Characterization of the pufferfish Otx2 cis-regulators reveals evolutionarily conserved genetic mechanisms for vertebrate head specification

The establishment of the body plan of the vertebrate embryo requires the expression of regulatory genes in precisely controlled spatial domains, such that their products can function appropriately in the patterning process. Elucidation of the underlying mechanisms involved in this regulatory cascade initially depends on the identification of cis-acting regions that mediate spatial and temporal regulation. Homeobox genes are thought to participate in such developmental processes, thereby providing an ideal system with which to examine the genetic regulatory mechanism. Notably, Hox genes play a crucial role in the body plan of the trunk region, caudal to the hindbrain(Krumlauf, 1994). An intriguing co-linearity exists between the order of vertebrate Hox genes on the chromosome, and their temporal and spatial expression(Krumlauf, 1994). Much is known regarding the genetic regulatory mechanism of Hox expression(Nonchev et al., 1997; Duboule, 1998). However, in the head region, the regulatory mechanism of homeobox gene expression remains uncertain (Logan et al., 1993; Kammandel et al., 1999; Zerucha et al., 2000).

The vertebrate head is a complex structure, which is primarily composed of the skull, specific sense organs, muscles, glands, skin and the brain, which can be further subdivided into distinct and complex domains. Therefore,vertebrate head development requires highly coordinated processes. An understanding of the genetic mechanisms that control regional specification and morphogenesis of the embryonic rostral brain has only recently come to light; homeobox genes, Fgf, Wnt and other signaling molecules play an important role in these processes (reviewed by Rubenstein et al., 1998). Among these genes, Otx2, a paired-type homeobox gene, has been isolated as a mouse cognate of orthodenticle in Drosophila (Finkelstein and Perrimon, 1991; Simeone et al., 1992). The protein sequence of the homeodomains, and their patterns of expression, have all been highly conserved during evolution of the animal kingdom. Notably, mouse Otx2 expression occurs earliest in the inner cell mass at the blastocyst stage, and in the visceral endoderm and epiblast at the prestreak stage (Acampora et al., 1995; Ang et al.,1996; Kimura et al.,2000; Kimura et al.,2001). During midstreak stages, the expression domain is restricted to the anterior region in all three germ layers corresponding to the presumptive rostral head (Simeone et al., 1993; Ang et al.,1994). At the pharyngula stage, Otx2 expression is present in developing sense organs, such as the eyes, olfactory epithelium and the inner ear, as well as in the prospective forebrain and midbrain regions(Simeone et al., 1993; Mallamaci et al., 1996).

Proper regulation and function of this gene is crucial, as the mutation of Otx2 leads to defects in the rostral head in several developmental processes (Acampora et al.,1995; Acampora et al.,1997; Matsuo et al.,1995; Ang et al.,1996; Suda et al.,1996; Suda et al.,1997; Suda et al.,2001; Tian et al.,2002; Hide et al.,2002). However, very little is known about how Otx2 gene expression is dynamically regulated. The expression of homeobox genes is regulated primarily at the level of transcription. Accordingly, cis-elements and trans-factors that govern spatiotemporal Otx2 expression may provide valuable insight into the genetic pathways that control anteroposterior axis formation, patterning of the head, and specification of the forebrain and midbrain. Previously, we identified cis-regulatory regions in the mouse Otx2 5′-flanking region, which are mainly responsible for the expression patterns in the head mesenchyme, anterior visceral endoderm and mesendoderm (Kimura et al., 1997; Kimura et al.,2000). However, we were not able to detect any cis-elements governing embryonic expression in the forebrain, midbrain and sense organs within the 80 kb mouse Otx2 locus (–65 kb to +15 kb) (C.K.-Y. and I.M., unpublished). This finding suggests that cis-elements of mouse Otx2 expression must lie outside of this surveyed 80 kb region.

To unravel the difficulty of identifying Otx2 cis-regulatory regions, we took advantage of the compact genome of the pufferfish Fugu rubripes (Fugu), which displays a genome of 365 Mb, which is approximately eight times smaller than the human and mouse genomes(Brenner et al., 1993; Aparicio et al., 2002). The pufferfish genome has a similar number of genes and this fish shares a common body plan for specialized functions in higher vertebrates; consequently, it provides an ideal system with which to investigate conserved mechanisms of Otx2 gene regulation. Coincidentally, it has been shown that several regulatory cis-elements have been functionally conserved between mammals and pufferfish (Marshall et al.,1994; Pöpperl et al.,1995; Aparicio et al.,1995; Kimura et al.,1997; Camacho-Hubner et al.,2000; Brenner et al.,2002).

In the present study, the Fugu Otx2 (Fotx2) gene was cloned. In addition, endogenous Fotx2 expression was analyzed during Fugu embryogenesis. Subsequently, cis-acting regulators throughout the entire Fotx2 genomic locus were surveyed by employing transgenic mice. Many distinct cis-regulators were identified, which directed temporally and spatially specific expression, mimicking endogenous Otx2expression. Additionally, these cis-regions possessed the ability to drive transgene expression in zebrafish embryos and contained evolutionarily conserved sequence elements. These results indicate that developmental domain-specific multiple cis-regulators control highly coordinated Otx2 expression during vertebrate rostral head development.

