The transcription factor Foxc1a in zebrafish directly regulates expression of nkx2.5, encoding a transcriptional regulator of cardiac progenitor cells

Cardiogenesis is a tightly controlled biological process required for formation of a functional heart. The transcription factor Foxc1 not only plays a crucial role in outflow tract development in mice, but is also involved in cardiac structure formation and normal function in humans. However, the molecular mechanisms by which Foxc1 controls cardiac development remain poorly understood. Previously, we reported that zebrafish embryos deficient in foxc1a, an ortholog of mammalian Foxc1, display pericardial edemas and die 9–10 days postfertilization. To further investigate Foxc1a's role in zebrafish cardiogenesis and identify its downstream target genes during early heart development, we comprehensively analyzed the cardiovascular phenotype of foxc1a-null zebrafish embryos. Our results confirmed that foxc1a-null mutants exhibit disrupted cardiac morphology, structure, and function. Performing transcriptome analysis on the foxc1a mutants, we found that the expression of the cardiac progenitor marker gene nkx2.5 was significantly decreased, but the expression of germ layer–patterning genes was unaffected. Dual-fluorescence in situ hybridization assays revealed that foxc1a and nkx2.5 are co-expressed in the anterior lateral plate mesoderm at the somite stage. Chromatin immunoprecipitation and promoter truncation assays disclosed that Foxc1a regulates nkx2.5 expression via direct binding to two noncanonical binding sites in the proximal nkx2.5 promoter. Moreover, functional rescue experiments revealed that developmental stage–specific nkx2.5 overexpression partially rescues the cardiac defects of the foxc1a-null embryos. Taken together, our results indicate that during zebrafish cardiogenesis, Foxc1a is active directly upstream of nkx2.5.

Cardiogenesis is a complex process consisting of a series cellular specification, differentiation, and morphogenesis. Similar to mammalian heart development, zebrafish cardiac progenitors emerge earliest from late blastula stage (1). Along with cardiac progenitor cell migration, differentiation, and proliferation, two-chambered zebrafish heart starts to work from 24 hpf 2 . Because they have rapid early heart development and can survive for almost 1 week without a functioning cardiovascular system, zebrafish have become an increasingly popular model for cardiovascular research.
Foxc1 is a member of the FOX transcription factors with a conserved forkhead domain (2). It not only plays important roles in vertebrate heart development, but also is involved in cardiac structure formation and normal function in humans (3). Mice carrying Foxc1 null alleles die at birth with hydrocephalus, skeletal abnormalities, and heart defects. Mouse Foxc1 and Foxc2 function redundantly in regulating the expression of Tbx1 in out-flow tract morphogenesis (4 -6). Zebrafish foxc1aknockout embryos display severe heart defects (7). In humans, some patients carrying the FOXC1 single allele mutation suffer from different kinds of cardiac anomalies, including mild dysplasia of the left ventricle, OFT, valvula tricuspidalis, and heart failure (8 -10). Recently, in the induction of human ES cells into cardiomyocytes, FOXC1 is essential for the differentiation by regulating the expression of MYH7 (11). Analysis of RNA profiles from human failing and non-failing heart suggests specific roles of FOX transcription factors (FOXC1, C2, P1, P4, and O1A) in modulating the human heart failure pathogenesis by regulating the expressions of key factors, such as MEF2, NKX, NF-AT, and GATA (12).
Although Foxc1 plays essential roles in cardiogenesis and cardiac pathology, little is known about the molecular mechanisms underlying its roles (5,13,14). Nkx2-5 is one of the most pivotal regulators during vertebrate cardiac progenitor cell (CPC) specification and cardiomyocyte differentiation. Nkx2-5 knockout mice die at embryonic day 10.5 with only a single ventricle and defects of OFT (15). Postnatal mice with conditional knockout of Nkx2-5 exhibit a disturbed heart conduction system (16,17). Mutation of nkx2.5 in zebrafish disrupts the cardiac morphogenesis, exhibiting a small, constricted ventricle and a dilated atrium (18). Zebrafish embryos with double null genes of nkx2.5 and nkx2.7 almost do not form the ventricle (19). Lots of NKX2-5 mutations have been discovered in human congenital heart disease patients (20 -22). This work was supported by the National Natural Science Foundation of China Grants 31471355 and 31671518 and Graduate Student Research and Innovation Program of Jiangsu Province Grant KYZZ15_0041. The authors declare that they have no conflicts of interest with the contents of this article. This article contains Figs. S1-S6, Tables S1-S4, and Movies S1 and S2. 1 To whom correspondence should be addressed. Tel.: 86-25-58641527; Fax: 86-25-58641500; E-mail: qingshun@nju.edu.cn.
The dynamic expression of NKX2-5 is known to be tightly controlled in vertebrate heart. In addition to signaling pathways, such as retinoic acid and WNT, that act upstream of Nkx2.5 to regulate CPC specification (23)(24)(25), transcription factors like FOXP1 and the post-transcriptional regulation mechanism are also involved in its expression regulation (26,27). In this study, by comprehensively characterizing the disrupted heart structure and function in foxc1a-null zebrafish embryos, we demonstrate that Foxc1a controls zebrafish cardiogenesis by directly regulating nkx2.5 expression.

