gbx2 Homeobox gene is required for the maintenance of the isthmic region in the zebrafish embryonic brain



We isolated cDNA clones for the zebrafish gbx2 gene, which is implicated in the establishment of the midbrain–hindbrain boundary (MHB) in other vertebrates. Spatially localized expression of gbx2 was observed at the MHB from 90% epiboly through to the hatching stage. Comparisons with the expression of otx2, wnt1, and krox20 showed that gbx2 is expressed in the anterior hindbrain. Ectopic expression of gbx2 by mRNA injection caused cyclopia or truncation of the fore- and midbrain and severely affected isthmic and cerebellar structures, while hindbrain formation was not significantly affected. At the molecular level, gbx2 suppressed the expression of otx2 in the fore/midbrain, six3 in the anterior forebrain, and MHB-specific genes such as eng2 and wnt1. In contrast, gbx2 did not down-regulate the expression of the hindbrain marker genes. Therefore, gbx2 specifically suppressed the formation of the entire fore/midbrain. Meanwhile, misexpression of otx2 suppressed the expression of gbx2 in the embryonic brain. Abrogation of gbx2 expression with an antisense morpholino oligonucleotide disrupted the midbrain/anterior hindbrain region, and these loss-of-function effects were rescued by activating the Gbx2 protein immediately after the end of gastrulation. Taken together, these results suggest that the zebrafish gbx2 gene is essential for the maintenance of MHB and/or the formation of the isthmic structure during somitogenesis, rather than for the MHB establishment during gastrulation. We also suggest that other factors, including gbx1, is required for the establishment of the MHB during gastrulation. Developmental Dynamics, 2003. © 2003 Wiley-Liss, Inc.


The three-dimensionally complex brain structure and the cellular diversity it displays during development are critical for the later higher-order function of vertebrate brains. The development of this structure starts with regional specification in the neural plate along the anteroposterior and dorsoventral axes in vertebrate embryos. Brain regionalization of this type appears to be dependent on planar signals that emanate from the signaling center within the neuroepithelium. One of the major local signaling centers identified to date is the junction between the presumptive midbrain and hindbrain, i.e., the midbrain–hindbrain boundary (MHB), which is also called the isthmic organizer (Joyner et al., 2000; Simeone, 2000; Rhinn and Brand, 2001). Transplantation experiments in chick embryos have suggested that the MHB region, which later develops to form the isthmus, organizes the formation of the midbrain and cerebellum (Martinez et al., 1991, 1995; Marin and Puelles, 1994).

The underlying mechanism for the establishment of the MHB and the organizer activities of the MHB during vertebrate brain development have attracted much attention. Around the end of gastrulation, several transcription-factor genes, which include Pax2/5/8 and En1/2, as well as secreted-factor genes, such as Wnt1 and Fgf8, are expressed around the MHB in mice and chicks (Joyner et al., 2000; Simeone, 2000). Recent studies have suggested that Fgf8 is the prime candidate molecule for the organizing activity (Crossley and Martin, 1995; Crossley et al., 1996; Martinez et al., 1999). Many of these MHB genes have been disrupted by gene targeting and have been shown to be involved in the brain formation around the MHB. Thus, null mutants for Pax2/5, Wnt1, and En1 lack the MHB structure and adjacent regions (Thomas and Capecchi, 1990; Wurst et al., 1994; Schwarz et al., 1997), and weak alleles of Fgf8 show similar phenotypes (Meyers et al., 1998). In zebrafish, the orthologous genes for Pax2/5, En, Fgf8, and Wnt1 are also expressed around the MHB in a sequential manner (pax2.1/5/8, eng1/2/3, fgf8, wnt1; Rhinn and Brand, 2001). Several mutants that were severely affected in the isthmic region, such as acerebellar (ace), no isthmus (noi), and spiel-ohne grenzen (spg), were obtained by large-scale mutagenesis screens (Brand et al., 1996; Schier et al., 1996). Of these, ace and noi turned out to be mutations in fgf8 (Reifers et al., 1998) and pax2.1 (Lun and Brand, 1998), respectively. Analyses of these mutations in zebrafish revealed the importance of networking among the MHB genes, which is consistent with previous findings that disruption of the MHB genes in mice always gives rise to defects in the MHB and/or neighboring brain structures. The third mutation, spg, was shown recently to encode zebrafish Pou2, which is apparently required for competence to Fgf8 and maintenance of the gene network around the MHB (Belting et al., 2001; Burgess et al., 2002). Thus, the basic process of brain formation around the MHB is conserved among vertebrates.

The identities of the upstream signals that activate the expression of the MHB genes are unclear at present, although the homeobox genes Otx2 and Gbx2 are considered to be the key regulatory genes. At the early stages of gastrulation in the mouse, chick, and Xenopus, Otx2 is expressed in the presumptive forebrain and midbrain, while Gbx2 is expressed in the anterior hindbrain, thereby forming a shared border at the level of the prospective MHB along the anterior–posterior axis. The patterns of expression of the two genes suggest that the formation of the expression boundary between Otx2 and Gbx2 activates a downstream gene network, thereby leading to MHB establishment (Wassef and Joyner, 1997; Joyner et al., 2000). This hypothesis is supported by the finding that the anterior brain rostral to the hindbrain rhombomere 3 is deleted in Otx2-null mutant mice (Acampora et al., 1995; Matsuo et al., 1995), whereas Gbx2 mutant mice lack the rostral hindbrain and show caudal expansion of the midbrain (Wassarman et al., 1997). The aberrant expression of Gbx2 in the posterior midbrain using the Wnt1 promoter suppresses Otx2 expression, thereby causing anterior expansion of the hindbrain (Millet et al., 1999). Meanwhile, Otx2 misexpression driven by the En1 promoter in the anterior hindbrain suppresses Gbx2 expression and causes the midbrain to be expanded caudally (Broccoli et al., 1999). Importantly, the MHB genes are induced at the interface of Otx2 and Gbx2 expression also in these transgenic mouse embryos. Similar mutual suppression of Otx2 and Gbx2, as well as the induction of MHB genes at the expression boundary of the two genes, have been shown in chick embryos using the electroporation technique (Katahira et al., 2000). Therefore, the early expression of Otx2 and Gbx2 and their mutually suppressive interactions seem to determine the position of the MHB.

In zebrafish, otx2 is expressed in the prospective fore/midbrain in mid-gastrulae, as in other vertebrates (Li et al., 1994; Mori et al., 1994), and gbx1, which is another member of the Gbx gene family, appears to be involved in MHB establishment (Rhinn and Brand, 2001; Reim and Brand, 2002). The expression of gbx2 was also reported recently (Su and Meng, 2002). Unexpectedly, however, the onset of the gbx2 expression was shown to be much later than that of otx2 (late gastrula stage, 90% epiboly), and the gbx2 domain was described as the midbrain, in contrast to the Gbx2 expression in the other vertebrates examined. Therefore, the role of gbx2 in zebrafish brain formation remains enigmatic.

In this study, we isolated cDNA clones of the zebrafish gbx2 gene, and re-examined the temporal and spatial patterns of gbx2 expression in zebrafish embryos. We then overexpressed gbx2 by mRNA injection and performed loss-of-function experiments that used antisense morpholino oligonucleotides to demonstrate that gbx2 is involved in the maintenance of the MHB and/or the formation of the isthmic structure in zebrafish embryos. The data obtained also suggest the presence of an additional factor(s) that is involved in the establishment of the MHB early during gastrulation at least in zebrafish embryos.


Cloning of the Zebrafish gbx2 cDNA

Reverse transcription-polymerase chain reaction (RT-PCR) using degenerate primers for the two conserved amino acid sequences in the Antennapedia-type homeodomain produced a partial cDNA fragment that encoded a protein with sequence similarities to other vertebrate Gbx2 proteins (data not shown). By using this PCR fragment as the probe, we isolated three cDNA clones from the 20 hours postfertilization (hpf) cDNA library (pGX1–3), all of which encoded the same protein. pGX1 represented the longest (2113-bp) cDNA, and encoded an open reading frame (ORF) of 342 amino acids (a.a.; Fig. 1A; deposited in the DDBJ/EMBL/GenBank databases with accession no. AB075028). The length of this cDNA was close to that of the shortest transcript detected in Northern analysis, as described below. The entire a.a. sequence of the deduced protein shared high-level (65–72%) identity with other vertebrate Gbx2 proteins (Fig. 1A,B). Therefore, we conclude that pGX1 encodes the zebrafish Gbx2, and we refer to this gene as gbx2. Recently, another group isolated a zebrafish cDNA clone (gbx-2), which encoded an a.a. sequence that is almost identical to that encoded by the gbx2 reported here, except that the gbx-2 product has a leucine at position 26 (L26) in place of the serine (S26) seen at the corresponding position in Gbx2 (Su and Meng, 2002). We also noticed that the genomic sequence and deduced a.a. sequence of the zebrafish gbx2 had already been deposited in the GenBank/EMBL/DDBJ databases (AF288762). Although this deduced a.a. sequence closely resembles our sequence, the V171 residue seen in our sequence is missing, and the GT-AG rule for normal introns predicts that our sequence is the real one.