Degenerate PCR primers were designed based on the amino acid sequences of the zebrafish and mouse Otx2 genes. The conserved amino acid sequences GQNKV and WKFQV were utilized to synthesize the oligonucleotide primers PF1 (5′-GGAATTCGGNCA(A/G)AA(T/C)AA(A/G)GT-3′) and PF2(5′-TTGGATCCAC(T/C)TG(A/G)AA(T/C)TTCCA-3′), respectively. A 590 bp PCR product was cloned into pBluescript (KS–) and sequenced. The Fugu genomic phage library(Kimura et al., 1997) was probed with the 590 bp Fugu Otx2 fragment described above. DNA sequence determination of the 3.5 kb Fugu Otx2 genomic region was performed by sequencing of both strands of double-stranded plasmid DNA.

Wild pufferfish embryos were maintained at 18°C, and sacrificed at subsequent developmental stages for RNA preparation and in situ hybridization as described (Suzuki et al.,2002). Primers for Fotx2 expression,5′-TTACGCGCGCCCAGTTAGACGTTTTGGAGG-3′ and 5′-GACGCCGGGGACTGGTTCAGATGGCTTGTG-3′, afforded a 586 bp product. Primers for β-actin expression, 5′-GGCACCGCTGCCTCCTC-3′ and 5′-GCCAGGATGGAGCCTCC-3′, yielded a 359 bp product. Whole-mount in situ hybridization was conducted according to the method of Jowett and Lettice(Jowett and Lettice,1994).

All constructs used in this study were generated using standard molecular cloning techniques (Sambrook and Russell,2001). Generation of construct F0placZ was described previously, as construct #11 (Kimura et al., 1997). Construct F1placZ was generated by insertion of a 5′ 4.0 kb EcoRV-NspV genomic fragment, blunted by T4 DNA polymerase, into a SmaI site of construct F0placZ. Construct F2placZ was produced by insertion of a 5′ 2.5 kb EcoRV-BamHI genomic fragment, blunted by T4 DNA polymerase,into a SmaI site of construct F0placZ. Construct F3placZ was generated by cloning of a 5′ SacII-BamHI genomic fragment into the SacII and BglII sites of construct F0placZ. Construct F4placZwas generated by insertion of a 5′ 4.0 kb NotI-BamHI genomic fragment into NotI and BglII sites of construct F0placZ. Construct F5placZ was generated by insertion of a 5′ 11.0 kb BamHI-BamHI genomic fragment into a BglII site of construct F0placZ. Construct F6placZwas generated by insertion of a 5′ 1.5 kb NotI-BamHI genomic fragment into NotI and BglII sites of construct F0placZ. Construct F7placZ was produced by insertion of a 3′ 2.5 kb BstEII-BamHI genomic fragment blunted by T4 DNA polymerase into a SmaI site of construct F0placZ. Construct F8placZ was generated by insertion of a 3′ 8.0 kb XbaI-BamHI genomic fragment into XbaI and BglII sites of construct F0placZ. F9placZ was generated by insertion of a 3′ 5.0 kb XbaI-BamHI genomic fragment into the XbaI and BglII sites of construct F0placZ. Construct F10placZ was produced by insertion of a 3′ 9.0 kb BamHI-NotI genomic fragment into NotI and BglII sites of construct F0placZ. Construct F11placZ was generated by insertion of a 3′ 5.5 kb NotI-BamHI fragment into NotI and BamHI sites of construct F0placZ. Construct F12placZ was generated by insertion of a 3′ 6.3 kb BamHI-BamHI fragment into a BglII site of construct F0placZ. Construct F13placZ was produced by insertion of a 4.8 kb NotI-BamHI fragment into NotI and BamHI sites of construct F0placZ.

In order to construct the F3hsplacZ vector, a 2.4 kb SmaI-EcoRV fragment of hsp-lacZ(Kothary et al., 1989) was first inserted into SmaI and EcoRV sites of construct pBSlacZIII (Kimura et al.,1997), yielding the pBShsplacZ vector. Construct F3hsplacZ was then generated by cloning of a 5′ SacII-BamHI genomic fragment into the SacII and BglII sites of construct pBShsplacZ. In order to construct F3pGFP and F8pGFP transgenes, the sequence of the translational start site of the lacZ gene in F3placZ and F8placZ was converted into a NcoI linker sequence by PCR-based mutagenesis, yielding F3placZ(Nco) and F8placZ(Nco), respectively. Subsequently, a 740 bp NcoI-EcoRI fragment of the plasmid pEGFP (Clontech)was inserted between the NcoI and EcoRI sites of the F3placZ(Nco) and F8placZ(Nco) vectors, respectively. Construct E6/7dplacZ was produced by insertion of three genomic fragments, the 660 bp NotI-ClaI (blunted), the 960 bp BsgI (blunted)-SphI (blunted) and the 916 bp SapI(blunted)-BamHI, into NotI and BamHI sites of construct F0placZ.