Zebrafish foxc1a-null embryos exhibit severe cardiovascular defects
Previously, we reported that foxc1a-null zebrafish embryos exhibit abnormal somitogenesis and heart development (7). Morphologically, the mutant embryos could be distinguished from their wild-type siblings earliest from 50 hpf due to their obvious pericardium edema (Fig. 1, A and B). Additionally, foxc1a mutants displayed shorter body length and smaller eyes at 72 hpf (Fig. S1). When reaching 108 hpf, the mutants exhibited more serious edema than their wild-type siblings ( Fig. 1, C and D). The edema index (EI; defining the extent of pericardium edema) (28) of foxc1a mutant embryos was significantly (p Ͻ 0.0001) higher than that of their wild-type siblings at 108 hpf (Fig. 1G). Histologically, the mutant hearts displayed thinner myocardium layers, shorter OFT (Fig. 1, E and F), and defective primitive valve leaflets in which the endothelial cells only accumulated in the atrio-ventricular canal (AVC) rather than forming leaflets and extending into the ventricle (Fig. 1, E, F, EЈ, and FЈ).
To evaluate whether the heart functions were affected in foxc1a mutants, we first examined the heart rates at different developmental stages. We found that the heart rates of mutant embryos were comparable with wild-type siblings at 50 hpf (p ϭ 0.7714) and 108 hpf (p ϭ 0.4358), respectively (Fig. 1H). However, the heart rates of mutant embryos at 132 hpf were decreased dramatically (p ϭ 0.0089) compared with their wildtype siblings (Fig. 1H). The results indicated that the reduction of heart rate was the consequence of heart defects but not the reason for cardiac malformation in foxc1a mutants.
We then analyzed the ventricle minor axis shortening fraction (SF) to validate the ventricle contraction ability of zebrafish embryos. The results showed that the ventricular SFs of foxc1a mutant embryos were significantly lower than that of their wild-type siblings at both 50 hpf (p ϭ 0.001) and 108 hpf (p ϭ 0.0071), respectively (Fig. 1I). These data elucidated that the heart contraction function was severely disrupted in foxc1anull zebrafish embryos from 50 hpf. Taken together, our data demonstrate that Foxc1a deficiency resulted in heart defects, including both disorganized structures and disrupted functions in zebrafish embryos.

Heart chamber morphogenesis and AVC specification were disrupted in foxc1a mutant zebrafish embryos
To explore when the heart malformation occurs in foxc1anull zebrafish embryos, we observed the heart morphogenesis in the embryos of Tg(myl7:eGFP) m225 ;foxc1a nju19 at 50 hpf. To rule out the size difference in different heart contract phases, we examined the heart sizes of the 50-hpf embryos at both ventricular diastole (VD) and systole (VS) phases. The results showed that both atrium and ventricle of foxc1a mutant embryos could form, but the size of the ventricle was smaller and that of the atrium was larger than in their wild-type siblings, respectively (Fig. 2, A, B, AЈ, and BЈ). Measuring the crosssectional areas of heart chambers at VD phase, we found that the ventricle area of the mutant embryos was significantly smaller (p Ͻ 0.0001), whereas the atrium area was significantly larger (p ϭ 0.0015) than in their wild-type siblings (Fig. 2D). Similar results were found at VS phase (Fig. 2C). On the other hand, we examined the expressions of the molecular markers for the formation of cardiac chambers in the embryos at 30 and 50 hpf, respectively. The results showed that the chamber-specific myosin heavy chain marker genes, including amhc (marker for atrial chamber) and vmhc (marker for ventricular chamber), were all expressed in foxc1a mutant embryos at both stages (Fig.  2E). However, the amhc expression area was larger, whereas the vmhc expression area was smaller in the mutant embryos than in their wild-type siblings (Fig. 2E). The results are consistent with the morphologic observation in the embryos of Tg(myl7: eGFP) m225 ; foxc1a nju19 transgenic zebrafish (Fig. 2, A-D, AЈ, and BЈ), suggesting that Foxc1a is essential for regulating the size of two cardiac chambers and that the myocardium malformation should emerge from the earlier period of chamber differentiation no later than 30 hpf.
Hemodynamic changes, such as the malformation of vasculogenesis and blood flow, are known to affect heart development (29,30). We therefore first investigated whether vasculogenesis and angiogenesis were normal in foxc1a mutants by observing endothelial cells in the embryos of Tg(kdrl:eGFP) s843 ; foxc1a nju19 . The results showed that the mutant embryos exhibited normal trunk vasculogenesis but with only an angiogenesis defect in the head at both 30 hpf (Fig. S2, A, B, AЈ, and BЈ) and 50 hpf ( Fig. S2, C, D, CЈ, and DЈ). They suggested that zebrafish foxc1a is critical for angiogenesis in the head region but dispensable in the trunk. Unlike the abnormal expression patterns of the marker genes for myocardium occurring in the foxc1a-null embryos at both 30 and 50 hpf (Fig. 2E), the expression pattern of kdrl (marker for endocardium) was similar to those of their wild-type siblings at 30 hpf (Fig. S2, A, B, AЈ, and BЈ). However, its expression level was significantly decreased in the mutant embryos at 50 hpf ( Fig. S2, C, D, CЈ, and DЈ).
It is reported that there is visible blood in Foxc1/2 mutant mice (4). Similarly, our study also found that blood flow could be observed in foxc1a mutant zebrafish embryos at 50 hpf (Movies S1 and S2). However, hemoglobin marker gene hbae1 expression was significantly decreased in foxc1a-null embryos at both 30 and 50 hpf (Fig. S3, C-F). o-Dianisidine staining also showed that the erythrogenesis was affected in foxc1a mutant embryos at 50 hpf (Fig. S3, A, B, AЈ, and BЈ). We next examined the possible changes of blood flow shear force by checking the expression of klf2a, a marker gene expressed in the AVC region, in the mutant embryos at 50 hpf (29). We found that both mutant embryos and their wild-type siblings exhibited similar expressions of klf2a in AVC (Fig. 2G). These results indicated that the angiogenesis defects in the head and the abnormal

