Genetic locus half baked is necessary for morphogenesis of the ectoderm



The zebrafish epiboly mutants partially block epiboly, the vegetalward movement of the blastoderm around the giant yolk cell. Here, we show that the epiboly mutations are located near the centromere of Linkage Group 7 in a single locus, termed the half baked locus. Nevertheless, except for the similar mutants lawine and avalanche, we find the epiboly traits of each of the alleles to be distinguishable, forming an allelic series. Using in situ analysis, we show that the specification and the formation of the germ layers is unaffected. However, during early gastrulation, convergence movements are slowed in homozygous and zygotic maternal dominant (ZMD) heterozygous mutants, especially in the epiblast layer of the blastoderm. Using triple-mutant analysis with squint and cyclops, we show that ablating involution and hypoblast formation in hab has no effect on the epiboly phenotype on the ventral and lateral sides of the embryo, suggesting that the hypoblast has no role in epiboly. Moreover, the triple mutant enhances the depletion of cells on the dorsal side of the embryo, consistent with the idea that convergence movements are defective. Double-mutant analysis with one-eyed pinhead reveals that hab is necessary in the ectodermal portion of the hatching gland. In ZMD heterozygotes, in addition to the slowing of epiboly, morphogenesis of the neural tube is abnormal, with gaps forming in the midline during segmentation stages; later, ectopic rows of neurons form in the widened spinal cord and hindbrain. Cell transplantation reveals that half baked acts both autonomously and nonautonomously in interactions among cells of the forming neural tube. Together, these results suggest that half baked is necessary within the epiblast for morphogenesis during both epiboly and neurulation and suggest that the mechanisms that drive epiboly possess common elements with those that underlie convergence and extension. Developmental Dynamics 233:390–406, 2005. © 2005 Wiley-Liss, Inc.


In teleosts, morphogenesis begins with epiboly, the process in which the blastoderm covers and engulfs the large yolk cell. Here, we report our analysis of the defects in the epiboly mutants identified in the Tübingen genetic screen for mutations that affect morphogenesis of the zebrafish that was carried out at the Max Planck Institut für Entwicklungsbiologie (Haffter et al., 1996). Epiboly is blocked in the mutants half baked (hab), lawine (law), avalanche (ava), weg, and volcano (Kane et al., 1996; Solnica-Krezel et al., 1996). In each of these mutants, epiboly begins normally in the blastula and early gastrula stage, but by 70% to 80% epiboly, approximately 1.5 to 2 hr after the onset of gastrulation, mutants begin to arrest their vegetalward spreading and begin to dissociate, usually with their blastoderm peeling off the yolk.

The mutants hab, ava, and law also display a zygotic maternal dominant (ZMD) effect that is expressed when both zygotic and maternal genomes are heterozygous for the mutant locus. These embryos, termed ZMD mutants, display an intermediate rate of epiboly between that of wild-type and homozygous mutant siblings and complete epiboly approximately an hour after wild-type siblings. Later, during somitogenesis, cells dorsal to the developing neural tube round up and detach from the embryo (Kane et al., 1996). In addition, hab mutants display a semidominant trait of an enlarged hatching gland. Although these mutants fail to complement one another, the dominant and ZMD-dominant traits made it difficult to assess whether they belong to one locus; therefore, they were named separately.

The array of processes that drive epiboly are not completely understood. Before epiboly begins, the zebrafish blastula is composed of three domains: the epithelial-like enveloping layer (EVL) that covers the blastoderm, the deep cells of the blastoderm, and the yolk cell (Kane et al., 1992). In Fundulus, after removing both the EVL and deep cell layer from the yolk cell, the syncytial layer of the yolk completes epiboly autonomously (Trinkaus, 1951). This movement is dependent on a microtubule-based motor. A highly organized microtubule array has been observed in the zebrafish yolk cell (Strähle and Jesuthasan, 1993; Solnica-Krezel and Driever, 1994), and treatments that disrupt this microtubule network slow or stop epiboly. At least part of the blastoderm appears to be towed by this yolk cell motor: the EVL is tightly attached to the yolk cell (Betchaku and Trinkaus, 1978), and when the yolk syncytial layer begins to move, the EVL follows in tandem. Curiously, in the epiboly mutants, the EVL and yolk syncytial layer are unaffected and both complete epiboly normally.

The contribution—or lack thereof—of other morphogenetic movements to epiboly is not known. In the mutants, only the movement of the deep cells is arrested. To date, there is no known role for the deep cells in epiboly. Starting as an amorphous mound at the animal pole, between the EVL and the yolk cell, the deep cell layers thin as they spread over the yolk cell (Warga and Kimmel, 1990; Wilson et al., 1995), and it is unclear if the deep cells are active participants in epiboly. Nevertheless, the epiboly mutants demonstrate that at least some aspects of the epiboly of the deep cells are under a separate genetic control from that of the yolk cell and EVL.

Midway through epiboly, deep cells at the margin of the blastoderm commence involution, separating the blastoderm into two layers: an outer epiblast layer, the future ectoderm, and an inner hypoblast layer, the future mesoderm and endoderm (Warga and Kimmel, 1990; Warga and Volhard-Volhard, 1999). Epiboly is normal in squint/cyclops double mutants, where the Nodal pathway is abolished, blocking mesoderm formation and involution, suggesting that these processes have no effect on epiboly (Feldman et al., 2000; Dougan et al., 2003). However, epiboly is arrested in lefty1/2 antisense knockdown embryos, where the Nodal pathway is up-regulated and the rate of involution is increased (Branford and Yost, 2002; Chen and Schier, 2002; Feldman et al., 2002).

Shortly after involution commences, cells begin to converge and migrate from lateral positions toward the dorsal side of the embryo (Warga and Kimmel, 1990; Schmitz et al., 1993; Kimmel et al., 1994), except for the ventral-most region of the embryo where convergence does not occur (Myers et al., 2002). At the cellular level, cohorts of cells extend anteroposteriorly and narrow mediolaterally as they move dorsally. Whether in the epiblast, where cells move as an epithelial sheet, or in the hypoblast, where cells move as individuals (Jessen et al., 2002), these cells mediolaterally intercalate among themselves as they enter the body axis (Warga and Kimmel, 1990; Sepich et al., 2000). Although convergence is occurring during epiboly, the mechanistic relationships between the two processes are not understood. However, these movements must be somewhat interdependent, as epiboly is slowed in frogs by treatments that block convergence (Hikasa et al., 2002).

Here, we document that the epiboly mutations are all located on a small region of Linkage Group 7, strongly suggesting that the mutants map to a single locus. Nevertheless, we find that each allele displays specific differences in their effect on epiboly, which correlates with differences in aspects of their dominant phenotypes. We then document that specification of the three primary germ layers and subsequent tissue differentiation is normal in hab mutants and that involution of the mesendoderm occurs normally. However, we find that the axis of homozygous mutants is wider, particularly in the epiblast layer, suggesting that convergence processes may be defective during gastrulation. In ZMD mutants, we also document that, during segmentation stages, the formation of the neural keel is wider and contains ectopic neural structures. Using triple-mutant analysis of hab with squint and cyclops, we confirm that involution and mesendoderm formation is unrelated to the hab epiboly phenotype on the ventral and lateral sides of the embryo but find a depletion of cells on the dorsal side of the triple mutants, further indicating that convergence movements are defective. Examination of double mutants of hab and one-eyed pinhead reveals that hab is necessary in the ectodermal portion of the hatching gland. Furthermore, cell transplantation studies reveal that the hab gene product acts autonomously as well as nonautonomously in the neuroectoderm to produce aspects of the neural tube defect. Altogether, our data suggest that hab is required in the epiblast of the embryo for morphogenetic movements that drive epiboly and neural morphogenesis.