Transgenic mice were generated by microinjection of fertilized eggs from CD-1, as described by Hogan et al. (Hogan et al., 1994). Transgenic mouse embryos were identified by PCR analysis as previously described (Kimura et al., 1997).

β-Gal staining was conducted as described previously(Kimura et al., 1997). For whole-mount views of 10.5 dpc to 13.5 dpc embryos, specimens were washed with PBS and dehydrated in graded ethanol. Specimens were then cleared in 1:2 benzyl alcohol/benzoate as described (Eng et al., 2001). For histological analysis, embryos were embedded in paraplast. Serial sections (10 μm) were produced and stained with Eosin.

Zebrafish embryos were obtained by natural mating(Westerfield, 1995) and were staged accordingly (Kimmel et al.,1995). F3pGFP and F8pGFP constructs were linealized by digestion with SacII and SalI, respectively. Digested DNA was resuspended in water and injected into the blastomere of early one-cell stage embryos as described(Kobayashi et al., 2001). Embryos were fixed in PBS containing 4% paraformaldehyde, and examined under GFP-plus filters on a MZFLIII microscope (Leica) equipped with a C5810 chilled CCD camera (Hamamatsu Photonics).

In order to identify the Fugu Otx2 (Fotx2) gene, PCR of Fugu genomic DNA was used to isolate a 590 bp fragment of the Fotx2 gene (see Materials and methods). Consequently, two genomic DNA clones, which contained the coding region and overlapped one another, were obtained from a Fugu genomic phage library(Kimura et al., 1997)(Fig. 3). A 3.5 kb PstI-BamHI fragment hybridized with the 590 bp probe was used as the source of the complete DNA sequence of the Fotx2 gene,which is available at GenBank (Accession Number AY 303542). The overall size of the Fotx2 gene, from initiation to termination codons, is 1901 bp,which is approximately twofold smaller than the mouse gene(Fig. 1A)(Simeone et al., 1992)(http://www.ensembl.org/Mus_musculus/). The coding regions of the Fotx2 gene, consisting of three exons, were nearly identical in size, 94, 152 and 624 nucleotides, respectively, to those in the mouse Otx2 gene (Fig. 1A). By contrast, the first and second introns (467 and 560 nucleotides, respectively) of Fotx2 are approximately threefold smaller than those of the mouse (Fig. 1A).

The general structure of the amino acid sequence of Fotx2 is quite similar to the sequences of previously described members of the Otxfamily; each of these species shares a short N-terminal domain, a paired-type homeodomain and a long C terminus, including two repeats of the Otx tail motif (Fig. 1B)(Furukawa et al., 1997). Notably, Fotx2 exhibits 96% homology with zebrafish Otx2,93% homology with mouse Otx2 and 90% homology with Xenopus Otx2, throughout the entire peptide sequence(Fig. 1B). Remarkably, the homeodomain sequence is fully conserved but for a single amino acid change at position 52 in Xenopus Otx2 (Fig. 1B). Further phylogenetic analyses have indicated that Fotx2 is closely related to other vetebrate Otx2 genes;however, these data reveal that Fotx2 is not closely related to either Otx1 or Crx genes(Fig. 1C).

In order to investigate Fotx2 expression during development,RT-PCR analysis was performed in wild-type Fugu embryos. The developmental stage of Fugu is substantially similar to that of other teleosts, such as zebrafish (data not shown). Fotx2 mRNA expression was not detected at 22 hours postfertilization (hpf), which is equivalent to the dome stage in zebrafish. However, it was observed initially as early as 30 hpf, equivalent to the shield stage in zebrafish(Fig. 2A). These data indicate that Fotx2 mRNA is not present in the oocyte and, furthermore, that it is zygotically transcribed following midblastula transition. Subsequently, Fotx2 transcripts were significantly upregulated around 54 hpf, an early somites stage (Fig. 2A). These transcripts were detected throughout the segmentation and pharyngula periods; additionally, transcripts were detected at 172 hpf when embryos had been hatched (Fig. 2A).

To examine Fotx2 expression during embryogenesis more precisely,whole-mount in situ hybridization of Fugu embryos was performed. We found Fotx2 expression in the developing rostral head(Fig. 2C-E). Fotx2expression was apparent in the anterior portion of the neural plate at 44 hpf,which is equivalent to the bud stage of zebrafish(Fig. 2C). In order to compare anteriorly localized Fotx2 expression, pufferfish no tailexpression was also examined (Fig. 2B). As expected, no tail expression was observed consistently at the opposite side to Fotx2 expression; it was detected in the posterior portion of the body, including in the notochord and tail bud at 44 hpf (Fig. 2B). By 54 hpf stage, the 5-6 somites stage, Fotx2 expression was elevated significantly in the prospective forebrain and midbrain regions(Fig. 2D). At the subsequent 78 hpf stage (18-20 somites stage), Fotx2 expression was evident throughout the entire rostral head (Fig. 2E). These results indicate that pufferfish Otx2 is expressed in the rostral head during Fugu embryogenesis.

