Patterning of proneuronal and inter-proneuronal domains by hairy- and enhancer of split-related genes in zebrafish neuroectoderm

Spatial regulation of neurogenesis is a fundamental process that controls development of the central nervous system. Little is known about the mechanism by which neurogenesis is spatially regulated in vertebrates. Zebrafish embryos, like other anamniotes, have two distinct populations of neurons -primary and secondary neurons. Primary neurons are rapidly developing early neurons, which include primary-motor, primary-inter and Rohon-Beard (RB)neurons, and they are required for the coordinate movements of larvae soon after hatching (reviewed by Appel,2000). Primary neurons are generated from proneuronal domains,also called neurogenic regions, that are established as three longitudinal stripes along the anteroposterior axis in the dorsal ectoderm during the late gastrulation and early segmentation periods. The proneuronal domains express basic helix-loop-helix (bHLH) proneural genes, neurogenin1(neurog1) (Blader et al.,1997; Kim et al.,1997), olig2 (Park et al., 2002) and neurod4 (previously called zath3)(Park et al., 2003; Wang et al., 2003), and a homeobox gene, pnx (Bae et al.,2003) in zebrafish. The proneuronal domains are separated by the inter-proneuronal domains, which do not express the proneural genes and do not exhibit neurogenesis when it occurs in the proneuronal domains.

As proposed in Drosophila(Campos-Ortega, 1993),proneural genes play an important role in neurogenesis. neurog1 is required for the formation of RB neurons(Cornell and Eisen, 2002), and olig2 is required for the formation of primary motoneurons(Park et al., 2002). Within the proneuronal domains, a subset of cells is selected to become neurons. This selection is controlled by the Notch-mediated lateral inhibition mechanism(Appel and Chitnis, 2002). In the context of lateral inhibition, cells that express proneural genes, such as neurog1, at high levels become neurons and simultaneously express the Notch ligands Delta or Serrate, which activate Notch signaling in neighboring cells (Appel and Eisen, 1998; Chitnis and Kintner, 1996; Haddon et al., 1998; Ma et al., 1996). These neighboring cells express Hairy and Enhancer of Split-related (Her)transcriptional repressors that inhibit neurogenesis(Takke et al., 1999). Although the mechanism that restricts the number of cells differentiating into neurons within the proneuronal domains is relatively clear, the mechanism that represses neurogenesis in the inter-proneuronal domains is largely unknown.

In Drosophila, Hairy functions differently from Enhancer of Split[E(spl)], which functions downstream of Notch signaling, in the context of sensory organ development (Orenic et al.,1993; Skeath and Carroll,1991). hairy is expressed in four longitudinal stripes in the Drosophila pupal legs, located between stripes expressing the proneural gene acheate. In the absence of hairy function, acheate expression expands into the inter-stripe regions that normally express hairy, fusing the acheate stripes and resulting in the disorganization of sensory organ bristles(Orenic et al., 1993),indicating that Hairy suppresses the neurogenesis in the inter-stripe region in the Drosophila leg. hairy expression is controlled by positional cues, and it is proposed to function as a prepattern gene that links positional information to the spatial regulation of neurogenesis(Kwon et al., 2004; Orenic et al., 1993; Skeath and Carroll, 1991) A similar role has been proposed for the Her/Hes-family genes in vertebrates. Xenopus ESR6e, which is expressed in the superficial layer of the ectoderm, is involved in protecting the ectodermal layer cells from becoming neurons (Chalmers et al.,2002), and mouse Hes1 represses neurogenesis in the olfactory placode independent of lateral inhibition(Cau et al., 2000). Her5 represses neurogenesis independently of Notch signaling in the midbrain-hindbrain boundary (MHB) in zebrafish(Geling et al., 2004). These reports suggest that Her/Hes-mediated prepattern mechanisms are involved in the spatial regulation of neurogenesis in vertebrates.