Roles of Foxc1a in zebrafish cardiogenesis
erythrogenesis were not the cause of heart malformation in foxc1a mutant embryos.
AVC is a specific heart region that connects the atrium and ventricle and gives rise to primitive leaflets that prevent the blood flow from retrograding (31). It is specified at 37 hpf shortly after the linear heart tube of zebrafish embryo starts to loop from 36 hpf (32). Due to the defective valve leaflets in foxc1a-null embryos, we examined whether AVC was normally specified in the mutant embryos. To do this, we first examined the expression of myl7, a marker gene for heart looping, in the mutant embryos. The results showed that the fox1a mutants exhibited an abnormal expression pattern of myl7 at 50 hpf, although its expression was normal at 30 hpf (Fig. 2E). Consistently, the chamber-specific gene nppa was expressed ectopically in the AVC of foxc1a-null embryos at 50 hpf (Fig. 2F). The results suggest that AVC malformation occurred in the mutant embryos. To confirm this conclusion, we examined the expression patterns of AVC-specific markers in the foxc1a mutant embryos and wild-type siblings at 50 hpf. The results revealed that the expressions of myocardium AVC-specific genes bmp4

Roles of Foxc1a in zebrafish cardiogenesis
and tbx2b (Fig. 2F) and endocardium AVC-specific genes notch1b and has2 (Fig. 2G) were all significantly reduced in foxc1a mutants. However, the expression of vcana was ectopically expanded into the ventricular chamber ( Fig. 2F) in the mutant embryos at 50 hpf. Taken together, our results demonstrated that the heart chamber differentiation and AVC specification are abnormal in foxc1a mutants, and the abnormal heart development should occur at an earlier developmental stage no later than 30 hpf.

The specification of cardiac progenitors was disrupted in foxc1a mutants
It has been reported that knockdown of FOXC1 leads to the decreased expression of mesoderm-specific genes (33). To test

Roles of Foxc1a in zebrafish cardiogenesis
whether Foxc1a play essential roles in mesoderm formation of zebrafish embryos, we examined the expressions of some germ layer differentiation genes in the mutant embryos. Results from whole-mount in situ hybridization of the representative marker genes for mesoderm (ta, gsc, dkk1b, noto, and gata5), endoderm (pou5f3, mixl1, sox32, and sox17), and ectoderm (zic1) (Fig. S4) showed that the formation of the three germ layers were normal in foxc1a mutant embryos at 6 hpf. Consistently, analysis on the transcriptome of the foxc1a-null embryos at 6 hpf further indicated that the expression of germ layer-patterning genes was unaffected in the mutant embryos compared with wild-type siblings at the gastrulation stage (data not shown).
Now that the heart defects in foxc1a-null embryos were found to occur no later than 30 hpf (Fig. 2E) and the germ layer formation, especially mesoderm formation, looked normal in the mutant embryos (Fig. S4), it was reasonable for us to hypothesize that Foxc1a contributes to heart development via affecting the specification of cardiac progenitors. Cardiac progenitors specified from lateral plate mesoderm will differentiate to cardiomyocytes in vertebrate. To test this hypothesis, we compared the transcriptome of foxc1a mutants with their wildtype siblings at 14 hpf when the cardiac progenitors are specified in the caudal end of the anterior lateral plate mesoderm (ALPM) of zebrafish embryos. The result showed that the expression of cardiac progenitor marker gene nkx2.5 was decreased significantly (Fig. 3A), and those of other well-known ALPM related genes such as etv2 and tal were mildly downregulated in foxc1a mutants (data not shown). To confirm the RNA-seq result, we examined the expression patterns of nkx2. 5, etv2, and tal1 in the foxc1a mutants and wild-type siblings. Consistent with the RNA-seq result, whole-mount in situ hybridization analysis revealed that the expression of nkx2.5 was significantly reduced in the caudal portion of ALPM (Fig. 3, G and GЈ), and those of etv2 and tal1 (for specifying the progenitors of endocardium, head endothelium, and primitive myelopoiesis) were decreased in medial LPM, although the two genes had almost no change in the rostral portion of ALPM in the foxc1a-null embryos (Fig. 3, H, J, HЈ, and JЈ). Quantitative realtime PCR analysis revealed that the expression levels of nkx2. 5, etv2, and tal1 were decreased to 25.0, 63.5, and 81.1% in the foxc1a-null embryos compared with their wild-type siblings, respectively (Fig. 3B).
To exclude the possibility that the abnormal expression patterns of nkx2.5, etv2, and tal1 were due to the defective specification of ALPM, we examined the expressions of ALPM marker genes, including gata4, gata5, and gata6, in the mutant embryos and their wild-type siblings. The results from both quantitative real-time PCR and whole-mount in situ hybridization revealed that the expression levels and patterns of all of the three genes in the foxc1a-null embryos were the same as their wild-type siblings at the six-somite stage (Fig. 3, B, C-E, and CЈ-EЈ). The results suggested that the ALPM was normally formed in foxc1a-null embryos. Consistent with the normal formation of ALPM, the results from whole-mount in situ hybridization demonstrated that the expressions of pax2a (Fig. 3, F  and FЈ), a marker gene for PLPM, and hand2 (Fig. 3, I and IЈ), a second key regulator for the specification of cardiac progeni-