In order to understand the cis-regulatory mechanism of Fotx2 gene expression, the activity of reporter gene constructs harboring its genomic DNA were surveyed. First, five lambda phage contigs of genomic DNA from–30.5 kb to +38.5 kb (Fig. 3) were cloned and mapped. These contigs largely cover the entire Fotx2 genomic locus (see Discussion). Next, 13 genomic fragments were tested for cis-activity via generation of transgenic mouse embryos carrying the reporter lacZ gene (Figs 3, 4). Transgene expression analysis was transiently conducted at 10.5 dpc; seven genomic fragments directing lacZ expression were identified (Figs 3, 4). All thirteen constructs contained a 2.4 kb Fotx2 promoter, which displays no cis-activity during embryogenesis (Fig. 3; F0placZ) (Kimura et al.,1997). For seven of the 13 DNA fragments tested, lacZexpression occurred in spatial and temporal patterns that mimicked a subset of the normal Otx2 expression pattern, with the exception of the F4 cis-region. This fine correlation suggests that these cis-regions can function independently of one another, and that their spacing relative to the promoter is not crucial for expression.

Most of the Fotx2 cis-acting regions that we defined activated expression in multiple domains and at several developmental stages (Figs 3, 4, 5, 6, 7, 8, 9, 10, 11), suggesting that the Fotx2 loci are composites of smaller more specific cis-regions. For example, the distant F3 region, spanning from –17 to –8.9 kb,directed transgene expression in the roof of the diencephalon and the medio-caudal telencephalon (Fig. 4A). The F4 region, which spans from –18 to –14 kb,directed lacZ expression in the premandibular, mandibular and hyoid arches (Fig. 4B; F4placZ). The further distant F5 region, which spans from –29 to –18 kb, governed lacZ expression in the lateral portion of the mesencephalon and the ventral diencephalons, including the zona limitans intrathalamica (ZLI) (Fig. 4C). The F8 region spanning +2.5 to +10.5 kb directed lacZ expression in the dorsal diencephalon and mediocaudal telencephalon(Fig. 4D; F8placZ). The F9 region (+10.5 to +15.5) governed lacZ expression in the cephalic mesenchyme (Fig. 4E; F4placZ) (Kimura et al.,1997). The further distant F11 region, spanning +22.8 to +26.5 kb,could drive lacZ expression in the trigeminal ganglions, cranial nerves, dorsal diencephalon, mesencephalon, rhombencephalon and spinal cord(Fig. 4F). The F12 region,harboring +26.5 to +32.8 kb, governed lacZ expression in the nasal pits, retina and first branchial groove(Fig. 4G).

In order to characterize Fotx2 cis-activity during embryogenesis more precisely, transgene expression was analyzed during embryogenesis. Each cis-acting region was found to drive transgene expression in a distinct subdomain of Otx2 expression (Figs 5, 6, 7, 8, 9). Transgenic mice were established; each strain carried one transgene from F3placZ, F5placZ,F8placZ, F11placZ and F12placZ. Subsequently, their F1 or F2 hemizygous offspring were examined with respect to lacZexpression.

lacZ expression governed by F3placZ carrying the–17 to –8.9 kb region was observed initially in the restricted domain of the anterior neural plate at 8.5 dpc(Fig. 5A,B). At 9.5 dpc, lacZ expression was upregulated in the prospective dorsocaudal telencephalon and dorsal diencephalon (Fig. 5C). By 10.5 dpc, lacZ expression was further restricted in the diencephalic roof and dorsocaudal telencephalon(Fig. 5D,H). Histological analysis revealed that lacZ expression was present in the most caudal and medial aspects of the dorsal telencephalon, surrounding the lamina terminalis (Fig. 5E,F). At the level of the diencephalon, lacZ expression, which was evident in the most dorsal neural tube, occurred as two stripes excluding the dorsal-most roof (Fig. 5G). At 12.5 dpc,cis-activity remained in structures that developed into the choroid plexus;however, activity was not apparent in the cortical hem(Fig. 5I-L). In actuality,expression governed by the F3 cis-region appears to be closely related to Bmp and noggin expression (Fig. 5O)(Furuta et al., 1997; Hebert et al., 2002). This activity remained at 13.5 dpc (Fig. 5M,N).

In order to assess whether the F3 fragment can drive similar expression patterns for a heterogeneous promoter, the mouse hsp68 promoter,which is widely used in transgenic mice, was employed in place of the pufferfish Otx2 promoter (Kothary et al., 1989). Subsequently, we found that the F3 fragment was also able to direct expression under the control of the hsp68promoter in the diencephalic roof and the dorsocaudal telencephalon at 10.5 dpc (n=4/5; F3hsp68lacZ; Fig. 5P). This finding indicates that lacZ expression patterns are induced primarily by the F3 cis-region, and are not due to the Fotx2 promoter activity itself.