Bmp activity has been shown to provide positional information for the posterior primary neurons (Barth et al.,1999; Nguyen et al.,2000). In embryos with mild defects in Bmp signaling, the RB neurons are absent, and the primary interneuron domains are expanded and located ventrally. In embryos with severe defects in Bmp signaling, both the RB neurons and primary interneurons are absent. These studies suggest that the position and width of the proneuronal and the inter-proneuronal domains are regulated directly or indirectly by Bmp signaling. It has been reported that the Xenopus Zic-related gene Zic2 is expressed in the inter-proneuronal domains and is involved in patterning the proneuronal and inter-proneuronal domains (Brewster et al., 1998). However, none of the zic-related genes displays a similar expression pattern in zebrafish(Grinblat and Sive, 2001) or any other vertebrate species. There should be other genes that are controlled by positional information, in which Bmp signaling is involved, and that spatially control neurogenesis along the dorsoventral axis in the zebrafish posterior neuroectoderm.

We report that zebrafish her3 and her9, which are expressed in the inter-proneuronal domains, repress the expression of proneural genes in the inter-proneuronal domains and thereby control the formation of the inter-proneuronal domains. her3 and her9expression is not regulated by Notch signaling, but rather is controlled by positional information, in which Bmp signaling is involved. The data indicate that her3 and her9 function as prepattern genes, which spatially control neurogenesis through a mechanism similar to that involving hairy in Drosophila.

Wild-type zebrafish (Danio rerio) embryos were obtained from natural crosses of fish with the Oregon AB or AB/India genetic background. The following zebrafish mutants were used: swr^tc300,snh^ty68a and mib^m178(Mullins et al., 1996; Schier et al., 1996).

Expression plasmids for Myc-tagged Her9 (Her9MT), Her9VP16, Her9EnR and Her3MT were constructed in pCS2+ as previously published(Bae et al., 2003). Expression plasmids for the activated form of zebrafish Notch5 (zN5ICD), the antimorphic form of Xenopus Delta-1 (xDlu^stu) and β-galactosidase have also been published (Bae et al.,2003). The RNA for Her9MT, Her9VP16, Her9EnR, Her3MT, zN5ICD or xDlu^stu was co-injected with 50 pg of β-galactosidase RNA into one blastomere of two- to four-cell stage zebrafish embryos. The embryos were fixed and stained with X-gal and antisense riboprobes, as reported previously(Bae et al., 2003). To visualize the in situ hybridization, BM purple AP substrate was used as the substrate for alkaline phosphatase. The neurod4(Wang et al., 2003), her2 (NM_131089), her4(Takke et al., 1999), her12 (NM_205619), her3(Hans et al., 2004) and hes5 (Raya et al.,2003) probes were amplified by PCR. pnx(Bae et al., 2003), her9 (Leve et al.,2001), olig2 (Park et al., 2002), neurog1(Kim et al., 1997), islet1 (Bae et al.,2003), deltaA (Haddon et al., 1998) and elavl3(Kim et al., 1996) were detected as described previously.

Embryos were incubated with 100 μM DAPT from 60% epiboly until the one-somite stage.

Morpholino oligonucleotides (MOs) were generated by Gene Tools. neurog1-MO and olig2-MO have been previously published(Cornell and Eisen, 2002; Park et al., 2002). The other MOs used here were: her9ATG-MO,5′-CTCCATATTATCGGCTGGCATGATC-3′; her9SD- MO,5′-GTGATTTTTACCTTTCTATGCTCGC-3′; her9_5mis-MO (the lower-case letters indicate mispaired sequences in the control morpholino),5′-CTCtATATgcTCGGCTGatATGATC-3′; her3-MO1,5′-CTGTTGGATGCTGTAGCCATTGTCC-3′; and her3-MO2,5′-TGCAGCCATTGTCCTTAAATGCTCA-3′ [the same sequence as published(Hans et al., 2004)]. We also used the standard MO, 5′-CCTCTTACCTCAGTTACAATTTATA-3′. We used two distinct MOs for her3 (her3-MO1 and her3-MO2) and her9 (her9ATG-MO and her9SD-MO), and obtained essentially the same results from them. Neither the her9_5mis control MO nor the standard MO elicited significant abnormalities in neurogenesis at the doses used.