Loci of the Epiboly Mutants Map Closely to the Linkage Group 7 Centromere

Because of the dominant effects of many of the epiboly mutants, the results from the initial complementation matrix were uninterpretable. Like the Minute loci of Drosophila (Lindsley and Grell, 1968) and the swirl-somitabun loci of zebrafish (Mullins et al., 1996), it was quite possible that the epiboly mutants were intra-allelic noncomplementing dominant mutations. To test if they represented different loci, we positioned the mutant loci onto the zebrafish genetic map. Initial mapping, using half-tetrad analysis (not shown), loosely linked all of the mutants to the centromere of Linkage Group 7. Unfortunately, this method was inadequate for fine mapping of the epiboly mutants because of scoring artifacts. The high hydrostatic pressure used for the half-tetrad method affects the microtubule network of the embryo, slowing epiboly and creating other unrelated defects (Hatta and Kimmel, 1993). This effect caused a fraction of the ZMD heterozygous embryos to resemble the homozygous recessive phenotype and a fraction of the wild-type embryos to resemble the ZMD phenotype. Moreover, many of the homozygous recessive mutants tended to prematurely dissociate and not be scored at all. Hence, the number of heterozygous embryos was overrepresented and the distance from the centromere to the hab locus exaggerated. Genetic distances also varied because of experimental variations of the treatment on different clutches of embryos from slightly different backgrounds.

Using haploid mapping panels to carefully map the individual mutant loci in relation to microsatellite markers, we first confirmed the locations of available markers, creating the mitotic map shown in Figure 1A. Because the mitotic events in a haploid panel occurred exclusively in females, this map has longer intramarker distances than the sex-averaged mitotic map from Boston (Shimoda et al., 1999) and resolved many areas of genetic compression. Furthermore, markers that had been linked previously to the center of Linkage Group 7 using the radiation hybrid panel were remapped, adding to the number of useful markers. We also used half-tetrad mapping panels that were generated for the mapping of unrelated (and unlinked) mutant loci and found no recombinants in a mapping panel of over 125 individuals between the markers z7958 and z9869 or either marker and the centromere; this finding established these markers as reliable markers of the centromere for Linkage Group 7 (Fig. 1B) and closer than markers that we and others have suggested previously (Kane et al., 1999; Mohideen et al., 2000).

Figure 1.

Mapping of the half-baked locus. A: Mitotic map for Linkage Group 7 constructed from haploid analysis using a panel of 2,916 individuals. B: Mitotic map constructed from half-tetrad analysis using a panel of 125 individuals. Numbering begins at the centromere and proceeds along each arm. C: Order of microsatellite markers and location of mutant loci. Numbers ± standard deviation indicates recombination events between individual markers or between markers and mutation. Black boxes indicate location of individual mutant loci and size of 95% confidence limits inferred from numbers of recombinants. The haploid mapping panel sizes are as follows: hab, 222; ava, 90; law, 163; and weg, 114. Markers not shown were monomorphic. D–F: Testing linkage of candidate genes to the hab locus by polymorphism in the haploid mapping panel. The first lane in each gel is a molecular weight marker, except in F, and the phenotype of each haploid individual is indicated above its lane on the gel. Arrows indicate recombination individuals for the hab locus and the candidate. D: Mapping of ventral neural cadherin (VNcad) to the hab locus using the size polymorphism in microsatellite marker Z3008. PAC clone 30C24 contains VNcad (not shown) as well as microsatellite marker Z3008, whereas a control PAC clone, 133F4, does not (lanes 2 and 3). E: Mapping of fibroblast growth factor4 (fgf4) to the hab locus using the size polymorphism in expressed sequence tag AI957973 on the recombinants from the hab haploid panel. Cutting with AciI revealed a restriction fragment length polymorphism between hab and wild-type embryos (lanes 3 and 4). F: Mapping of fibroblast growth factor3 (fgf3) to the hab locus using a single-stranded conformation polymorphism. WT, wild-type.

As part of a related project, to identify the molecular nature of the hab locus, linked candidate genes were eliminated based on their recombination with the hab locus. These genes became valuable markers in the process of mapping the hab locus. Typically, using the 3′ region of a cloned fragment, we identified a size polymorphism that could be visualized on an agarose gel and then checked if the candidate was recombinant in individuals that were known to have recombination events close to the hab locus. Such an example is shown in Figure 1E, where we identified recombination events between hab and the closely linked gene FGF4. In some cases, when no polymorphism to the gene was present in the test cross, we identified a clone containing the gene of interest from a large-insert genomic PAC library (Amemiya and Zon, 1999) and then identified a polymorphism in the ends of the clone or used a closely linked microsatellite marker on the same clone. The microsatellite marker was then used to test linkage of the candidate gene to the hab locus as is shown in Figure 1D, where VN cadherin, which was linked to the marker z3008 on PAC clone 30C24, was subsequently shown to be over 20 cM from the hab locus. When no polymorphisms could be found using simple size polymorphisms on agarose gels, we used single-stranded conformation polymorphisms (SSCP) methodology, which successfully identified polymorphisms in approximately 90% of cases attempted. Such an example is shown in Figure 1F, where we identified recombination events between hab and the very closely linked gene FGF3.

Finally, using the individual haploid mapping panels, we established that each of the mutant loci map between marker z1239 and the centromeric marker z7958, a distance of approximately 20 cM. Additional mapping placed all the mutants at approximately the center of this interval, suggesting that they belong to the same locus, which we term the half baked locus (Fig. 1C). Because hab and ava were tightly linked to marker z6852, linkage to this marker subsequently was used to genotype individual embryos for the hab locus.

Epiboly Mutants Show Allele Specific Epiboly Phenotypes

The epiboly mutants display an array of dominant phenotypes that are specific to each individual allele, and we have suggested previously that the recessive epiboly phenotypes may display differences specific to each allele as well (Kane et al., 1996). To measure the fluid changes of epiboly, we staged clutches at the eight-cell stage; as the first embryos of the clutch reached the 100% epiboly stage, the phenotypes of the remainder of the clutch were quickly scored as to the degree of epiboly using the photographs in Figure 2A as guides. Afterward, we genotyped the embryos based on the segregation of closely linked markers that flank the hab locus.

Figure 2.

The epiboly phenotypes of the epiboly mutant alleles. Comparison of epiboly phenotypes between the different epiboly mutants, scoring the range in epiboly defects observed within a single clutch at 100% epiboly stage, grouping embryos by genotype. A: Photographs of “guide embryos” used for scoring. Phenotype “0” indicates that embryos dissociated before 100% epiboly, usually because the blastoderm peeled off of the yolk cell. B: Variation in epiboly of wild-type (WT) strains. C: Variation of the weg allele. D: Variation of the hab allele. E: Variation of the ava allele. Note that, in C, D, and E, the results are compiled from multiple clutches, and the embryos are selected before genotype or phenotype is known. Note that the expected ratio is 1:2:1 for wild-type: ZMD heterozygotes, zygotic maternal dominant homozygotes.

Analysis of three of the “wild-type” strains maintained in the laboratory showed that, as the first embryos reach the end of epiboly, less than 10% were delayed, and in the puma strain, the usual outcross strain for the epiboly mutants, less than 2% are delayed (Fig. 2B). Analysis of the weg allele showed that the mutant phenotype was completely recessive and that no slowing of epiboly could be seen in either heterozygous or homozygous wild-type embryos (Fig. 2C). Interestingly, the early dissociation phenotype of the weg−/− mutant is not evident in this work; this represents a shift in the phenotype of the mutant, perhaps caused by moving the allele to the puma background. (The name “Weg” originates from a German idiom for “gone” or “disappeared”.)

When homozygous, the epiboly alleles expressed a slowing of epiboly followed by a retraction of the blastoderm (Fig. 2D,E). In these cases, the mutants attained approximately 60 to 70% epiboly and then their blastoderms begin to retract. This phenotype was less severe in ava and law compared with hab and only rarely expressed in the weg phenotype. During this process, many embryos dissociated before completing epiboly, giving the “0” phenotype in Figure 2. This dissociation phenotype was variable: typically the blastoderm peeled off the yolk cell, but sometimes large numbers of cells detached from the blastoderm and the yolk cell lysed. When heterozygous, the ZMD mutants also expressed a slowing of epiboly, reaching 70% to 90% epiboly when wild-type siblings were completing epiboly; however, in these cases, the mutant blastoderm did not retract, and the mutants slowly completed epiboly. ZMD mutants only rarely dissociated.