Figure 1.

A: Comparison of the Gbx2 proteins from zebrafish and other vertebrates. The Gbx2 proteins from zebrafish, human (AF118452), mouse (U74300), chick (AF022151), and Xenopus (L47990) are aligned for comparison. The hyphens represent gaps that were introduced to maximize similarities. The three conserved domains (CD1–3) and the homeodomain (HD) are marked overhead with thick lines. The overall identities are also shown. B: Comparison of the zebrafish Gbx2 homeodomain with the homeodomains of Gbx2 proteins from other vertebrates, those of Ovx1/Gbx1 (human, S32510; carp, X99910), and those of Hox proteins. Identities to the zebrafish Gbx2 homeodomain are shown on the right. C: Temporal expression pattern of gbx2 during zebrafish embryogenesis. Expression of the gbx2 transcript was quantified by reverse transcription-polymerase chain reaction (PCR). The expression of EF-1α was quantified in parallel as the control. The PCR products were detected by the Southern method for gbx2 and by ethidium bromide staining for EF-1α. [Color figure can be viewed in the online issue, which is available at]

Given that Gbx2 has been identified in species from fish to mammals, it is possible to define regions that have been highly conserved during vertebrate evolution. Alignment of the Gbx2 a.a. sequences from different vertebrates revealed three highly conserved domains (CD1–3), in addition to the homeodomain (HD; Fig. 1A,B). The Gbx2 homeodomains of human, mouse, chick, and zebrafish were found to share 100% identity (Fig. 1B), with only the Xenopus Xgbx-2 showing a single a.a. substitution. The CD1 and CD2 regions corresponded to the 63-a.a. and 13-a.a. N-terminal regions, respectively, in the Gbx2 proteins. Identities of 83–88% were seen for CD1, and 100% identity was obtained for CD2. The CD1 regions of all the examined Gbx2 proteins contained stretches of prolines at their carboxyl ends. Furthermore, a relatively high frequency of proline residues was seen in the N-terminal region (22/81 of the N-terminal a.a. in zebrafish Gbx2). Finally, the 42-a.a. C-terminal region (CD3) also showed high-level similarity (93–98% identities) among the different vertebrates.

The Gbx2 homeodomains of zebrafish and other vertebrates had 96% identity to those of the other Gbx-subgroup proteins Gbx1/Ovx1 (human, mouse, chick, zebrafish, and carp; Fig. 1B; data not shown). The CD3 of Gbx2 also showed a relatively high-level of identity (75–81%) to the C-terminal regions of Gbx1/Ovx1 (data not shown), and CD1 showed limited similarity to the N-terminal regions of Gbx1 from mouse and zebrafish (data not shown). Meanwhile, the CD2-related sequence and the proline stretch seen at the C-terminus of CD1 were not seen in the Gbx1 proteins. The CD3 sequence is also partially conserved in Unplugged, the Gbx homologue in Drosophila (Chiang et al., 1995). The highest level of similarity to the Gbx2 homeodomain (apart from the Gbx group proteins) was seen for the Hox proteins, although the latter levels of identity did not exceed 64% (Fig. 1B; data not shown).

Temporal and Spatial Expression of gbx2

The RT-PCR analysis (Fig. 1C) showed that gbx2 transcripts existed maternally at a relatively low level, whereas zygotic expression started at the shield stage and persisted at relatively constant levels from the end of gastrulation to 48 hpf. In Northern blot analysis, three transcripts of 5.4 kb, 4.3 kb, and 2.2 kb were detected throughout development (data not shown) and may have arisen as the result of alternative splicing, initiation, or termination.

Recently, it was reported that gbx-2 was expressed initially around the MHB during late gastrulation (90% epiboly), and subsequently in the otic primordia, dorsoposterior telencephalon, rostral branchial arches, pronephric duct, and median fin fold (Sun and Meng, 2002). However, our observations differed from previous results in certain aspects. We also found that gbx2 was initially detectable at 90% epiboly as a pair of bilateral transverse stripes in the prospective neural region by the whole mount in situ hybridization (data not shown). Sun and Meng concluded that this region corresponded to the posterior midbrain, because the expression domains of gbx-2 and pax2.1 overlapped. They assumed that pax2.1 was a marker for the posterior midbrain, although this gene recently is considered as an MHB marker (Rhinn and Brand, 2001). Furthermore, gbx2 cognate genes in other vertebrates are known to be expressed in the anterior hindbrain. Therefore, we compared the expression of gbx2 at the bud stage with that of otx2, which is expressed in the prospective forebrain and midbrain at this stage (Li et al., 1994; Mori et al., 1994). Two-color in situ hybridization showed that the anterior boundary of gbx2 corresponded to the posterior boundary of otx2 expression (Fig. 2A). Double in situ hybridization with wnt1, which is expressed in the posterior midbrain (Kelly and Moon, 1995), also showed that gbx2 was expressed in the anterior hindbrain (Fig. 2B–D). Comparison with the expression of krox20 at the bud stage, which is expressed in rhombomeres 3 and 5 (r3 and r5; Oxtoby and Jowett, 1993), showed that gbx2 was expressed mainly in the r1 region of the hindbrain.

Figure 2.

Expression of gbx2 in zebrafish embryos, as revealed by whole-mount in situ hybridization. A–D,F,I,L: Dorsal views, with anterior to the top (A–D) or to the left (F,I,L). G,H,K: Lateral views, with anterior to the left and dorsal to the top. gbx2 expression in the anterior hindbrain, otic placode/vesicle, neural crest cells/pharyngeal arches, and neurons in the hindbrain are marked with black arrowheads, thin arrows, thick arrows, and white arrows, respectively. White arrowheads show the endomesoderm (A,E) or dorsal thalamus in the diencephalon (H–L). Red arrowheads show wnt1 expression in the posterior midbrain. A: Two-color in situ hybridization was conducted at the bud stage for otx2 (red) and gbx2 (purple). B–D: Double in situ hybridization was conducted for wnt1 plus gbx2 (B), gbx2 plus krox20 (C), and wnt1 plus krox20 (D). The anterior boundary for gbx2 expression and/or posterior boundary for wnt1 are marked with dashed lines. Rhombomeres 3 and 5 (r3, r5) are indicated. E: Cross-section of a bud-stage embryo at the level marked with the dashed line in A. F: A 10-somite stage embryo showing gbx2 expression in the migrating neural crest cells, in addition to the anterior hindbrain and otic vesicle. G: A 26-somite stage embryo. H–L:gbx2 expression in the head at 36 hours postfertilization (hpf; H,I) and at 48 hpf (K,L). J: Cross-section of the head of a 36-hpf embryo at the level shown in I with a line, showing gbx2 expression in the dorsal thalamus. Scale bars = 100 μm in A–D,F–I, K,L. Scale bars = 50μm in E,J.

The expression of gbx2 in the anterior hindbrain persisted at least until 60 hpf, in contrast to the previous finding that gbx-2 expression in the corresponding region was down-regulated by 48 hpf, although the expression was restricted progressively to a narrower band (Fig. 2F–I,K,L; data not shown for 60 hpf), with a dorsal-to-ventral intensity gradient (Fig. 2G,H). We confirmed the previous report that gbx2 was expressed in the otic primordia (Fig. 2A–C,E–I,K,L), rostral branchial arches (Fig. 2G–I), pronephros, and median fin fold (data not shown). In addition, we made the novel observation that neural crest cells originating from r4 expressed gbx2 in mid-somitogenesis (six-somite stage; Fig. 2F). We also noticed gbx2 expression in the endomesoderm underlying the anterior hindbrain (Fig. 2A,E). In contrast to the previous report, we did not observe gbx2 expression in the telencephalon during somitogenesis (Fig. 2G). Meanwhile, we observed expression in the dorsal thalamus of pharyngula embryos (Fig. 2H–L) and in neuron-like cells of the hindbrain from late somitogenesis until the hatching stage (Fig. 2G–I,K,L; additional data not shown), which has not been described previously for the zebrafish. Although there are several differences between our data and the previous report, which cannot be explained at present, the overall patterns of expression appear to be consistent.

