florescence microscope. Total RNA was isolated from embryos and hearts using Trizol reagents (Invitrogen). First-strand cDNA was synthesized using the Superscript II RT system (Invitrogen), followed by a standard PCR.

For endocardial cell labeling, one-cell Tg(fli1:nuEGFP) or Tg(kdrl:EGFP) transgenic embryos were microinjected with 0.96 ng of tnnt2aMO to stop heart beating. For live imaging, embryos were embedded in 1.5% low-melting-temperature agarose with 0.16 mg/ml tricaine (Sigma) in E3 water. When the agarose cooled completely, extra agarose were removed leaving the embryo head facing E3 water with tricaine. The embryos were imaged with Zeiss 700 confocal microscope, following up to 10 hours at 5–6 minutes time intervals using a 20× water immersion objective. Z-stack movies were assembled in 3D time lapse using Imaris7.0 (Bitplane).

BrdU labeling for DNA synthesis and phosphorylation of histone 3 for cell cycles were carried out to determine endocardial cell proliferation. For BrdU pulse chasing, sco^te382 mutants and siblings were incubated with 10 mM BrdU for 10 hours, and then subjected to immunostaining with anti-BrdU (1∶500, Sigma) as described (Laguerre et al., 2005). For detecting phosphorylation of histone 3, staged sco^te382 mutants and siblings were fixed and subjected to immunostaining with anti-pH3 (1∶500, Millipore).

For anti-ZO1 antibody staining, all embryos were put into 0.12% tricaine (Sigma, St. Louis, USA) to stop and relax the heart before fixation. Embryos were fixed in 4% paraformaldehyde (PFA) at room temperature for 2 hours, embedded in 3% low melting agarose gel and sectioned into 100 µm slices with a vibratome (Leica VT 1,000s, Bannockburn, USA). Embryos were blocked with a blocking solution [10% normal goat serum (NGS) plus 2% blocking reagent (BR) (Roche Diagnostics, Indianapolis, USA) in maleic acid buffer (MAB) (150 mM maleic acid, 100 mM NaCl, pH 7.5)]. Immunostaining with anti-ZO1 (Invitrogen, Eugene, USA) (1∶100) antibody was performed in an incubation solution (2% NGS plus 2% BR in MAB) at 4°C overnight. Goat anti-rabbit IgG(H⁺L) (Vector Laboratories, Burlingame, USA) (1∶2000) was used as secondary antibody. Images were taken under the confocal microscope (Zeiss LSM 5 Pascal or Zeiss LSM510).

We acknowledge Dr Jinrong Peng and members of Dr Xiong's laboratory for comments on this manuscript and Dr Iain Bruce for the English editing.