To reveal the mechanism by which proneuronal and inter-proneuronal domains are established, we sought genes that are specifically expressed in these domains. The bHLH proneural genes neurog1 and neurod4 and the homeobox gene pnx display spotty expression within the proneuronal domains (Bae et al.,2003; Blader et al.,1997; Kim et al.,1997; Park et al.,2003; Wang et al.,2003) (Fig. 1A,A′; Fig. 2A, data not shown), and olig2 is also expressed in the proneuronal domains, which correspond to the primary motoneurons(Park et al., 2002)(Fig. 1G,G′). We found that the zebrafish Her genes displayed two distinct expression profiles in the neural plate (Fig. 1J). her2, her4, her12 and hes5 are expressed in the proneuronal domains (Takke et al., 1999)(Fig. 1D,D′, data not shown), similar to the expression of the proneural genes. By contrast, her9 and her3 were expressed in the inter-proneuronal domains (Fig. 1B,B′,E,E′,H,H′). It has previously been reported that her9 is expressed as three longitudinal stripes(Leve et al., 2001), but it was not clear whether these expression domains corresponded to the proneuronal domains or the inter-proneuronal domains. Co-staining of early segmentation-stage embryos with her9 and pnx or neurog1 probes revealed that the expression of her9 was complementary to that of neurog1 and pnx(Fig. 1C,C′, data not shown), and detected between the primary motoneurons and interneurons, and between the primary interneurons and RB neurons. Co-staining with her3 and hes5 or olig2 probes showed that her3 was expressed between the primary motoneurons and interneurons(Fig. 1F,F′,I,I′). The her3 expression domains were more restricted than those of her9 in the inter-proneuronal domain between the primary motoneurons and interneurons. These data suggest that her3 and her9 are expressed differently and have one or more different functions from other Her genes that are expressed in the proneuronal domains.

The positions of the primary neurons are known to be controlled by Bmp in the zebrafish (Barth et al.,1999; Nguyen et al.,2000). It has previously been reported that the RB neurons are lost and primary interneuron domains expanded in snailhouse(snh)/bmp7(snh^ty68) and that both the RB neurons and interneurons are lost in swirl(swr)/bmp2b (swr^tc300) mutant embryos(Nguyen et al., 2000). In our experiments, the snh mutant embryos showed a similar phenotype to that described in the previous publication, and the lateral proneuronal domains that contain neurod4-expressing RB neurons and hes5-expressing cells were lost and the primary interneuron domains were expanded and located ventrally in these embryos(Fig. 2B,E). However, both the RB neurons and primary interneurons were lost in the swr mutant embryos (Fig. 2C,F). This phenotype was similar to that of swr mutant embryos into which chordin RNA was injected (Nguyen et al., 2000). The severity of the phenotypes of zebrafish mutants often depends on the genetic background, the age of the parent heterozygotes and the condition of the culture. The swr^tc300 mutant embryos in our experiments displayed more severe Bmp signaling-defect phenotypes than did those reported previously(Nguyen et al., 2000). In these snh and swr mutant embryos, the inter-proneuronal spaces were expanded. In correlation with this, the her3 and her9 expression was expanded ventrally. The her3 and her9 expression did not reach the most ventral part of the snh mutant embryos (Fig. 2H,K), but in the swr mutant embryos, their expression was expanded and reached the most ventral part of the embryos(Fig. 2I,L). These data indicate that the position and width of the her3 and her9expression is regulated by positional cues, in which Bmp signaling is involved directly or indirectly. The formation of the proneuronal and the inter-proneuronal domains is co-regulated by the positional information.

