Overexpression of the forebrain-specific homeobox gene six3 induces rostral forebrain enlargement in zebrafish

Understanding the organization of the vertebrate embryonic forebrain is one of the most important issues in biology, especially since, unlike in the case of more posterior regions of the central nervous system (CNS), little is known about the molecular mechanisms underlying forebrain organization. Recent progress towards elucidating the control of embryonic forebrain development has been made through the identification of regionally specific genes whose combinatorial expression may direct the development of distinct regions in the forebrain. Rubenstein and his colleagues proposed that the forebrain is subdivided into six transverse domains named prosomeres (reviewed in Puelles and Rubenstein, 1993; Rubenstein et al., 1994). The prosomeres can be grouped into the diencephalon (p1 to p3) and the secondary prosencephalon (p4 to p6). P6, the most rostral subregion of the forebrain, includes some distinct tissues such as optic stalk, olfactory bulb, commissural plate and chiasm. Recent studies have shown that homeobox genes, such as members of the Emx and Otx family, have a central role in the patterning of more caudal regions of the vertebrate forebrain and other homeobox genes may participate in the regionalization of its most rostral sections. The newly identified homeobox gene Six3, which is expressed in the most rostral aspect of the forebrain in the mouse and chicken, is a candidate for such a function (Oliver et al., 1995a; Kawakami et al., 1996b; Bovolenta et al., 1998).

Mammalian Six family genes were identified by homology to the Drosophila sine oculis (so) gene (Oliver et al., 1995a,b) and by the binding of a Six protein to the promoter of the Na,K-ATPase α1 subunit gene (Kawakami et al., 1996a,b). The so gene encodes a homeodomain protein and is essential for Drosophila eye formation (Cheyette et al., 1994; Serikaku and O’Tousa, 1994). Five members of the mouse Six gene family have been characterized. Of these, Six1 and Six2 are expressed in head and body mesenchyme, limb muscles and tendons (Oliver et al., 1995b); Six3 is expressed in the anterior forebrain and in the eyes (Oliver et al., 1995a; Kawakami et al., 1996b); Six4/AREC3 is expressed in neural tissues and encodes the transcription factor regulating the Na,K-ATPase α1 subunit gene (Kawakami et al., 1996a; Ohto et al., 1998), and Six5/mDMAHP is expressed in a wide variety of mouse tissues (Boucher et al., 1995; Kawakami et al., 1996b; Heath et al., 1997). The human SIX5/DMAHP gene maps immediately 3’ to the CTG repeat which is known to associate with myotonic dystrophy (Boucher et al., 1995; Klesert et al., 1997; Thornton et al., 1997). It has been reported that mutations in Drosophila so genes lead to defects of the adult and/or larval visual system (Cheyette et al., 1994; Serikaku and O’Tousa, 1994) and that ectopic expression of mouse Six3 in medaka embryos promotes ectopic lens formation in the area of the otic vesicle (Oliver et al., 1996). A recent report by Pignoni et al. (1997) has demonstrated that ectopic co-expression of so with eyes absent (eya) leads to ectopic eye tissues in the Drosophila antennal, wing and leg discs. These findings suggest a conserved function for vertebrate Six3 and Drosophila so in visual system development. In Drosophila, the structure of the brain is also affected by the so mutation (Serikaku and O’Tousa, 1994), implying that Six3 may have a function in brain formation.

Organization and subdivision in the vertebrate forebrain remain more controversial than those in the posterior brain. Molecular studies of the embryonic forebrain in lower vertebrates such as the zebrafish, Danio rerio, may reveal evidence of such subdivision. In this article, we report the cloning of the zebrafish six3 gene and a study of its expression pattern during embryonic development, demonstrating conservation of Six3 sequence and expression pattern among the zebrafish, mouse and chicken. Overexpression of six3 in zebrafish embryos induced an enlargement of the rostral region of the forebrain including the optic stalk, providing evidence for a role of six3 in the development of the rostral forebrain.

