Fgf signalling through MAPK cascade is required for development of the subpallial telencephalon in zebrafish embryos

The telencephalon is the anterior-most structure in the CNS. Though its morphologies are quite divergent in different species, organisation of telencephalic subdivisions, subpallial and pallial territories, are basically conserved between species (Wilson and Rubenstein, 2000). The subpallial telencephalon constitutes most of the basal ganglia and the pallial telencephalon contains the progenitors in the cerebral cortex in mammals. Fate-mapping studies have revealed that the telencephalon derives from cells at the anterior margin of the neural plate (Rubenstein et al., 1998; Whitlock and Westerfield, 2000). Experiments with mice and zebrafish have demonstrated that the tissue located in the ANB plays an important role in induction and patterning of the telencephalic region (Shimamura and Rubenstein, 1997; Houart et al., 1998). Fgf8 expressed in the ANB has been implicated in these processes (Shimamura and Rubenstein, 1997; Reifers et al., 1998; Shanmugalingam et al., 2000). The phenotypes of zebrafish ace mutants, in which no functional Fgf8 is produced, reveal the function of fgf8 in the regionalisation of the telencephalon; some subpallial markers are downregulated in the mutants (Reifers et al., 1998; Shanmugalingam et al., 2000). Furthermore, mutant mice in which fgf8 function is reduced, show the loss of olfactory bulb and reduction of the forebrain (Meyers et al., 1998). However, owing to overlapping expression and functional redundancy of Fgf family genes (Maruoka et al., 1998; Xu et al., 1999), a precise role of Fgf signalling in development of the telencephalon remains unclear.

We address this question in zebrafish embryos by direct observation of the response to Fgfs. Fgf receptors are members of the receptor tyrosine kinase (RTK) superfamily (Klint and Claesson-Welsh, 1999; Ornitz, 2000). At present, four members are known and each of them has several splicing isoforms (Klint and Claesson-Welsh, 1999). In general, RTKs trigger, by way of Ras, a sequential activation of the mitogen-activated protein kinase (MAPK) (or extracellular regulated kinase (ERK)) signalling cascade (Cobb and Goldsmith, 1995). ERK is activated by dual phosphorylation of threonine and tyrosine residues by MAPK/ERK kinase (MEK) (Crews et al., 1992). A monoclonal antibody raised against a dually phosphorylated form of ERK (dpERK) has been successfully applied to Drosophila and Xenopus, and reveals a dynamic dpERK staining pattern in early embryos (Gabay et al., 1997a; Gabay et al., 1997b; Christen and Slack, 1999). In Drosophila, several different RTKs participate in ERK activation, while, in Xenopus, the Fgf family seems to be responsible for the full pattern of dpERK in early development as activation can be blocked by the expression of a dominant-negative form of Fgf receptor (Christen and Slack, 1999).

We find that ERK is activated in the zebrafish ANB. Like Xenopus, the activation in the ANB is attributable to an Fgf signal. Interestingly, ERK activation in the ANB is maintained in ace mutants, suggesting other Fgfs function in this region. We also examined the function of Fgf signalling in zebrafish forebrain patterning by use of several inhibitors of the Fgf/”Ras/MAPK signalling cascade, such as dominant-negative form of Ras (Ras^N17) and the Fgf receptor, and a chemical inhibitor of Fgfr, SU5402. In treated embryos, the induction of telencephalic territory proceeds normally but the development of the subpallial telencephalon is suppressed, indicating that Fgf signalling is required for the regionalisation within the telencephalon. Finally, we show using antisense that both Fgf3 and Fgf8 are required for normal development of the subpallial region.

Zebrafish, Danio rerio, were maintained at 26°C and embryos were collected from natural crosses of wild-type fish. Homozygotes of ace^ti282a were obtained from crosses of heterozygotes. Collected embryos were maintained in 1/3 Ringer’s solution (39 mM NaCl, 0.97 mM KCl, 1.8 mM CaCl₂, 1.7 mM Hepes at pH 7.2) at 28.5°C. Embryos were staged according to hours postfertilisation (hpf) at 28.5°C and morphological criteria (Kimmel et al., 1995).

Capped sense RNAs were synthesised using the mMESSAGE mMACHINE™ large scale in vitro transcription kit (Ambion) from the plasmid containing full-length human Ras^N17 (Deng and Karin, 1994), Xenopus bΔFR4 (Hongo et al., 1999), and zebrafish fgf3, fgf3-myc, GFP and lacZ. GFP and lacZ RNA were used as a control. The mRNAs were diluted to 0.8-1.0 μg/μl with distilled water and 200-500 ng/μl (Ras^N17), 400 ng/μl (bΔFR4), 5 ng/μl (fgf3-myc) or/and 5 or 100 ng/μl (fgf3) mRNAs were injected into one- to two-cell stage embryos. Morpholino oligonucleotides were solubilised in water at the concentration of 50 μg/μl. The resulting stock solution was diluted to working concentrations in water before injection into one- to two-cell stage embryos. Injected embryos were cultured in 1/3 Ringer’s solution until use. The sequences of morpholino oligonucleotides used in the present study were as follows (see Fig. 9N): fgf8-MO, 5′-TCAACCGTGAAGGTATGAGTCTC-3′; fgf3-MO, 5′-TATAACCATTGTGGCATGGCGGGAT-3′; 4-mis-MO, 5′-TATTACCTTTGTGGCATGCCGGCAT-3′; and fgf3-MO’, 5′-CAACAAGAGCAGAATTATAACCATT-3′.