The ZMD epiboly phenotypes of hab was the most severe of all the dominant alleles (compare D to E in Fig. 2); indeed, the similarity between the phenotype distributions of the ZMD hab−/+ mutant to the simple recessive weg−/− mutant is striking. Such correlations were also seen among the neural tube phenotypes of the mutants, e.g., the dominant phenotype of hab is more severe than that of law and ava. However, there was a curious exception to this correlation of severity: whereas ava−/−, law−/−, and weg−/− mutants never survive past the five-somite stage, some hab−/− embryos survived through segmentation stages and did not dissociate until after 24 hr.

Note that, for all of the epiboly-dominant alleles, the wild-type progeny of heterozygous females expressed a slight slowing of epiboly (Fig. 2), a dominant maternal effect. Regardless of the genotype, progeny from heterozygous males and homozygous wild-type females never expressed a slowing of epiboly (data not shown).

half baked Mutants Have a Widened Dorsal Axis in Ectodermal Structures

To determine whether the epiboly arrest was the result of a misallocation of cell fate, we examined patterns of RNA expression in wild-type and mutant embryos during the epiboly and segmentation stages, a period extending from approximately 4 to 20 hr of development. Fate map experiments using lineage tracing techniques were not attempted, because the early lethality of mutants hindered documentation. Our studies can be roughly divided into an analysis of germ layer specification during gastrulation and analysis of tissue differentiation during somitogenesis. Note that, in all the experiments, the wild-type embryos are siblings of the mutant embryos and were produced from heterozygous females. However, we have not found any differences in either such wild-type embryos or ZMD heterozygote embryos before 75% epiboly, and we believe that the expression patterns in these embryos appear completely wild-type.

To investigate axial tissues from all three germ layers, we examined the expression of goosecoid (Schulte-Merker et al., 1994a; Thisse et al., 1994) and axial/forkhead1/FoxA2 (Strähle et al., 1996; Odenthal and Volhard-Volhard, 1998) gene products. At 40% epiboly, goosecoid expression in the dorsal axis was identical between hab−/− and wild-type embryos (Fig. 3A), indicating that the axial layers of the mutant were normal in early epiboly. However, by late epiboly, the expression pattern of FoxA2 (Fig. 3B) was noticeably shorter anteroposteriorly and slightly wider mediolaterally compared with wild-type siblings. In contrast, the nonaxial endodermal expression domain of axial appeared normal.

Figure 3.

Specification in hab mutants before and during gastrulation. In the following experiments, all hab−/+ embryos are zygotic maternal dominant (ZMD) embryos; during these stages, in situ analysis cannot distinguish between the wild-type and dominant phenotypes. Embryos were genotyped after completion of photography as outlined in the Experimental Procedures section. All embryos are dorsal views with the animal pole to the top, and arrows indicate the width of the axial expression pattern, except C, which is a side view with dorsal to the right and arrows indicate the anteroposterior extent of expression. Arrowheads indicate the edge of the blastoderm margin in mutant embryos. A: Expression of goosecoid, 40% epiboly, in wild-type and hab−/−, dorsal view. B: Expression of FoxA2/axial/forkhead 1, 75% epiboly in wild-type and hab−/−. C: Expression of gata2, 75% epiboly in wild-type and hab−/− (C). Arrows indicate the expression domain. D:sonic hedgehog, tail bud stage in wild-type and hab−/−. Note the absence of expression in the anterior neural plate of the mutant. E,F:FoxB1/mariposa/forkhead 3, 75% epiboly and tail bud stage in ZMD hab−/+ and hab−/−. E′ and F′: Arrows indicate widening of the midline in the mutants. F′: Note gaps in the midline and absence of anterior forebrain expression. G,H:no tail, 75% epiboly and tail bud stage in wild-type and hab−/−. Ectopic forerunner cell clusters are indicated by asterisks. Wild-type embryos in G are from hab−/+ mothers. I,J:spadetail, 75% epiboly and tail bud stage in wild-type, ZMD hab−/+ and hab−/−. K,L:paraxial protocadherin, 75% epiboly and tail bud stage in wild-type, ZMD hab−/+and hab−/−. Note that, in the hab−/− mutants, expression of both spadetail and paraxial protocadherin in the prechordal plate is retained longer.

We examined ectodermal fates in hab−/− mutants using the expression of gata2 (Read et al., 1998), forkhead 3/mariposa/FoxB1 (Odenthal and Volhard-Volhard, 1998; Varga et al., 1999), and sonic hedgehog (Krauss et al., 1993) gene products. At 75% epiboly, gata2 is expressed in the ventral non-neural ectoderm; this domain appeared smaller yet more-intensely stained in hab−/−mutants (Fig. 3C). sonic hedgehog, which is expressed in the ventral neuroectoderm, revealed related but more-severe defects. There was a failure of the posterior neuroectoderm to reach the midline, causing the axis to be bifurcated posteriorly, with each posterior half wrapping around the yolk cell at the approximate location of the germ ring (Fig. 3D). Also, expression of sonic hedgehog was absent anteriorly, detected in what is normally the forebrain region (compare Fig. 3D with 3F), indicating perturbations in anterior forebrain patterning. FoxB1 expression, which is expressed in the dorsal neuroectoderm, reveals that neural fates were mediolaterally wider in hab−/−mutants by late epiboly (Fig. 3E) and markedly wider by tail bud (Fig. 3F). Altogether, in the ectoderm, there is a general impression of a failure of convergence to move fate domains toward the dorsal side of the embryo. Also, subtle defects in patterning at the animal pole may be present.

We examined mesodermal tissues with the expression of no tail, the homologue of the Brachyury gene (Schulte-Merker et al., 1994b), spadetail (Griffin et al., 1998), and its downstream target gene paraxial protocadherin (Yamamoto et al., 1998). Expression of no tail mRNA illustrated that the notochord anlage was shorter anteroposteriorly in hab−/−mutants by 75% epiboly (Fig. 3G). By tail bud, these defects were more marked (Fig. 3H) and the notochord anlage was slightly wider mediolaterally. Complementary defects were seen in the paraxial mesoderm with the expression of both spadetail (Fig. 3I,J) and paraxial protocadherin (Fig. 3K,L).

The expression of no tail (Fig. 3G,H) also revealed a defect in a population of cells, termed the dorsal forerunner cells, that normally migrate as a cohesive group just vegetal of the deep cell margin (Cooper and D'Amico, 1996; Melby et al., 1996). It has been documented that at least a portion of these forerunner cells appear to migrate ahead of the blastoderm in volcano mutants, a weg-like mutant isolated in the Boston Screen (Solnica-Krezel et al., 1996). We found that, rather than forming a single group of cells, forerunner cells in the hab mutants coalesce into multiple clusters (asterisks in Fig. 3G,H′), which migrate ahead of the blastoderm margin. Interestingly, the forerunner defect segregates as a maternal phenotype: all embryos derived from habdtv43/+ females, regardless of genotype, have multiple forerunner clusters. This finding is shown in a homozygous wild-type embryo in Figure 3G.

We examined the effects of the hab mutation on later development using the expression patterns of the gene products hatching gland gene (Thisse et al., 1994), Krox 20 (Oxtoby and Jowett, 1993), forkhead 6/FoxD3 (Odenthal and Volhard-Volhard, 1998), alpha-collagen 2a (Yan et al., 1995), deltaA (Appel and Eisen, 1998; Haddon et al., 1998), and Zn-12 (Trevarrow et al., 1990), all of which mark differentiated tissues. The expression of hatching gland gene in the hatching gland cells of the prechordal plate of hab−/−mutants at the five-somite stage indicated that these cells have begun to differentiate normally. However, the area of expression was larger and the borders uneven, and often, there were ectopic cells posterior to the normal region of expression (Fig. 4A). Such changes in distribution may be related to the semidominant enlarged hatching gland trait in heterozygous siblings at 24 hr. Krox 20 expression in rhombomeres 3 and 5 of the mutant also revealed that segmentation of the hindbrain was occurring properly. Similarly, at later stages, expression of forkhead 6/FoxD3, alpha-collagen 2a, and deltaA (Fig. 4B–D) revealed that somites, floor plate, notochord, hypochord, and many specific neuroectodermal structures differentiate properly, albeit within the context of the arrested epiboly phenotype. For example, somites and spinal cord, which have lateral and ventral progenitors, always bifurcated and extended both ways around the equator of the embryo. However, the notochord, which originates from midline progenitors, extended randomly to either side of the midline, along the equator. Moreover, in surviving 1-day hab−/− mutants, some individual neurons formed axon tracks extending around this germ ring, revealed by the Zn12 antibody (Fig. 3D). These neurons must form normal synapses on muscle tissue, for such survivors display normal rhythmic twitching movements.