Degenerate oligonucleotide primers against the conserved regions among mouse Six1, Six2, Six3 and Six4/AREC3 (Oliver et al., 1995a,b; Kawakami et al., 1996a) were synthesized, and RT-PCR was performed using these primers and total RNA from zebrafish embryos. The following oligonucleotide primers were used (see also Fig. 1A,C): CCGGATCCAT(A/C/T)TGGGA(C/T)GG(A/C/G/T)GA-(A/G)GA(A/G)AC (sense strand corresponding to IWDGEET plus EcoRI linker), CCGGATCCGT(A/C/G/T)TA(C/T)TG(C/T)TT(C/T)-AA(A/G)GA(A/G)AA (sense strand corresponding to VYCFKEK plus EcoRI linker), and CCGTCGACAACCA(A/G)TT(A/C/G/T)-(A/T)(G/C)(A/C/G/T)AC(C/T)TG(A/C/G/T)GT (antisense strand corresponding to TQVSNWF plus SalI linker). Conditions of PCR were the same as described previously (Taira et al., 1992). PCR product was purified from an agarose gel and was subcloned into pBluescript II KS+. The zebrafish six3 cDNA was isolated from a λZap shield stage cDNA library (Rebagliati et al., 1998) using these PCR-derived clones as probes.

Adult fish were kept at 28.5°C on a 14-hour light/10-hour dark cycle, and embryos were obtained by natural matings (Westerfield, 1995). Some embryos were raised in 0.2 μM phenylthiocarbamide (Sigma) to inhibit pigment formation, as described previously (Hyatt et al., 1992). For sections, embryos were embedded in Technovit 7100 (Heraeus Kluzer GmbH, Germany) after fixation with 4% paraformaldehyde/PBS. Transverse sections were prepared at 10 μm, and were stained with hematoxylin. RNA injection was performed as described previously (Toyama et al., 1995b).

The six3 injection construct was made by inserting the coding region of zebrafish six3 into the vector pCS2+ (Rupp et al., 1994) to make pCS2six3. For the deletion constructs, six3 cDNA in pBluescript II KS+ was digested with MscI-HindIII, HindIII-KpnI, or KpnI-NotI, followed by being digested both protruding ends with Mung Bean nuclease or filled recessed 3’ ends with Klenow fragment, and was religated with T4 DNA ligase. The coding regions of these deleted six3 were inserted into the pCS2+ to generate pCS2six3d1, pCS2six3d2 and pCS2six3d3, respectively. All constructs were verified by sequencing. Synthetic capped RNA was transcribed from these plasmids or pSP64-Xβm (Krieg and Melton, 1984) using the SP6 Megascript in vitro transcription kit (Ambion) with m⁷G(5’)ppp(5’)G (Boehringer Mannheim).

Embryos were analyzed by whole-mount in situ hybridization as described by Toyama et al. (1995b). The antisense six3 probe contained the entire region of six3 cDNA. Digoxigenin-labeled RNA probes were prepared according to the instruction of the manufacturer (Boehringer-Mannheim). We used RNA probes prepared from cDNAs of the zebrafish emx2 (Morita et al., 1995), dlx3 (Akimenko et al., 1994), gsc (Stachel et al., 1993), Islet-1 (Inoue et al., 1994), lim3 (Glasgow et al., 1997), lim5 (Toyama et al., 1995a), ntl (Schulte-Merker et al., 1992), otx2 (Mori et al., 1994), pax2 (Krauss et al., 1991) and shh (Ekker et al., 1995). Frozen sections of whole-mount-stained embryos were prepared at 10 μm. Whole-mount immunostaining using antibody against acetylated α-tubulin (Sigma, 1:500 dilution) and the Vectastain Elite ABC kit (Vector Laboratories) was performed by the protocol of Miyagawa et al. (1996).