Roles of Foxc1a in zebrafish cardiogenesis
tors, were both normal in foxc1a mutants. Additionally, results from both quantitative real-time PCR and whole-mount in situ hybridization revealed that the expression of gata1a, a marker gene for the formation of red blood progenitors (Fig. 3, B, K, and KЈ), and fli1a, the endothelial cell marker gene (Fig. 3, B, L, and LЈ), had no obvious change in ALPM and PLPM of foxc1a-null embryos. Taken together, our results demonstrate that foxc1a depletion does not affect the formation of mesoderm but influences the specification of cardiomyoblast in the caudal end of ALPM of zebrafish embryos by reducing the expression of nkx2.5.

Zebrafish foxc1a is co-expressed with nkx2.5 in the caudal end of ALPM
Zebrafish foxc1a is a zygotic gene that is expressed from 30% epiboly (data not shown). When the embryos grow to the 50% epiboly stage, the expression of foxc1a is mainly in the marginal zone of the meso-endoderm (Fig. S5A), and then the mesoderm in the embryos at 75% epiboly (Fig. S5B). At the somite stage, foxc1a is expressed in the paraxial mesoderm (Fig. S5, C and  D). When the embryos reach 24 hpf, the foxc1a is dominantly expressed in the endothelial cells of the head and trunk and the eyes. At 50 hpf, the embryos exhibit weak expression of foxc1a in the heart in addition to its expression in the endothelial cells.
To investigate whether Foxc1a regulates the expression of nkx2.5 in the caudal part of ALPM directly, we first examined whether the two genes were co-expressed in zebrafish embryos at somite stage. By visualizing the expressions of nkx2.5 and foxc1a in the wild-type embryos at 12 and 14 hpf using the Cy5 (observed by red color) channel and fluorescein (observed by green color) channel of the confocal microscope, respectively (ta was co-stained by Cy5), we found that some green fluorescent cells also harbored red fluorescence (Fig. 4, H-M and  HЈ-KЈ). The results suggested that foxc1a and nkx2.5 were coexpressed in at least part of the cells in the caudal portion of ALPM.

Foxc1a controls the expression of nkx2.5 by binding to its promoter in zebrafish embryos at somite stage
To determine whether nkx2.5 is a direct target gene of Foxc1a that controls heart development, we next determined whether Foxc1a regulates nkx2.5 expression by directly binding to its promoter. Performing a Dual-Luciferase assay to examine the activities of truncated promoters of zebrafish nkx2.5 with different lengths in response to Foxc1a, we found that the nkx2.5 promoter with diverse lengths (1908, 1462, 1120, 629 -1120, and 629 bp) could all be activated by overexpression of wild-type foxc1a mRNA (Fig. 5A). In contrast, the 1120-bp promoter activity could not be regulated by foxc1a mutant mRNA (Fig. S6A). The results suggest that there are binding sites of Foxc1a in the 1120-bp fragment.
To identify the binding sites of Foxc1a, we performed ChIP assay on the chromatin isolated from zebrafish embryos at the 10-somite stage using antibody against zebrafish Foxc1a. The results showed that Foxc1a was significantly enriched in the fragment of S1 (Ϫ1 to Ϫ172 bp) and S5 (Ϫ624 to Ϫ782 bp) region of zebrafish nkx2.5 promoter (Fig. 5B). To confirm the results, we performed the Dual-Luciferase assay on the mutated 1120-bp promoter in which either S1 or S5 was deleted or both  S1 and S5 were deleted. Consistent with the ChIP results (Fig.  5B), the mutated promoters with deletion of either S1 or S5 were still activated by Foxc1a, although their activities were significantly (p Ͻ 0.05) lower than that of the wild-type 1120-bp promoter (Fig. 5C), respectively. However, the mutated promoter with double deletion of S1 and S5 had no response (p Ͼ 0.05) to Foxc1a (Fig. 5C).