Effects of gbx2 Overexpression on Brain Formation in Zebrafish Embryos

We examined the effects of ectopic gbx2 expression on zebrafish embryogenesis. After the injection of gbx2 mRNA into single blastomeres at the one- to four-cell stage, a broad range of brain defects were noted at 24–28 hpf (Fig. 3; Table 1), which appeared to depend on the amount of mRNA injected. Although mRNA from the full-length cDNA gave moderate effects, even at the highest concentration used (gbx2-FL, 500 pg/embryo), deletion of both the 5′- and 3′-untranslated regions (UTRs; gbx2ΔUTR) significantly enhanced this activity (Table 1), which suggests that either the 5′-UTR or 3′-UTR effects the translational efficiency or stability of gbx2 mRNA. In the experiments described below, we used gbx2ΔUTR mRNA as the gbx2 mRNA.

Figure 3.

Effects of gbx2 overexpression on zebrafish development. A–C,F,H–M: Lateral views, with anterior to the left and dorsal to the top. D,E,G: Dorsal views, with anterior to the left. A–F: Zebrafish embryos observed with differential interference optics at 24 hours postfertilization (hpf). Embryos that were injected with egfp mRNA (250 pg/embryo) show normal morphology of the head (A,D). Embryos that were injected with a low level of gbx2 mRNA (50 pg/embryo) show defects in the isthmic region, although they retain the forebrain and optic vesicles (B,E). Embryos that were injected with a high level of gbx2 mRNA (250 pg/embryo) show truncation of the forebrain and midbrain (C) or fusion of the eyes (F, cyclopia). G: An embryo that was injected with gbx2 mRNA (250 pg) and lacZ mRNA (50 pg), and stained for β-galactosidase at 26 hpf. The eye (black arrowhead) and the tail are severely affected on the left side that incorporated the mRNA. H,I: Embryos (3.5 days postfertilization) that were injected with egfp mRNA (250 pg, H) and gbx2 mRNA (250 pg, I). J–M: Embryos that were injected with 100 pg mRNA for egfp (J,L) or gbx2 (K,M) were stained with acridine orange and viewed by fluorescence microscopy at the five-somite stage (J,K) or at 26 hpf (L,M). White arrowheads and arrows show midbrain–hindbrain boundary/isthmic regions and otic vesicles, respectively. Scale bars = 100 μm in A (applies to A–C), D (applies to D,E), F, G, H applies to H,I), J (applies to J,K), L (applies to L,M).

Table 1. Effects of gbx2 Overexpression on Zebrafish Embryogenesis
mRNA (pg/embryo)2505025050050100250250
  • a

    The entire gbx2 cDNA from pGX1 was cloned into pCS2+ and used as a template for mRNA synthesis.

  • b

    The coding region of gbx2 cDNA was amplified by polymerase chain reaction, cloned into pCS2+, and used for mRNA synthesis.

  • c

    The percentages are calculated for the respective phenotypes of 24 hours postfertilization embryos that were injected with mRNA; the total may be more than 100%.

  • d

    Embryos that lacked the forebrain and midbrain or those showing amorphous fore/midbrain.

  • e

    Embryos that retained the fore/midbrain, despite the lack of well-formed isthmic structures.

  • f

    Embryos that developed short or twisted tails.

Total embryos (n)27849807961635371
Normalc (%)99847161251308
Abnormalc (%)        
 Anterior truncationd0125108217677
 MHB defectse00010302461
 Tail defectsf1491470130
 Epiboly arrest001150060

At high doses of injected gbx2 mRNA, the entire fore/mid-brain was truncated (Table 1; Fig. 3C), or severely affected, thereby producing an amorphous fore/midbrain (data not shown), while the hindbrain remained intact. No distinct isthmic constriction or cerebellar structures were formed in embryos with severely affected anterior brains, despite the presence of the hindbrain, while otic vesicle formation was apparently normal in most cases (Fig. 3B,C,I). Eye formation was affected in the mildly affected embryos; some showed cyclopic eyes as a result of a fusion of both eyes (Fig. 3F), whereas others retained a single eye on one side (Fig. 3G). Embryos with severe head defects developed with apparently normal posterior structures until at least 3.5 days postfertilization (Fig. 3H,I). Aside from the brain defects, a few embryos had rather short or kinked tails (Table 1; Fig. 3G). Of interest, isthmus formation was specifically affected (30% of the embryos) at a relatively low dose of gbx2 mRNA (50 pg/embryo), leaving the more-anterior brain relatively intact (Table 1; Fig. 3B,E).

To locate gbx2 misexpression, the mRNA for the β-galactosidase gene (lacZ) was coinjected with the gbx2 mRNA. The expression of lacZ was always observed where gbx2 effects were seen; thus, the lacZ expression was detected in the severely affected anterior heads (data not shown), or in the posterior parts of injected embryos that showed abnormal posterior structures (Fig. 3G). Notably, only the eye on the side injected with the mRNA was affected (Fig. 3G). Therefore, the effects of gbx2 were restricted to the tissues in which gbx2 was overexpressed.

We then asked if disruption of the anterior brain was due to degeneration. Embryos were injected at one- to two-cell stage with 100 pg/embryo of gbx2 mRNA and allowed to develop until the bud stage, or 26 hpf, when the embryos were stained for 1 hr with acridine orange. No difference was observed at the bud stage with respect to the density of apoptotic cells that stained with acridine orange, whereas at 26 hpf, significantly more degenerating cells were detected in the anterior brain (Fig. 3J–M).

Molecular Characterization of Embryos Injected With gbx2 mRNA

To further analyze the effects of gbx2 at the molecular level, we examined the expression, at 24–28 hpf, of marker genes that are involved in brain formation. The expression of otx2 in the diencephalon and midbrain (Mori et al., 1994; Li et al., 1994) of the injected embryos was decreased (Fig. 4A,B), and six3 expression in the anterior forebrain of normal embryos (Kobayashi et al., 1998) was also drastically repressed (Fig. 4C,D). The expression of wnt1 in the dorsal midbrain and in the posterior ring-like region of the midbrain (data not shown; Kelly and Moon, 1995) and those of eng2 and pax2.1 at the MHB (Fig. 4E–H; Krauss et al., 1991; Ekker et al., 1992) were abrogated.

Figure 4.

Expression of brain regional marker genes at the pharyngula stage in embryos that were injected with gbx2 mRNA. The expression levels of otx2, six3, eng2, pax2.1, fgf8, krox20, and shh in 26–28 hours postfertilization embryos were examined by whole-mount in situ hybridization. Lateral views, with anterior to the left and dorsal to the top. A,C,E,G,I,K,M: Control embryos that were injected with egfp mRNA (250 pg/embryo). B,D,F,H,J,L,N: Embryos that were injected with gbx2 mRNA (250 pg/embryo). The white arrowheads mark the midbrain–hindbrain boundary/isthmic region, and the black arrowheads show the expression of marker genes in the anterior forebrain (C,I) or optic stalk (G). B: The remnant otx2 expression in the anterior end of the affected brain is marked with an arrow. C,I,K,M: The optic vesicles were removed for the visualization of gene expression in the anterior brains. FP, shh expression in the floor plate; r3 and r5, rhombomeres 3 and 5. Scale bar = 100 μm in N (applies to A–N).

Meanwhile, fgf8 expression in the anterior portion of the hindbrain (Fürthauer et al., 1997) was essentially intact, although it was down-regulated in the dorsal telencephalon and anterior diencephalon (Reifers et al., 1998; Fig. 4I,J). Likewise, the expression of krox20 in r3 and r5 was not affected (data not shown), or was expanded, together with the expression gap between r3 and r5, which corresponds to r4 (Fig. 4K,L). The expression of sonic hedgehog (shh) was suppressed in the ventral diencephalon but was unaffected in the floor plate of the hindbrain and spinal cord (Fig. 4M,N; Krauss et al., 1993). Finally, wnt1 expression in the dorsal hindbrain and spinal cord was hardly affected (data not shown). Taken together, gbx2 overexpression specifically suppresses the formation of the forebrain, midbrain, and MHB region, but allows hindbrain formation.

To study the effect of gbx2 overexpression immediately after the regionalization of the neural plate during gastrulation, we examined the expression of brain region-specific markers in injected embryos at the bud or early somitogenesis stages (Fig. 5). The expression of otx2 in the prospective forebrain and midbrain was again severely down-regulated (Fig. 5A,B), and six3 expression was also suppressed in the prospective anterior forebrain (Fig. 5C,D). The expression of wnt1 in the future posterior midbrain, and eng2 expression at the prospective MHB were also abrogated (Fig. 5E–H). Meanwhile, the expression of pax2.1 at the MHB shifted anteriorly (Fig. 5I,J), and fgf8 expression in the anterior hindbrain expanded anteriorly (Fig. 5K,L). At this stage, the krox20 stripe in r3 was essentially intact (Fig. 5M,N).