To search for zebrafish Six genes, we designed two sets of degenerate primers based on conserved sequences in the Six domain and the homeodomain (Kawakami et al., 1996a; Oliver et al., 1995a,b). RT-PCR analyses using these primers were performed using total RNA prepared from zebrafish embryos of various stages as templates and the resulting PCR products were subcloned. The nucleotide sequences of these clones revealed the existence of at least seven zebrafish Six genes (data not shown). RT-PCR analyses with specific primer pairs based on these sequences showed that four different Six genes, including a Six3 homolog, are expressed at the shield stage. A shield stage cDNA library was screened with a mixture of these four Six probes at low stringency, resulting in 83 positives among 2×10⁶ phage, three of which were independent isolates of the zebrafish six3 homolog; the other zebrafish Six genes will be discussed elsewhere. An in-frame termination codon exists 105 bp upstream of the putative translation initiation codon in the zebrafish six3 cDNA sequence (data not shown), making it likely that this methionine is the actual N terminus of the protein. The amino acid sequence of zebrafish six3 protein has striking similarity with that of Six3β, deduced from one of the three known alternatively spliced Six3 mRNAs in the mouse (Fig. 1A; Kawakami et al., 1996b). The overall identity between mouse Six3β and zebrafish six3 is 76% at the nucleotide level and 85% at the amino acid level. The only prominent difference between these two proteins is that zebrafish six3 does not have the glycine tracts in the N-terminal region present in mouse Six3β (Fig. 1A). Intermediate numbers of glycine residues in the chicken homolog cSix3 suggests that this glycine tract may expand during evolution for an unknown reason (Fig. 1A; Bovolenta et al., 1998). While all Six family proteins, as well as the entire bicoid class (Bürglin, 1994), share a lysine at position 50 of the homeodomain, sequence similarity between zebrafish six3 and other family members is much lower than with mouse Six3β and chicken cSix3 (Fig. 1B,C). It is notable that the nematode clone W05E10.3 encodes a Six-related protein with a homeodomain highly homologous to vertebrate Six3 (Fig. 1B,C), while clone C10G8.7 encodes a protein more similar with mouse Six1 (data not shown). Thus it appears that divergence within the Six family is ancient in metazoans, and that Six3 homeodomain sequences have been conserved during a long period of evolution.

The expression patterns of zebrafish six3 during embryogenesis were examined by whole-mount in situ hybridization. The six3 transcript was undetectable until the shield stage (data not shown). six3 is expressed in hypoblast cells at the shield stage (Fig. 2A), and at the anterior edge of involuting axial mesendoderm during gastrulation (Fig. 2B). As gastrulation proceeds, six3 mRNA-expressing cells become increasingly limited to the anterior, the presumptive prechordal plate, not overlapping with ntl mRNA-expressing cells which form the presumptive notochord (Fig. 2C,G; Schulte-Merker et al., 1992). The expression pattern in the axial mesendoderm is similar to that of gsc (Stachel et al., 1993; Schulte-Merker et al., 1994; Thisse et al., 1994), although six3 is more limited to the most anterior edge (Fig. 2D-F). Six3 expression in the mesendoderm has been also observed in the chicken (Bovolenta et al., 1998).

six3 starts being expressed at the anterior edge of the presumptive neuroectoderm at the 80-85% epiboly stage (data not shown). At the beginning of the segmentation period, six3 is expressed in the rostral end of both mesendodermal and ectodermal cells (Figs 2D, 3A,D), with the former fading during somitogenesis, leaving six3 mRNA-expressing cells confined to the rostral region of the prospective forebrain (Figs 2H,I, 3B). These cells will develop into the telencephalon, retina and part of diencephalon according to the fate map of Woo and Fraser (1995). A gap between six3 and pax2 mRNA-expressing cells implies that most of the prospective midbrain is six3 negative (Fig. 2H). In the prospective forebrain, six3 is expressed in superficial cells in the neural rod and in the overlying ectoderm at the 6-somite stage (Fig. 3B,E) and, by the 16-somite stage, becomes confined to the lateral surfaces of the retina and lens (Fig. 3C,F).

In the 24 hour embryo, six3 is expressed in rostral cells of the telencephalon and ventral diencephalon (Fig. 4A). These regions correspond to p6, the most rostral subregion of the forebrain, as defined by Rubenstein and colleagues in the mouse (Puelles and Rubenstein, 1993; Rubenstein et al., 1994). Ross et al. (1992) proposed that the rostral surface of the brain bends downward at the transverse ventral flexure, making a dorsal surface to be the most anterior portion of the embryo; this rostral surface is essentially identical to the six3 mRNA-expressing region. This broad rostral expression pattern is transient and, by 36 hours, six3 mRNA are confined to specific areas within the rostral surface. To identify six3-positive cells at this stage, we double-stained embryos with six3 probe and with antibody against acetylated α-tubulin to visualize the axonal tracts (Chitnis and Kuwada, 1990). six3 is expressed in medially located cells of the telencephalon, the ventral tip of the olfactory bulb and the optic stalk (Fig. 4B,C), and also in the pituitary anlage visualized as lim3-positive cells (Fig. 4D,E; Glasgow et al., 1997). The cells around the postoptic commissure (POC) and those around the intersection of the supraoptic tract (SOT) and the anterior commissure (AC) are also six3 positive (Fig. 4B,C). These tissues were shown to have a common origin from the anterior neural ridge (ANR) in Xenopus (Eagleson et al., 1995). six3 starts being expressed in the ventral midbrain tegmentum at 24 hours and the expression becomes stronger at later stage (Fig. 4A,B,D). six3 is expressed in the entire eyes at 24 hours, with the expression fading later except in the lens and its neighbors (Fig. 4C; data not shown).