For construction of fgf3-myc, the full-length of fgf3 cDNA was amplified by PCR and inserted into pCS2+ containing 6× Myc epitopes such that the Myc epitope is placed in the C terminus of Fgf3 (kindly provided by Dr M. Hibi, Osaka University).

Mouse recombinant Fgf8b protein (R&D Systems) or BSA was used at the concentration of 0.25 μg/μl. Preparation of the beads soaked in Fgf8b or BSA were performed as described in Makita et al. (Makita et al., 1998). Dechorionated embryos at tailbud stage were placed in 3% methylcellulose in 1/3 Ringer’s solution and transplanted into the forebrain using a tungsten needle. After transplantation, embryos were cultured at 28.5°C in 1/3 Ringer’s solution until use.

For injection, 10 mg/ml SU5402 (Calbiochem) in DMSO was used. Dechorionated embryos were placed in 1.5% methylcellulose in 1/3 Ringer’s solution and injected into thick head region at tailbud stage. DMSO was injected as a control. Alternatively, dechorionated embryos at proper stages were soaked for 10 minutes in SU5402/1/3 Ringer solution at a concentration of 0.1 mg/ml, and washed several times with 1/3 Ringer’s solution. Treated embryos were cultured at 28.5°C in 1/3 Ringer’s solution until use. We found that the activity of SU5402 varied a great deal, depending on the lot number. Thus, the incubation was sometimes carried out at a concentration of 0.2 mg/ml.

Apoptosis in zebrafish whole-mounts was detected according to a protocol given by the manufacturer with some modifications (Dead End™ Colorimetric Apoptosis Detection System, Promega). After fixation overnight in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS), embryos were transferred in methanol and then rehydrated in PBST (PBS/0.1% Tween 20). Subsequently, embryos were digested in 5 μg/ml proteinase K in PBS for 5 minutes and postfixed for 20 minutes in 4% PFA in PBS. Then the embryos were immersed in acetone for 7 minutes at –20°C and incubated in the equilibration buffer (provided in the kit) for 10 minutes at the room temperature. After incubation for 3 hours at 37°C in working strength terminal deoxynucleotidyl transferase (TdT) enzyme, the DNA end-labelling reaction using biotinylated dUTP was stopped by washing in 2× saline sodium citrate (SSC) and PBST. Biotin was detected by horseradish-peroxidase-labelled streptavidin with diaminobenzidine (DAB).

Digoxigenin-labelled probes were synthesised by in vitro transcription using T3, T7 and SP6 polymerases. Whole-mount in situ hybridisation was performed by the standard protocol with some modifications (Schulte-Merker et al., 1992).

For histological analysis, the specimens were embedded in Technovit 8100 (Heraeus Kulzer, Wehrheim) and cut at 7 μm.

For whole-mount immunostaining, embryos were fixed with 3.7% formaldehyde/0.2% glutaraldehyde/PBS for 1 hour at room temperature. After washing with PBS, they were dehydrated with methanol and transferred to PBS. Then, they were washed with MABT (MAB/0.1% TritonX-100; MAB, 100mM maleic acid and 150 mM NaCl) three times for 10 minutes and MABDT (MAB/0.1% Triton X-100/1% DMSO) twice for 30 minutes. After blocking with 2% FCS/MABDT, the embryos were incubated in the blocking solution containing 1:10000 anti-di-phosphorylated ERK1 and ERK2 (MAPK-YT) antibody (Sigma) for overnight at 4°C. They were then washed with MABDT three times for 5 minutes, four times for 30 minutes, and incubated in blocking solution again for 30 minute, followed by incubation with the second antibody (1:500 anti-mouse IgG biotin conjugated antibody, Vector Laboratory) for 2 hours at room temperature. After the washing with MABDT as described above, the signals were detected with ABC staining kit according to the manufacturer’s instructions (Vector Laboratory). To examine ERK activation in ace homozygous mutants at the five-somite stage, embryos obtained from two heterozygous ace carriers were cut into two pieces at the level of the hindbrain, and they were fixed immediately. The anterior and posterior halves were then processed for dpERK or myoD staining, respectively.

Western blot analysis was performed following the standard method for ECL western blotting detection reagent (Amersham Pharmacia Biotech). Proteins from three embryos were separated by 12.5% polyacrylamide gels and transferred to Hybond™ ECL™ membranes (Amersham Pharmacia Biotech) by electroblotting. Monoclonal antibody against dpERK was used at the same concentration as used in immunostaining. The homogenates were prepared from whole embryos at 75% epiboly, bud and the six- to seven-somite stage. ace homozygotes could be distinguished by morphology by the six-somite stage.