Figure 5.

Expression of brain regional marker genes around the end of gastrulation in zebrafish embryos that were injected with gbx2 mRNA. The expression levels of otx2 (A,B), six3 (C,D), wnt1 (E,F), eng2 (G,H), pax2.1 (I,J), fgf8 (K,L), krox20 (M,N), and gsc (O,P) were examined by whole-mount in situ hybridization. Dorsal views, with anterior to the top. A,C,E,G,I,K,M,O: The control embryos were injected with egfp mRNA (250 pg). B,D,F,H,J,L,N,P: Embryos that were injected with gbx2 mRNA (250 pg). L: fgf8 expression at the midbrain–hindbrain boundary was broadened anteriorly (arrowhead), and faint expression appeared in the prospective fore/midbrain region (asterisk). M,N: At the early bud stage, krox20 is expressed only in rhombomere 3 (r3). Scale bar = 100 μm in P (applies to A–P).

Thus, specification of the forebrain and midbrain, as well as the establishment of the MHB region, was specifically suppressed by gbx2 misexpression already around the end of gastrulation, while the prospective anterior hindbrain, which was characterized by fgf8 expression, expanded anteriorly. The more-posterior hindbrain seemed to be relatively unaffected at this stage. Specification of the hindbrain was demonstrated by the expression of krox20 at this early stage, which indicates that the apparently normal formation of the hindbrain at 24–28 hpf is not the result of recovery after gastrulation. Because it is well known that the formation of the anterior brain is dependent on the underlying prechordal plate (PCP), we examined the expression of goosecoid (gsc) during gastrulation in embryos that were injected with gbx2 mRNA. At both 60% epiboly (data not shown) and 75% epiboly (Fig. 5P), gsc expression in the PCP was not affected by gbx2. The expression of the no tail (ntl) gene in the notochord primordium (Schulte-Merker et al., 1994) was not affected significantly, either, during epiboly (data not shown). Thus, gbx2 does not abrogate the formation of the axial mesoderm.

Regulation of gbx2 Expression in Embryos

Because Otx2 is known to down-regulate Gbx2 expression in mice and chick embryos, we asked if gbx2 expression was also suppressed by otx2 in zebrafish embryos. The expression of gbx2 was significantly down-regulated in embryos that were injected with 250 pg/embryo of otx2 mRNA, both at the bud stage (Fig. 6A,B) and at 25 hpf (data not shown). Therefore, as is the case in other vertebrates, otx2 negatively regulates gbx2 in zebrafish embryos.

Figure 6.

Regulation of gbx2 expression in zebrafish embryos. The expression of gbx2 at the bud stage was evaluated by whole-mount in situ hybridization. Dorsal (A–C) or anterodorsal views (D,E), with anterior to the top. A,B: Embryos that were injected with mRNA (250 pg/embryo) for egfp (A) or otx2 (B). C–E: Control embryos (C) or embryos that were treated for 1 hr with retinoic acid (RA) at 10−7 M (D) or 10−6 M (E) during the shield stage. The expression of gbx2 in rhombomere 1 and in the otic placode are shown with thick arrows and arrowheads, respectively. The stage of embryos in C–E is a little earlier than that of the embryos in A,B. Scale bars = 100 μm in A (applies to A,B), C (applies to C–E).

It is known that patterning of the hindbrain along the anteroposterior axis is regulated by retinoic acid (RA; Begemann and Meyer, 2001). Consistent with this notion, RA treatment caused anterior expansion of the hoxb1b expression in zebrafish embryos (Kudoh et al., 2002) and Xgbx-2 expression in Xenopus embryos (von Bubnoff, 1996), while it down-regulated the expression of otx2 in embryos of the zebrafish and chick (Bally-Cuif et al., 1995; Kudoh et al., 2002). We examined the expression of gbx2 at the bud stage in embryos that had been treated with 10−7 M or 10−6 M RA at the shield stage (Fig. 6C–E). Contrary to the previous findings in Xenopus, gbx2 expression in r1 was down-regulated in a dose-dependent manner, while gbx2 expression in the otic placode showed anterior expansion, which indicates that RA specifically suppresses gbx2 expression in the anterior hindbrain of zebrafish embryos.

Knockdown of gbx2 Function by Morpholino Oligonucleotides

We extended the functional analysis of gbx2 in zebrafish embryogenesis by examining the effects of an antisense morpholino oligonucleotide (MO) directed against gbx2 (MO-gbx2). When MO-gbx2 was coinjected into zebrafish embryos (1.5 pmol/embryo) with mRNA for the gbx2-GFP hybrid gene (50 pg/embryo), the fluorescence due to the hybrid protein was efficiently abrogated at 30% epiboly (Fig. 7A–D) and at 24 hpf (data not shown), which confirms the efficacy of MO-gbx2. Rescue of the loss-of-function phenotypes by forced expression of the wild-type genes, i.e., by coinjecting gbx2 mRNA, was not possible in this instance, because both gbx2 overexpression (see above) and gbx2 knockdown (see below) disrupted the midbrain and/or isthmic region. Therefore, we resorted to the hormone-inducible gbx2, which encoded the ligand-binding domain of the glucocorticoid receptor that was fused to Gbx2 (gbx2-GR). The product (Gbx2-GR) is activated by glucocorticoid and translocated into the nucleus, where it acts as a transcriptional regulator (Tada et al., 1997). In this experiment, when injected with gbx2-GR mRNA and allowed to develop in the absence of dexamethasone (Dex), normal development was observed until the hatching stage. Meanwhile, when embryos loaded with gbx2-GR mRNA were treated with 2 μM Dex from the beginning of gastrulation (6 hpf), they showed phenotypes that were similar to those of embryos that were injected with gbx2 mRNA (unpublished data). In contrast, when embryos that harbored gbx2-GR mRNA and MO-con (the control morpholino oligonucleotide) were treated with Dex at 11 hpf and allowed to develop until 26 hpf, they showed morphologies that were essentially normal (Fig. 7E,F), which shows that sensitivity to the actions of gbx2 is lost after gastrulation, as has been shown for Xenopus embryos (Tour et al., 2002b).

Figure 7.

Morphologic phenotypes induced by the knockdown of gbx2 in zebrafish embryos. A–D: The 30% epiboly embryos that were injected with gbx2-GFP mRNA together with the control morpholino oligonucleotide (MO-con; A,B) or with the antisense MO directed against gbx2 (MO-gbx2; C,D). A,C: Brightfield views. B,D: Epifluorescence views showing the expression of green fluorescent protein (GFP). E–L: Effects of gbx2 knockdown by MO-gbx2 and rescue by activation of Gbx2-GR with dexamethasone (Dex) treatment. E,G,K: Lateral views, with anterior to the left and dorsal to the top. F,H–J,L: Dorsal views, with anterior to the left. E,F: The 28 hours postfertilization (hpf) embryos that were injected with Gbx2-GR mRNA and MO-con at the one-cell stage and treated with Dex from 11 hpf to 28 hpf, showing normal morphology. G–J: The 28-hpf embryos that were injected with Gbx2-GR mRNA and MO-gbx2 at the one-cell stage and not treated with Dex show incomplete isthmic constriction (G,H; asterisks) or defects in the entire midbrain/anterior hindbrain structure (I). J: Hindbrain region with a left otic vesicle (arrowhead); the right otic vesicle is absent. K,L: The 28-hpf embryos that were injected with Gbx2-GR mRNA and MO-gbx2 at the one-cell stage and then treated with Dex from 11 hpf to 28 hpf show essentially normal morphology. Scale bars = 300 μm in A–D or 100 μm in E–L.

Importantly, when the embryos that were coinjected with MO-gbx2 and gbx2-GR mRNA were allowed to develop until 26 hpf in the absence of Dex, they showed severe defects in the midbrain–isthmic region; some (28% of the embryos) showed incomplete isthmic constriction (Table 2; Fig. 7G,H), whereas others (34%) showed defective midbrains and anterior hindbrains (Table 2; Fig. 7I). Some of the embryos lacked otic vesicles on both sides (data not shown) or retained only a single otic vesicle (Fig. 7J; Table 2), while tail elongation was often incomplete (data not shown). Essentially, the same phenotypes were obtained when the MO-gbx2 alone was injected into embryos (data not shown). In contrast, when coinjected embryos were treated with Dex at 11 hpf and allowed to develop until 26 hpf, the midbrain–hindbrain region recovered morphologically (Fig. 7K,L; Table 2). In addition, defects in the otic vesicles and tails were also decreased by the activation of Gbx2-GR (Table 2). Thus, gbx2 is required for the formation of the isthmic structure and otic vesicles, and the loss of gbx2 function may be rescued by activation of Gbx2 after gastrulation.

