UDP-Glucose Dehydrogenase Required for Cardiac Valve Formation in Zebrafish

See allHide authors and affiliations

Science  31 Aug 2001:
Vol. 293, Issue 5535, pp. 1670-1673
DOI: 10.1126/science.293.5535.1670


Cardiac valve formation is a complex process that involves cell signaling events between the myocardial and endocardial layers of the heart across an elaborate extracellular matrix. These signals lead to marked morphogenetic movements and transdifferentiation of the endocardial cells at chamber boundaries. Here we identify the genetic defect in zebrafish jekyll mutants, which are deficient in the initiation of heart valve formation. The jekyll mutation disrupts a homolog of Drosophila Sugarless, a uridine 5′-diphosphate (UDP)–glucose dehydrogenase required for heparan sulfate, chondroitin sulfate, and hyaluronic acid production. The atrioventricular border cells do not differentiate from their neighbors in jekyll mutants, suggesting that Jekyll is required in a cell signaling event that establishes a boundary between the atrium and ventricle.

Cardiac valves form at chamber boundaries and function to prevent retrograde blood flow through the heart. Extensive work in a chick explant system has revealed some of the cellular interactions necessary for valve formation (1). Endocardial cells at the boundary between the atrium and ventricle are prepatterned to receive a signal from the overlying myocardial cells. This myocardial signal induces the endocardial cells to undergo an epithelial-to-mesenchymal transition, thereby initiating the formation of prevalvular cushions that are later remodeled to form the valves proper. Recent work has implicated transforming growth factor–β family members in the myocardial-to-endocardial signaling event that induces endocardial cushion formation (2–5). However, the mechanism by which myocardial cells at the atrioventricular (AV) boundary acquire the competence to send that signal is not known.

Large-scale screens in zebrafish have identified several mutations that affect cardiac valve formation, the most severe of which is the recessive mutation jekyll (6).jekyll mutant embryos exhibit pericardial edema and toggling of blood between the two chambers of the heart (compare Fig. 1, A and B). Together these phenotypes are generally indicative of defective AV valve function and are consistent with previous observations that jekyll mutant hearts lack valve tissue at 48 hours postfertilization (hpf) (6). To analyze endocardial morphology in vivo, we generated a line of jekyll heterozygotes bearing an integrated mouse tie2 promoter driving green fluorescent protein (GFP) expression in the developing endocardium (7). In wild-type embryos, endocardial cells cluster at the AV boundary at the onset of valve formation at 43 hpf (Fig. 1C). However, in mutant hearts this clustering fails to occur (Fig. 1D), indicating that jekyll function is required for this early endocardial morphogenetic event.

Figure 1

Comparison of wild-type andjekyll mutant heart morphology. Wild-type (A) and mutant (B) embryos at 48 hpf; lateral view, anterior to the left. Pericardial edema surrounds the jekyll mutant heart and blood has accumulated in both chambers [at this stage the ventricle is anterior (to the left) and in this embryo, the atrium is slightly obscured by a skin melanocyte]. Arrows point to the heart.tie2-GFP allows visualization of wild-type (C) and mutant (D) endocardial morphology at 43 hpf; lateral view, anterior to the top and dorsal to the left. Arrows indicate the AV boundary where endocardial cells cluster in wild-type embryos at the onset of valve formation. wt, wild-type. Bars, 140 μm.

To gain further insight into the jekyll valve defect, we isolated the disrupted gene by synteny cloning. We localizedjekyll to a centromeric region of linkage group 1 using bulk segregant analysis (8) of embryos genotyped for polymorphic CA repeat markers. We then performed fine mapping of the region with 15 polymorphic markers on 200 wild-type haploid embryos and found close linkage between jekyll and three of these markers. Next, we genotyped an additional 1150 affected diploid embryos with those three markers as well as for a polymorphism in the 3′-untranslated region ofldb3 (9). These studies allowed us to further narrow the jekyll interval to a 0.5-centimorgan region (Fig. 2A).