Many Her genes are known to be targets of Notch signaling in various zebrafish tissues (Henry et al.,2002; Itoh et al.,2003; Oates and Ho,2002; Pasini et al.,2004; Raya et al.,2003; Takke and Campos-Ortega,1999; Takke et al.,1999). We therefore examined the regulation of her3 and her9 by the Notch signal. mind bomb (mib) embryos have a mutation in the gene encoding a ubiquitin ligase for Delta and display a strong reduction in Notch signaling(Itoh et al., 2003). We also used a chemical compound, DAPT, which is a γ-secretase inhibitor that inhibits Notch signaling by preventing the generation of NotchICD(Geling et al., 2002), and RNAs for an activated form of Notch5 (zN5ICD)(Bae et al., 2003) or an antimorphic form of Xenopus Delta (xDl^stu), which inhibits Notch signaling (Chitnis et al.,1995). In the mib mutant embryos and embryos treated with DAPT, hes5 expression was strongly reduced or abolished, and the neurod4 expression became homogenous within the proneuronal domains(Fig. 3E,H,I,L). This is a typical expression pattern seen in Notch-defective embryos(Appel et al., 2001; Itoh et al., 2003) and is consistent with hes5 being a target of Notch signaling(Raya et al., 2003). By contrast, the expression of her3 and her9 was not significantly affected at the early segmentation stage in these Notch-defective embryos (Fig. 3F,G,J,K). Expression of zN5ICD strongly induced the ectopic expression of hes5 and abolished the neurod4 expression(Fig. 3M,P). By contrast,expression of xDl^stu suppressed the hes5 expression and induced a homogenous expression of neurod4 within the proneuronal domains (Fig. 3Q,T). However,neither zN5ICD nor xDl^stu affected the her9 expression(Fig. 3O,S). The her3expression was suppressed by zN5ICD, as reported(Hans et al., 2004), but was not affected by xDl^stu (Fig. 3N,R). These data indicate that neither the her3 nor the her9 expression requires Notch signaling at the early segmentation stage. The expression of her3 and her9 was partly dependent on Notch signaling at later stages (data not shown).

The her3 and her9 expression is complementary to that of the proneural genes, and the expression of her3, her9 and the proneural genes is co-regulated in embryos that are defective in Bmp signaling. These data raise the possibility that her3 and her9 expression is negatively controlled by the proneural genes or established upon the formation of the proneuronal domains. Positional information-dependent her3 and her9 expression might be mediated through the proneural genes or their downstream genes. To address this issue, we examined the expression of her3 and her9 in embryos that misexpressed or had defects in the function of proneural genes. Misexpression of neurog1 induced the ectopic expression of neurod4 and the postmitotic neuronal marker elavl3(Kim et al., 1997) (data not shown), suggesting that the misexpression of neurog1 could induce ectopic proneuronal domains. Nevertheless, the misexpression of neurog1 did not affect the expression of her9(Fig. 4B). The neurog1misexpression reduced the expression of her3(Fig. 4A), probably through the upregulation of delta and the activation of Notch signaling(Takke et al., 1999). Consistent with this, the ectopic expression of hes5 was induced by the misexpression of neurog1 (Fig. 4C).

The injection of antisense morpholino oligonucleotides (MOs) for neurog1 and olig2 inhibits the formation of RB neurons and primary motoneurons, respectively (Cornell and Eisen, 2002; Park et al.,2002). Consistent with this, the co-injection of neurog1-MO and olig2-MO inhibited the expression of islet1 and elavl3 in RB neurons and primary interneurons(Fig. 4F,I; data not shown),but did not affect the expression of her3 and her9(Fig. 4D,E,G,H), indicating that the regulation of her3 and her9 expression is independent of the proneural genes and is not established upon the formation of the proneuronal domains.

To reveal the functions of her3 and her9, we injected the her3-MO and/or her9-MO into wild-type or mib mutant embryos. We used two different MOs for each of Her genes and obtained essentially the same results (see Materials and methods). We show the data from the her9ATG-MO and her3-MO1 in Figs 5 and 6. In the her9-MO-injected embryos, neurod4 was detected in the inter-proneuronal domains between the RB neurons and primary interneurons,whereas neurod4 was not expressed in this region in the control embryos at the early segmentation stage(Fig. 5A,B,E,F). In the her9-MO-injected wild or mib mutant embryos, the neurod4-expressing intermediate and lateral domains were fused and became homogenous (Fig. 5C,D,G,H) (94%, n=129 for Fig. 5C,G; and 96%, n=45 for Fig. 5D,H),indicating that her9 is required for the suppression of neurod4 expression in the inter-proneuronal domains between the RB neurons and the primary interneurons. her9 is not required for restricting the number of neurod4-expressing cells within the proneuronal domains. In the her9-MO-injected embryos, the fourth ventricle in the hindbrain became smaller and the domain containing the atoh1-expressing neural precursors in the subventricular area was reduced at the pharyngula stage (88%, n=67 at 30 hours post fertilization, hpf), compared with the control embryos. The data suggest that Her9 is involved in the generation or maintenance of neural progenitor cells in the hindbrain, which are located in the ventricular zone of the fourth ventricle. As reported previously, injection of the her3-MO induced ectopic neurog1 expression only in rhombomeres 2 and 4(Hans et al., 2004) (86%, n=52 for Fig. 5M,N). These data suggest redundant roles for Her3 and Her9 in the inter-proneuronal domains that are located between the primary motoneurons and interneurons, and in the position posterior to rhombomere 4.