Table 2. Effects of gbx2 Knockdown on Morphogenesis in Zebrafish Embryos and Rescue by Gbx2 Activation at the Early Somitogenesis Stagea
Treatmentgbx2-GR + MO-congbx2-GR + MO-gbx2gbx2-GR + MO-gbx2
Dex 11 hpfbDex 11 hpfb
  • a

    Embryos were injected at the one-cell stage with gbx2-GR mRNA (200 pg/embryo) alone or together with morpholino oligonucleotides (1.5 pmol/embryo) and allowed to develop until 24 hours postfertilization at which stage the morphologies were scored.

  • b

    Dexamethasone (Dex; 2 μM) was added to the culture at 11 hours postfertilization as applicable. This treatment alone did not give rise to apparent morphological defects (data not shown).

  • c

    The numbers of embryos are shown. The percentages of abnormal embryos are shown in parentheses.

  • d

    Abnormal midbrain–hindbrain regions with incompletely expanded ventricles.

  • e

    Embryos with weakly constricted isthmi.

  • f

    The otic vesicles are reduced in size or missing.

  • g

    Incomplete elongation of the tail.

  • h

    Severe and nonspecific defects, such as amorphous embryos.

Total embryos (n)405858
Alivec39 (98)51 (88)52 (90)
Normalc36 (90)4 (7)21 (36)
 Midbrain–hindbraind0 (0)20 (34)1 (2)
 Isthmuse1 (3)16 (28)13 (22)
 Otic vesiclef0 (0)24 (41)6 (10)
 Tailg1 (3)44 (76)10 (17)
 Nonspecifich0 (0)6 (10)2 (3)

The expression of regional markers (otx2, six3, wnt1, eng2, pax2.1, fgf8, and krox20) in the brain was examined in embryos that were injected with MO-gbx2. At the bud stage, essentially normal expression of all these genes was observed in embryos that were loaded with MO-gbx2 (Fig. 8A–C,A′–C′; data not shown), which shows that the MHB region was formed in the absence of the gbx2 function. Consistent with the apparent lack of sensitivity to gbx2 overexpression described above, ntl expression was also intact (Fig. 8A,A′). At the three-somite stage, the expression levels of otx2, six3, pax2.1, fgf8, and krox20 were not affected by MO-gbx2 (Fig. 8D,F,D′,F′; data not shown), while down-regulation of wnt1 (20%, n = 25; data not shown) and eng2 (40%, n = 28; Fig. 8E,E′) was observed in some of the embryos. At 24 hpf, the expression of otx2, six3, pax2.1, and krox20 was essentially intact (Fig. 8H,I,K,H′,I′,K′; data not shown). Meanwhile, significant numbers of embryos showed down-regulation around MHB for eng2 (48%, n = 59; Fig. 8J,J′), wnt1 (28%, n = 25; Fig. 8I,I′), and fgf8 (30%, n = 90; Fig. 8L,L′). Therefore, it seems that gbx2 function is not required for the establishment of MHB during gastrulation but is involved in the maintenance of MHB and/or the formation of the isthmic structure during somitogenesis.

Figure 8.

Effects of gbx2 knockdown on regionalization of the embryonic brain. The expression levels of the following marker genes were examined by whole-mount in situ hybridization: six3, eng2 (en) and ntl (A,A′), otx2, and krox20 (B,D,H,B′,D′,H′), fgf8 (C,L,C′,L′), eng2 (E,G,J,E′,G′G′′,J′), pax2.1 (F,K,F′,K′), and wnt1 and krox20 (I,I′). A–F,H–L,A′–F′,H′–L′: Embryos that were injected with the control morpholino oligonucleotide (MO-con; Con; A–F,H–L) or the antisense MO directed against gbx2 (MO-gbx2; A′–F′,H′–L′) and allowed to develop until the bud stage (A–C,A′–C′), three-somite stage (3-s; D–F,D′–F′), or 26 hours postfertilization (hpf; H–L,H′–L′). G–G′′: Normal eng2 expression is observed in embryos that were injected with gbx2-GR mRNA plus MO-con and treated with dexamethasone (Dex) from 11 hpf and 13 hpf (eight-somite; G), while eng2 expression is significantly down-regulated at 13 hpf in embryos that were injected with gbx2-GR mRNA plus MO-gbx2 and raised in the absence of Dex (G′). In contrast, normal eng2 expression is seen in embryos that were injected with gbx2-GR mRNA plus MO-gbx2 and treated with Dex from 11 hpf to 13 hpf (G′′). A–D,F–G,A′–D′,F′–G′,G′′: Dorsal views, with anterior to the top. E,H–L,E′, H′–L′: Lateral views, with dorsal to the right and anterior to the top (E,E′), or with dorsal to the top and anterior to the left (H–L,H′–L′). The white arrowheads indicate the midbrain–hindbrain boundary in control embryos. Scale bar = 100 μm in A (applies to A–G′′), H (applies to H–L′).

We sought to confirm molecularly the rescue of the MO-gbx2 effects by Gbx2 activation at 11 hpf. The expression of eng2 was essentially normal in embryos that were injected with gbx2-GR mRNA plus MO-con and treated with Dex at 11–13 hpf (Fig. 8G), whereas the embryos that were injected with gbx2-GR mRNA and MO-gbx2 showed down-regulation of eng2 at 13 hpf (8-somite stage), as described above (Fig. 8G′). However, embryos that were injected with gbx2-GR mRNA and MO-gbx2 regained normal eng2 expression at 13 hpf after treatment with Dex for the last 2 hr (11–13 hpf), which is consistent with the morphologic observation, confirming the specificity of MO-gbx2.

Involvement of Another gbx-Family Gene of the Zebrafish, gbx1, in Brain Formation

Recently, another Gbx-family gene, gbx1, was suggested to be important for the MHB development in zebrafish, although the expression of gbx1 during early to mid-phase of epiboly, during which gbx2 is not detected by in situ hybridization, has not been reported in detail yet (Rhinn and Brand, 2001; Reim and Brand, 2002). Therefore, to further examine the role of gbx1 in MHB formation, we compared the expression of gbx1 during epiboly with that of otx2. As was mentioned previously, gbx1 expression was observed early during epiboly in the prospective posterior neural plate, and it was apparently complementary with that of otx2. However, narrow gaps of expression were always seen between the domains of otx2 and gbx1 (Fig. 9A–C). Later, the expression was excluded from r1, while seen in more posterior rhombomeres (Fig. 9D, see also Zebrafish Information Network).

Figure 9.

Expression of gbx1 in normal embryos and the anomalies of brain formation observed in gbx1-misexpressing embryos. A–C: Expression of otx2 (red) and gbx1 (purple) was examined by two-color in situ hybridization at 75% epiboly (A), 90% epiboly (B), and the bud stage (C). Narrow gaps of expression (white arrows) were seen between the domains of the two genes. D:gbx1 expression in the hindbrain at 25 hours postfertilization (hpf). E–K:egfp (E,H,J) or gbx1 (F,G,I,K) were overexpressed in embryos, which were examined morphologically at 25 hpf (E–G) or molecularly at the bud stage (H–K). For overexpression, mRNA of 250 pg/embryo (E,G–K) or 50 pg/embryo (F) was microinjected into eggs. Expression of otx2 and krox20 (H,I) or that of pax2.1 (J,K) was examined by in situ hybridization. A,B,D,J,K: Dorsal views, with anterior to the top (A,B,J,K) or anterior to the left (D, flat mount). C,H,I: Anterodorsal views with anterior to the top. E–G: Lateral views with anterior to the left, and dorsal to the top. White arrowheads mark the isthmic constrictions. Scale bars = 100 μm.

Effects of gbx1 overexpression on brain formation was also examined by mRNA injection. Morphologically, gbx1 gave rise to anomalies of brain formation such as anterior truncation of the brain (Table 1; Fig. 9G) at a high dose of mRNA (250 pg/embryo) or lack of the isthmic constriction at a lower dose (Fig. 9F; 50 pg/embryo). Expression of otx2 in the anterior neural plate was significantly down-regulated (Fig. 9I), while krox20 expression in the hindbrain was little affected (Fig. 9I) or anteriorly expanded (data not shown). Meanwhile, pax2.1 expression at the MHB was shifted anteriorly (Fig. 9K). Therefore, in terms of the effects of overexpression on brain formation, gbx1 and gbx2 show similar activities.