Figure 2

The jekyll locus encodes Ugdh. (A) Integrated genetic and radiation hybrid (RH) maps of thejekyll region illustrate the relationships between genes and CA-repeat “z” markers. Numbers below the line indicate the number of recombination events seen in 1150 diploid and 176 haploid embryos tested. Four zebrafish ESTs were chosen for RH mapping because they encode homologs of genes near SLIT2, LDB3, and UCHL1 on human chromosome 4p. One of the four, a clone encoding a protein with 84% amino acid identity to human UGDH, maps between slit2 and ldb3, two genes that flank thejekyll locus. (B) Sequencing of ugdhcDNA revealed a T to A change at base pair 992 in the jekyllmutant. This mutation results in an Ile-to-Asp substitution at residue 331. (C) Ile-331 sits in a nonpolar pocket of the enzyme as illustrated in these rasmol images of the bovine UGDH crystal structure (12, 38). The Ile is colored in green, the nonpolar amino acids are in white, and the polar ones are in blue. (D to I) Antisense “knock-down” ofugdh phenocopies the jekyll mutation [(D to F) uninjected wild-type; (G to I) morpholino-injected]. (D and G) Injection of ugdh morpholino into sensitized embryos causes a failure of AV valve formation, here illustrated by a failure of endocardial clustering at the AV boundary in 43 hpf embryos (arrows indicate the AV boundary). Examination at later stages reveals that valve formation does not occur in morpholino-injected embryos, showing that the defect at 43 hpf does not reflect a delay [(E and H) 72 hpf; (F and I) 96 hpf]. wt, wild-type; mo, morpholino-injected. Bars, 20 μm. The zebrafish ugdh cDNA sequence was deposited in GenBank with the accession number AF361478.

Examination of the emerging map of the jekyll region revealed a striking conservation of synteny with a region of human chromosome 4p. Taking advantage of this conserved synteny, we mapped four zebrafish expressed sequence tags (ESTs) with sequence similarity to human genes in this 4p region. One of these ESTs, corresponding to a homolog of the sugarless gene (known as UDP–glucose dehydrogenase in humans and by convention, referred to as ugdhhereafter), was found by radiation hybrid mapping to lie between two markers closely flanking the jekyll locus. Sequence analysis of cDNA prepared from wild-type and mutant embryos revealed a T to A change at base pair 992 in the mutant allele. Genotyping for this change revealed no recombination between jekyll and the observed lesion in 2870 meioses (Fig. 2, A and B) (10). The T to A change results in an Ile-to-Asp substitution at residue 331 (11). Ile-331 is conserved in Drosophila, human, and zebrafish Ugdh and is situated in a pocket of nonpolar amino acids in the “hinge” of the omega loop “gate” that allows UDP-glucose access to the active site of the enzyme (Fig. 2C) (12). This Ile-to-Asp substitution is likely to affect enzyme activity.

Ugdh enzymatic activity is required for the conversion of UDP-glucose into UDP–glucuronic acid, a critical component of hyaluronic acid, heparan sulfate, and chondroitin sulfate glycosaminoglycans (13). Other mutations that affect the production of heparan sulfate proteoglycans in vertebrates result in defects during gastrulation. These include a targeted mouse mutation in the heparan sulfate glycosyltransferase, EXT1(14), and the zebrafish knypek mutation, which disrupts a Glypican homolog (15). Thejekyll mutation, which should affect the production of heparan sulfate at the earliest step in the pathway, shows no obvious phenotype until organogenesis stages. One explanation for this incongruity is that zebrafish ugdh mRNA is provided maternally. Indeed, whole-mount in situ hybridization analyses reveal the presence of ugdh mRNA at the four-cell stage (11), whereas zygotic transcription begins at the 1000-cell stage. Expression domains in the otic vesicle, heart, and branchial arches in older embryos (30, 37, and 48 hpf, respectively) are consistent with the jekyll mutant phenotypes reported here and previously (16).

Morpholino antisense “knockdown” (17) ofugdh translation phenocopies the jekyll mutation (Fig. 2, D to I). Interestingly, this phenocopy can be achieved only by genetically sensitizing the injected embryos. Such sensitization can be attained in two ways: by halving the maternal product through the use of embryos generated from a cross between a jekyllheterozygote female and a wild-type male, in which case 90% of the antisense-injected embryos displayed the jekyll phenotype, or by halving the zygotic transcription, i.e., use of embryos from ajekyll heterozygote male and a wild-type female, in which case 35 to 50% of the antisense-injected embryos were affected. This sensitization is likely due to the decrease of early ubiquitous maternal and zygotic expression of the jekyll gene. The lack of an early phenotype in antisense-injected embryos suggests that in addition to ugdh mRNA, Ugdh protein is also provided maternally, or that another protein compensates for the lack of Ugdh, or even that some maternal mRNA is protected from antisense-mediated translational inhibition.