We also examined whether Her9 functions as a transcriptional repressor for the formation of the inter-proneuronal domains. To address this issue, we co-injected RNAs for Myc-tagged Her9 (Her9MT), a fusion protein of Her9 and the transcriptional activator VP16 (Her9VP16), or a fusion protein of Her9 and the transcriptional repressor domain of Drosophila Engrailed(Her9EnR), with β-galactosidase RNA into one blastomere of two- to four-cell stage embryos. In the majority of the her9MT RNA-injected embryos, the cells that received the her9MT RNA were localized in the dorsal part of the embryos. These cells did not localize to the proneuronal domains, but rather to the inter-proneuronal domains in the neural plate(Fig. 5O, Table 1). However, in those cells that localized to the proneuronal domains, neurod4 expression was reduced or absent (Fig. 5P, Table 1). Similar results were obtained in her9EnR RNA-injected embryos(Table 1). The cells that received the her9VP16 RNA were evenly distributed. The expression of Her9VP16 disrupted the proneuronal domain-specific neurod4 expression and induced an ectopic and spotty pattern of expression(Fig. 5Q). We also found that the misexpression of Her3 repressed the expression of neurod4 in the proneuronal domains (Fig. 5R, Table 1). These data indicate that Her9, like Her3 (Hans et al.,2004), functions as a transcriptional repressor to inhibit neurogenesis in the inter-proneuronal domains.

As the expression of her3 and her9 overlaps in the inter-proneuronal domain between the primary motoneuron and interneuron, it is likely that her3 and her9 function redundantly in the inter-proneuronal domain. We co-injected her3-MO and her9-MO into wild-type or mib mutant embryos, and examined the expression of neurog1, neurod4, deltaA and elavl3. Co-injection of her3-MO and her9-MO induced the ubiquitous expression of neurog1, deltaA, neurod4 and elavl3 within the neural plate,but not in the midline region that includes the floor plate[Fig. 6C (96%, n=78), Fig. 6G (94%, n=59), Fig. 6K (97%, n=66), Fig. 6O (99%, n=71)],indicating that Her3 and Her9 function redundantly in the inter-proneuronal domains. These data also show that Her3 and Her9 are required to repress the proneural genes and their downstream events in the inter-proneuronal domains.

As the ectopic expression of neurog1, deltaA, neurod4 and elavl3 was detected in the position of the inter-proneuronal domains of the her3/her9-MO-injected embryos, it is possible that some population of differentiated cells was increased or ectopically generated in the position of the inter-proneuronal domains. We examined the expression of islet1 and tlx3a, which are expressed in the primary motoneurons and RB neurons, and in the RB neurons, respectively(Andermann and Weinberg, 2001; Inoue et al., 1994; Langenau et al., 2002), in the her3/her9-MO-injected embryos. The islet1 and tlx3aexpression was not affected in the her3/her9 morphant embryos[Fig. 6R (100%, n=55), Fig. 6T (100%, n=63)]. Furthermore, there was no prominent change in the expression of the interneuron markers, lim1 (Nguyen et al., 2000), hlx1/2(Fjose et al., 1994; Seo et al., 1999), eng1b (Higashijima et al.,2004), evx1 (Thaeron et al., 2000), vsx1/2(Passini et al., 1997),sax2 (Bae et al., 2004)and pax2.1 (Mikkola et al.,1992) at the early and late segmentation stages, or at the pharyngula stage (data not shown). These data suggest that, although ectopic neurogenesis took place in cells located in the inter-proneuronal domains of the her3/her9 morphant embryos, they did not undergo differentiation to specific neuronal cell-types at the early segmentation stage or they differentiated into neurons that are not recognized by the genetic markers.