lacZ activity driven by F8placZ, carrying the +2.5 to+10.5 kb region, was detected initially in the dorsal diencephalon and in restricted areas of the dorsocaudal telencephalon at 9.5 dpc(Fig. 6A,B). Subsequently, at 10.5 dpc, lacZ expression was upregulated in the dorsal diencephalon,including in the prospective epithalamus and the mediocaudal telencephalon(Fig. 6C,D). Further histological examination demonstrated that lacZ activity was present in the dorsomedial portion of the caudal telencephalon, corresponding to the edge of the embryonic cerebral cortex, which is referred to as the cortical hem (Fig. 6E-G)(Grove et al., 1998). From 9.5 to 10.5 dpc, the expression domains governed by the F8 cis-region appear to display partial overlap with those governed by the F3 cis-region. Later, at 12.5 and 13.5 dpc, lacZ activity was detected in both the diencephalon and the telencephalon, including the cortical hem; however,activity was absent in the choroid plexus, which is directed by F3placZ as described (Fig. 6H-K, Fig. 5). The expression pattern generated by the F8 cis-acting region appears to be similar to that of Wnt3a (Fig. 6L) (Grove et al.,1998; Lee et al.,2000).

lacZ expression governed by the F5 (–29.0 to –18.0 kb)cis-region was observed initially in the ventral portion of the diencephalon at 9.5 dpc (Fig. 7A). By 10.5 dpc, lacZ expression was detected in the ZLI, the ventral diencephalon and the lateral portion of the mesencephalon, with a longitudinal direction as described (see Fig. 7B,C). Further histological examination indicated activity, at 11.5 dpc, in the lateral mesencephalon, ZLI, and ventral portion of P2 and P3 diencephalon, including in the retromammillary region(Fig. 7D-G). Even at 13.5 dpc, lacZ expression remained in these structures(Fig. 7H,I).

At 9.5 dpc, lacZ expression mediated by the F11 (–22.8 to–26.5 kb) cis-region was detected initially in the dorsal portion of the mesencephalic neuroepithelium, corresponding to the mesencephalic trigeminal ganglions, oculomotor nerves and a rostral aspect of trigeminal nerves(Fig. 8A). The active domains were extended both rostrally and caudally by 10.5 dpc in these structures(Fig. 8B, D-G). Additionally,expression was also evident in the dorsal region of the pretectum(Fig. 8B). With respect to branches of the trigeminal nerve, the F11 region drives expression exclusively in the opthalmic and maxillar branches, but not in the mandibular branch(Fig. 8B). Precise histological examination indicated that lacZ expression was present throughout the trigeminal ganglions and dorsal mesencephalic neural tube at 10.5 dpc(Fig. 8D-G). Subsequently, most lacZ expression in cranial nerves was diminished considerably by 11.5 dpc (Fig. 8C). Consequently, lacZ activity covered the dorsal pretectum, and the inferior and superior collicullus, by 13.5 dpc (Fig. 8H-J). Moreover, lacZ expression was observed consistently, from 9.5 dpc, in ventral domains of the neural tube at the level of the hindbrain, where Otx2 protein is also localized(Fig. 8A-C,E-G)(Mallamaci et al., 1996). Additionally, lacZ expression at the spinal cord level was observed ectopically in the transgenic embryos (Fig. 8A,B).

F12 (+26.5 to +32.8 kb) cis-regulatory activity was detected initially in the rostral first branchial groove and nasal portions as early as 9.5 dpc(Fig. 9A). lacZactivity in the retina commenced by 10.5 dpc(Fig. 4B,C). Subsequently, at 11.5 dpc, lacZ activity in the caudal portion of the mandibular arch extends to the cranial portion of the hyoid arch; moreover, activity was evident in both epithelium and mesenchyme, which corresponds to the auricular hillocks (Fig. 9E). This activity remained detectable in the external acoustic meatus of the ears at 12.5 dpc (Fig. 9F-I). Close histological examination revealed the presence of lacZ activity in the olfactory epithelium, as well as the in the outer layer of the optic cup corresponding to the prospective pigment layer of the retina at 12.5 dpc(Fig. 9J,K). Moreover, lacZ activity was observed in the developing inner ear, cochlea and saccule (Fig. 9H,I).

These expression patterns indicate that the aforementioned cis-regulators,with the exception of F4, mediate Otx2 expression in spatiotemporally distinct domains during rostral head development. Many cis-regulatory regions of genes have been shown to govern overlapping expression domains in a redundant manner; however, the seven cis-regulators currently identified displayed no redundant expression patterns (Figs 4, 5, 6, 7, 8, 9, 10). Furthermore, we concluded that the Fotx2 cis-regions function in a domain-specific fashion,rather than in a developmental stage-specific fashion. Their activities most likely depend on the expression of position-specific transactivators, which are expressed over time, spanning several subdomains within a given lineage.

In order to test whether the cis-regulators identified here can direct transgene expression in fish, cis-activity in zebrafish embryos was investigated. Among the seven cis-regulators, F3 and F8 regions faithfully govern expression in distinct domains of the dorsal diencephalon and mediocaudal telencephalon (Figs 4, 5, 6, 10). Thus, the activity of these two cis-regulatory regions was transiently examined in transgenic zebrafish. All constructs were tested via injection of DNA into zygotes;subsequently, the resulting mosaic zebrafish embryos were analyzed for GFP activity. First, the Fugu Otx2 promoter fused with the EGFP vector(pGFP) alone afforded no GFP signals throughout zebrafish embryogenesis (data not shown; n=0/74). The F3pGFPconstruct, which harbors the –17 to –8.9 kb genomic fragment fused with the pGFP vector, was initially expressed specifically at the anterior ectoderm, and in the posterior portion of zebrafish embryos ectopically from 80% epiboly onward (Fig. 11A). At early somites stage, GFP activity was observed specifically in the prospective forebrain region(Fig. 11B). Subsequently, at the 18 somites stage, GFP activity was consistently detected in the diencephalon and the telencephalon (Fig. 11C,D; n=38/51). At the later pharyngula, 32 hpf, stage,GFP activity remained in evidence in the dorsal diencephalon(Fig. 11E).