Explant experiments with embryonic chick tissues have shown that valve formation requires precise patterning of the myocardium and endocardium, as well as an inductive signal from the myocardium to the endocardium (1). To determine how jekyll affects cardiac valve development, we assessed the expression of early differentiation markers of the valve-forming region: bone morphogenetic protein (bmp) 4, notch1b (18), and br146 [a zebrafish versican homolog (19)]. Initially all three genes are expressed throughout the anteroposterior extent of the heart. Later expression of these genes is restricted to the valve-forming region (bmp4 andbr146 in the myocardium at 37 hpf and notch1b in the endocardium at 45 hpf). jekyll mutant embryos show defects in the expression of these genes (Fig. 3). Although bmp4 and br146 expression domains become restricted from the atrium and largely from the ventricle, no heightened expression of these genes is detectable at the AV boundary of jekyll mutant hearts at any stage assayed from 36 to 48 hpf (48 hpf shown in Fig. 3, B and D and F and H). notch1bexpression becomes initially restricted from the atrial endocardium of mutant hearts. However, at 45 hpf, notch1b expression is elevated in the AV boundary of wild-type but not jekyllmutant embryos (20). At 48 hpf, atrial restriction fails to be maintained in mutant hearts, and notch1b expression is again observed throughout the atrial and ventricular endocardium (Fig. 3L). Thus, jekyll functions early in the process of AV valve formation and is required specifically in patterning the myocardium and endocardium at the AV boundary. The myocardial patterning defects seen in jekyll mutants are likely direct effects of the loss of Jekyll function because they are the first molecular defects to be observed, whereas the later endocardial defects may be secondary to the lack of bmp4 restriction in the myocardium.

Figure 3

Molecular analyses of AV valve development reveal early defects in jekyll mutant embryos. Schematized representations are shown to the left of the actual data. Initially expressed throughout the anteroposterior extent of the heart (20), bmp4, br146, andnotch1b become restricted to the AV boundary before the onset of valve formation. At 48 hpf, bmp4 (A andB) and br146 (E and F) are expressed in the myocardium and notch1b (Iand J) in the endocardium. RNA in situ hybridization (39) for these genes reveals defects in jekyllmutant hearts. (B, D, F, andH) Although expression of bmp4 andbr146 does become restricted from the atrium and largely from the ventricle, there is no heightened expression at the valve-forming region as in the wild-type. (J andL) notch1b expression is not increased at the AV boundary and at this stage is no longer restricted from the atrium. A, atrium; V, ventricle; wt, wild-type. Bar, 20 μm.

These data lead us to hypothesize that jekyll is required for cell signaling events that set aside the valve-forming region as distinct from atrium and ventricle. Perhaps similar to other boundary-formation events in development, the cells at the border between the atrium and ventricle are specified to adopt a tertiary cell fate, that of the valve-forming region. In the Drosophilawing disc, for example, differences in engrailed expression between the anterior and posterior compartments restrict expression of the Hedgehog signal to the posterior and the Hedgehog receiving apparatus to the anterior. Thus, only cells on the anterior side of the compartment border can receive the Hedgehog signal coming from the posterior (21). A similar situation exists in the vertebrate limb bud, where differences in fringe expression set up a domain of specific Notch signaling at the dorsoventral border (22, 23). The cells at that boundary differentiate to form the apical ectodermal ridge, which is required for further outgrowth of the limb.

Atrial and ventricular fates are properly assigned injekyll mutants, because expression of chamber specific markers appears normal (11). Therefore, the first apparent molecular defect in jekyll mutants is a lack of boundary-restricted bmp4 expression in the myocardium. A similar defect is seen in cloche mutant embryos, which lack endocardium altogether (24, 25). Analysis of these two mutants suggests a model in which a primary endocardial signal to the overlying myocardium sets up a border between atrial and ventricular cells. A secondary signal within the myocardium could then produce at the AV boundary a tertiary myocardial cell type with the competence to signal to the underlying endocardium. jekyllmay be required for the primary endocardial signal or for the secondary myocardial signal that results in the tertiary cell fate.

In Drosophila, Sugarless has been shown to be important for Fgf and Wg signaling (26–29). We have analyzed the currently available zebrafish mutations that affect these signaling pathways and have not found any overt defects in cardiac valve formation. By contrast, the similarity of the morphological defects in the formation of the jaw elements injekyll, pipetail/wnt5a, andknypek/glypican mutants (11, 15) suggests that in this process Jekyll is required for Wnt signaling, and that one substrate for Jekyll is Glypican. In the case of Jekyll's function in the heart, the similarity of the jekyll valve phenotype with that seen in mutants for the chondroitin sulfate proteoglycan gene Versican (30) suggests that Versican is one substrate for Jekyll in this process. However, mice deficient for Hyaluronan synthase–2 also exhibitjekyll-like valve defects (31), suggesting thatugdh may also function in valve formation through its requirement for hyaluronic acid (HA) synthesis. We propose that a signaling pathway, heretofore not identified to require glycosaminoglycan production, depends on Jekyll function during valve formation. Neuregulin is expressed in the ventricular endocardium, and loss of Neuregulin signaling leads to several phenotypes that are also observed in jekyll mutants, including the absence of myocardial proliferation in the ventricle and hypoplastic endocardial cushions (31–34). It will be important to investigate the respective role of HA, proteoglycans, and their associated signals during cardiac valve formation.

  • * To whom correspondence should be addressed. E-mail: didier_stainier{at}


View Abstract

Navigate This Article