In the her3/her9-MO-injected embryos, neurog1,deltaA, neuroD4 and elavl3 still displayed spotty expression patterns, suggesting that the cells expressing these genes in the inter-proneuronal domains were still subjected to lateral inhibition. Consistent with this, in the her3/her9-MO-injected mibembryos, all of these genes were expressed ubiquitously and homogenously within the neural plate, except in the midline region[Fig. 6D (100%, n=31), Fig. 6H (96%, n=22), Fig. 6L (100%, n=26), Fig. 6P (100%, n=27)],further indicating that Her3/Her9 and Notch signaling play different roles in neurogenesis. Notch signaling functions in lateral inhibition to restrict the numbers of neuronal cells in the proneuronal domains. Her3 and Her9 function as prepattern genes that spatially repress neurogenesis and thereby generate the inter-proneuronal domains.

We demonstrated that there are two types of Her genes that are regulated by different mechanisms and exhibit distinct roles in neurogenesis in the zebrafish posterior neuroectoderm. her2, her4, her12 and hes5 are regulated by Notch signaling, whereas her3 and her9 are regulated by positional information (Figs 2, 3). There are several signaling molecules that are involved in the formation of the dorsoventral body axis and regulate the formation of neuroectoderm, including Bmps, Bmp antagonists and Fgfs. However, only the reduction of Bmp signaling affects the position and width of the proneuronal and inter-proneuronal domains(Barth et al., 1999; Nguyen et al., 2000). It is not clear how Bmp signaling regulates the periodic stripe expression of the proneural genes her3 and her9. Turing's `reaction diffusion'mechanism is proposed to be a good model to explain the formation of stripes,such as skin stripes (Kondo,2002; Meinhardt,2003). In this model, a certain chemical activator (such as a growth factor) induces a feedback inhibitor that can diffuse over a long distance, and the summation of the activator and the inhibitor generates a stripe of high activity of the activator. In the case of Bmp signaling, there are several feedback regulators reported, but none of them has been shown to form stripes of activity or to be involved in the stripe formation of the proneuronal domains, so far. Furthermore, periodic (striped) Bmp signaling activity (e.g. the phosphorylation of Smad1) has not been reported. Bmp signaling might regulate the position of the proneuronal and inter-proneuronal activity through the regulation of other signaling cascades. Alternatively,each stripe of her3 and her9 (and proneural gene) expression may be regulated by different mechanisms. A combination of dorsal and ventral signals might act on the distinct regulatory elements of the her3 and her9 genes, and generate the striped expression. Promoter analysis of her3 and her9 in transgenic zebrafish will be required to clarify the mechanism for their striped expression. Drosophila hairyalso shows a periodic expression pattern in the pupal leg(Orenic et al., 1993). It has been reported that hairy expression is regulated in part by Dpp, a Drosophila Bmp (Kwon et al.,2004). In this sense, there is some conservation between the regulation of Drosophila hairy and zebrafish her3/her9expression.

It has previously been reported that her3 is negatively regulated by Notch signaling (Hans et al.,2004). We also observed the repression of her3 expression by the expression of NotchICD or the misexpression of Neurog1. As the misexpression of neurog1 upregulates the hes5 expression(Fig. 4C), the suppression of her3 is likely to be mediated through Delta-Notch signaling. However,the expression of her3 was not affected in mib mutant,DAPT-treated or antimorphic Delta-expressing embryos, indicating that Notch signaling is not required for the endogenous expression of her3. The situation is similar to that of the expression of her5, which is proposed to function as a prepattern gene (described below) in the zebrafish mid-hindbrain boundary. her5 expression is repressed by NotchICD expression but is not affected in Notch-defective embryos(Geling et al., 2004). It is possible that NotchICD activates a gene cascade(s) that is not activated under physiological conditions, and renders the neuroectoderm incompetent to express her3. In any case, the striped expression of her3 and her9 does not require Notch signaling.

her2, her4, her12 and hes5 are expressed in the proneuronal domains, where the proneuronal genes and pnx are also expressed (Fig. 1). The expression of these Her genes is absent or strongly reduced in Notch-defective embryos (Fig. 3, data not shown), indicating that they are strictly regulated by Notch signaling. This is in contrast to the homogenous expression of neurog1, neurod4 and deltaA in the Notch-defective embryos(Fig. 6). As the proneural gene neurogenin activates the delta expression and subsequently activates Notch signaling in the neighboring cells(Takke et al., 1999), the cells expressing these Her genes are different from those expressing the proneural genes within the proneuronal domains, although their expression may overlap transiently during the lateral inhibition. All of these data indicate that the regulation of her3 and her9, and that of the other Her genes, is different: the latter are controlled by Notch signaling and the former by positional information.