In contrast to the F3pGFP transgene, F8pGFP, which contains the –17 to –8.9 kb genomic fragment fused with the pGFP vector, was initially expressed from 12 to 14 hpf, and from the 6 to 10 somites stage (data not shown). Subsequently, GFP activity directed by F8pGFP was detected consistently in the diencephalon and in the telencephalon at 18 hpf (18 somites stage)(Fig. 11F; n=20/69). Later, at 30 hpf, an early pharyngula stage, F8pGFP expression remained in the telencephalon and the diencephalon(Fig. 11G). These expression patterns indicated that the activity of F3pGFP consistently occurs earlier, and is stronger and more widely distributed, than that of F8pGFP, suggesting that each fragment mediates distinct expression in the forebrain regions of the transient transgenic zebrafish embryos (Figs 5, 6, 11). These data indicate that the spatiotemporal expression governed by these two cis-regions is conserved between transgenic mouse and zebrafish, which suggests that an evolutionarily conserved mechanism controls these two cis-regions of Otx2 gene.

Recent reports of large-scale sequence analysis, combined with transgenic mouse approaches, reveal that evolutionarily conserved non-coding genomic sequence elements tend to control gene expression(Hardison, 2000). Thus, in order to identify highly conserved, putative regulatory sequence elements, a comprehensive comparative analysis of the intergenic regions flanking the mouse and pufferfish Otx2 genes was performed by PipMaker analysis(Schwartz et al., 2000). Consequently, we identified eight highly conserved sequence elements within the pufferfish cis-regulators (Fig. 12). The F4, F9 and F12 cis-regions possess conserved elements,E4, E5 and E8, respectively. The F5 cis-region possesses three conserved elements, E1, E2 and E3. F11 contains two conserved regions, E6 and E7. In addition, the corresponding eight conserved elements in mouse were located in an order similar to that of pufferfish; however, these elements were within an approximately eightfold extended region up to 550 kb. These data further support the notion that genetic mechanisms of Otx2 gene regulation are evolutionarily conserved among vertebrates. Moreover, as homologous regulatory elements present in different species represent direct descendants of a common ancestral regulatory element, individual cis-acting regions appear to have developed their distinct regulatory elements independently.

In order to determine whether these evolutionarily conserved elements play crucial roles with respect to cis-activity, a mutant F11 construct, in which both elements, E6 and E7, were specifically deleted (E6/7dplacZ) was generated (Fig. 3). The pufferfish E6 and E7 elements are located at the 3′ end of the Otx2 gene, at +23.5 and +25 kb, respectively(Fig. 12B). The corresponding mouse E6 and E7 sequences, although also located at the 3′ end, are at+175 and +196 kb, respectively, which is distant from the Otx2 coding region. Subsequently, lacZ expression of transient transgenic embryos carrying the E6/7dplacZ construct was analyzed as described. In contrast to F11placZ transgenic embryos, lacZ expression was not detected in E6/7dplacZ transgenic embryos at 10.5 dpc(n=0/9; Fig. 3, Fig. 12C,D). This finding demonstrates that conserved elements between mouse and pufferfish, E6 and E7,are essential for the transgene expression in mouse embryos. Moreover, this conserved function of regulatory elements between mouse and pufferfish suggests that the compact Fugu genome could be a potentially powerful and useful system for the rapid analysis of transgene expression.

We have described the isolation and characterization of the pufferfish Otx2 gene. Furthermore, we have identified multiple cis-regulators of the Fotx2 locus in transgenic embryos. The current data indicate that Fotx2 is regulated by distinct and physically separable cis-acting sequences that reside in a sequence of more than 60 kb surrounding the Fotx2 coding region. Each transgene harboring different cis-acting regions is reproducibly expressed in subdomains of Otx2 expression in mouse and fish embryos. Given the relatively simple pattern of Otx2expression, it was surprising to find such a complex set of regulatory elements controlling Otx2 expression in the developing rostral head. A likely explanation regarding the complexity of regulation of Otx2expression is that each cis-regulator might mark a distinct domain related to local signaling centers and to the development of specific organs. Moreover,the expression patterns of Fugu cis-regions in transgenic zebrafish,and the presence of conserved sequence elements between the pufferfish and mouse, suggest that the genetic mechanism of vertebrate head specification is highly conserved between fish and mouse.