Roles of Foxc1a in zebrafish cardiogenesis
To further identify the core sequences of Foxc1a-binding sites in the nkx2.5 promoter region, we dissected the S1 and S5 region into three regions and then performed a Dual-Luciferase assay to examine the truncated promoter activities in response to Foxc1a, respectively (Fig. 5D). The results showed that S1-2 and S5-3 should contain binding sites of Foxc1a (Fig. 5E). Previously, it was reported that the core sequence of the Foxc1a-binding sites is (T/G)(G/C)(T/R)(T/ Y)T(A/G)TTT (34). Performing bioinformatics analysis, we found that there were two putative non-canonical binding sites, namely SA (TATTGTTTGGGT) in S5-3 and SB (CTGTTTAGTTT) in S1-2, respectively (Fig. 5F). To verify that they were the real core sequence of Foxc1a-binding sites in zebrafish nkx2.5 promoter, we made new promoter constructs by deleting S1-2 and S5-3 or SA and SB (Fig. 5G) and then performed the Dual-Luciferase assay to examine their responses to Foxc1a, respectively. The results revealed that the luciferase activities were significantly decreased (p Ͻ 0.05) in the S1-2 and S5-3 double deletion or the SA and SB double deletion group compared with the 1120-bp wild-type promoter group, respectively (Fig. 5G). . S1-S6, the PCR amplification region in the ChIP-PCR assay (B). TSS, transcription start site. The black lines indicate different lengths of nkx2.5 promoter that were cloned into the pGL3-basic vector for the Dual-Luciferase assay. B, ChIP-PCR assay indicating that S1 and S5 may be the two possible Foxc1a-binding regions in the nkx2.5 promoter. C, Dual-Luciferase assay showing the activities of the 1120-bp promoter, 1120-bp promoter without S1 region, 1120-bp promoter without S5 region, and 1120-bp promoter without either S1 or S5 region in response to Foxc1a. D, schematic diagram shows dissection of nkx2.5 promoter S1 region into S1-2, S1-2, and S1-3 and the S5 region into S5-1, S5-2, and S5-3. The red asterisk in D and G shows the two putative Foxc1a-binding sites SA and SB present in nkx2.5 promoter regions. E, the Dual-Luciferase assay was performed with mutated promoters of S1-1 and S5 deletion, S1-2 and S5 deletion, S1-3 and S5 deletion, S1 and S5-1 deletion, S1 and S5-2 deletion, and S1 and S5-3 deletion in wild-type zebrafish embryos. F, the core sequences of the conserved Foxc1 transcription factor binding sequence in vertebrates and in zebrafish nkx2.5 promoter regions of SA and SB. G, Dual-Luciferase assay was performed with 1120-bp promoter, S1-2 and S5-3 deletion, and SA and SB deletion in wild-type zebrafish embryos. Black histogram, control activity that was normalized as 1.0; white histogram, activity derived from the embryos co-microinjected with 20 pg of foxc1a mRNA plus the promoter. *, p Ͻ 0.05; **, p Ͻ 0.01; ns, not significant. Error bars, S.D.

Overexpression of nkx2.5 could partially rescue the cardiac defects in foxc1a mutant embryos
To verify that the heart defects in foxc1a mutant zebrafish embryos result (at least partially) from the reduction of the nkx2.5 expression in ALPM of foxc1a-null embryos at the somite stage, we microinjected the transgenic construct comprising the hsp70l (heat shock protein 70-like) promoter to drive the expression of Nkx2.5-P2A-eYFP (Fig. 6A) into the embryos at the one-cell stage derived from incross of Tg(kdrl: eGFP) s843 ; foxc1a nju19/ϩ zebrafish and then performed heat shock (37°C) for 1 h from 10 hpf (18). Up to 14 hpf, the embryos with yellow fluorescence were selected for further experiment, and the embryos without heat shock activation were selected as the control group (Fig. 6, B and C). When they grew to 50 hpf, the embryos were measured to calculate their areas of atrium and ventricle at the VS and VD stage (Fig. 6A). The results showed that the size of the ventricle in foxc1a mutant embryos was increased significantly (p Ͻ 0.0001) after overexpression of nkx2.5 and became similar to that of wild-type siblings overexpressed with nkx2.5 (p ϭ 0.1689) at VS stage (Fig. 6D). However, the size of the atrium in foxc1a mutant embryos was not changed (p ϭ 0.1121) at the VS stage after overexpression of nkx2.5 (Fig. 6E). Similarly, the size of ventricle in foxc1a mutant embryos was increased significantly (p Ͻ 0.0001) due to overexpression of nkx2.5, which became comparable with that of wild-type siblings (p ϭ 0.8564) at the VD stage (Fig. 6DЈ). The size of the atrium in foxc1a mutant embryos was not changed (p ϭ 0.5430) at the VD stage after overexpression of nkx2.5 (Fig.  6EЈ). These results indicated that the reduction of ventricle size in foxc1a-null embryos may be due to the decrease of nkx2.5 expression at cardiac progenitor specification stage, whereas there may be another mechanism that causes the increase of atrium size.
When the heat-shocked embryos grew to 108 hpf, they were measured to calculate their EI and SF. The results showed that the EI was significantly reduced (p ϭ 0.0012) in foxc1a-null embryos, although it was significantly (p Ͻ 0.0001) increased in their wild-type siblings overexpressed with nkx2.5 (Fig. 6F). In terms of SF, overexpression of nkx2.5 by heat-shock activation significantly (p ϭ 0.005) rescued the SF of foxc1a-null embryos but did not change (p ϭ 0.1030) the SF of their wild-type siblings (Fig. 6G). Taken together, the results demonstrated that developmental stage-specific overexpression of nkx2.5 partially rescued the ventricle morphogenesis and ventricular contraction defects caused by Foxc1a depletion.