In Drosophila, hairy and enhancer of split have different roles in neurogenesis. E(spl) functions downstream of Notch signaling and is involved in the lateral inhibition mechanism, whereas hairydoes not function downstream of Notch signaling, but rather as a prepattern gene that acts in the interface between positional information and neurogenesis (Davis and Turner,2001; Fisher and Caudy,1998). hairy shows a periodic longitudinal expression pattern in the Drosophila leg and determines the position of sensory organs by repressing the proneural gene acheate(Orenic et al., 1993; Skeath and Carroll, 1991). Although the structure of the zebrafish neural plate is very different from that of the Drosophila leg, the role of Her3 and Her9 in the zebrafish neural plate is similar to that of Hairy in the Drosophilaleg. We propose that Her3 and Her9 function as prepattern genes and control the position of the proneuronal and inter-proneuronal domains through a conserved mechanism, by which Hairy also controls the position of sensory organs in the Drosophila leg. Among the Her genes in vertebrates,only a few are reported to function in a similar way. Mouse Hes1 is expressed in the olfactory placodal domains independent of Mash1 activity,thus Hes1 expression is suggested to be Notch signal independent(Cau et al., 2000). Combined disruption of Hes1 and the Hes5 gene, whose expression is dependent on Mash1 and thus possibly controlled by Notch signaling, leads to a strong upregulation of neurog1 expression in the olfactory epithelium(Cau et al., 2000). The situation is similar to the effect of the combined inhibition of Her3/Her9 and Notch signaling in the zebrafish posterior neuroectoderm(Fig. 6). Zebrafish her5 and him, which are expressed in the MHB, function to repress neurogenesis in the MHB (Geling et al., 2004; Ninkovic et al.,2005). Although the mechanism that induces the expression of her5 and him is not clear, the maintenance of her5and him1 in the MHB involves Pax2.1, Eng2/3 and Fgf8, which provide positional information in the MHB (Geling et al., 2003; Geling et al.,2004; Ninkovic et al.,2005). In this sense, the function of her3 and her9 is similar to that of her5 and him. Mouse Hes1 and zebrafish her5 function downstream of positional information related to the anteroposterior (AP) axis and control neurogenesis in the specific position of the AP axis, whereas her3 and her9 function as prepattern genes that control neurogenesis in the context of the dorsoventral (DV) axis in the neuroectoderm. Intriguingly, the inhibition of Her5 and Him function leads to ectopic neurog1expression in the MHB, while leaving a neurog1-negative domain in the lateral region of MHB (Ninkovic et al.,2005). her3 is expressed in the lateral region of the MHB(Fig. 1)(Hans et al., 2004). It is possible that her3 contributes to the repression of proneural genes in the lateral region of the MHB. her3, her5 and him might redundantly function in this region, as the inhibition of Her3 function did not lead to ectopic neurog1 expression in the MHB(Fig. 5). All of these data support the role of a subset of Her genes in the prepatterning that functions downstream of the positional information linked to the DV and AP axes.

We found a difference in the activity of Her3 and Her9 by misexpression studies (Fig. 5). Cells expressing exogenous her3 contributed to the proneuronal domains and repressed the expression of neurod4 in the proneuronal domains. When cells that expressed exogenous her9 were located in the proneuronal domains, neurod4 expression was repressed. However, the majority of exogenous her9-expressing cells was located in the inter-proneuronal domains and did not contribute to the proneuronal domains. These data suggest that Her9 may play an additional role other than the repression of neurogenesis (which it shares with Her3). Her9 may confer on the cell its ability to be localized to the inter-proneuronal domains. Her9 may be involved in the regulation of genes that encode cell-adhesion molecules, or repulsive or attractant molecules. This Her9-mediated activity to localize cells to the inter-proneuronal domains is attributable to the clear separation of the proneuronal and the inter-proneuronal domains.