Comparisons of Gbx2 Proteins From Different Vertebrates

In this study, we cloned the cDNA for the zebrafish Gbx2. Structural comparisons revealed four regions (CD1–3 and the homeodomain) that are conserved among the vertebrate Gbx2 proteins. Because the homeodomain, which is established as the DNA-binding domain, is almost 100% conserved, it seems likely that the function of Gbx2 and its target cis elements are highly conserved among vertebrates. CD1 and CD3 are not only highly conserved among different vertebrates but are also conserved in Gbx1, which suggests that they contribute to functions shared by Gbx1 and Gbx2. Because no domains homologous to CD1 or CD3 were found in proteins other than the Gbx family proteins by BLAST searching, it seems that CD1 and CD3 represent novel functional domains. Although proline-rich sequences, such as those seen in the N-terminal region of Gbx2, are often assumed to be transcriptional activation domains (Mermod et al., 1989), Gbx2 operates both as a transcriptional activator (Kowenz-Leutz et al., 1997) and suppressor (Tour et al., 2002b). This issue is currently under investigation in zebrafish embryos.

gbx2 Gene Suppresses the Formation of the Anterior Brain and Isthmic Region in Zebrafish Embryos

The MHB is a well-defined signaling center that organizes the formation and polarization of the midbrain and cerebellum during somitogenesis (Joyner, 1996; Wassef and Joyner, 1997). Gene-targeting experiments in mice, as well as mutant analyses in zebrafish, have shown that MHB-specific genes, such as Pax2, Wnt1, Fgf8, and En1/2, are involved in the development of the MHB/isthmic region. Further mutant analyses in zebrafish revealed the presence of two phases (Rhinn and Brand, 2001): (1) the establishment phase during gastrulation, in which several regulatory genes, including wnt1, pax2.1, and fgf8, are activated independently of each other; and (2) the maintenance phase, in which these genes become mutually dependent to stabilize the positive feedback loop, thereby promoting formation of the isthmus by activation of downstream genes, such as eng1/2/3, pax5/8, and fgf17 (Lun and Brand, 1998; Reifers et al., 2000).

In contrast, the mechanism of MHB positioning in the neuroectoderm at earlier stages is not well defined. However, it is known for several vertebrate species that Otx2 and Gbx2 are expressed early during gastrulation in the anterior neural plate (Bouillet et al., 1995; Wassarman et al., 1997) and prospective anterior hindbrain (Niss and Leutz, 1998; Shamim and Mason, 1998), respectively, and that the expression boundary between the two genes later forms the MHB. Gene targeting (Wassarman et al., 1997; Acampora et al., 1998) as well as misexpression experiments (Broccoli et al., 1999; Millet et al., 1999; Katahira et al., 2000) have suggested that the mutually suppressive interaction of Otx2 and Gbx2 at the boundary contributes to the establishment of the MHB (Joyner et al., 2000; Simeone, 2000). Furthermore, it has been shown that an isthmic organizer-like tissue is generated when the midbrain and r1 tissues are juxtaposed, which underlines the importance of the interface between the Otx2 and Gbx2 expression domains (Irving and Mason, 1999).

In the present study, we show that zebrafish gbx2 is also expressed in the anterior hindbrain at the end of gastrulation. This result prompted us to address the question of whether gbx2 function in brain formation is similar to that of Gbx2 in other vertebrates. Morphologically, the overexpression of gbx2 in zebrafish embryos by mRNA injection gave rise to significant suppressive effects on the formation of the forebrain and midbrain. Absence or hypotrophy of the anterior brain was confirmed by down-regulation of the anterior brain marker genes in pharyngula stage embryos. Furthermore, the expression of these markers was abrogated as early as the end of epiboly, which shows that gbx2 suppresses the specification of the anterior brain formation during epiboly by down-regulating the genes that specify the anterior brain. Acridine orange staining showed that the absence of the anterior brain at the pharyngula stage was at least partially due to apoptosis, while enhancement of apoptosis was not observed at the early somitogenesis stages. This finding suggests that the down-regulation of the anterior marker genes observed during the early stages is due to a change in the fate of the neural plate. In keeping with this assumption, anterior shift and expansion of expression were observed for pax2.1 and fgf8, respectively, around the end of gastrulation.

Coinjected lacZ mRNA was always in tissues that showed the effects of gbx2, which confirms that gbx2 function is essentially autonomous, i.e., gbx2 appears to suppress anterior brain formation within the neuroepithelium. In fact, the PCP and notochord required for brain induction were essentially unaffected by gbx2 overexpression, as revealed by the expression of gsc and ntl. The optic vesicles were relatively sensitive to gbx2, and cyclopia was induced frequently, which may be attributable to the suppression of shh by gbx2 in the ventral forebrain.

In contrast, the morphologic and molecular analyses revealed that the hindbrain was formed almost normally, or even expanded to some extent. Indeed, gbx2 overexpression broadened fgf8 expression at the bud stage and caused expansion of the r3–r5 region, which was characterized by krox20 expression, at later stages. Consistent with these results, the otic vesicles, whose formation depends on signals from the hindbrain (Mendonsa and Riley, 1999), were also formed normally. Thus, gbx2 specifically suppresses the formation of the entire forebrain and midbrain, while allowing or promoting hindbrain formation. It should be mentioned that both the MHB/isthmic region and cerebellar structures were also severely affected, even in the presence of the hindbrain. Actually, the expression of eng2 and wnt1 around the MHB was suppressed. In addition, low-level overexpression of gbx2 tended to abrogate isthmic constriction, even when the anterior brain was essentially intact. These findings are consistent with the view that the isthmus and cerebellum are formed by appropriate interactions between the fore/midbrain and hindbrain. Furthermore, gbx2 knockdown, which will also affect the interaction of genes at the MHB, disrupted the isthmic structure and down-regulated MHB genes such as eng2, wnt1, and fgf8. Recently, overexpression of Xgbx2 in Xenopus embryos was shown to produce morphologic anomalies in head formation and suppression of En2, which appears to corroborate our results (Tour et al., 2001, 2002a).

Although the molecular mechanism underlying the effects of gbx2 overexpression on brain formation, especially in terms of the target genes of gbx2, remains to be clarified, it is possible that different regional specification genes are directly regulated by gbx2. Alternatively, most of these effects may be explained by the suppression of otx2, because the expression of otx2 precedes that of most other genes that are involved in brain regionalization (Rhinn and Brand, 2001). In fact, the effects of gbx2 overexpression observed here were similar to those observed in Otx2-null mutant mouse embryos (Acampora et al., 1995; Matsuo et al., 1995). Whatever the mechanism, the role of gbx2 in the regulation of MHB genes is complex, because different MHB genes showed different responses to gbx2 overexpression. Generally speaking, pax2.1 is relatively refractory to gbx2 overexpression, while eng2 and wnt1 are rather sensitive to the same treatment. It is possible that sensitive genes are directly suppressed by gbx2, while the gbx2 overexpression only indirectly affects the pax2.1 expression. Expansion of the fgf8 expression is probably ascribed to its restriction to the anterior hindbrain that is allowed to develop or even expanded in gbx2-overexpressing embryos. Consistent with these possibilities, wnt1, pax2.1, and fgf8 are known to be activated independently during epiboly, showing different transcriptional regulation (Lun and Brand, 1998). Another possibility, although not exclusive with those mentioned above, is that pax2.1 represents the early stage of MHB development that is relatively stable, whereas eng2 and wnt1 are involved in the later stage of the isthmic development that is fragile and sensitive to gbx2 overexpression. This possibility might also explain the difference in sensitivity among different genes that was observed in the knockdown experiment. Thus, the low level of gbx2 expression that could have escaped the effect of the morpholino oligonucleotide might be sufficient for the early stage but not for the following isthmic development.

Involvement of gbx2 in the Maintenance of MHB and Formation of the Isthmic Region

In addition to the suppressive effect of gbx2 on otx2 expression, we observed down-regulation of gbx2 by otx2, as shown previously in mice and chicks (Broccoli et al., 1999; Katahira et al., 2000). Therefore, our present data are apparently consistent with the model in which Gbx2 is involved in the establishment of the MHB through a mutually suppressive interaction with Otx2. Because transplantation experiments have shown that the fate of the MHB region is already determined at the mid-gastrula stage in the prospective neuroepithelium of zebrafish embryos (Miyagawa et al., 1996), such interaction between otx2 and gbx2 is predicted to occur before mid-gastrulation. In fact, as in other vertebrate embryos, otx2 is expressed early during epiboly in the prospective fore/midbrain of zebrafish (Li et al., 1994; Mori et al., 1994).