Discussion
FOXC1, a bidirectional regulated transcription factor, plays important roles in many biological processes, such as eye development, circulatory system development, somitogenesis, and skeleton development, and its abnormal expression is involved in the progression of multiple cancers (4,(35)(36)(37). Previously, we reported that zebrafish foxc1a-knockout mutants displayed defective heart development (7). By analyzing the morphologic defects in details, we found that the foxc1a-null embryos exhibited disrupted heart structures and heart functions occurring from 50 hpf (Fig. 1) in addition to shorter body length and smaller eyes at 72 hpf (Fig. S1). The morphological defects were somehow similar to the phenotype of the heart, skeleton, and eye malformation in Foxc1 knockout mice and patients carrying mutated FOXC1 mutations (9). The results support the fact that zebrafish foxc1a is an ortholog of mammalian Foxc1, and they share conserved functions during vertebrate evolution.
It has been reported that zebrafish foxc1a has a duplicated copy named foxc1b. Double knockdown of foxc1a and foxc1b in zebrafish embryos results in loss of most vascular structure due to the reduced expression of the hemangioblast master regulator etv2 and tal1 in the rostral portion of ALPM of the morphants (38,39). In this study, we found that the expressions of these two master regulators were almost not affected in the rostral portion of ALPM (Fig. 3B, H, HЈ, J, and JЈ); nor was the expression of endothelial marker gene fli1a (Fig. 3, B, L, and LЈ). Consistently, the artery, vein, and intersegment vessels in the trunk were formed normally in foxc1a-null embryos (Fig. S2). However, the angiogenesis in head was disrupted in foxc1a-null embryos (Fig. S2). The results suggested that foxc1a/b played redundant roles in the formation of only some vascular structure. Therefore, the vascular mechanism underlying the heart disorder of foxc1a mutants could not be completely ruled out.
Although there is visible blood flow in foxc1a-null zebrafish embryos from 30 hpf, erythrogenesis was disrupted with decreased expression of hemoglobin marker gene hbae1 (Fig.  S3, C-F). Considering the reduced expression of tal1 in medial LPM (Fig. 3, B, J, and JЈ) of foxc1a-null embryos, we hypothesize that Foxc1a may regulate erythrogenesis by affecting the expression of the upstream master transcription factor tal1.
Nkx2-5 is one of the master transcriptional factors controlling vertebrate heart development. In human embryonic stem cells, knockdown of FOXC1 leads to the decreased expression of marker genes for cardiac mesoderm, such as Mef2c, Isl1, and Nkx2-5 (33). Consistent with this observation, the expression of the CPC marker gene nkx2.5 was significantly decreased in the foxc1a-null embryos at 14 hpf (Fig. 3A), when cardiac progenitors are specified. By examining the expressions of gata4/5/6 in the foxc1a-null embryos, we found the mutants had normal formation of ALPM (Fig. 3, C-E and CЈ-EЈ). Additionally, the expressions of both pax2a, a marker gene for posterior lateral plate mesoderm, and hand2, a second key regulator for the specification of cardiac progenitors, were normal in foxc1a mutants (Fig. 3, F, FЈ, I, and IЈ). These results are consistent with the observation that the germ layers were normally formed in foxc1a-null embryos, which suggests that Foxc1a is not involved in mesoderm formation but directs the specification of cardiac progenitors. Together with the finding that overexpression of nkx2.5 is able to partially rescue the defective function of foxc1a-null embryos, our results demonstrate that foxc1a depletion affects the specification of cardiomyoblast in the caudal end of ALPM of zebrafish embryos by reducing the expression of nkx2.5.
Previously, researches have revealed that zebrafish nkx genes are essential for maintenance of ventricular identity (19). In mice, Nkx2-5 play a restrictive role in atrial myocyte proliferation. Nkx2-5 deficiency in the atria causes massive enlargement of working and conduction myocardium, leading to hyperplastic atrial myocardium (17). Consistent with nkx2.5 mutants that

Roles of Foxc1a in zebrafish cardiogenesis
have defects of chamber differentiation, a dilated atrium and narrowed ventricle (19), we found in this study that the foxc1anull zebrafish embryos also displayed a smaller ventricle and more inflated atrium, although the defective phenotype was not as severe as in nkx2.5 mutants (Fig. 2, A-D). The mitigatory phenotype could be due to the residue expression of nkx2.5 in foxc1a-null embryos. However, the abnormal atrial enlargement was not rescued by nkx2.5 overexpression (Fig. 6). Taken