In amniotes, there is no discernible primary neuron. However, some Hes/Her genes show a spatially restricted expression pattern in the spinal cord during a specific developmental period(Hatakeyama et al., 2004; Wu et al., 2003), and these Hes/Her genes may function in the spatial regulation of neurogenesis similar to her3 and her9. As observed for mouse Hes1 in the olfactory placode(Cau et al., 2000), a specific Her/Hes gene may be regulated by the positional information early and controlled by Notch signaling later. Consistent with this idea, the expression of her3 and her9 also becomes dependent on Notch signaling at the later stages (data not shown). Notch-mediated Her9 function may be involved in later processes of neurogenesis, such as the reduction of atoh1-expressing neural precursor cells in the hindbrain region(Fig. 5L).

Although neurog1, deltaA, neurod4 and elavl3 were ectopically expressed in the inter-proneuronal domains in the her3/her9 morphant embryos, we did not detect ectopic/aberrant expression of the genes that are expressed in primary motoneurons(islet1 and islet2), primary interneurons (lim1),RB neurons (islet1, islet2, tlx3a) or of other interneuronal markers(hlx1/2, eng1b, evx1, vsx1/2, sax2 and pax2.1)(Fig. 5, data not shown). There are two possible explanations for this result. First, the cells ectopically expressing neurog1, deltaA, neurod4 and elavl3 may not undergo differentiation to specific types of neurons at the early segmentation stage. Alternatively, these cells may become neurons that are not detected by the markers we used, or neurons that do not normally exist. The motor-, inter-and sensory-neuronal markers used in this study correspond to the chick and mouse genetic markers that cover the most of the neuronal types along the DV axis, in particular, mid-ventral interneurons, in the spinal cord. Therefore,it is more likely that these cells do not differentiate into specific types of neurons until the physiological timing of the differentiation. These data suggest that Her3 and Her9 are not involved in the specification of neurons. We also noticed that glial populations, which were marked by Gfap, Blbp (brain lipid-binding protein) and MBP, were not prominently affected in the her3/her9 morphant embryos, whereas they were strongly reduced in the mib mutant embryos (data not shown). These data support the idea that Her3 and Her9 do not function in the specification of glia versus neurons. Considering these observations together, Her3 and Her9 are involved in the delay of the neuronal differentiation, and the inter-proneuronal domains,which require the function of Her3 and Her9, provide the field that maintains undifferentiated neural cells.

It is not yet clear what types of neurons and glia are generated from the inter-proneuronal domains. In Notch-defective embryos, the numbers of primary neurons are increased and the secondary neurons are strongly reduced(Appel et al., 2001). This is consistent with the finding that Notch-defective embryos show increased elavl3-expressing neuronal cells and a strong reduction of elavl3-negative cells within the proneuronal domains(Itoh et al., 2003)(Fig. 6). Thus, it is conceivable that both the primary and the secondary neurons are generated from the proneuronal domains, and the inter-proneuronal domains give rise to only the non-neuronal populations - glia. However, inhibition of the inter-proneuronal domains by her3/her9-MO did not strongly affect the glial population in the spinal cord (data not shown). Therefore, it is more likely that the inter-proneuronal domains generate the secondary neurons after the expression of her3 and/or her9 is reduced or terminated at later periods. Notch signaling could function in the inter-proneuronal domains in the absence of Her3 and Her9 function(Fig. 6). Therefore, Notch signaling may function in the specification of glia versus neurons in the inter-proneuronal domains at the later stages. Precise fate mapping of cells in the inter-proneuronal domains will be required to clarify the fates of the cells in the inter-proneuronal domains.

In summary, her3 and her9 function as prepattern genes,the expression of which is controlled by positional information linked to the dorsoventral polarity of the posterior neuroectoderm, and their expression spatially controls neuronal differentiation at the beginning of neurogenesis. The Her-mediated prepattern mechanism contributes to the establishment of the central nervous system through the spatially coordinated regulation of neurogenesis.

We thank D. Tautz, Y. Imai, M. Kobayashi, H. Okamoto, E. Weinberg, C.-H. Kim, A. Kawakami and A. Chitnis for providing the DNA samples and the fish; C. Hirata-Fukae and H. Akiyama for fish care; A. Katsuyama for making histological sections; and members of the Hibi laboratory for critical discussion. Y.-K.B. is supported by the Special Postdoctoral Researcher program of RIKEN. This work was supported by Grants-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science, Sports and Technology.