Figure 4.

Differentiation in hab mutants after gastrulation. Embryos were genotyped after completion of photography as outlined in the Experimental Procedures section. A:Expression of hatching gland gene and Krox20, five-somite stage in wild-type (A) and hab−/−(A′), dorsoanterior view. Note ectopic hatching gland cell (arrow) below pollster in the mutant. B:FoxD3/forkhead 6, 10-somite stage in zygotic maternal dominant hab−/+ and hab−/−. Side view except for B″, which is a dorsal view. Note weak expression in mutant tail bud. Dorsal view shows bilateral somites (indicated by asterisks) forming on either side of the germ ring. C:alpha-collagen2a, 14-somite stage in wild-type (side view) and hab−/− (dorsal view of kinked notochord). C′: Higher magnification image of the boxed regions. D:delta A,18-somite stage in wild-type (dorsal view, D) and hab−/− (dorsal view and side view, D′). E: Zn12 antibody, 24 hr in wild-type (E) and rare hab−/− escaper (E′), dorsal view. Note bifurcation of axis in mutant. cg, cranial ganglia; fp, floor plate; hb, hindbrain; hcd, hypochord; llf, lateral longitudinal fascicle; mb, midbrain; ncd, notochord; ncr, neural crest; os, optic stalk; ov, otic vesicles; pol, polster; r3, rhombomere3; r5, rhombomere5; rb, Rohan–Beard cells; sc, spinal cord; tbm, tail bud mesenchyme; tg, trigeminal. Arrows indicate the width of the axis.

In summary, although epiboly is arrested during gastrulation, the effects on specification of germ layers or differentiation of tissues in hab−/− mutants are subtle. Indeed, the differentiation of posterior structures occurs in situ at positions reminiscent of their origin in the fate map. The slight perturbations observed in spatial patterns, such as the shortened but widened axis, appear to be the result of morphological changes of two sorts in mutant embryos: the block in epiboly, which stops the vegetalward extension of tissues around the yolk, and a reduction in convergence movements, which broadens the axis mediolaterally. Moreover, the defects in convergence seem to appear earlier and be more severe in the epiblast than in the hypoblast.

Ectoderm Is Perturbed in the Enlarged Hatching Gland

hab−/+ mutants exhibit an enlarged and ragged hatching gland, a trait that segregates as a semidominant phenotype (Fig. 5A,B). In the zebrafish, the hatching gland is composed of two cell populations: mesodermally derived hatching gland cells, which express the hatching gland gene (hgg), and ectodermally derived support cells (Kimmel et al., 1990; Thisse et al., 1994). Because the number of hatching gland cells is unaffected in hab−/+ mutants (Kane et al., 1996), it is possible that the defect is due to loss of hab function in the ectodermal portion rather than the mesodermal portion of the hatching gland. Therefore, we made double mutants of hab to one-eyed pinhead (oep), a mutation that perturbs mesendoderm formation, lacks the prechordal plate, and lacks hatching gland cells (Hammerschmidt et al., 1996; Schier et al., 1997a; Strähle et al., 1997; Warga and Kane, 2003). At 24 hr of development, we examined the embryos for the presence of the hatching gland defect in the double mutants, individuals that should be devoid of hatching gland cells. From a cross of an oeptz257/+; hab+/+ female to an oeptz257/+; habdtv43/+ male, one quarter of the progeny segregated with the oep phenotype, and half of these, the expected number, displayed the enlarged hatching gland-dominant phenotype (Fig. 5A–D). All the embryos were assayed for hgg expression to verify the lack of hatching gland progenitors. Siblings from both of the oep classes lacked hgg-positive cells (Fig. 5C′,D′), including those that displayed the hab enlarged “hatching gland.” This result indicates that the hgg-positive mesodermal portion of the hatching gland is not necessary for the enlarged hatching gland phenotype, and strongly suggests that hab is necessary in the support cells of the hatching gland, cells that are ectodermally derived.

Figure 5.

half baked acts in the ectoderm of the hatching gland. A–D: Twenty-four hour progeny of oep/+ female × oep/+; hab/+ male: wild-type (A), hab/+ (B), oep (C), and hab/+; oep (D). Note that hab/+ embryos are not zygotic maternal dominant mutants. A′–D′: Expression of hgg at 24 hr in the hatching gland of similar embryos after in situ hybridization; ventral view. Arrows indicate semidominant enlarged hatching gland trait of the hab−/+ mutant. Note absence of expression in the oep mutants and the double mutants.

Involution of Marginal Deep Cells and Epiboly of the EVL Is Unaffected in hab

The block of epiboly in hab−/− mutants might result from an increase in the number of cells that involute during epiboly as seen in the lefty1/2 antisense knockdown embryos (Branford et al., 2000; Feldman et al., 2002). Although such a phenotype should cause a marked reapportionment of cells between the ectoderm and mesendoderm—which is seen in the lefty1/2 knockdowns but not in hab−/− mutants—we checked this possibility using four-dimensional (4D) Nomarski time-lapse analysis. The involution of marginal epiblast cells was recorded just after the onset of involution. Because each time point contained a stack of 10 to 15 focal planes, we could compute the x, y, and z coordinates for the nuclei of each individual cell, and using these data, geometrically reorient the embryos to show a cross-section at the blastoderm margin. This strategy is shown for a period of 30 min for the hab−/− and wild-type embryos (Fig. 6A,B, and Supplementary Figure S1, which can be viewed at These experiments found no defects in the involution of marginal mesendodermal cells (n = 3 hab and 3 wild-type embryos).

Figure 6.

half baked is not necessary for involution of the mesendoderm or epiboly of the epithelial-like enveloping layer (EVL). A,B: Cell movement during involution, showing 30 min, approximately from germ ring to shield stage in hab and wild-type (WT) siblings. Face view recorded Z-stacks were geometrically reoriented to appear in optical cross-section; EVL cell trajectories (thick yellow), yolk syncytial layer nuclei (thick blue–green), and deep cells (thin multicolored; blue, 5 μm deep to red, 50 μm deep). Inset in B indicates location of recordings; arrows indicate general direction of movement of like-colored tracings; ordinate (red numbers) are percentage epiboly. C: Animal pole view of a shield stage hab embryo showing thickness of the germ ring. Slashes indicate location of measurements at dorsal (top), dorsolateral, and ventral positions on the blastoderm. D: Average ± standard error germ ring thickness in hab−/− (clear), zygotic maternal dominant (ZMD) hab−/+ (red), and wild-type (green) embryos. E–H: Lateral views of progeny from a sqt/+; cyc/+; hab/+ incross at tail bud stage: wild-type (E), sqt; cyc (F), hab (G), sqt; cyc; hab (H). Arrows indicate the extent of epiboly. G′,H′: Dorsal views of G and H. I: Face view of EVL trajectories from four hab−/− embryos. The interval is 30 minutes, and marks along the axes (red numbers) indicate latitude (percentage epiboly) and longitude (degrees from dorsal). Inset in I indicates location of recording; arrow indicates direction of movement. J: Average ± standard error rate of EVL epiboly (y-vector) movement vs. time postfertilization in hab−/− (open squares), ZMD hab−/+ (red triangles), and wild-type (green circles) embryos.