Fugu has the most compact genome among vertebrates. Comparative genomic analysis of human, mouse and pufferfish genome sequences identifies the coding region of the genome. However, the function of non-coding regions for gene expression remains elusive. The compact nature of the pufferfish genome is primarily due to the smaller size of the non-coding region; on average there is one gene per 10 kb in the pufferfish genome(Aparicio et al., 2002). Therefore, the Fugu genome may provide an ideal system with which to analyse the function of non-coding regions; in particular, to analyse those regions with a pivotal role in gene expression. In the present investigation,in excess of 69 kb of the entire Fotx2 locus (the Otx2 locus occupies ∼700 kb in the mouse genome) was examined. Coincidentally,cis-regulatory activities were not observed within the 70 kb spanning from–50 to +20 kb of the mouse Otx2 locus, with the exception of the 5′ 1.8 kb promoter sequence (C.K.-Y. and I.M., unpublished). It is evident from these findings that cis-regions regulating Otx2expression are scattered over extremely large genomic distances from the coding region.

Coincidentally, genes immediately upstream and downstream of Fotx2include the Brain secretory protein Sec10 gene located at –38 kb, and the pellino 2 gene located at +43 kb(http://fugu.hgmp.mrc.ac.uk/). Comparison with the mouse Otx2 locus reveals that these immediate gene orders are conserved between pufferfish and mouse(http://www.ensembl.org/Mus_musculus/). By contrast, the mouse ortholog Brain secretory protein Sec10 gene is located approximately –350 kb, and pellino 2 is located approximately+500 kb, downstream of the mouse Otx2 locus. These genome sequence data indicate that the Otx2 locus is an extremely gene-bare region when compared with the average of one gene per 10 kb in the pufferfish genome. Together, these suggest that the 69 kb Fugu sequence from –30.5 to +38.5 kb may encompass most of the cis-regulatory regions of the Fugu Otx2 locus.

Consistently, the Fugu cis-acting regions we identified possess sequence elements that are highly conserved with those that are located within the 550 kb region of the mouse Otx2 locus(Fig. 12). Moreover, the conserved elements E6 and E7 were shown to be crucial for F11 cis activity(Fig. 12D). Curiously, most cis-regulators identified in the Fotx2 locus do not direct redundant expression patterns, which implies that cis-regulatory regions for redundant or overlapping expression may be exclusively abolished in the pufferfish so as to reduce its genome to as small a size as possible.

We have described pufferfish Otx2 gene expression during wild Fugu embryogenesis. Developmental patterns of the Fuguteleost are essentially identical with those of other teleosts such as zebrafish and medaka. Accordingly, Fotx2 expression appears to be substantially similar to zebrafish Otx2 expression(Li et al., 1994); briefly, a common aspect of vertebrate Otx2 gene expression in anterior region of the body is shared. Additionally, we have also identified multiple pufferfish cis-acting regions in transient transgenic zebrafish embryos. This finding implies that the cis-elements and the trans-acting factors, which are involved in Otx2 expression, had fully evolved in the common ancestor of teleosts and mammals, and have been conserved during 400 million years of evolution.

The F3, F5 and F8 cis-regulatory regions identified in this study direct transgene expression that nearly coincides with local signaling centers in the rostral brain, as defined by the expression Bmps, Shh and Wnts, respectively(Figs 4, 5, 6, 7, 10). The F3 cis-acting region governed expression in the most medial and caudal aspects of the telencephalon, and in the roof of the diencephalon, coincident with the prospective site of choroid plexus development (Figs 5, 10). The expression domains afforded by the F3 cis-region are strikingly colocalized with Bmps, noggin, Msx1 and transthyretin (Ttr)(Furuta et al., 1997). Indeed, Otx2 is expressed in the dorsocaudal telencephalon and in the dorsal diencphalon, including in the prospective choroid plexus and the `cortical hem' (see below) in mouse embryos (Simeone et al., 1993; Boncinelli et al., 1993; Stoykova et al.,1996). In order to establish whether the F3 cis-region is controlled directly by Bmp signaling, experiments were performed involving neural plate explants with BMP beads as described(Furuta et al., 1997). The neuroectoderm was isolated at the level of the forebrain from F3placZtransgenic embryos at 10.5 dpc. Recombinant BMP2-coated beads were transplanted into these forebrain explants, and the explants were then cultured for 24 hours. However, no ectopic lacZ expression induced by the BMP2-coated beads was detected (C.K.-Y. and I.M., unpublished). These data suggest that the F3 cis-regulator may not be regulated by BMP signaling. Alternatively, the Otx2 expression from this F3 cis-regulator might participate in the expression of Bmp molecules in the prospective choroid plexus. By contrast, the F8 cis-acting region directed transgene expression in the mediocaudal telencephalon where Wnt molecules are co-expressed, defining a zone termed `cortical hem' (Figs 6, 10)(Parr et al., 1993; Grove et al., 1998). Therefore, Wnt signals might control the Otx2 expression mediated by the F8 cis-region in the cortical hem and dorsal telencephalon. Additional transgenic zebrafish studies have indicated that F3 and F8 cis-regions directed considerably conserved expression in the forebrain(Fig. 11). Concomitant with this finding, zebrafish Wnt8b is expressed in the dorsal forebrain(Kelly et al., 1995). These data suggest that local signaling centers play an essential role in forebrain development in zebrafish embryos and are evolutionarily conserved among vertebrates.