For detection of Fgf3-Myc, 15 blastoderms taken from sphere-stage embryos injected with RNAs were loaded in each lane. The same staining protocol was applied except for that 9E10 (1:100 dilution, kindly provided by Dr Tatsumi Hirata, National Institute of Genetics) and 2% skim milk were used as the first antibody and as a blocking solution, respectively.

To analyse cell mitosis at early segmentation stages, about 0.5 nl of 25 mM of BrdU (SIGMA) was injected into the yolk at tailbud stage. After 3 hours’ incubation, injected embryos were fixed with 4% PFA/PBS for 2 hours at room temperature. After washing with PBS, the embryos were dehydrated with 25%, 75% and 100% methanol, and rehydrates in PBST. They were then treated with proteinase K (5 μg/ml in PBS) for 5 minutes at 37°C and with glycine-HCl (2mg/ml, pH 2.2) for 5 minutes at room temperature, followed by washing with PBST and post-fixation with 4%PFA/PBS for 20 minutes at room temperature. After washing with PBST, the embryos were treated with 2N HCl for 20 minutes at room temperature to relax chromatin and facilitate immunodetection of incorporated BrdU. Acid-treated embryos were then washed in PBSDT (1%DMSO, 0.1%Tween20 in PBS), incubated with PBSDNT (PBSDT+2% foetal calf serum) for 30 minutes at room temperature, followed by overnight incubation with anti-BrdU antibody (1:200 dilution in PBSDNT, Sigma) at 4°C. After intense washing with PBSDT and blocking with PBSDNT, the embryos were incubated overnight with the second antibody (1:500 dilution in PBADNT, anti-mouse IgG antibody biotin-conjugated, Vector Laboratory) at 4°C. After intense washing in PBSDT, the signals were detected with ABC staining kit as described above.

For cell count, the stained embryos were embedded in Technovit 8100 and serially sectioned at 5 μm. Three sagittal sections (more than 10 μm apart to each other) at the midline of the telencephalon from each treated embryo were selected, and the total number of positive and negative cells in each section was counted (six to seven embryos for each treatment). The ‘ventral’ corresponds approximately to the region that is positive for ERK staining.

ERK is one of well-known downstream mediators of Fgf receptor (Gotoh and Nishida, 1996; Chirsten and Slack, 1999). We examined the pattern of ERK activation in whole embryos, focusing on the anterior neural region. During gastrulation, strong activation is detected in the blastoderm margin that gives rise to the mesoderm (Fig. 1A). This activation is finally confined to the tailbud region by the end of gastrulation. The midbrain/hindbrain boundary (MHB) becomes positive at the end of gastrulation, and the activation reaches a peak level at the three-somite stage (Fig. 1B). After the activation in the MHB, the anterior part of the forebrain becomes positive at the three-somite stage, and the strongest staining is observed at the five-somite stage (asterisk in Fig. 1C). Histological sections of the five-somite stage embryos reveal that ERK activation in the forebrain is confined to the ANB (Fig. 1D). The activation gradually decrease (Fig. 1E,F), and become undetectable by the stages later than 12 somites. The activated pattern closely resembles that reported in Xenopus (Christen and Slack, 1999).

As ERK activation is triggered by downstream targets of various RTKs (see Introduction), we examined the contribution of Fgf signalling toward ERK activation by use of inhibitors to Fgf signalling cascade. We first tested whether or not ERK activation at segmentation stages is Ras-dependent by injecting synthetic RNAs encoding dominant-negative form of human Ras (Ras^N17), in which serine at position 17 had been substituted by asparagine (Deng and Karin, 1994). The synthetic RNAs were injected at one- to two-cell stages, and stained for dpERK at segmentation stages. In Ras^N17-injected embryos, all positive staining disappeared (43/48, Fig. 2A,B). We then examined the effects of XFD (Amaya et al., 1991) and bΔFR4 (Hongo et al., 1999), a dominant-negative form of Xenopus Fgfr1 and Fgfr4a, respectively. Both types of dominant-negative Fgfrs effectively reduced ERK staining (24/24 for XFD and 30/30 for bΔFR4; Fig. 2C), although the effect of XFD tended to be weak. Finally, developing embryos were treated with a chemical inhibitor to Fgfr, SU5402. SU5402 has been reported to inhibit specifically the kinase activity of nearly all types of Fgfr (Mohammadi et al., 1997). We injected 10 mg/ml of SU5402 into the anterior thick region of the brain near the ANB at tailbud stage (see Fig. 2D), and the embryos were fixed at the five-somite stage, followed by anti-dpERK staining. Again, the positive staining in the ANB was greatly reduced in injected embryos (19/19, Fig. 2E,F). Western blot analysis confirmed this result. As shown in Fig. 2I,J, one major band (50 kDa) was detected by anti-dpERK antibody. The intensity of this band much decreased (about 20% of the control) after injection of Ras^N17 or bΔFR4 RNAs and SU5402 treatment (Fig. 2I,J). However, the effect of XFD RNAs did not last long, and the activation level recovered to more than 50% by the six-somite stage (Fig. 2I).