To quantify involution in a larger number of embryos, we measured the thickness of the germ ring at shield stage (Fig. 6C), a method that was used to demonstrate an increase in the rate of involution in lefty1/2 antisense knockdown embryos (Feldman et al., 2002). The thickness of the germ ring was measured at the dorsal midline, as well as at 45 degrees and 180 degrees from dorsal. No significant differences were found in the thickness of the germ ring between hab−/−, ZMD hab−/+, and wild-type embryos, regardless of location (Fig. 6D), indicating that involution and mesendoderm formation are morphologically normal at this stage.

In squint; cyclops double mutants, mesendodermal tissues fail to be specified normally and involution movements are absent. Also, the axes of such double mutants are shortened and do not extend to the animal pole. Still, except for a small notch that forms on the dorsal side, they complete epiboly normally (Feldman et al., 1998, 2000). To test whether a reduction in involution movements affects the epiboly phenotype of hab−/− mutants, we made triple mutants of hab with squint and cyclops (Fig. 6E–H). In the triple mutants, the absence of involution had no effect on the hab epiboly phenotype on the ventral and lateral sides of the blastoderm (arrows in Fig. 6G,H), suggesting that involution is not necessary in these regions for expression of the hab phenotype.

However, on the dorsal side of the triple mutants, the epiblast margin arched toward the animal pole (Fig. 6H′) and the large gap was covered only by the thin epithelium of the EVL. This phenotype is a very strong enhancement of the notching phenotype expressed in the sqt; cyc double mutants at the same stage (not shown). The widening of the notch on the dorsal side may be related to the perturbations in mesendodermal cell fates in the triple mutants; however, it also may be due to an impairment of convergence. In any case, in the local region of the dorsal side, squint; cyclops function is necessary for the normal hab phenotype and lack of squint; cyclops enhances the hab−/− epiboly defect.

Using low-power magnification recordings of the shape of the embryo, we previously noted that the yolk cell continued epiboly normally in hab mutants, and by inference, that the EVL was normal as well (Kane et al., 1996). To confirm directly that the EVL was moving normally over the epiblast, we analyzed the epiboly of this layer in hab−/−, ZMD hab−/+, and wild-type embryos using 4D Nomarski time-lapse analysis. After recording, analyzing, and geometrically re-orienting the embryo with the animal pole to the top as described above, we measured the rate of migration of cells at approximately the equator of the blastula. A face view is displayed in Figure 6I, showing the combined trajectories of EVL cells from four different hab−/− mutants. We could discern no difference in the rate of EVL epiboly between the 50% and 90% epiboly stages in hab−/−, ZMD hab−/+, and wild-type embryos (Fig. 6J), which agrees with our earlier observations (Kane et al., 1996).

Abnormal Neural Tube of the ZMD Phenotype

ZMD hab−/+ mutants often exhibit gaps along the midline of the neural keel, with dorsal detached cells rounding up and accumulating along the trunk and tail of the embryo in the early segmentation stages (Kane et al., 1996). These dorsal detached cells are a very reliable indicator of the heterozygous genotype and, in certain backgrounds, are useful for sorting embryos to be raised for stocks. Although subtle, other aspects of the ZMD neural tube defect could be identified in mid-segmentation stages. Before the tail bud begins to evert, when the embryo is still wrapped around the yolk cell, the head to tail distances were longer in ZMD mutants and the embryos correspondingly shorter (Fig. 7A,B). In ZMD siblings, rather than the oval shape of the wild-type neural tube (Fig. 7A″), the anterior neural tube of the mutant appeared triangular in shape when viewed in deep optical cross-section and the dorsal portion of the neural tube was flatter (Fig. 7B″) When viewed from the dorsal side of the embryo, the neural keel appeared wider (Fig. 7B‴). Similar defects were also observed for the somites, albeit to a lesser degree. Both the widening of the neural tube and the somites could also be visualized with probes to N-cadherin mRNA (Fig. 7C), which is expressed in the neural tube and the somites at this stage (Bitzur et al., 1994). This marker further revealed the gaps that form along the midline of the mutant neural tube (Fig. 7C′, white arrowhead), a phenotype not observed in the convergence mutants trilobite, silberblick, or knypek. In ZMD law−/+ and ZMD ava−/+ mutants, the deformities in neural tube morphogenesis, as well as the detached cell phenotype, occurred most strongly near the tail, approximately at the location of the yolk plug, suggesting a connection between the retardation of epiboly and the neural tube defects. However, in ZMD hab−/+ mutants, these defects were much more extensive and often occurred the entire length of the neural tube. Correlating with the severity of the neural phenotype, other ectodermal structures, such as the eye and ear placodes, tended to be smaller and misshapen in more severe ZMD mutants (Fig. 7A′,B′).

Figure 7.

Convergence and extension of the neural axis is defective in zygotic maternal dominant (ZMD) heterozygotes. A,B: Wild-type (WT, A) and severely affected (B) ZMD ava−/+ sibling embryos at 10-somite stage; side view; white arrowheads demarcate the head to tail distance of the embryo. A″,B″: Dorsal view focused deep, showing the neural and somitic tissue in optical cross-section. A‴,B‴: Dorsal view focused shallowly, showing the width of the neural (arrows) and somitic (arrowheads) tissue along the mediolateral axis. A′,B′: Outline of the notochord (n), neural keel (nt), and somites (s), as traced from view in orientation (A″,B″), and of the eye (e), as traced from view in orientation A and B. C: Expression of N-cadherin (cdh2), in wild-type (C) and ZMD hab−/+ (C′) embryos, 10-somite stage, dorsal view. Arrows indicate width of the neural tube and arrowheads width of somitic mesoderm. White arrowhead points out the gap in the midline of the neural tube. D,E: Expression of deltaA in wild-type (D) and ZMD hab−/+ (E) embryos, 10-somite stage, dorsal view. Arrow indicates an extra row of cells in the midline of the neural tube, and the arrowhead points out the gap in the midline of the neural tube. D′–E‴: Transverse cross-sections as indicated for wild-type (D′,D″) and ZMD hab−/+ (E′,E″,E‴). Outlines below the cross-sections demarcate the boundaries of deltaA staining in each neural cross-section. For reference, the dorsal detached cells and notochord, which do not express deltaA, are also outlined. F,G: Anti-acetylated tubulin staining (αAT), in wild-type (F) and ZMD hab−/+ (G) embryos, 24 hr, dorsal view. Arrows indicate extra rows of cells in the midline of the neural tube.

The neural tube phenotype expresses considerable variability among ZMD mutants, both between the alleles themselves and between individual clutches of embryos. The variability of the phenotype could be seen in survival rates: whereas survival to adulthood could be as high as 50% of the ZMD mutants, usually it was less than 5%, especially with the hab allele. Weak ZMD law−/+ and ZMD ava−/+ mutants slowly recovered, so that the neural tube defects were relatively mild by 24 hr. Most of these embryos survived to adulthood; in a manner, this finding is similar to the silberblick homozygote, which suffers mild defects in dorsal convergence but later recovers (Heisenberg et al., 2000). In aquaculture, we have noticed no apparent changes in the behavior of the adult heterozygotes, e.g., in feeding or evasion to capture.

In ZMD hab−/+ individuals that displayed stronger phenotypes, we examined the pattern of deltaA (Appel and Eisner, 1998; Haddon et al., 1998), which at 24 hr, labels most neurons weakly and nascent neurons strongly. This experiment showed that the neural tube of the strong ZMD mutant was much broader than normal (Fig. 7D,E), especially in the trunk and tail regions of the axis. Moreover, the width of the neural keel was variable within individual embryos, sometimes being 50% wider in some regions compared with others. Closer examination of the cells that strongly express deltaA revealed dramatic disorganization of the neural tube, which can also be seen in hand sections through various levels of the axis. In particular, there were often ectopic rows of cells in the midline of the neural tube (Fig. 7E). Staining with an antibody to acetylated tubulin, which visualizes differentiating neurons, confirmed these observations (Fig. 7F,G). Most of these strong ZMD mutants do not survive; thus, it is unclear to what extent the neural tube ultimately closes and, subsequently, what morphological consequences might be retained at the larval stage of development.