In summary, six3 is expressed at the anterior end of the mesendoderm and in the prospective forebrain in early development, and in the rostral surface of the brain and its derivatives in later development.

With the goal of studying six3 function during development, we injected synthetic six3 mRNA into zebrafish embryos at the 2-cell stage, resulting in mosaic but fairly even distribution of the injected mRNA by early somitogenesis. Injected embryos exhibited specific morphological defects of the brain area (Fig. 5A,B). In control embryos, three ventricles were apparent in the brain, the diencephalic, mesencephalic and rhombencephalic ventricles. These ventricles were not well formed in six3 mRNA-injected embryos, despite the obvious presence of a lumen within the neural tube. In addition, the constriction normally present at the midbrain-hindbrain boundary was absent. Most six3 mRNA-injected embryos showed defects in the head region while tissues in the trunk such as notochord, somites and spinal cord appeared normal (Fig. 7B; data not shown).

six3 mRNA-injected embryos showed an enlargement of the telencephalon between AC and epiphysis as seen by Islet-1 staining (Fig. 5C,D; Inoue et al., 1994). In contrast, the region between epiphysis and otic vesicles, including the midbrain and anterior hindbrain, was shortened and broadened. Immunostaining with antibody against acetylated α-tubulin (Fig. 5E-H) revealed that the axonal tracts of AC, POC and SOT are elongated in six3 mRNA-injected embryos, and the preoptic area which is surrounded by these axonal tracts is enlarged (Fig. 5G,H); in contrast, axons of the posterior commissure (PC) were reduced. Telencephalic neuronal clusters were enlarged and disorganized (e.g., in Fig. 5F).

Transverse sections at the level of the prospective forebrain of six3 mRNA-injected embryos at early somitogenesis stages show that cell number in the dorsal but not ventral neural tube was increased (Fig. 6A,B) while, at the level of the midbrain, the structure was disorganized, but cell number appeared largely unchanged (Fig. 6C,D). These observations suggest that one reason for the disorganization of the head may be an excessive accumulation of cells in the anterior/dorsal neural tube, enlarging the forebrain and compressing the midbrain and anterior hindbrain into a short and broadened shape.

Both the Six domain and the homeodomain are conserved among vertebrate and invertebrate Six proteins (Fig. 1). The C-terminal half of the Six domain and the entire homeodomain of mouse Six4/AREC3 have been shown to be essential for sequence-specific DNA binding (Kawakami et al., 1996a). In addition, the Six domain of Drosophila so binds to eya protein, a presumptive transcriptional co-activator, and may mediate their synergistic function (Pignoni et al., 1997). To examine whether these domains are required for six3 function, we overexpressed six3 derivatives with deletions in the Six domain and/or the homeodomain in zebrafish embryos (Fig. 7A). Northern blots revealed that the stability of injected six3d1, six3d2 and six3d3 mRNAs was similar to that of wild-type mRNA at 24 hours (data not shown). About 70% of wild-type six3 mRNA-injected embryos showed head-specific defects, while no embryos displayed such a phenotype after mRNA injection of either six3 derivatives, even when the amount of injected mRNA was increased five-fold (Fig. 7B). General defects seen in some six3d1 mRNA-injected embryos may due to nonspecific effects of the homeodomain. These results not only confirm that the phenotype in six3 mRNA-injected embryos is due to overexpression of six3 gene product but also demonstrate that both the Six domain and the homeodomain are essential for six3 gene function.