Roles of Foxc1a in zebrafish cardiogenesis
together, the results suggest that the atrial enlargement occurring in the foxc1a-null embryos might be caused by both the absence of nkx2.5 expression and of cardiac dysfunction.
Although the expression of nkx2.5 was significantly reduced in the caudal portion of ALPM in the foxc1a-null embryos, the expression area and mRNA level of hand2, a second marker gene of cardiac progenitor cells, was not changed in the ALPM (Fig. 3, B, I, and IЈ). The results suggest that Foxc1a deficiency would not affect the number of cardiac progenitor cells. Additionally, the previous research demonstrated that Nkx2.5 deficiency neither enhances the proliferation of atrial cells nor alters the patterns of cell death in embryos aged from 26 to 52 hpf (19). Therefore, similar to nkx2.5 mutant zebrafish embryos, the cardiac chamber morphogenesis defects in foxc1a mutant zebrafish embryos probably result from the transdifferentiation of some ventricular cells into atrial cells. Considering that foxc1a is widely expressed in mesoderm and mesodermderived tissues (Fig. S5) and that the conventional knockout model of zebrafish foxc1a has its obvious limitation for investigating the molecular mechanism responsible for the heart defects, it will be very helpful to use a conditional knockout zebrafish to study the role of Foxc1a in zebrafish heart development in future studies.
Traditionally, FOX transcription factors regulate target genes by binding to the conserved core cis-element RYMAAYA in their promoter region (40). In zebrafish, Foxc1a is demonstrated to regulate the expression of etv2 by binding to the canonical core sequence of (T/G)(G/C)(T/R)(T/Y)T(A/G)TTT in its enhancer region (39). Although there is a classic binding site, TGTTTGTTT, in the Ϫ506 bp position upstream of the zebrafish nkx2.5 transcription start site, our results from the Dual-Luciferase activity assay revealed that deletion of this classic binding site did not affect the activation of Foxc1a on nkx2.5 promoter (Fig. S6B). Performing a truncated promoter activity assay and ChIP assay in wild-type zebrafish embryos, we demonstrated that there were two non-canonical binding sites present in the zebrafish nkx2.5 promoter (Fig. 5). Together with the results from double fluorescence in situ hybridization revealing that the expressions of foxc1a and nkx2.5 were co-localized in some cells at ALPM of the embryos at both 12 and 14 hpf, we concluded that Foxc1a regulates the expression of nkx2.5 directly.
It has been reported that Foxc1-and Foxc2-deficient mice exhibit a defective out-flow tract due to reduced Tbx1 expression (40). Human patients carrying FOXC1 mutations suffer from cardiac anomalies, such as mild dysplasia of left ventricle, OFT, valvula tricuspidalis, and heart failure (10,14,18). These events suggest an important role of FOXC1 in second heart field development. The existence of the second heart field in zebrafish also has been verified (41,42), and we also observed a shortened out-flow tract and dysplasia ventricle in the foxc1a mutant zebrafish embryos. These results indicate the conserved function of FOXC1 in the regulation of the second heart field among different species. However, we did not detect a change in tbx1 expression at 14 hpf by transcriptome sequencing in foxc1a mutants (data not shown). This means that there may be another factor mediating the function of Foxc1a in zebrafish second heart field development. It has been reported that zebrafish second heart field development relies on the normal nkx2.5 function (43), so we speculate that, compared to the mammalian Foxc1, zebrafish Foxc1a functions in second heart field development by directly regulating nkx2.5 expression.
Even so, more experiments should be performed to investigate whether the regulation of nkx2.5 by Foxc1 is conserved in mammals.

Ethics statement
The experimental protocols involved in using zebrafish were approved by the institutional animal care and use committee at the Model Animal Research Center, Nanjing University. All animal experiments were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. Tubingen strain, transgenic zebrafish lines including Tg(kdrl:eGFP) s843 , Tg(myl7:eGFP) m225 , and foxc1a nju19/ϩ mutant lines were used in this study.

Live imaging and quantification
To observe the morphology and count heart rates, we treated embryos with 0.04 mg/ml MS-222 (Sigma) at room temperature for 5 min and then mounted them with 3% methyl cellulose (Sigma). To examine the EI, we measured the distance from the heart center to the heart boundary (a) and from the heart center to the pericardium (b), as reported previously (28). The EI was calculated by the expression, b/a. The ventricular SF and heart chamber area were quantified and calculated in accordance with the reported method (44). Embryonic pictures were taken at VD and VS phases, respectively. The SF was computed by the equation, SF ϭ (width of ventricle at VD Ϫ width of ventricle at VS)/width of ventricle ϫ 100.
Heart rates of embryos were counted under a dissecting microscope (Leica). The frequency of heartbeat in each embryo was counted for 20 s, and the heart rate (beats/min) was calculated by multiplying by 3. By opening the files of the embryo

Roles of Foxc1a in zebrafish cardiogenesis
photos taken at different phases with Photoshop software, we measured the chamber area by marking the ventricle and atrium region.