Cell Autonomy of half baked

To determine the site and mode of hab gene function, we moved cells of different genotypes by transplantation either into ZMD mutants or wild-type embryos. Previously, we had transplanted cells between hab−/− and wild-type embryos and then examined the resulting chimerae during late gastrula stages and at 24 hr. However, no differences were found between these experiments and their controls (data not shown). In the following experiments, we focused our analysis on the altered morphogenesis of the neural tube in ZMD ava−/+ mutants, selecting heterozygous females that produced embryos with extreme dorsal detached cell phenotypes. We used the ava allele for these experiments because of the local differences where the neural tube phenotype was expressed, i.e., in the tail and not in the trunk, and because of the increased survival of the ZMD ava heterozygotes.

We first tested whether wild-type cells transplanted into ZMD ava−/+ mutants could incorporate into the dorsal detached cell cluster (Fig. 8A). Of a total of 21 ZMD ava−/+ mutant chimerae, 7 of these contained wild-type donor cells within the detached cell clusters (Fig. 8B–D). In these cases, donor wild-type cells were always intermingled among the mutant cells (Fig. 8B′–D′). The mutant dorsal detached cells must have originated, at least in part, from the deep cell domain, because EVL cells are not moved or created in transplant operations (Ho and Kimmel, 1993). Moreover, in these experiments, wild-type cells were capable of acquiring aspects of the heterozygous phenotype when transplanted into ZMD ava−/+ mutants, a nonautonomous result.

Figure 8.

half baked acts autonomously and nonautonomously. A: Experimental approach for moving wild-type (WT) cells into zygotic maternal dominant (ZMD) ava−/+ embryos. B–D: Three examples of wild-type cells transplanted into ZMD ava−/+ embryos. B′–D′: Chimeras are shown at the 10-somite stage, and boxes indicate the detached cell area, shown at higher magnification. E: Experimental approach for moving ZMD ava−/+ and ava−/− cells into wild-type embryos. Note that, in contrast to the wild-type hosts, the wild-type donors are produced from ava−/+ parents. F,G: Two examples of ava−/− mutant cells transplanted into wild-type embryos. F′,G′: Chimeras are shown at the 10-somite stage, and boxes indicate the detached cell area, shown at higher magnification. In A, the hosts were genotyped based on phenotype. In E, the genotypes of the donors were determined as described in the Experimental Procedures section.

To test whether ava mutant cells could integrate into tissues of a wild-type host, we moved cells from embryos derived from an incross of heterozygous parents into wild-type embryos (Fig. 8E). Of 24 chimerae, 6 contained ava−/− mutant cells, and 2 of these had small clusters of dorsal detached cells (Fig. 8F,G). Whereas the detached cells were composed primarily of mutant donor cells, in both cases, they included wild-type cells from the host. At 24 hr, the remaining mutant donor cells (those that did not detach) were found primarily in ectodermally derived tissues. These results suggest that the ava gene product can act autonomously, most probably in the ectoderm, to produce the detached cell phenotype. The mutation also acts nonautonomously to recruit wild-type cells into the detached cell cluster. We also observed that the mutant donor cells can integrate normally into the host embryo, a result that has been noted previously (Kane et al., 1996).


half baked locus

Our mapping of the epiboly mutants places the genes near the centromere of Linkage Group 7. This region is in a neighborhood of many important genes that act in early development, including sonic you (sonic hedgehog), engrailed 2a, cyclin E, cyclin D1, FGF3, and FGF4, sox17, tc4, apolipoprotein Eb, achaete-scute complex-like 1b, and ubiquitin-conjugating enzyme E2I (Postlethwait and Talbot, 1997; Shimoda et al., 1999; Postlethwait et al., 2000; Woods et al., 2000). We propose that the epiboly mutants are at a single locus, termed the half baked locus, named after the first mutant isolated. The mutations must be in a single gene or in a complex of closely linked, functionally similar genes. If in a single gene, there are three arrays of phenotypes that have been separated previously on the basis of their dominant phenotypes: weg, which has no dominant phenotype; law and ava, which have ZMD phenotypes; and hab, which has an extreme ZMD phenotype and partial dominant phenotypes. In this report, we have added to these characterizations, showing that these three arrays of phenotypes extend to expression of the recessive epiboly phenotype.

Although not the normal case, Drosophila has many loci that have multiple phenotypes. For example, the Notch locus has an intimidating array of diversely named alleles, including deletions that express dominant haplo-insufficient phenotypes, and nucleotide transversions some that express dominant antimorphic phenotypes and some that express recessive hypomorphic phenotypes, and remarkably, some of these alleles even complement each another. In fact, even after the nucleotide transversions of alleles at the Notch locus are identified, it takes a complex series of genetic tests to sort out the hypomorphic and hypermorphic natures of each allele (Go and Artavanis-Tsakonas, 1998).

If a single gene, how would the epiboly alleles be ordered? One possibility is that hab, lab = ava, and weg represent a series of hypomorphic alleles, with hab being the strongest, an explanation based on the hypothesis that the hab locus would be haplo-insufficient. In this case, the hab allele itself may not be null, for true null alleles may be heterozygous lethals and not represented in the mutant collection. Another possible hypothesis is that weg is a hypomorph and law, ava, and hab are hypermorphic gain-of-function alleles, which are normally dominant. At the present time, we cannot distinguish among these or other possibilities.

The combination of recessive and ZMD phenotypes forms an impressive phenotypic series, and such a series is a useful genetic tool in judging the response of the hab locus to other mutants or treatments. In order of increasing severity, the recessive mutants would be ordered as weg, ava = law, and hab. The ZMD mutants would be ordered as ZMD ava = ZMD law, and ZMD hab. However, because the ZMD hab appears similar to the weg recessive, the entire series, from weakest to strongest, would be ZMD ava = ZMD law, ZMD hab = weg, ava = law, and hab.

half baked Has Minor Effects on Tissue Specification and Differentiation

Based on in situ hybridization with a variety of markers encompassing all three germ layers, specification appears largely unaffected in hab mutants. Hence, hab falls into a class of mutants such as trilobite and knypek, which are necessary for morphogenesis of the embryo but tend to have subtle secondary effects on other developmental processes. For example, changes such as the shortened and widened axis are a likely consequence of a slowing of convergence. Similar changes in gene expression were seen in volcano, another epiboly mutant not yet shown to be allelic to hab (Solnica-Krezel et al., 1996). Because of the epiboly arrest in hab mutants, lateral and posterior tissues differentiate in situ around the germ ring, reminiscent of their origins in the early gastrula stage fate map (Kimmel et al., 1990; Warga and Volhard-Volhard, 1999). Moreover, the defects in the morphogenesis of the neural tube in the ZMD mutants indicate that hab also acts in later types of cell movement, because convergence movements are necessary during neural tube formation (Keller et al., 1992; Schmitz et al., 1993; Kimmel et al., 1994; Papan and Campos-Ortega, 1994; Elul et al., 1997; Concha and Adams, 1998; Goto and Keller, 2002).

Differentiation of tissues around the germ ring brings to mind the curious idea of “concrescence,” proposed for the teleost embryo by T.H. Morgan (Morgan, 1895), who suggested that the axis of the embryo formed independently as two separate fields, and this “germ ring” then zippered together at the midline, beginning at the shield, by the combined actions of epiboly and convergence of lateral regions to the posterior dorsal midline. In mutants such as hab, where there is a cessation of deep cell epiboly, the two posterior halves of the body axis do indeed develop independently of each other on separate halves of the yolk cell, lending support to such a hypothesis. Interestingly, these germ ring tissues of the hab mutants lack important interactions normally obtained at the midline; consequently, they are not normal in terms of size or shape.