Fgf-dependent activation of ERK was further examined by transplantation of Fgf-soaked beads into the anterior head region. Ectopic activation of ERK was induced around the transplanted beads that had been soaked in PBS containing recombinant mouse Fgf8b (15/15, Fig. 2H), while no such activation was detected around BSA-soaked bead (12/13, Fig. 2G). Taken together, we conclude that under our experimental conditions, ERK activation in the ANB was entirely dependent on Fgf signalling. Though ERK activation may not represent all the activity of Fgf signalling, it is reasonable to assume that the positive staining is an indicator for the activation of Fgf signalling.

We then examined ERK activation in ace mutants. The ace mutation is a G to A transition in the splice site after the second exon of the fgf8 gene, which leads to incorrect splicing generating a truncated non-functional protein (Reifers et al., 1998). Because, at the five-somite stage, homozygous ace mutants can not be distinguished morphologically from wild-type embryos, we examined the expression pattern of myoD in the trunk region. It has been reported that the adaxial cells of homozygous ace mutants show reduced and fragmented expression of myoD (Reifers et al., 1998). Although the intensity of ERK staining varies to some degree and tends to be weaker as compared with wild-type siblings, a substantial amount of positive reactions are reproducibly observed in the ANB (7/10, Fig. 3A,B) of ace mutants. Western blot analysis revealed that the level of ERK activation was about half of the wild type in ace mutants. These results indicated that Fgf signalling remains activated in the ANB in the absence of Ace/Fgf8.

The temporal and spatial pattern of ERK activation correlated well with that of fgf3 expression in the ANB (compare Fig. 1C, D with Fig. 4C,D) (Shanmugalingam et al., 2000). Zebrafish fgf3 starts to be expressed at the end of the gastrulation in the anterior head region, the anterior forebrain and MHB (Fig. 4A), and its expression persists until around the 10-somite stage (Fig. 4C). Histological sections reveal that the anterior staining is located in the ANB (Fig. 4B,D), where fgf8 is also expressed (Shanmugalingam et al., 2000). Importantly, fgf3 expression is maintained in the ANB of ace homozygous embryos at the tailbud stage (10/10, Fig. 1E) and mid-somite stage (10/10), suggesting that Fgf3 may compensate for the function of Fgf8 in the ace ANB. Indeed, Fgf3 is a potent activator of ERK: fgf3 RNA injection into one- to two-cell stage embryos induced ERK activation in entire embryo (25/25, Fig. 4F), while this activation was abolished by co-injection of Ras^N17 RNAs (38/38, Fig. 4G).

To explore the roles of Fgf signalling in zebrafish forebrain development, we examined the effects of Ras^N17 injection. Morphologically, Ras^N17-injected embryos showed obvious defects in the MHB and tail region. In most cases, the tail and yolk tubes did not extend well and the segmented structures in the trunk was hardly observed, indicating that the disruption of Fgf signalling cascade affected mesodermal tissues. Furthermore, we observed that the forebrain and midbrain region tended to be enlarged, probably owing to reduced posteriorising signals derived from the non-axial mesoderm (Woo and Fraser, 1997; Koshida et al., 1998). In addition to the MHB and tail, in more than half of Ras^N17-injected embryos the telencephalic region became turbid at 33 hpf, especially in the ventral part of the telencephalon (Fig. 5A,B). DNA fragmentation analysis revealed that apoptotic cells were induced by overexpression of Ras^N17 (8/11, Fig. 5C,D). Apoptotic cells were frequently observed in the ventral part of telencephalon but less in the dorsal. However, at 26 hpf, there was little apoptotic cell detected in the entire telencephalon of Ras^N17-injected embryos (18/18, Fig. 5E,F).

To determine whether the telencephalic patterning was altered before the appearance of apoptosis, we stained injected embryos with forebrain markers at different developmental stages between bud to 26 hpf. In situ hybridisation analyses with 1-day-old embryos revealed that, in Ras^N17-injected embryos, development of the subpallial telencephalon was severely affected. In zebrafish, the distal-less related gene, dlx2, is normally expressed in the subpallial telencephalon and diencephalon (Akimenko et al., 1994). Overexpression of Ras^N17 reduced or abolished dlx2 expression in the telencephalon (42/60, Fig. 5G,H), while the expression in the diencephalon was unaffected. Overexpression of Ras^N17 also reduced the expression of zebrafish nk2.1b (Rohr et al., 2001), which is a basal telencephalic marker related to mouse Nkx2.1 (Shimamura et al., 1995; Rubenstein et al., 1998) (61/64; Fig. 5I,J). Contrary to the subpallial markers, the expression domains of emx1 (Morita et al., 1995), eomesodermin (eom) and T-brain-1 (tbr1) (Mione et al., 2001), which are normally confined to the pallial region of the telencephalon at 26 hpf, nearly covered the entire region of the telencephalon in Ras^N17-injected embryos (32/37 for emx1, Fig. 5K,L; 25/25 for eom, Fig. 5M,N; 24/24 for tbr1, Fig. 5O,P).