Histological section and staining
Zebrafish embryos were fixed with 4% paraformaldehyde at 4°C for Ͼ24 h. The fixed embryos were dehydrated with gradient ethanol, cleared with xylene, and then embedded in paraffin. Embedded embryos were cut as series of sections with 5-m thickness, as in our previous report (30). The sections were placed on glass slides and stained with hematoxylin and eosin. Photomicrographs were taken using a DP70 digital camera under a dissecting microscope (Olympus). At least three consecutive sections were observed to ascertain that the sections contained the embryonic heart structures.
With regard to o-dianisidine staining, embryos were collected at 50 hpf and incubated with o-dianisidine solution for 15 min in the dark. o-Dianisidine solution contained 0.6 mg/ml o-dianisidine, 0.01 M sodium acetate (pH 4.5), 0.65% H 2 O 2 , and 40% ethanol (v/v). Embryos were embedded with glycerol, and photographs were taken with a DP70 digital camera (Olympus).

Whole-transcriptome deep sequencing
Total RNA was isolated from single offspring of foxc1a nju19/ϩ zebrafish incross using TRIzol reagent (Invitrogen). The cDNA synthesized from total RNA isolated from a single embryo by reverse transcriptase (Vazyme) was used as the template of PCR to genotype foxc1a alleles, as we reported previously (7). After genotyping, 5 g of total RNA was collected from wild-type siblings and foxc1a mutants for RNA deep sequencing, respectively. The transcriptome sequencing was performed by Novel-Bio Bio-Pharm Technology Co. (Shanghai, China). Gene expression levels were quantified by RPKM (reads per kilobase of transcript per million mapped reads) arithmetic. MapSplice software was used for data alignment, and EB-Seq arithmetic was used for the screening of differential expression genes.
The method of double fluorescence ISHs was used as in the previous report (46). Briefly, the wild-type embryos were fixed at 12 and 14 hpf with 4% paraformaldehyde overnight. After dehydration and protease K treatment, the embryos were then hybridized with nkx2.5 and ta RNA probe labeled with digoxin and foxc1a RNA probe labeled with fluorescein at 65°C overnight. After the removal of probe, the anti-fluorescein conjugated with peroxidase antibody (1:500; Roche Applied Science) along with TSA plus fluorescein solution (1:50; PerkinElmer Life Sciences) was added. Then the anti-digoxin conjugated with peroxidase antibody (1:1000; Roche Applied Science) along with TSA plus Cy5 solution (1:50; PerkinElmer Life Sciences) was added. After inactivation of the primary antibody, the embryos were then flatted and placed on the glass slides, and images were photographed with a TCS SP5 confocal microscope (Leica).

In vitro synthesis of foxc1a mRNA
To synthesize mRNA in vitro, we first cloned the full-length coding sequences of wild-type and mutant foxc1a (NM_ 131728.3) from wild-type zebrafish and foxc1a mutants by RT-PCR. The cDNAs were cloned into pGEM-T easy vector (Promega) using Phanta Super-Fidelity DNA Polymerase (Vazyme), and then their identities were confirmed by direct sequencing from both ends. The capped and tailed mRNA was synthesized from the linearized vector by NcoI (Thermo) using the mMES-SAGE mMACHINE Sp6 Ultra Kit (Ambion). The synthesized mRNAs were further purified with the MEGAclear TM Transcription Clean-Up Kit (Ambion) to remove the free nucleotides.

Promoter construction and Dual-Luciferase assay
Zebrafish nkx2.5 promoters with different lengths were individually cloned into pGL3 basic vector linearized with NcoI and XhoI (Thermo). To test the promoter activity in response to Foxc1a, we microinjected 50 pg of luciferase vectors, 1 pg of Renilla expression vectors, and 20 pg of wild-type foxc1a mRNA or 20 pg of foxc1a mutant mRNA into wild-type

Roles of Foxc1a in zebrafish cardiogenesis
embryos at the one-cell stage. Embryos were then collected at

ChIP-PCR
To perform the ChIP assay, we collected the wild-type zebrafish embryos at the 8 -10-somite stage. 5 g of FOXC1 antibody (GTX25079, GeneTex) was used to immunoprecipitate all genomic DNA crossed with Foxc1a protein, and 1 g of mouse IgG was used as control. The ChIP-PCR assay was performed using the EZ-ChIP chromatin immunoprecipitation kit (Millipore) following the manufacturer's instructions. The semiquantitative PCR was performed as follows. The PCR conditions were 95°C for 5 min; 35 cycles of 95°C for 30 s, 52°C for 30 s, 72°C for 15 s; and 72°C 10 min. The PCR products were then subjected to separation by 2% agarose electrophoresis. The primers for PCR are listed in Table S3.

Statistics
Data are presented as mean Ϯ S.D. Statistical significance was determined using the unpaired two-tailed t test. A value of p Ͻ 0.05 (*) was considered statistically significant, and p Ͻ 0.01 (**) and p Ͻ 0.001 (***) were considered statistically very significant.