Relationship Between Epiboly and Other Morphogenetic Movements

Besides the widened axial structures shown in our in situ analysis, the results with the sqt; cyc; hab triple mutants lend further support to the idea that convergence is slowed in hab mutants. This experiment, done to examine the effects of the ablation of involution on the hab phenotype as discussed below, unexpectedly revealed that the phenotype of the triple mutant expressed a large gap on the dorsal side of the epiblast. Formally, this gap is a local slowing of epiboly and demonstrates that sqt; cyc function is necessary on the dorsal side of the gastrula to compensate for the lack of hab function. In sqt; cyc mutants, the Nodal pathway is down-regulated, causing the ablation of the majority of the mesendoderm. Moreover, the residual morphogenetic movements on the dorsal side seem affected more severely than elsewhere in the double mutant, with dorsal cells moving away from the midline in a seemingly haphazard way (Warga and Kane, 2003). Nevertheless, lateral cells persist in their movement toward the dorsal side, creating a large mound of cells centered approximately in the position of the future hindbrain. This finding suggests that, in the double mutants convergence, a lateral cell behavior is normal, but, in the triple mutants, it is not. An absence of convergence on the lateral side (from the hab phenotype) eliminates the “pile of cells” on the dorsal side, and in conjunction with the abnormal behavior of cells on the dorsal side (from the sqt; cyc phenotype), a large gap forms at the midline. No such gap forms in the hab mutant. Therefore, Nodal-dependent movements must compensate for the lack of hab function on the dorsal side of hab mutants. Altogether, the analysis of the triple mutants suggests two points. First, an interconnection may exist between the hab and Nodal pathways or the movements they control. Second, some of the components of morphogenesis that power epiboly on the dorsal side may be different than those on the lateral and ventral sides.

Disruption of convergence would be an ideal candidate to explain an epiboly phenotype where the epiblast is slowed but the EVL and yolk are not. Convergence in the epiblast is thought to be powered by mediolateral intercalation, a movement that would act autonomous to the tissue that it shapes. In Xenopus, late convergence movements were suggested to be sufficient to drive blastopore closure (Keller et al., 1985), and indeed, experiments in Xenopus suggest an interconnection between convergence and extension movements and epiboly. For example, the overexpression of Xror2, a receptor tyrosine kinase that affects convergence movements through the noncanonical Wnt pathway, also causes an epiboly arrest (Hikasa et al., 2002). In contrast, fish mutants that specifically slow convergence—many of which disturb the WNT signaling pathway—do not slow epiboly. However, this comparison is not completely appropriate, as the convergence defects in these convergence mutants tend to occur toward the end of or after the epiboly period.

Convergence could help propel the latter advance of epiboly through a type of purse-string movement: the forces acting in the epiblast, which move cells dorsally, also could help to contract the marginal region of the blastoderm, pinching the blastoderm over the ventral yolk cell, engulfing it (Keller and Trinkaus, 1987; Fink and Cooper, 1996). Indeed, a cortical belt has been observed in the zebrafish EVL, which could facilitate such a model (Zalik et al., 1999). Although no such belt has been found in the epiblast, the cells of the ventral side of the embryo are aligned parallel with the blastoderm margin, as if under tension, and when they divide, their divisions tend to align along those presumed lines of force (Concha and Adams, 1998). Hence, if chains of cells wrapping around the ventral side of the blastoderm are the string of the purse, it is possible that such a mechanism might be operating in the epiblast of the blastoderm.

A model we have discarded is one where the epiblast is somehow attached to the EVL–yolk junction and pulled along by means of the yolk cell motor of Trinkaus (1951). This process would involve a connection between the site where involution into the hypoblast is occurring in the blastoderm and the site where the EVL attaches to the yolk cell membrane. Because of the complexity of this interaction, comprising cells that are participating in an epithelial to mesenchymal transition as they move from the epiblast into the hypoblast, it is difficult to speculate on the myriad of attachment possibilities at this particular location. Nevertheless, this model was considered based on the analysis of Lefty1/2 morpholino knockdown embryos (Feldman et al., 2000), which magnify involution and slow epiboly, notwithstanding that, in this case, the balance between epiblast and hypoblast is extremely upset, and the slowing effect on epiboly could actually result from the diminutive epiblast. We find that involution and epiboly seem independent, based on several experiments and observations. First, when involution movements were examined in the hab mutants, they are normal. Second, the expression of mesoderm and endodermal markers is normal in hab mutants, suggesting that the involuted germ layers contain the normal amount of tissue. Third, mutants that completely block involution such as squint/cyclops double mutants (Feldman et al., 2000) or MZoep mutants (Gritsman et al., 1999; Carmany-Rampey and Schier, 2001) have no effect on epiboly, and we have shown here, using double- and triple-mutant combinations of these mutants with hab, that the mutant involution phenotype is independent of the hab epiboly phenotype, at least on the ventral side. Last, in ZMD heterozygotes, where the zone of involution lags far behind the yolk–EVL junction, the blastoderm is slowed but it continues to move vegetally, suggesting that it is not necessary that involution occur at the EVL–yolk junction, as one might imagine for the involution zone to be towed.

Widened Neural Tube: The Result of Lack of Convergence?

Compared with the wild-type neural tube, the ZMD mutant neural tube is up to 50% wider. The hab homozygote neural plate is wider still, at least in the head regions, and posterior of the hindbrain region, the neural plate bifurcates along the midline with both halves extending to opposite sides of the arrested germ ring. There are interesting consequences on cell specification in both of the mutants. In the ZMD mutant, ectopic rows of neurons form near the midline, suggesting that the widened neural tube has perturbed the short-range morphogenetic fields that pattern the neural tube. In the case of the bifurcated neural tube of the homozygous mutant, only one side receives the proper signals from the notochord, which only extends along one side. In both cases, studies of the changes in fate made in the forming neural tube will be instructive as to pattern formation in the CNS.

Lack of cell convergence in the epiblast is the most likely explanation for the neural tube defects in both the hab homozygote and the ZMD heterozygote. Normally, cells of the lateral regions of the neural plate converge to the midline and then sink into the forming neural keel. If either of these processes do not occur, then the neural plate would remain mediolaterally splayed out. Defects in such a movement would be autonomous to the neural tube. It seems that the cells of the neural tube must be defective in cell adhesion, because based on our transplantation experiments in the ZMD mutants, the detached cells originate from the neural tube.

half baked Acts in the Ectoderm, Autonomously and Nonautonomously

One objective of these studies was to pinpoint the location where hab is required. We now favor the hypothesis that hab is acting in the epiblast of the gastrula, the future ectoderm, an idea supported by our epistasis experiments with one-eyed pinhead (oep) and squint; cyclops, and our transplantation experiments. oep mutants lack hatching gland cells (Hammerschmidt et al., 1996; Schier et al., 1997b; Strähle et al., 1997). Here, we show that, when the semidominant hab−/+ is genetically combined with oep, which lacks mesodermally derived hatching gland cells, the double mutant retains the enlarged “hatching gland” phenotype in the absence of the mesodermal component. Hence, the normal function of hab must be required in the ectodermally derived support cells of the hatching gland and not in the mesodermally derived cells. In the case of the hab; squint; cyclops triple, the mutant lacks almost all mesoderm. There is no hypoblast apparent and little involution, if any, occurring at the margin of the epiblast. Yet, the hab epiboly arrest phenotype is unchanged, again suggesting that the formation and presence of the hypoblast is not required to express the mutant phenotype.

Also, the cell transplantation experiments indicate that hab is acting in the epiblast. In this work, we report on the behavior of cell transplantations into ZMD mutants, focusing on the presence of wild-type donor cells in the dorsal detached cell clusters. These experiments illustrate that the detached cells of the ZMD hab−/+ mutants are most probably of ectodermal origin, because this is the predominant fate of the remaining transplanted cells in these particular experiments. This finding was the logical expected result, as ectoderm is the most superficial germ layer. The experiments further demonstrate that the dorsal detached cells are not derived exclusively from EVL cells, because, in our experience, EVL cells are never transplanted in these types of experiments.

Because the transplanted wild-type deep cells integrate into detached cell clusters, mutant host cells must change the fate of neighboring wild-type cells, a nonautonomous action. On the other hand, when hab homozygous mutant cells are transplanted beneath wild-type EVL cells, they do not change their fate, but assume the detached cell phenotype seen in the ZMD heterozygote, an autonomous action. Similar mixes of results are seen in oep (Warga and Kane, 2003), trilobite (Jessen et al., 2002), and glass onion (Malicki et al., 2003), evoking the idea that, like all of these other genes, hab acts at or near the cell surface.