The F5 cis-acting region directs expression in the ZLI, and in ventral portions of the diencephalon and the lateral mesencephalon that may be related to longitudinal columns termed `midbrain arcs' (Figs 7, 10)(Agarwala et al., 2001; Sanders et al., 2002). Notably, mouse Otx2 mRNA and Otx2 protein are expressed in the ZLI,and in the ventral diencephalon and mesencephalon in mouse embryos(Simeone et al., 1993; Mallamaci et al., 1996; Stoykova et al., 1996). Importantly, expression in the ZLI, ventral diencephalon and midbrain arcs are proposed to be closely related to Shh signaling, which is essential for the dorsoventral patterning of the ventral neural tube. Therefore, these expression patterns suggest the possibility that Shh signaling controls the Otx2 expression mediated by the F5 cis-acting region.

lacZ activity directed by the F9, F11 and F12 cis-regions appears to be closely related to the developing organs. Notably, it has been suggested that Otx2 is required for the development and specification of these specific organs. The F9 cis-region governs expression in the cephalic mesenchyme (Kimura et al.,1997). Consistently, Otx2 is expressed in the cephalic mesenchyme, and functions in the formation of premandibular and mandibular skulls (Matsuo et al., 1995; Kimura et al., 1997; Hide et al., 2002). The F11 region governs expression in the dorsal pretectum and mesencephalon, including in the posterior commissure, the mesencephalic trigeminal neurons, the oculomotor nerve, the first and second branches of the trigeminal nerve, and the trigeminal ganglions (Figs 8, 10). In actuality, Otx2 mRNA is expressed in the dorsal pretectum and mesencephalon, in the trigeminal ganglions and opthalmic branch(Simeone et al., 1993; Ang et al., 1994)(Fig. 8M). Furthermore, as Otx2 heterozygousmutant mice display anomalies of these neurons and cranial nerves, Otx2 may be required for the proper development of these tissues cell-autonomously (Matsuo et al., 1995). The F12 cis-region drives expression in the developing sense organs, the olfactory epithelium, the pigment layer of the retina, and the developing inner and outer ears. Otx2 expression is consistently found in these developing sense organs(Simeone et al., 1993; Mallamaci et al., 1996; Morsli et al., 1999)(Fig. 8K). Moreover, Otx2 is required for normal development of the inner ear and pigment epithelium, cooperatively with the Otx1 gene(Morsli et al., 1999; Cantos et al., 2000; Martinez-Morales et al.,2001).

We have surveyed the entire pufferfish Otx2 locus; furthermore, we observed the regulation of Otx2 expression by multiple, yet distinct,cis-acting regions. The regulatory mechanism of this gene is much more complex than previously anticipated. It is certain that these regions transactivate Otx2 expression independently. One possible assumption regarding the presence of such long-distance and multiple cis-acting regions entails the ability of each cis-region to transactivate Otx2 expression upon adjacent cis-elements without mutual interference. Current reports have shown that Hox genes are clustered into complexes; moreover, their positions within the cluster correlate with their time of expression and with the position of the AP boundaries of their expression domains(Duboule and Morata, 1994). Similarly, β-globin gene clusters are controlled in order of their expression during ontogeny(Grosveld, 1999). In the present study, Fotx2 cis-regulators are clustered into the 70 kb genomic sequence of the Fotx2 locus; however, their positions relative to the coding region, and the order of the 5′ to the 3′side, do not correlate with their time of expression, or with the expression domains along the anteroposterior or dorsoventral axis. This observation suggests that these cis-regulators may direct multiple promoters located within this genomic locus. Consistent with this notion, a novel gene, annoted as NM172600, which is located over +150 kb from the mouse Otx2 coding region, is a homolog of Fugu Q9NX78(Fig. 12B), and is expressed in the rostral brain, in which Otx2 expression occurs during embryogenesis (C.K.-Y. and I.M., unpublished). This finding supports our assumption that genes located over 500 kb from this locus may be coincidentally controlled by these distant cis-regulators. Curiously, the corresponding conserved sequence elements in mouse are also located in an order similar to that in pufferfish; however, these elements are within an approximately eightfold extended region, up to 550 kb(Fig. 12). Thus, we propose that, during evolution in the common ancestor of teleosts and mammals, the Otx2 promoter might have acquired multiple region-specific cis-regulators in order to control the highly coordinated processes of rostral head development. Consequently, the Otx2 homeobox gene could play multiple and pivotal roles in rostal head development by means of these cis-elements. Investigation of the mechanism of non-vertebrate Otx2gene expression is of keen interest(Oda-Ishii and Saiga,2003).

We are grateful to the Animal Resources and Genetic Engineering Team at the RIKEN Center for Developmental Biology for housing the mice. We thank Drs K. Shimamura and C. Ragsdale for helpful comments on the manuscript; Drs A. McMahon and R. Harland for in situ probes; and H. Nakano for his assistance. C.K.-Y. is supported by a special postdoctoral researchers' program. This work was supported in part by grants-in-aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports Science and Technology,Japan.