However, we have shown here that, although gbx2 expression starts zygotically at the shield stage (early gastrula), restricted expression is initiated in the prospective anterior hindbrain at the late gastrula stage. Thus, in contrast to all the other vertebrates examined, gbx2 expression starts after the establishment of the MHB in zebrafish embryos together with the MHB genes. Other differences between zebrafish and other vertebrates are seen in the expression of Gbx2/gbx2 in the brain: (1) gbx2 expression is restricted to the anterior hindbrain from the initiation of expression in zebrafish embryos, while Gbx2 expression is observed broadly in the more-posterior neural plate, in addition to anterior hindbrain, at the early stages in mice (Bouillet et al., 1995; Wassarman et al., 1997) and chicks (Niss and Leutz, 1998; Shamim and Mason, 1998); and (2) RA treatment causes anterior expansion of the Xgbx-2 expression in Xenopus embryos (von Bubnoff, 1996), while the same treatment suppresses gbx2 in zebrafish embryos. These differences together point to functional diversification during the molecular evolution of Gbx2/gbx2.

It is possible that the low level of gbx2 expression that was detected by RT-PCR in early gastrulae is sufficient for the establishment of the MHB. However, in zebrafish embryos with a knockdown of gbx2, the expression of all the MHB markers was not affected at the bud stage, which argues against the involvement of gbx2 in MHB establishment during epiboly. MHB marker expression was down-regulated slightly at the early stages of somitogenesis, and the effect was evident at the pharyngula stage (26 hpf), at which time the isthmic structure was severely affected morphologically. Therefore, we assume that gbx2 is involved not in the establishment but in the maintenance of the MHB and is essential for the formation of the isthmic structure during somitogenesis. This view is consistent with our result showing that the effects of the gbx2 knockdown could be rescued by Gbx2 activation, even after gastrulation.

Recently, another member of the zebrafish gbx group, gbx1, was implicated in MHB establishment during epiboly (Rhinn and Brand, 2001; Reim and Brand, 2002). Here, we reexamined the expression of gbx1 during epiboly by comparison with otx2 expression, showing that gbx1 expression starts from the early stage of gastrulation together with otx2 just like Gbx2 of other vertebrates, and that the expression domain is almost complementary to that of otx2 as is mentioned previously (Rhinn and Brand, 2001). This seems to support the proposed view that, in zebrafish, gbx1 contributes to the establishment of MHB together with otx2 like the Gbx2 gene of other vertebrates (Rhinn and Brand, 2001). Meanwhile, during somitogenesis (data not shown) and at the pharyngula stage (25 hpf), gbx1 expression is excluded from r1 and seen in more posterior hindbrain (this work and Zebrafish Information Network). It should also be mentioned that the expression of Ovx1, which appears to be the carp orthologue of gbx1, was seen on the dorsal side of embryos at 50% epiboly and around the MHB at early somitogenesis stages and then faded away caudally (Stroband et al., 1998). Thus, in fish embryos, gbx1/Ovx1 is expressed from the early phase of epiboly in the anterior hindbrain, but later during somitogenesis, gbx1/Ovx1 expression is absent from r1, whereas gbx2 expression starts in the anterior-most hindbrain. Thus, the combined expression pattern of zebrafish gbx2 and gbx1/Ovx1 seems similar to the expression of Gbx2 in other vertebrates. Furthermore, although the Gbx1 and Gbx2 proteins show limited structural differences despite their overall similarity, we observed in this study that overexpression of gbx1 gave rise to effects that were similar to those of gbx2 on brain formation. We also obtained similar effects on brain formation in zebrafish embryos for chick Gbx1 and Gbx2 (unpublished data).

However, it should also be mentioned that, despite the apparent complementary expression of otx2 and gbx1 during epiboly, we always observed a narrow gap between the two genes. It might be because the possible low level of expression of the two genes at the boundary is below the detection limit in our hands. Alternatively, another gene(s) could be involved in the establishment of MHB together with otx2 at least in zebrafish embryos. Furthermore, although gbx2 knockdown did not significantly affect the establishment of MHB in this study, we cannot completely exclude the possibility that a very low level of gbx2, which was detected at early epiboly stage by RT-PCR, cooperates with and modify the function of gbx1 in MHB establishment. Although this issue should be further addressed in the future, based on its temporal and spatial expression pattern as well as the effects of overexpression on brain formation, we favor the view that gbx1 is involved in the MHB establishment early during epiboly.

Taken together, we assume that functional division of labor occurs for MHB development in fish among different genes that include gbx1/Ovx1 and gbx2. In this scheme, gbx1/Ovx1 functions in MHB establishment, while gbx2 is involved in the maintenance and/or promotion of the formation of the isthmus and cerebellum. Of interest, Gbx1 is expressed transiently in the mouse embryonic forebrain (Bulfone et al., 1993), which suggests that functional shuffling took place during vertebrate evolution.

Possible Roles of gbx2 in Other Aspects of Vertebrate Embryogenesis

In addition to the MHB region, gbx2 is expressed in a variety of regions during the later stages of embryogenesis. In the otic region, gbx2 is expressed from the bud stage through to the pharyngula stage. The expression of Gbx2 in otic vesicles is also known in other vertebrates (Bouillet et al., 1995; von Bubnoff et al., 1996; Hidalgo-Sanchez et al., 2000), and the development of the vestibular organ in the ear is affected in Gbx2 mutant embryos (Wassarman et al., 1997). Although the structure of the otic vesicle was not disrupted by gbx2 overexpression, it was severely affected by MO-gbx2, which could be rescued by Gbx2 activation at the early somitogenesis stages; this shows that gbx2 is also required for ear development during somitogenesis in zebrafish.

We show for the first time in zebrafish that gbx2 is expressed in the dorsal thalamus at the later stages of development, as has been shown for murine and avian embryos (Fig. 2; Bulfone et al., 1993; Niss and Leutz, 1998). The forebrain is composed of several compartments along the anteroposterior and dorsoventral axes (Bulfone et al., 1993; Hauptmann and Gerster, 2000). The expression of gbx2 in a restricted domain in the thalamus suggests that the molecular mechanism governing specification of the dorsal thalamus is conserved among vertebrates. Because gbx2 expression is also detected in neuron-like cells in the hindbrain in pharyngula-stage embryos, it may be that gbx2 is involved in the specification and/or differentiation of neurons in the hindbrain, although the identity of these gbx2-expressing cells is undefined.

In zebrafish embryos at the early somitogenesis stage, the cranial neural crest cells begin to migrate ventrally along at least three pathways that lie adjacent to r1/r2, r4, and r6 (Akimenko et al., 1994). Of these different neural crest populations, gbx2 appears to be expressed in the cells that migrate anteriorly to the otic placode (r4) at the early somitogenesis stage. It is likely that these cells differentiate into gbx2-expressing cells in the pharyngeal arches at later stages of somitogenesis, as cranial neural crest cells enter the pharyngeal arches and participate in the formation of a variety of tissues, including cartilage and connective tissues (Schilling and Kimmel, 1997). The expression of Gbx2 in pharyngeal arches has also been demonstrated in mouse, chick, and Xenopus embryos (Bouillet et al., 1995; von Bubnoff et al., 1996; Shamim and Mason, 1998), which indicates that the Gbx2 roles in pharyngeal arch development are highly conserved among vertebrates.

Aside from the ectoderm, gbx2 expression was detected in the endomesoderm that underlies the prospective anterior hindbrain. The expression of Gbx2 in the mesoderm and/or endoderm during gastrulation occurs during amniote embryogenesis (Bouillet et al., 1995; Shamim and Mason, 1998), although to the best of our knowledge this is the first report of the spatially close relationship between gbx2 expression in the endomesoderm and neuroectoderm. This spatial correlation of gbx2 expression suggests an interaction between these germ layers that is mediated by molecules whose expression requires gbx2.



Adult zebrafish (Danio rerio) were maintained at 27°C in a 14-hr light/10-hr dark cycle. Embryos were raised at 28.5°C until the appropriate stages. Morphologic features, hours postfertilization, and days postfertilization were used to stage embryos (Kimmel et al., 1995). When necessary, 0.2 mM phenylthiourea (Nakarai) was added to culture to prevent melanization. For treatment with RA, the embryos were kept in the presence of RA (all-trans-retinoic acid; Sigma) from 6 hpf to 7 hpf, washed three times with water, and allowed to develop to the bud stage.

Cloning and Sequencing of Zebrafish gbx2 cDNA

Screening of the cDNA library from zebrafish 20-hpf embryos (kindly donated by Dr. H. Okamoto, RIKEN) was performed by hybridization with the partial gbx2 cDNA that was generated by RT-PCR (see below). The DNA sequences of both strands were determined by using the ALFexpress DNA sequencer with the ThermoSequenase fluorescently labeled primer cycle Sequencing Kit and 7-deaza-dGTP (Amersham Pharmacia Biotech).


Total RNA was prepared from embryos by the acid guanidinium thiocyanate-phenol-chloroform method (Chomczynski and Sacchi, 1987). The cDNA was prepared by reverse transcription of total RNA with the oligo(dT)17 primer and M-MLV reverse transcriptase (GibcoBRL). To obtain a partial cDNA clone of the zebrafish homeobox genes, the cDNA from 1k-cell stage embryos was amplified by PCR using the following degenerate primers, which were designed based on the two conserved amino acid sequences in the Antp-type homeodomain (ELEKEF and KIWFQN; Burglin, 1994): the sense primer incorporating an EcoRI linker (underlined), 5′-GCGAATTCGA(A/G)(T/C)T(A/G/T/C)GA(A/G)AA(A/G)GA(A/G)TT-3′; and the antisense primer incorporating an EcoRI linker (underlined), 5′-GCGAATTCCTT(T/C)TG(A/G)AACCA(A/G/T)AT(T/C)TT-3′. The target sequence was amplified for 30 PCR cycles that consisted of denaturation at 94°C for 1 min, annealing at 37°C for 45 sec, and extension at 72°C for 1 min, with the exception of the last cycle, during which extension was allowed to proceed for 5 min. The PCR product was subcloned into the EcoRI site of pBluescript SK(−).

For the analysis of gbx2 expression, 5 μg of total RNA from embryos at different stages were treated with RNase-free DNase I (Takara) and used for the synthesis of cDNA as described above. The derived cDNA was used for PCR with the gbx2-specific sense (5′-CACACCGTTCATGATGATGC-3′) and antisense (5′-GACAATCATCCTCCTTCGAG-3′) primers to generate a PCR product of 521 bp. The following PCR conditions were used: 29 cycles of 94°C for 1 min, 50°C for 30 sec, and 72°C for 1 min, followed by a final cycle of 94°C for 1 min, 50°C for 30 sec, and 72°C for 10 min. The RT-PCR products were subjected to Southern analysis with the full-length gbx2 cDNA as the probe. As a control for RNA integrity, EF-1α cDNA was amplified for 25 PCR cycles in parallel, using the same cDNA template and with the primers: EFS1, 5′-CAAGTCTGTTGAGATGCACC-3′; and EFAS1, 5′-AGCCTTCTGTGCAGACTTTG-3′ (Nordnes et al., 1994). Preliminary experiments showed that the levels of PCR products for gbx2 and EF-1α were increased in a quantitative manner, at least until the 30th and 25th cycles, respectively.

The cDNA for zebrafish gbx1 (Reim and Brand, 2002; AF288763) was obtained by RT-PCR using specific primers and subcloned into pBluescript SK(−) for in situ hybridization.

Whole Mount In Situ Hybridization (WMISH)

Digoxigenin (DIG)-labeled and fluorescein-labeled RNA probes were synthesized by using the T3 and T7 RNA polymerases (Stratagene) with the DIG RNA Labeling Mix and Fluorescein RNA Labeling Mix (Roche Diagnostic), respectively, according to the manufacturers' protocols. Single-color WMISH was performed, essentially as described previously (Schulte-Merker et al., 1992). Two-color in situ hybridization was performed, essentially as described by Jowett (2001). For closer inspection, stained embryos were embedded in JB-4 (Polysciences, Inc.) according to the manufacturer's protocol, sectioned at a thickness of 10–20 μm, placed on glass slides, and observed under the microscope.

Construction of Plasmids for Gene Overexpression

The full-length gbx2 cDNA (gbx2-FL) was excised by EcoRI digestion from pGX1, and subcloned into the EcoRI site of pCS2+. To exclude interference by the UTR, the ORF of gbx2 cDNA (gbx2ΔUTR) alone was amplified by PCR, and subcloned into the EcoRI site of pCS2+. For otx2 overexpression, otx2 cDNA was excised from pM2 (Mori et al., 1994), and subcloned into the EcoRI site of pCS2+. The cDNA for zebrafish gbx1 obtained as described above was subcloned into pCS2+ at the StuI/XbaI site (pCS-gbx1). For the synthesis of egfp mRNA, the egfp gene was excised from pEGFP-1 (Clontech) by digestion with Notl followed by blunting and second digestion with BamHI, and subcloned into the BamHI and StuI sites of pCS2+ (pCS-EGFP). To construct the fusion between the gbx2 and egfp genes (gbx2-GFP), the gbx2 cDNA for the entire coding region, but lacking the 3′-terminal stop codon, was amplified by PCR, and subcloned at the BamHI site of pCS-EGFP in-frame immediately upstream of the egfp gene (pGbx2-GFP). For the construction of the hybrid gene between gbx2 and the ligand-binding domain of the glucocorticoid receptor (gbx2-GR), the region of the gbx2 cDNA that encodes the N-terminal 301 a.a. was amplified by PCR, and subcloned into the EcoRI/XbaI site of pCS2+MT (pMT-NH). Then, the DNA for the ligand-binding domain of the human glucocorticoid receptor in pSP64T-Xbra-GR-HA (kindly donated by Dr. M. Tada; Tada et al., 1997) was amplified by PCR and subcloned into pMT-NH at the XbaI site downstream to the gbx2 sequence in frame (pGbx2-GR). The product is Myc-tagged at the N-terminus due to the sequence in pCS2+MT, thereby making the hormone-inducible gene refractory to the effect of the morpholino oligonucleotide against gbx2 (MO-gbx2) described below. All of the constructs obtained were verified by sequencing.

mRNA Synthesis and Microinjection Into Embryos

For the synthesis of capped mRNA, the template plasmids were linearized with appropriate restriction enzymes and transcribed with SP6 RNA polymerase, using the MEGAscript SP6 Kit (Ambion) according to the manufacturer's protocol. Capped mRNA was pressure-injected into single blastomeres of one- to four-cell stage embryos, which were allowed to develop to the appropriate stages in the presence of 10 U/ml penicillin G and streptomycin (GibcoBRL). The mRNA for egfp, which was synthesized from pCS-EGFP, was injected into the embryos as negative control, while mRNA for lacZ that was synthesized from pCS2+cβ-gal (gift from Dr. D. Turner) was used as a lineage tracer.

Detection of the β-Galactosidase Activity in Embryos Injected With lacZ mRNA

Embryos that were injected with lacZ mRNA were fixed with 4% paraformaldehyde in PBS for 30 min at room temperature, washed four times with PBS that contained 0.1% Tween 20 (PBST) for 10 min each time, and washed once for 10 min with 17.5 mM K3[Fe(CN)6], 17.5 mM K4[Fe(CN)6], and 1 mM MgCl2 in PBS (Buffer A). Color staining was performed by using 0.4 mg/ml X-gal (Wako) in Buffer A for 2 hr at 37°C. The samples were washed three times with PBST for 5 min each and refixed with 4% paraformaldehyde in PBS at 4°C overnight. The embryos were then processed for in situ hybridization.

Acridine Orange Staining

To detect degenerating cells, live embryos were dechorionated and placed for 1 hr in PBS that contained 2 μg/ml acridine orange (Wako). After two brief washes in PBS, the embryos were mounted in methylcellulose and viewed by fluorescence microscopy using the fluorescein isothiocyanate filter set.

Morpholino Oligonucleotides

The morpholino oligonucleotide (Gene-Tools, Inc.) that was targeted to the 5′-terminal sequence of the coding region of gbx2 (MO-gbx2) was injected into one-cell stage embryos (1.5 pmol/embryo). MO-gbx2 had the sequence 5′-ACGGTGTGCTGAAAGCTGCACTCAT-3′; and the control morpholino oligonucleotide (MO-con) had the sequence 5′-CCTCTTACCTCAGTTACAATTTATA-3′.


We thank Drs. M. Fürthauer, P. Ingham, T. Jowett, M. Kobayashi, R. Moon, H. Mori, H. Nakamura, A. Obinata, H. Okamoto, H. Takeda, and D. Turner for providing us with plasmids. We also thank Dr. H. Okamoto for the cDNA library, Drs. K. Shimamura and K. Hashimoto for advice on the anatomy of the brain, and S. Nagayoshi for technical assistance. K.Y. was funded by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.