Whereas our experiments demonstrate that hab acts in the ectoderm, the gene product of hab may be autonomously required in tissues other than the epiblast. For example, the defects seen in forerunner and prechordal plate morphogenesis may reflect a requirement for hab in the mesendoderm. Of interest, in both of these cases, the tissues are more epithelial in nature compared with the loose organization typical of the early mesendoderm. Hence, perhaps hab is autonomously required for epithelial tissues, which would of course include the entire epiblast field. Alternatively, the effects seen on nonectodermal sites could be the failure of interactions between the mesoderm and the ectoderm; this finding would be a nonautonomous effect.

In hab embryos, the defects in epiboly, convergence, and neural tube formation provide the first genetic evidence of a common element connecting these forms of morphogenesis. What is the hab gene product? If hab acts at the cell surface, it would seem that cadherins, nerve cell adhesion molecules, integrins, and other cell surface molecules would be likely candidates and would be expected to have many effects on different morphogenetic processes throughout the embryo. Whereas we eliminated VN-cadherin, there are other cadherins that are present. Of note is E-cadherin, which has been shown recently to have antisense knockdown phenotypes similar to hab (Babb et al., 2001; Babb and Marrs, 2004). We are working currently to identify the hab gene product. Understanding the function of this gene will be essential for understanding the control of cell movement at the molecular level and how this control choreographs cells to play morphogenesis.


Zebrafish Strains

The habdtv42, avatm94, lawts18, and wegtx230 mutations were isolated in a large-scale mutagenesis screen (Haffter et al., 1996) and initially out-crossed to the polymorphic WIK (L11) strain of wild-type fish for mapping. Subsequent generations were out-crossed to the puma wild-type strain of fish, selecting individuals that were prescreened for polymorphisms in closely linked microsatellite markers on Linkage Group 7. The ZMD hab; oep double mutants were produced by mating males doubly heterozygous for habdtv43 and oeptz257 to females heterozygous for oeptz257. The hab; squint; cyclops triple mutants were produced by mating identified pairs of fish that were triply heterozygous for habdtv43, squintcz35, and cyclopsmm294.

Mapping Panels

The epiboly mutants were mapped to the centromere of Linkage Group 7 by half-tetrad analysis (Streisinger et al., 1986; Johnson et al., 1995). The half-tetrad mapping panel was generated by subjecting a heterozygous female to EP-parthenogenetic reproduction as described in Westerfield (1995). Half-tetrad mapping panels were also generated from unrelated and unlinked mutations to provide additional information about the centromere on Linkage Group 7. Mapping resolution was increased using a haploid mapping panel for each of the mutations. Haploid embryos were generated as described by Westerfield (1995).

DNA Extraction

Half-tetrad and haploid embryos were allowed to develop to 10 hr of development, sorted by phenotype, and then harvested for DNA by placing in 50 μl of lysis buffer (1.5 mM MgCl2, 10 mM Tris-HCl pH 8.3, 50 mM KCl, 0.3% Tween 20, 0.3% Triton X-100). The extraction protocol includes incubating at 98°C for 20 min, followed by incubating at 55°C for 1.5–2 hr, while treating with 1 mg/ml proteinase K (Roche Biochemicals), and finally incubating at 98°C for 20 min. All DNA extractions were performed in individual wells of a 96-well polymerase chain reaction (PCR) plate (LPS Plastics), sealed with PCR tape (Nunc) using a Hot Bonnet thermal cycle machine (MJ Research). The resulting extraction was diluted directly into sterile distilled H2O at 1/200.

Linkage Analysis

In general, microsatellite markers were mapped as simple sequence length polymorphisms (SSLP). PCR reactions were carried out in 30-μl volumes containing: 10 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 0.0001% gelatin, 100 μl/ml bovine serum albumin, 100μM concentrations of each dNTP, 1 μM concentrations of each primer, 0.06 μl of Taq DNA polymerase, and 5 μl of the diluted DNA (5–50 ng). PCR: initial denaturation, 94°C, 5 min; 45 cycles: denaturation, 94°C, 30 sec; annealing, 55°C, 30 sec; extension, 72°C, 1 min; final extension, 72°C, 7 min. Polymorphisms were scored by gel electrophoresis on 2.5 to 4% agarose gels.

Many of the zebrafish genes and expressed sequence tags (ESTs) were mapped as restriction fragment length polymorphisms. These were amplified and scored as above for SSLP, only before loading onto a gel, 12 μl of the PCR reaction was digested for 3 hr in a final volume of 25 μl, with the appropriate restriction enzyme for revealing a polymorphism. Subsequently, the digest was loaded onto a 2.5% agarose gel. Some microsatellite markers, genes, and ESTs were mapped as SSCP. These were amplified and scored on a polyacrylamide gel using the protocols described for SSCP (Fornzler et al., 1998), except that ∼4 Ci/mmol of 32P dATP was added to the dNTP mixture per reaction instead of being used to label one primer.

Primers for microsatellite markers designed and used as described (Knapik et al., 1996, 1998; Shimoda et al., 1999) and can be found at Primers for Danio rerio genes and ESTs were designed from sequence deposited at NCBI and are available upon request.

Immunohistochemistry and RNA In Situ Hybridization

Antibody staining was performed as described in Warga and Nüsslein Volhard (1998). RNA in situ hybridization was carried out as described in Thisse et al. (1993). Embryos were cleared in 70% glycerol and photographed on a Zeiss AxioPhot.

Genotyping of Individual Embryos

Genomic DNA extraction and PCR reactions were carried out as described above for SSLP, except that, after in situ hybridization experiments, DNA was not diluted. Embryos were genotyped using a closely linked z-marker, Z6852 (F-CAT GTG GTA CAG TTG AAG GGG, R-ATC ATT GGA AAC CAT CCA TAC A), which is less than 0.5 cM from the hab locus. cyclops and squint embryos were genotyped as described in Warga and Kane (2003).

Cell Transplantation

Cell transplantations were carried out between sphere and 30% epiboly stages as previously described in Ho and Kane (1990); also, some transplantations were performed on a dissection scope, using ×50 power. Donors were examined at tail bud stage for phenotype and then later genotyped. Hosts were examined between 6- and 10-somite stages for the presence of dorsal detached cells. Embryos were recorded using Nomarski and fluorescence microscopy as described in Warga and Kane (2003).

4D Nomarski Time-Lapse Analysis

Embryos were mounted and time-lapse recordings were performed as described (Kane et al., 1996), except that images were digitally stored. Embryos were mounted before germ ring stage in 0.08% agarose between coverslips and sealed with Vaseline. Low-magnification recordings were acquired initially to determine the orientation of the embryos, and then they were recorded using a Zeiss DIC ×40 or ×63 water immersion lens at 10 to 15 planes using a 6 to 8 μm z-spacing for periods of 3 to 4 hr. In a single experiment, groups of 10 to 15 individuals were recorded using a computer controlled stage, and each location was visited every 150 to 180 sec. Afterward, low magnification recordings were again acquired to determine the latitude and longitude of the recording. Later, the embryos were carefully removed from between the coverslips and their genotypes were determined by segregation of linked markers or by phenotype. Using NIH Image 1.63, movements of individual cells were traced throughout the analysis period as has been previously described in Warga and Kane (2003). EVL cells, yolk syncytial layer nuclei, and deep cells were followed beginning at the start of involution movements for a total of 40 ± 2 min. For a “global” view, data from multiple embryos were combined, using the latitude and longitudes of the recordings to place and orient the cells in the field.


We thank Lynne Angerer, Cheeptip Benyajati, and Fred Hagen for comments on early versions of this manuscript, and Lyndsay Field and Esther Liu for their contributions to Figure 7. We also thank Yi-Lin Yan, Sharon Amacher, Ashley Bruce, and Yun-Jin Jiang, as well as many other members of the zebrafish community for reagents. D.A.K. was funded by the Pew Charitable Trust and the National Institutes of Health.