The loss of the subpallial telencephalon was detected as early as segmentation stages. In normal embryos, nk2.1b expression in the telencephalon starts to be detected around the 14-somite stage (Rohr et al., 2001). However, nk2.1b expression was not detected throughout development in Ras^N17-injected embryos (20/20; Fig. 5Q,R). Consistently, dlx2 expression in the telencephalon, which starts to be activated at around the 15-somite stage (Akimenko et al., 1994), was not observed in Ras^N17-injected embryos at any stages (30/36, data not shown). In normal embryos, emx1 is first expressed in the telencephalic region, and, by the 10-somite stage, the expression is downregulated in the anteriormost of the forebrain, which may be fated to the subpallial telencephalon (Morita et al., 1995; Shanmugalingam et al., 2000). This downregulation, however, was never observed in Ras^N17-injected embryos (29/29; Fig. 5T). Thus, the regionalisation within the telencephalon was stopped at its early phase (at least as early as 10-somite stage of development) when Ras/MAPK cascade was blocked. During these early stages, inhibition of Ras pathway may affect cell growth in the telencephalic region. To test this, we injected BrdU into Ras^N17-injected embryos at tailbud stage and examined the proportion of labelled cells after 3 hours’ incubation (8-10 somite stage). However, we did not find any significant difference in cell proliferation in the telencephalon, either in ventral or dorsal region (Fig. 5Y,Z; Table 1).

In addition to the region-specific markers, Ras^N17 affected markers for neuronal differentiation in the telencephalon. islet-1 expression around the anterior commissure (Inoue et al., 1994) (Fig. 5U) was not observed in Ras^N17-injected embryos at 26 hpf (19/20; Fig. 5V). zash1a expression in the telencephalon was also undetectable in Ras^N17 injection (20/24; Fig. 5W,X). zash1a, one of the zebrafish genes related to Drosophila achaete-scute (Campuzano and Modolell, 1992), starts to be activated in the part of the subpallial telencephalon at 12-somite stage (Allende and Weinberg, 1994). These data demonstrated that Ras^N17 suppressed the differentiation of putative subpallial neurones.

In contrast to the regionalisation within the telencephalon, early overall patterning of the forebrain appeared normal in Ras^N17-injected embryos. As described above, emx1, an early telencephalic marker, was normally activated at the tailbud stage in the prospective telencephalic region of the Ras^N17-injected embryos (21/22; Fig. 6A,B). At this stage, we observed normal expression of gsc (Stachel et al, 1993) (21/21; Fig. 6C,D), dkk-1 (Shinya et al., 2000) (28/28) and hlx1 (Fjose et al., 1994) (20/20) in injected embryos, indicating that Ras^N17 did not affect the development of the underlying prechordal plate that is known to be important in establishment of the telencephalic region (Shinya et al., 2000). The expression of another early telencephalic marker, bf-1, which is expressed in the telencephalon and a part of eyes (Toresson et al., 1998), was not altered by Ras^N17 injection when examined at the five-somite stage (20/20; Fig. 6E,F) and 24 hpf (41/41, data not shown). Furthermore, downregulation of otx2 at the mid-somite stage in the region fated to the telencephalon and ventral diencephalon (Mori et al., 1994) normally occurred in injected embryos (33/34; Fig. 6G,H). Thus, the establishment of telencephalic and diencephalic territories normally takes place even when the Ras/MAPK cascade is suppressed.

We also confirmed that Ras^N17 did not affect the development of other parts of CNS: for example, the expression of sonic hedgehog (shh) in the ventral spinal cord and ventral diencephalon (Krauss et al., 1993) was normal in Ras^N17-injected embryos (18/18, data not shown).

When specific inhibitors to Fgfr-mediated signal were employed, essentially the same results were obtained (Table 2). In bΔFR4-injected embryos, expression of dlx2 (Fig. 7A,B), nk2.1b (Fig. 7C,D) and islet-1 (20/30, Fig. 7E,F) in the subpallial telencephalon at 26 hpf was suppressed while pallial markers, emx1 (Fig. 7G,H), eom (23/29) and tbr1 (Fig. 7I,J) expression covered the entire telencephalon. These effects were already observed at mid-somite stage. In the bΔFR4-injected embryos, nk2.1b expression was suppressed (15/18, Fig. 7L,M) at the 14-somite stage and emx1 was expressed throughout the telencephalon at the 15-somite stage (43/44, Fig. 7N,O). By contrast, the territory of the telencephalon was established normally, judging from the expression pattern of emx1 at tailbud stage, bf-1 at the five-somite stages and otx2 at the 15-somite stage (data not shown). The same phenotypes were obtained when XFD RNAs were injected (Table 2; Fig. 7K) (7/13 for tbr1).