While no changes were observed in gsc expression in the anterior axial mesendoderm at the 75% epiboly or tail bud stages (data not shown), six3-overexpressing embryos exhibited modified expression of subregion-specific marker genes in the CNS during later embryogenesis. During somitogenesis, lim5 mRNA forms a tightly delimited stripe in the diencephalon in normal embryos (Toyama et al., 1995a), which moved posteriorly in six3 mRNA-injected embryos especially in its dorsal aspect (Fig. 8A,B), indicating an expansion of the rostral forebrain. Such expansion was also indicated by the pattern of otx2, which is expressed in the dorsal diencephalon and optic tectum in normal 24-hour embryos (Li et al., 1994; Mori et al., 1994); overexpression of six3 led to an enlarged rostral otx2-negative region and a shortened otx2-positive region (Fig. 8C,D). The domain of shh expression in ventral regions of the brain (Krauss et al., 1993; Ekker et al., 1995) showed only slight expansion after six3 mRNA injection (Fig. 8E,F). In the case of emx2, which is expressed both dorsally in the telencephalon and ventrally in the hypothalamus (Morita et al., 1995), overexpression of six3 caused an expansion of the expression domain in the dorsal but not ventral forebrain, even though both regions appear enlarged (Fig. 8G,H). Genes expression in areas posterior to the otic vesicles was unaffected, such as Islet-1 expression in the spinal chord and shh expression in the floor plate and notochord (data not shown). Furthermore, no obvious changes were observed in lim3 or dlx3 expression in the pituitary anlage and olfactory placodes, respectively (data not shown).

Thus, the expression of marker genes indicates an expansion of the rostral forebrain in six3 mRNA-injected embryos especially in the dorsal domain (Fig. 8), but gross morphological appearance (Fig. 8) and the modified arrangement of axonal pathways (Fig. 5G,H) suggests that ventral regions are enlarged as well.

Because evolutionary considerations implicate the six3 gene in visual system development, we tested pax2 expression in six3 mRNA-injected embryos. During segmentation, pax2 is expressed primarily in the optic stalk, midbrain-hindbrain boundary and otic vesicles (Krauss et al., 1991). Of these major expression sites, pax2 staining in the prospective optic stalk at the 10-somite stage was enlarged in six3 mRNA-injected embryos, while the signal at the midbrain-hindbrain boundary and in the otic vesicles appeared unaffected (Fig. 9A,B). In normal 24-hour embryos, the optic stalk as visualized by pax2 staining, was located ventrally because of the bending of the rostral forebrain (Fig. 9C). In contrast, the optic stalk in six3 mRNA-injected embryos was displaced to a frontal location (Fig. 9D), indicating that the flexure of the forebrain had failed to occur. The base of the optic stalk in the vicinity of the choroid fissure at the anterior retina was dramatically enlarged in six3 mRNA-injected embryos (Fig. 9E-H), indicating that six3 is involved in optic stalk formation. Though the region of pax2-negative cells in the retina remained essentially unchanged, the extension of pax2-positive regions caused an enlargement of the retina.

The amino acid sequence of the zebrafish six3 protein is highly similar to mouse Six3β and chicken cSix3 (Fig. 1). We suggest that zebrafish six3 is the ortholog of mouse Six3 and chicken cSix3 by the following criteria. (1) The entire amino acid sequence of these proteins is highly conserved and is distinct as compared to other Six proteins. (2) The expression pattern is similar for zebrafish six3, mouse Six3 and chicken cSix3 (Oliver et al., 1995a; Bovolenta et al., 1998); these genes are first expressed at early gastrula and later are limited to the rostral forebrain.

The amino acid sequence of Drosophila sine oculis (so) is more similar to those of mouse Six1 and Six2 than of Six3. Although the predicted amino acid sequence of so is not closely similar to vertebrate Six3, there are some similarities in the expression pattern and suggested function of these genes. Thus, no clear orthology relationships can be proposed between so and vertebrate Six proteins. It is therefore somewhat surprising that the nematode, Caenorhabditis elegans, has both a Six1-related and a Six3-related gene in its genome. In this context, it is possible that there is another Six3-related so gene in Drosophila.

Zebrafish six3 is expressed in the most rostral regions of the forebrain throughout early development. On the basis of explant culture experiments in mouse embryos, Shimamura and Rubenstein (1997) have proposed that an anterior domain, the ANR, is a local organizing center for the patterning of the rostral forebrain. Similarly, a recent report by Houart et al. (1998) has implicated a signal from a small group of anterior neuroectodermal cells in zebrafish embryos at the 75% epiboly stage in the induction of the anterior forebrain-specific genes and patterning of the rostral brain. Since six3 expression in the neuroectoderm begins at the 80-85% epiboly stage, and overexpression of six3 induces the expansion of rostral forebrain, six3 might be a downstream target and possible mediator of this putative rostral brain organizing signal.

While it is not clear whether six3 or any other member of the Six family is an ortholog of Drosophila so, there are important functional similarities between so and six3. Serikaku and O’Tousa (1994) have shown that the most prominent defect in the so^mda allele is the absence of an optic stalk, which prevents a physical connection between the developing eye and the brain; thus, so appears to be essential for optic stalk formation. This conclusion is consistent with our results based on overexpression experiments, implicating six3 in the formation and patterning of the rostral forebrain, especially of the optic stalk.