Although the size of telencephalon tended to be smaller as compared with overexpression of Ras^N17 and bΔFR4, the injection of SU5402 (10 mg/ml) into the anterior head region at the tailbud stage (see Fig. 2D) gave rise to similar phenotypes in gene expressions of dlx2 (20/26; Fig. 8A,B), emx1 (9/13; Fig. 8C, D) and islet-1 (11/18, data not shown).

bΔFR4 and SU5402 also induced apoptosis in the ventral part of telencephalon, but the effect was weaker when compared with that in Ras^N17-injected embryos.

Among the RNAs injected, XFD seems to be less effective (Table 2). Especially, in the case of dlx2, only five out of 18 embryos injected with XFD RNAs showed reduced expression (Table 2). This could reflect the fact that the inhibitory effect of XFD on ERK activation does not last long (Fig. 2I). In spite of these, all inhibitors showed the similar tendency and, thus, we conclude that Fgf signalling through Ras/MAPK cascade is required for the development of the subpallial telencephalon.

Anti-dpERK staining revealed that ERK activation reached a level peak at the five-somite stage and, thereafter, gradually decreased. To address when Fgf signalling might be required for development of the subpallial telencephalon, we treated embryos with SU5402 at tailbud, five-somite, 10-somite and 15-somite stages, and its effect on dlx2 expression was examined. In this series of experiments, we soaked embryos in 0.1 mg/ml or 0.2 mg/ml SU5402 in 1/3 Ringer’s solution for 10 minutes. This worked well, and the resulting phenotype in the telencephalon was identical to that obtained with injection of SU5402.

As summarised in Table 3, early treatment (bud to 5-somite) with SU5402 led to loss or severe reduction of dlx2 in the subpallial telencephalon (Fig. 8E). By contrast, treatment from 10- and 15-somite stage gave rise to normal dlx2 expression in this region (Fig. 8F). Thus, Fgf signalling up to the five-somite stage is crucial for the establishment of the subpallial telencephalon but, thereafter, dlx2 expression can be maintained without Fgf signalling.

As described above, fgf8 and fgf3 are co-expressed in zebrafish ANB, and ERK activation is maintained in ace mutants. This led us to analyse the role of Fgf3 in the development of the subpallial telencephalon. To do this, we blocked the function of Fgf3 by using antisense morpholino-modified oligonucleotides (morpholino) that were targeted to the zebrafish fgf3 gene (fgf3-MO; Fig. 9N). Morpholinos have been reported to work well as effective and specific translational inhibitors in zebrafish (Nasevicius and Ekker, 2000; Sakaguchi et al., 2001). Indeed, injection of fgf8-targeted morpholino (fgf8-MO) at a concentration of 20 μg/μl (about 10 ng injected) resulted in a phenocopy of ace mutants, in which nk2.1b expression in the subpallial telencephalon was reduced (29/29; Fig. 9B), although the phenotype tended to be milder (compare Fig. 9B with 9F). Injection of 10 μg/μl fgf3-MO (about 5 ng injected) also reduced the nk2.1b expression (28/29; Fig. 9C). However, co-injection of fgf8-MO and fgf3-MO resulted in more severe reduction in nk2.1b expression (29/29; Fig. 9D). These results indicate a co-operative action of Fgf3 and Fgf8 on nk2.1b expression.

In the MHB, the effects of fgf8-MO and fgf3-MO were different from each other. Injection of fgf8-MO reduced the expression of en2 (11/11; Fig. 9B′) (Ekker et al., 1992) phenocopying the ace mutant, while that of fgf3-MO did not affect the expression (23/23; Fig. 9C′).

To test the specificity of the phenotype, we made another fgf3-MO designed against more 3′ sequence (fgf3-MO’; Fig. 9N). fgf3-MO’ exhibits similar effects on nk2.1b expression in wild-type (32/41) and fgf8-MO-injected (22/26) embryos. We then constructed a cDNA encoding Fgf3 tagged with Myc epitope at the C terminus, because an antibody against zebrafish Fgf3 is not available. Western blot analyses were performed with the embryos injected with either fgf3-myc RNAs alone or fgf3-myc and fgf3-MO. Co-injection of fgf3-MO completely blocked the translation of injected fgf3-myc RNAs, while the degree of inhibition much decreased in the case of 4mis-MO in which a four-base mismatch was introduced into fgf3-MO (Fig. 9M). We also confirmed that 4mis-MO, when injected, did not affect nk2.1b expression (Fig. 9A). Furthermore, the effect of fgf3-MO was diminished when fgf3 RNAs (5 ng/μl, about 2.5 pg per embryo) were co-injected (9/18, Fig. 9E). We were unable to obtained 100% rescue because the increased amount of fgf3 RNAs suppressed the head development in injected embryos even in the presence of fgf3-MO, probably owing to strong posteriorising and dorsalizing effects of Fgf3 at the gastrulation stage (S. K., M. S., M. Nikaido, N. Ueno, S. Schulte-Merker, A. K. and H. T., unpublished). Finally, we injected fgf3-MO (10 μg/μl) in embryos obtained from ace heterozygous crosses. Seventeen out of 58 injected embryos showed no or faint nk2.1b expression (Fig. 9H), while such enhanced phenotype was not observed by injection of fgf8-MO (54/54; Fig. 9G).

Finally, we examined emx1 and dlx2 expression in fgf3-MO-injected ace homozygous embryos. The expression of these two markers are relatively normal in ace mutants at 26 hpf. The ace homozygous embryos were identified by a lack of en2 expression in the MHB in double-stained embryos, as fgf3-MO alone does not affect en2 expression (Fig. 9C). Like SU5402 treatment, injection of fgf3-MO resulted in ventral expansion of emx2 (9/9; Fig. 9J) and reduction of dlx2 (11/18 for severe and 7/18 for mild reduction, Fig. 9L) expression in ace mutants, while no such effect was observed in control injections (18/18 for emx1 and 11/13 for dlx2; Fig. 9I,K). Thus, the expression of exm1 and dlx2 is more dependent on Fgf3.

In the present study, we report for the first time the spatial pattern of ERK activation in fish development. Strong ERK activation is detected in the mesoderm, ANB, MHB and tailbud in developing zebrafish embryos. Like Xenopus embryos (Christen and Slack, 1999), ERK activation at zebrafish segmentation stages depends on Fgf signalling. We then examine the role of Fgf signalling in forebrain development by use of inhibitors to Fgf signalling. Shanmugalingam et al. (Shanmugalingam et al., 2000) have shown that, in ace embryos, differentiation of the basal telencephalon and some putative telencephalic neurones were disrupted. In our study, the phenotype obtained by suppression of Fgf signalling encompasses that reported in ace mutants. In addition, we observe overall loss of subpallial fate and increased apoptosis in the ventral telencephalon, which are not seen in ace embryos. These data demonstrate that Fgf signalling in the ANB is required for the development of the subpallial telencephalon.

As in other vertebrates, zebrafish fgf8 is expressed during gastrulation in the mesoderm, presumptive MHB and telencephalon. (Fürthauer et al., 1997; Reifers et al., 1998). Expression in the CNS is prominent in the MHB and in the anteroventral telencephalon at segmentation stages (Reifers et al., 1998; Shanmugalingam et al., 2000). ERK activation appears to follow the expression pattern of fgf8 up to mid-somite stage, suggesting that fgf8 is responsible for ERK activation. However, the analysis with ace mutants, in which no functional Fgf8 is produced (Reifers et al., 1998), suggests that Fgf8 is not a sole Fgf responsible for ERK activation. Although the level of activation tends to be weaker, a substantial amount of ERK activation persists in the ANB of ace homozygous mutant. Expression analysis and overexpression experiments suggest that Fgf3 is one of the candidates involved in ERK activation in the ANB.

After mid-somite stage, the level of ERK activation gradually decreases in the MHB and then in the ANB, whereas fgf8 expression remains high in these regions. This suggests that Fgfs, albeit expressed, do not necessarily activate their signalling pathway. Alternatively, ERK activation may not represent all Fgf activity, or antibody labelling may not detect low levels of ERK activation. In fact, Fgfs are known to use various signalling cascades other than Ras/MAPK such as phosphatidylinositol 3′ kinase (Klint and Claesson-Welsh, 1999).

Analysis with telencephalic markers reveals loss of the subpallial fate after treatment with inhibitors to Fgf receptors. In addition to region specific markers, disruption of Fgf signalling reduces or abolishes the expression of differentiation markers specific to putative subpallial neurones. The effects seem direct because the underlying prechordal plate develops normally in treated embryos as judged by gene expression. The present results, therefore, indicate that Fgf signalling in the ANB is required for differentiation of subpallial neurones as well as establishment of the subpallial region.

There are at least three possible explanations for the phenotype in affected embryos: (1) the ventral telencephalon fails to acquire subpallial fate and develops into the pallium; (2) the subpallial region is established but disappears due to cell death; and (3) cells in the subpallium fail to proliferate and remains in a small number, while development of the pallial telencephalon proceeds normally. As cell death in the telencephalon is detected only after loss or reduction of the subpallial markers, the second possibility seems less likely. The detailed analysis of cell proliferation and movements during normal development of the subpallial telencephalon has not been done in fish. Thus, we cannot rule out either the first or third explanations. However, the following two observations favour the first possibility. First, inhibition of Ras pathway does not affect cell proliferation in the telencephalon at early segmentation stages (Table 1). Second, in affected embryos, subpallial markers such as dlx2 and nk2.1b are never detected at any stages examined, suggesting that the subpallial fate is not induced from the beginning.