Oliver et al. (1996) reported that ectopic expression of mouse Six3 from DNA expression constructs in medaka, a fish distantly related to zebrafish, induced ectopic lens formation within the otic vesicles in 2.5% of injected embryos. In our experiments, the effect of mRNA injection was more drastic, generating effects in the retina/optic stalk and the forebrain at high penetrance (Fig. 7). While no embryos with an ectopic lens in the otic vesicles or in any other part of the body were observed among more than one thousand six3 mRNA-injected embryos, it is not excluded that we failed to observe a low-level or delayed effect on the lens in the face of the high-frequency effects reported here. Instability of injected mRNA may also be involved, although we found that the level of injected six3 mRNA in 24-hour embryos was three times higher than that of endogenous mRNA.

Overexpression of six3 leads to an increase in pax2 expression in the optic stalk; therefore, the question arises whether the six3 effect is mediated by pax2, and whether six3 directly regulates pax2 expression. An analysis of pax2-defective mutants in zebrafish and in the mouse has shown that optic stalk cells were intact but failed to intercalate across the midline (Torres et al., 1996; Macdonald et al., 1997), suggesting that six3 may carry out functions in optic stalk formation beyond the induction of pax2. Furthermore, overexpression studies have suggested that shh can regulate optic stalk formation (Ekker et al., 1995; Macdonald et al., 1995), yet shh expression is not substantially affected in six3 mRNA-injected embryos (Fig. 8) and six3 expression is normal in cyclops mutant (unpublished results) even though shh expression is defective (Krauss et al., 1993; Ekker et al., 1995), suggesting that six3 and shh influence optic stalk formation separately.

Six3 may be involved not only in optic stalk formation but also in the formation of other tissues in the anterior forebrain. In experiments using explant culture from mouse embryo brain, Shimamura and Rubenstein (1997) showed that the expression of BF1, a winged helix transcription factor that is essential for telencephalon and eye development (Xuan et al., 1995), depends on a signal from the ANR. Fgf8, known to be expressed in the ANR (Heikinheimo et al., 1994; Ohuchi et al., 1994; Fürthauer et al., 1997), is capable of mediating this signal (Shimamura and Rubenstein, 1997). Since Six3 is expressed in the anterior forebrain at an earlier stage than Fgf8 (Crossley and Martin, 1995; Oliver et al., 1995a; Fürthauer et al., 1997), it is possible that Six3 induces Fgf8 expression, thereby regulating patterning during the formation of the forebrain.

Injection of six3 mRNA generated defects only in the anterior CNS, mostly the forebrain, and, even in this region, no duplications, formation of ectopic optic stalk or ectopic domains of pax2 expression were seen. These results suggest that the competence to respond to six3 overexpression is limited to those domains in which six3 functions normally. Such a restriction of competence could be due to a requirement for cooperating factors for six3 function that are limited to the rostral forebrain. Studies on so function in Drosophila are consistent with this suggestion, in that no dominant gain-of-function phenotypes associated with ectopic expression of so were observed without co-expressing eya (Serikaku and O’Tousa, 1994; Pignoni et al., 1997). The result that the Six domain, demonstrated to be required for eya binding in Drosophila, is also essential for achieving a head-specific phenotype in zebrafish (Fig. 7) suggests that a cooperating factor for six3 could be a vertebrate eya homolog. Three eya homologs have been isolated in the mouse, but their expression is not limited to the anterior forebrain (Xu et al., 1997; Zimmerman et al., 1997). It will be interesting to isolate zebrafish eya homologs and to test for overexpression phenotypes together with six3 in zebrafish embryos.

We thank Ms K. Miura and R. Wu for help in fish maintenance. We thank Drs Y. Kikuchi, H. Okamoto for providing fish, M. Rebagliati for the cDNA library, and M.-A. Akimenko, D. J. Grunwald, S. Krauss, M. Mishina, H. Okamoto, S. Schulte-Merker, A. R. Ungar for probes. We thank Drs E. Glasgow, T. Kudoh and H. Okamoto for technical suggestions and discussions. This work was supported by Jichi Medical School Young Investigation Award, and by grants from the Naito Memorial Foundation, the Sumitomo Foundation, the Nissan Science Foundation and from Ministry of Education, Science, Sports and Culture of Japan.