In the present study, inhibition of the Fgf signalling affects only a part of the telencephalon, the subpallial region. Cells in the subpallium are shown to derive from the ANB. The row 1 cells in the ANB contributes to the anteriormost subpallial telencephalon (Houart et al., 1998). The precise lineage analysis by Whitlock and Westerfield (Whitlock and Westerfield, 2000) has revealed that cells in the ventral telencephalon come from the anterior margin of the neural plate at the five-somite stage, whereas more posterior margin is fated to the dorsal telencephalon. Their fate map of the ventral telencephalon well fits with ERK activation domain that is narrower than emx1-positive domain (a maker for the entire telencephalon at early stages). Thus, it is reasonable to say that the effect of Fgf inhibition was observed in the region where Ras/MAPK pathway is activated.

Brief treatment of developing embryos with SU5402 reveals that Fgf signalling is not required for subpallial development after 10-somite stage. The crucial period coincides with the stage when the strongest activation of ERK is observed. However, it has been reported that the same treatment at 18-somite stage still causes some defects in axon guidance between anterior commissure and post-optic commissure (Shanmugalingam et al., 2000). Thus, it is likely that a crucial period for requirement of Fgf signalling varies, depending on the region and developmental processes of the telencephalon.

Studies in mice have provided evidence that the anterior neural ridge (ANR), which is located in the anterior boundary between the neural and non-neural ectoderm, plays an important role in induction of the telencephalon. Shimamura and Rubenstein (Shimamura and Rubenstein, 1997) demonstrated by in vitro culture method that the ANR in mice is necessary and sufficient for the induction of telencephalic bf-1 expression. In zebrafish, the most anterior neural cells in the ANB, called row 1 cells, were found to possess the similar inducing ability to the mouse ANR (Houart et al., 1998). The ablation of row 1 cells resulted in the loss of telencephalic markers and apoptosis in the anterior head region (Houart et al., 1998). As in vitro experiments in mice have demonstrated that Fgf8-soaked beads mimic nearly all activity of the ANR in mammals (Shimamura and Rubenstein, 1997), Fgf signalling has been proposed to account for the inducing ability of the ANR.

However, we observe in the present study that the establishment of the telencephalic and diencephalic territories normally proceeded in the Ras^N17-injected embryos, as indicated by normal expression patterns of bf1, emx1 and otx2 at early segmentation stages. This suggests more restricted functions of Fgf signalling in telencephalic development. Indeed, no expression of any fgf or activation of ERK is detected in the ANB at the gastrula stage when the overall patterning of the CNS is specified. Consistent with this, SU5402 treatment at mid-gastrula (75%-epiboly) stage did not affect emx1 expression at tailbud stage (data not shown). Therefore, although involvement of low level of Fgf signal can not be ruled out, our findings suggest that, in zebrafish, Fgf signalling is mainly required for regionalisation but not for induction of the telencephalon.

Our antisense experiments suggest that Fgf3 co-operates with Fgf8 in the ANB. The antisense experiments seemed to work well in the present study. We are able to generate a phenocopy of ace mutant by injecting fgf8-MO, and fgf3-MO is able to inhibit translation of injected fgf3 RNAs in a sequence-specific manner. Furthermore, we recently demonstrated that casanova (226D7)-MO injection phenocopied casanova mutant, a complete loss of endodermal precursors (Sakaguchi et al., 2001).

Although the injection of fgf8-MO or fgf3-MO alone results in slight reduction of nk2.1b expression, severe reduction of nk2.1b is observed after co-injection of fgf8-MO and fgf3-MO. The results demonstrate that both Fgf8 and Fgf3 are required for normal nk2.1b expression of the subpallial telencephalon. It has also been revealed that emx1 and dlx2 expressions are more dependent on Fgf3. These data may account for relatively mild defects in the subpallial telencephalon of ace mutants. In the MHB, however, zebrafish Fgf8 has a major role as indicated by the phenotype of ace mutant, even though fgf3 expression is detected in the region (Fig. 4A,C). In support of this, the expression of en2 persists in fgf3-MO-injected embryos while the expression is reduced in fgf8-MO-injected embryos. Thus, Fgf8 and Fgf3 possess overlapping and distinct functions in the ANB and MHB, respectively. Considering that there are certainly further Fgfs in zebrafish that have yet to be identified, further molecular and genetic analyses will be required to elucidate full functions of Fgf signalling in the developing CNS.

We thank Drs S. Wilson and C. Houart for useful discussions and critical reading of the manuscript, and Dr K. Rohr for providing an unpublished probe for zebrafish nk2.1b. We are also grateful to Drs M. Mishina, H. Mori, H. Okamoto, S. Krauss, M. Hibi and M. Mione for providing probes, Dr E. Amaya for XFD cDNA, Dr H. Okamoto for bΔFR4 cDNA, Drs T. Deng and M. Karin for Ras^N17 cDNA, and Dr M. Kobayashi for providing ace mutants. This work was supported in part by Pioneering Research Project in Biotechnology from the Ministry of Agriculture, Forestry and Fisheries of Japan, and by grants-in-aids from the Ministry of Education, Science, and Culture of Japan. M. S. is supported by the Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists.