Isthmus-to-midbrain transformation in the absence of midbrain-hindbrain organizer activity

It is an intriguing question, how neuronal diversity is ultimately generated in the developing nervous system. Work on several model systems has revealed that the initial crude anteroposterior subdivision of the vertebrate neuraxis into prosencephalon, mesencephalon, hindbrain and spinal neural tube is refined by local organizing centers. The best characterized local organizers involved in the refinement of the patterning of the nervous system are the floor plate and roof plate, the anterior neural ridge/row1, and the isthmic (midbrain-hindbrain) organizer (reviewed by Altmann and Brivanlou, 2001; Briscoe and Ericson, 2001; Liu and Joyner, 2001; Rhinn and Brand, 2001; Simeone, 2002; Wilson et al., 2002).

During gastrulation, the boundary between the prospective midbrain and hindbrain can be defined as the interface of a rostral Otx2 and a caudal Gbx2 (gbx1 in zebrafish) expression domain in the neural plate of both amniotes and zebrafish(Broccoli et al., 1999; Millet et al., 1999; Rhinn et al., 2003) (reviewed by Simeone, 2000). Later on at this interface, activation of a genetic network composed of various transcription factors triggers localized expression of a secreted organizer signal (Fgf8), which in turn determines the development of the surrounding tissue (reviewed by Liu and Joyner,2001; Rhinn and Brand,2001; Wurst and Bally-Cuif,2001). Molecularly, expression of fgf8 is controlled by distinct regulators, a combinatorial interaction between inductive and modulatory factors (Reifers et al.,1998; Lun and Brand,1998; Ye et al.,2001). Analysis of noi/pax2a mutants in zebrafish demonstrates that induction of fgf8 is independent of pax2a(Lun and Brand, 1998). In spite of the fact that the isthmic organizer develops at the interface of the otx2 and Gbx expression domains(Broccoli et al., 1999; Millet et al., 1999; Rhinn et al., 2003) (reviewed by Simeone, 2000), these factors are only involved in the maintenance and refinement of fgf8expression and not in its induction (Ye et al., 2001; Martinez-Barbera et al., 2001; Li and Joyner,2001). Beside positive autoregulatory circuits, Fgf8 triggers expression of the Fgf target gene sprouty, a negative feedback modulator of Fgf signaling at the MHB(Fürthauer et al., 2001). Many of the MHB cascade genes are initially induced independently of Fgf8, and become dependent on Fgf8 activity only around the mid-somitogenesis stages(Reifers et al., 1998). By contrast, expression of the ETS transcription factors erm, pea3 and gbx2 are tightly dependent on Fgf8, and may mediate Fgf8 responses during early MHB development (Raible and Brand, 2001; Roehl and Nusslein-Volhard, 2001; Rhinn et al., 2003). Fgf8 thus fulfills multiple functions during development. Emitted from a perpendicular narrow stripe in the rostral hindbrain gbx expression domain, Fgf8 is required for self-maintenance of the MHB domain(Reifers et al., 1998; Lun and Brand, 1998). Fgf8 also controls the morphogenetic events leading to the formation of the anatomical isthmic constriction (isthmus rhombencephali) that separates the midbrain and hindbrain domains macroscopically(Brand et al., 1996; Reifers et al., 1998). Fgf8 has a crucial role in polarizing the midbrain tectum and defining ordered ingrowth of retinotectal axons (Lee et al., 1997; Picker et al.,1999). It also strongly influences the patterning of the dorsal metencephalon, a part of which eventually gives rise to the cerebellum(Reifers et al., 1998).

Zebrafish acerebellar (ace) mutants have a point mutation in the fgf8 gene (Brand et al.,1996; Reifers et al.,1998; Araki and Brand, 2002). In ace mutants a number of important regulatory genes prefiguring the position of the future anatomical isthmic constriction are initially present, but their expression is later abrogated. Consequently, proper patterning of the midbrain/hindbrain along the rostrocaudal axis is disturbed in ace mutants. The isthmic constriction fails to form between the midbrain and rhombomere 1 (r1), and a separate cerebellar anlage is not recognizable in the mutants. Interestingly,the mutants have no obvious truncation along the rostrocaudal extent of the mesencephalic/hindbrain alar plate. Rather, the mutants appear to have a caudally enlarged tectum in place of the cerebellum. This raises the question of whether the cells in ace located at equivalent rostrocaudal positions along the neuraxis as the isthmic and cerebellar primordia in wild types, retain their original fate, or adopt a new one in both domains in mutant embryos. In the present study we investigated whether the special morphological features of the ace mutants are due to a simple dysmorphology or whether they are associated with fate alteration. To distinguish between these scenarios we analyzed the molecular and positional identities of the morphologically reorganized tectal compartment of mutant embryos by comparing tectal/cerebellar specific gene expression patterns,neuronal subtypes, and cell proliferation, cell death and cell lineage characteristics of both wild-type and mutant embryos. We provide evidence that the primordial cells of the rhombencephalic isthmus undergo marked cell fate changes, demonstrated by their rostralized gene expression pattern and by cell-lineage analysis. Implantation of Fgf8-protein coated beads suggests that the observed cell fate transformation and lack of cerebellar development is due to the missing polarizing and patterning activity of the organizer signal Fgf8.

Zebrafish (Danio rerio) were raised and kept under standard conditions at 27°C (Westerfield,1994; Brand et al.,2002), and carriers heterozygous for the ace locus were identified by random intercrosses. To obtain homozygous ace mutants,heterozygous carriers were crossed. Freshly laid fertilized eggs were harvested into E3 embryo medium. Embryos were raised at 28.5°C,occasionally with 1-phenyl-2-thiourea (PTU) to prevent melanization. Embryos were staged according to Kimmel et al.(Kimmel et al., 1995).

Embryos were fixed in 4% paraformaldehyde (PFA) at the required stage of development. After dechorionation, embryos were dehydrated in 100% methanol and stored at -20°C until use. Whole-mount in situ hybdridization was performed as described (Reifers et al.,1998).

Membrane-bound ephrin A molecules were detected with an Epha3-AP fusion protein as described (Picker et al.,1999).

Zebrafish larvae were anaesthetized and, after fixation in 4% PFA, were cryoprotected in 30% sucrose and embedded in OCT compound (Sakura). Sections were then cut at 15 μm thickness on a Microm cryostat. Sections were mounted onto positively charged microscope slides (Superfrost Plus). In situ hybridization was performed according to a protocol established by Henrique(Henrique et al., 1995).

For zebrin II/aldolase C immunolocalization, sections were incubated with a 1:1000 dilution of the primary antibody (anti-zebrin II mouse monoclonal;kindly provided by M. Mione, UCL, London) for 48 hours at 4°C. The imunoreactivity was detected with a biotinylated horse anti-mouse antibody(Vector), avidin-biotin-peroxidase complex (ABC Elite Vectastain kit, Vector)and DAB chromogen.

Cell proliferation was detected by whole-mount immunohistochemical detection of phosphorylated histone H3, using a rabbit polyclonal (IgG)antibody, according to the manufacturer's instructions (Upstate Biotechnology). To detect apoptotic cells an in situ nick-end labeling procedure was performed using a commercial kit (Roche).

To inhibit Fgf signaling a pharmacological inhibitor, SU5402 (Calbiochem),was used at a final concentration of 24 μM as described previously(Reifers et al., 2000a).

Bead implantation was carried out as described by Reifers et al.(Reifers et al., 2000a). Cell lineage analysis was as follows: ace mutant embryos were distinguished from their siblings by morphology at the 5-somite stage(Brand et al., 1996) and manually dechorionated. Mutant and wild-type embryos were fixed in the desired position by attaching them with one side to a stripe of 3% methyl cellulose on a microscope cover slip in Ringer's embryo solution. A glass capillary was pulled, covered with crystalline DiI (Molecular Probes), and inserted into the embryo at the level of the midbrain-hindbrain boundary. The capillary was left inside for not more than 5 seconds. The first pictures were taken at the age of 10 somites, after the embryos had completely recovered from the labeling procedure. Images were captured on an Olympus BX61 microscope equipped with a Spot RT camera and Metamorph imaging software. Measurements were made using Metamorph. The system was calibrated using a micrometer calibration slide. Images were taken at 10× or 20× magnification. Fluorescent images were captured using a standard rhodamine filter set.

The anatomical features of the ace mutants are easily recognizable on living embryos (Brand et al.,1996) (Fig. 1A,B). In ace mutants the isthmic indentation fails to form and, as we demonstrated previously, although several markers of the molecular MHB are initially induced, their expression is not maintained(Brand et al., 1996; Reifers et al., 1998). To illustrate development of the isthmic primordium in ace mutants, we demonstrate expression of another set of genes that also prefigure the position of the future anatomical isthmic constriction during normal development and that behave in a similar fashion to the previously described ones (Brand et al., 1996; Reifers et al., 1998). A previously uncharacterized zebrafish homolog of the Drosophila region specific homeotic zinc-finger gene spalt (spa1a; GenBank Number AJ293862) (E. M. Camp and M. T. Lardelli, unpublished), similar to medaka and mouse Spalt homologs, specifically labels the MHB(Koster et al., 1997; Carl and Wittbrodt, 1999; Ott et al., 2001) during normal development. In ace mutants, the MHB expression domain of spalt1a (spa1a) is completely missing, as revealed at the 9-to 10-somite stage (Fig. 1C,D). Expression of pax8, a member of the pax2/5/8 subgroup of Pax genes (Pfeffer et al., 1998),is initiated weakly at the MHB at early somite stages in the mutant embryos(Fig. 1E,F) and later lost in ace mutants (Fig. 1G,H). sprouty4 (spry4) is one of the vertebrate homologs of the Drosophila receptor tyrosine kinase (FGFR, EGFR)inhibitor sprouty (Fürthauer et al., 2001). As spry4 is a direct target of Fgf8, its expression is not initiated in the mutant embryos(Fig. 1I,J). The above examples combined with our previous observations suggest that, in ace mutants,preceding the lack of the isthmic indentation, the cells of the isthmic primordium change their gene expression profile at around early- to mid-somitogenesis stages.

The gradual loss of MHB domain markers in ace mutants(Brand et al., 1996; Reifers et al., 1998) (present study) raises the question of how identities of the confronting midbrain (Otx expressing) and hindbrain (Gbx expressing) domains react to the loss of isthmic identity. To test this, we first analyzed the expression of markers that are restricted to the mesencephalic side of the (molecular) boundary,which are therefore not expressed in the cerebellar anlage of wild-type embryos. In vertebrates, Otx2, a well-characterized homolog of Drosophila orthodenticle (otd), marks the caudal limit of the developing mesencephalon (Millet et al., 1996), abutting the expression domain of another homeodomain protein, Gbx2 (Millet et al.,1999). [The latter function is exerted by gbx1 in zebrafish (Rhinn et al.,2003)]. Otx2 is expressed from early gastrulation and its expression is maintained through embryogenesis(Simeone et al., 1993). In ace embryos, otx2 expression is expanded caudally(Fig. 2B,H,J), whereas its expression stops at the caudal end of the mesencephalon in wild-type embryos(Fig. 2A,G,I). The caudal expansion of otx2 remains prominent throughout the somitogenesis period (Fig. 2F,H). By applying a pharmacological inhibitor of Fgf signaling, SU5402, aspects of the ace mutant phenotype can be phenocopied. SU5402 blocks the ATPase domain of Fgfr1 (Mohammadi et al.,1997). In wild-type embryos treated with SU5402 from tailbud to 7-somite stage, the caudal expansion of otx2 expression can be recapitulated (Fig. 2C,D,I,J). dmbx1 (diencephalic-mesencephalic homeobox 1) is an evolutionarily conserved homolog of Drosophila aristaless that has recently been described (Martinez-Barbera et al., 2001; Kawahara et al.,2002; Ohtoshi et al.,2002; Gogoi et al.,2002). As in amniotes (mouse and chick), the caudal limit of dmbx1 expression coincides with the caudal end of the mesencephalic anlage, excluding the anterior portion of the MHB fold (posterior mesencephalic lamina) (Fig. 2K,M; and data not shown). dmbx1 behaves in a similar way to otx2 in ace embryos: its expression is dramatically expanded caudally by the 7-somite stage, the earliest stage investigated(Fig. 2K-N). The expansion of otx2 and dmbx1 is coincident with the fading of the MHB markers (see also Brand et al.,1996; Reifers et al.,1998) at the position relative to where the isthmic constriction should form later on. In vertebrates, members of the Wnt family, such as wnt1 or wnt4, are expressed in overlapping domains along the mesencephalic tectum (Hollyday et al.,1995; Ungar et al.,1995). However, their expression does not expand to the hindbrain side of the isthmic constriction. In wild-type zebrafish embryos, wnt4 expression is enriched between the fore- and midbrain domains at the 5-somite stage (Fig. 2O). By the 20-somite stage, wnt4 expression gradually extends along the dorsal midline toward the caudal end of the mesencephalic tectum(Fig. 2Q). However, in ace mutant embryos, there is a marked caudal shift in wnt4expression toward the mesencephalic tectum by the 5-somite stage(Fig. 2P). Later on, expression of wnt4 narrows down to the dorsal midline of the mesencephalic tectum, and extends toward the caudal end of the enlarged tectal compartment(Fig. 2R).

Taken together, our results reveal a marked rostralization in gene expression at the dorsal MHB primordium and cerebellar anlage in acemutants.

Having seen that dmbx1, otx2 and wnt4 appear at ectopic caudal locations, we analyzed the expression of genes that mark the upper rhombic lip where the cerebellar anlage forms in wild-type embryos. zath1 (zebrafish atonal homologue 1; atoh1 -Zebrafish Information Network) is a marker of the upper rhombic lips(Koster and Fraser, 2001). In ace mutants, we consistently find that zath1 is missing from the upper rhombic lip region, when compared with wild-type siblings(Fig. 3A-D).

As the upper rhombic lips showed a marked rostralized gene expression profile in ace embryos, we analyzed whether cerebellar development could still be initiated in spite of this. To monitor cerebellar development,we used markers expressed by cerebellar granule and Purkinje cells. Of these genes, gap43 (Reinhard et al.,1994) and tag1/axonin (cntn2 - Zebrafish Information Network) are expressed by migrating cerebellar granule cell precursors (Fig. 3E,G)(Console-Bram et al., 1996; Wolfer et al., 1994; Lang et al., 2001). In mutant embryos, no expression of these markers can be detected in the region that would give rise to the cerebellum, at any of the stages investigated(Fig. 3F,H).

Likewise, staining for the bHLH transcription factor neurod, a marker expressed by cerebellar granule cells(Fig. 3I,K)(Miyata et al., 1999; Lee et al., 2000; Mueller and Wullimann, 2002),fails to detect a cerebellar compartment in mutant embryos(Fig. 3J,L). In addition,Purkinje cells of the cerebellum cannot be detected by anti-zebrin II/aldolase C (Brochu et al., 1990)immunostaining of mutant embryos (Fig. 3M,N). Thus, complementary to the caudal expansion of gene expression domains normally excluded from the dorsal metencephalon, specific markers for the developing cerebellar primordium, or for later granule and Purkinje cells, are not expressed in ace embryos.

Rhombomeres are transiently appearing segmental structures during hindbrain development. Of the 7 (8 in amniotes) rhombomeres, the rostral rhombomere 1(r1) gives rise to the cerebellum and other important structures of the vertebrate central nervous system, such as the Locus coeruleus (LC)(Morin et al., 1997). As our results revealed that the cerebellar plate derived from rostral r1 lost its identity, we investigated the expression of further specific marker genes characteristic of r1 in wild-type embryos. Expression of epha4a(formerly known as rtk1) marks caudal r1(Fig. 4A,C), and its expression is not diminished in ace embryos(Fig. 4B,D; small arrow). However, the orientation of its expression domain is altered: instead of being perpendicular to the rostrocaudal axis of the hindbrain, it becomes slanted(Fig. 4B). Moreover, double detection of otx2 and epha4a(Fig. 4E,F) reveals that their expression domains now abut, forming a new interface in ace embryos(Fig. 4F).

Precursors of the LC catecholaminergic neurons expressing phox2aare born in the rostral r1 (Morin et al.,1997; Guo et al.,1999). Subsequently, descendants of these cells occupy a ventrolateral position in that rostral segment expressing tyrosine hydroxylase(TH), the rate-limiting enzyme of catecholamine synthesis(Fig. 4G). In aceembryos, TH message is not detectable at the pontine flexure, indicating that the LC is absent (Fig. 4H)(Guo et al., 1999).

The presence of epha4a in caudal r1, and the abutting of the epha4a and otx2 domains in conjunction with the lack of cerebellar and LC development, suggest that the lack of Fgf8 in acemutants affects the rostral and caudal part of r1 unequally. The effect of the loss of organizer-derived Fgf8 function is restricted to rostral r1. Our analysis supports the notion raised by C. Moens' and I. Mason's laboratories that r1 is a bipartite structure, where the rostral part of this rhombomere should be considered as a separate entity designated as rhombomere 0(Waskiewicz et al., 2002; Walshe et al., 2002). In ace mutants, rhombomeres rostral or caudal to r4 appear to be often reduced in width along the rostrocaudal axis of the hindbrain, as judged by rhombomere marker analysis (Maves et al.,2002) (Fig. 4A-H). This is most probably due to the impaired r4 signaling activity in ace (Maves et al.,2002), rather than to the loss of isthmic organizer function. However, beyond the smaller rostrocaudal extent of certain rombomeres in ace, the generation of rhombomere-specific branchiomotor neuron patterns, including r2 and r3 trigeminal motoneurons, is largely normal(Maves et al., 2002).

As previously shown, the narrow perpendicular stripe of the isthmic organizer region emits rostrocaudal patterning signals along the whole extent of the MHB region. The lack of functional Fgf8 signaling dramatically alters the morphology of the isthmic and dorsal metencephalic alar plate region in ace embryos. We also examined how the basal plate (tegmental) region is affected in ace mutants. To address this, we analyzed expression of zash1b (ashb - Zebrafish Information Network), gap43,tag1/cntn2 and epha4a. These genes are expressed at ventral aspects of the diencephalon and hindbrain leaving a gap at the tegmental part of the MHB and midbrain. In ace embryos we observe a fusion of the diencephalic and hindbrain expression domains of zash1b,gap43 and tag1/cntn2, filling the above mentioned gap(Fig. 4I,J, Fig. 3E,F and Fig. 3G,H, respectively). In the case of epha4a, a caudal expansion of the diencephalic domain can be observed (Fig. 4B,D). These findings are consistent with the recently reported caudal expansion of fgfr3 expression along the basal plate of the midbrain, and with the consequent fusion of the diencephalic and hindbrain expression domains in ace embryos (Sleptsova-Friedrich et al., 2002). Expression of twhh, instead of being expanded, shows a dramatic reduction in the ventral mesencephalon in ace mutants(Fig. 4K,L). The alteration in expression of twhh may reflect the expansion of more rostral, caudal diencephalic or rostral mesencephalic fates toward caudal co-ordinates of the mesencephalon. Taken together, these changes in gene expression are compatible with the observed rostralization taking place at the dorsal metencephalic alar plate, and are indicative of a caudal-to-rostral transformation of the basal plate of the MHB. However, the degree of rostralization along the mesencephalic/metencephalic basal and alar plates may be different.

Various mechanisms could account for the successive rostralization and loss of isthmic indentation in ace mutants, such as regionally distinct cell proliferation or apoptosis, or transformation of cell fate in the affected tissue. We therefore analyzed the mitotic behaviour of the dorsal mesencephalic (tectal) and MHB primordial cells using an antibody raised against the proliferation marker phospho histone H3 (PH3). H3 phosphorylation has previously been described to correlate with mitosis in mammalian cells, Xenopus and Tetrahymena. Before the actual formation of the anatomical isthmic constriction, marked differences cannot be detected in cell proliferation between wild-type and mutant embryos(Fig. 5A-D). However, at later stages [36, 57 and 72 hours postfertilization (hpf)], the immunoreactivity of PH3 is markedly reduced at the caudal-most edges of ace tecta, and the cerebellar proliferative zone typical for wild-type embryos is absent in ace (Fig. 5E-G; data not shown). Increased proliferation, therefore, is very likely not the cause of the enlarged tectum. However, the decreased proliferation observed in young ace mutant larvae, may account for the overall smaller than wild-type size of ace mutants at later stages, at day 5, for example(Fig. 3I,J)(Picker et al., 1999).

Next, we analyzed whether a decreased rate of apoptotic cell death could contribute to the altered MHB development in ace mutants. Although pronounced rostralization is evident in gene expression during early somitogenesis stages in ace embryos, marked deviation from the wild-type cell death pattern cannot be detected by TUNEL staining(Fig. 5H,I). However, from the 12-somite stage on, a higher number of dead cells can be detected in ace embryos than in wild-type siblings, as demonstrated by DIC images of wild type and ace mutants (Fig. 5J,K). In accordance with this, an increased nick-end labeling can be observed at the 15-somite stage in ace embryos, indicating a higher number of apoptotic events in mutant tecta in comparison with wild-type structures (Fig. 5L-O). As marked differences in cell death cannot be detected at early segmentation stages, when pronounced differences in gene expression are observed, cell death is therefore likely to be a secondary consequence of the earlier,abnormal mis-patterning of the MHB area in ace mutants.

The rostralized gene expression and the lack of cerebellar cell types in ace mutants, along with the results of the proliferation and the cell-death studies, all suggest that the isthmic and cerebellar primordium might have already adopted a new tectal identity by early- to mid-somitogenesis stages. To address this issue directly, we performed cell lineage tracing experiments to reveal the fate of the MHB primordial cells. We investigated whether MHB primordial cells in ace mutants are retained in a position that corresponds to the MHB compartment of wild types, or whether they end up elsewhere, acquiring a new fate. We labeled groups of cells at the level of the prospective MHB with the lipophilic dye DiI during early somitogenesis (5-somite stage) in wild-type and ace embryos. The morphological MHB has not yet formed at this stage, but MHB primordial cells can be targeted by their position along the rostrocaudal axis relative to the posterior edge of the optic vesicle. By comparing the position of DiI-labeled cells at the 10-somite stage(Fig. 6A,D), and at 24 hpf(Fig. 6B,C,E,F), in wild-type and ace mutant embryos, we were able to compare the fate of these marked cells. Table 1summarizes the results of our labeling experiments. In all wild-type cases,the labeled cells ended up in either the posterior tectum or in the cerebellum/r1, i.e. in the MHB region (Fig. 6B,C). In all ace cases, cells ended up in the posterior part of the enlarged tectum (Fig. 6E,F). Moreover, we did not observe a loss of the labeled cell population in ace embryos, when compared with wild-type siblings. These results suggest that cells that are normally fated to become MHB tissue in the wild-type are transformed into tectal cells in the acemutants.

In ace mutants, the molecular and anatomical identity of the isthmic and cerebellar primordial cells is not maintained, and they acquire a more rostral fate as we have visualized by various markers. At later stages,lack of tectum polarization in ace mutants leads to a somewhat variable, but always severely distorted, retino-tectal map formation,including altered expression of ephrin A2 and ephrin A5b(Picker et al., 1999). To test the reversibility of the alterations in the affected brain structures,functional Fgf8 protein was applied by implanting Fgf8-coated beads into wild-type and mutant embryos, and their MHB/tectal morphology was then evaluated. Implantation of Fgf8-coated beads between the 5- and 20-somite stage, into the prospective MHB of ace embryos, rescues the MHB phenotype (n=9/11), reconstituting the MHB fold and the upper rhombic lip (data not shown). Moreover, the coated beads induce an ectopic MHB fold when implanted in the posterior diencephalon of wild-type embryos(n=5/7). Although the induced phenotypes are stronger after early bead implantation, reversion of the ace phenotype and induction of an ectopic MHB in wild-type embryos are possible even when implantation is performed very late, at the 20-somite stage, the latest stage we tested (data not shown). To assay the restorative and inductive abilities of Fgf8 on polarized marker expression, Fgf8 beads were implanted at the 15-somite stage into wild-type and ace mutant embryos. As a read-out, expression of ephrin A proteins was assessed at 30 hpf, detected using an Epha3-AP-fusion protein recognizing all known ephrin A proteins(Fig. 7A)(Brennan et al., 1997; Picker et al., 1999). In wild-type embryos, distribution of ephrin A proteins shows an increasing rostral-to-caudal gradient at this stage(Fig. 7A). However, this gradient is completely absent in ace embryos(Fig. 7B). Upon implantation of Fgf8-coated beads, graded ephrin A expression is restored in aceembryos on the operated side (n=5/6; Fig. 7D). Moreover, by implanting Fgf8 beads at various ectopic anteroposterior locations within the diencephalon and tectum of wild-type individuals, we find that Fgf8 is able to induce an ectopic MHB fold, and is sufficient to establish a mirror gradient of ephrin A proteins (n=4/4; Fig. 7C) by polarizing the diencephalon and tectum. Apparently, Fgf8 induces ephrin A protein expression in locations experiencing low or no Fgf8 concentration, corresponding to wild-type diencephala and ace tecta,respectively. From these experiments, we conclude that the expanded tectal tissue retains its ability to respond to Fgf8, and that the transformation of the MHB and cerebellar primordium is a reversible process, as scored by morphological and molecular criteria.

We have analysed the mechanisms underlying cerebellar agenesis in the zebrafish fgf8 mutant acerebellar. Our analysis revealed that in ace mutants: (1) there is a marked rostralization in gene expression profiles of the midbrain, MHB and cerebellar primordium; (2) the cerebellar developmental program is not initiated because the cells of the MHB and cerebellar primordium are transformed, acquiring a more rostral, tectal identity; and (3) isthmic folding and tectal polarization can be elicited in ectopic locations, and can be restored in ace mutants, upon local application of an Fgf8 protein source, suggesting that the absence of the inductive organizer signal Fgf8 is responsible for the observed fate alteration in these mutants.

The examination of MHB region specific marker gene expression enabled us to investigate the processes underlying the special anatomical features of ace embryos (Brand et al.,1996; Reifers et al.,1998). In ace mutants, expression of specific genes that prefigure the future anatomical constriction (pax2a, her5, wnt1, eng2,eng3, erm, pea3, spry4, spa1a and pax8) is perturbed(Brand et al., 1996; Reifers et al., 1998; Raible and Brand, 2001; Fürthauer et al., 2001)(present study). Consequently, in ace embryos a caudal shift of the expression of forebrain and mesencephalic markers (otx2, dmbx1, wnt4)is evident in positions where the isthmic primordium and upper rhombic lips should differentiate. From the expression of wnt4, it can also be estimated how rostral the molecular identity of the structures in aceembryos is. This gene normally starts to be expressed at the junction of the caudal diencephalon and the rostral tectum(Ungar et al., 1995). Its appearance at more caudal positions indicates that ace mutants show gene expression profiles characteristic for the rostral tectum of wild-type embryos. Pharmacological inhibition of Fgf signaling in wild-type embryos faithfully mimics the expansion of otx2 expression seen in ace mutants. The rostralized gene expression profiles of the mutants,along with the lack of the rhombic lip/early cerebellar marker zath1and other markers of the developing cerebellum (gap43, tag1/cntn2, neurod,zebrin II), suggest a possible fate transformation of the isthmic and cerebellar regions preceding the agenesis of the isthmic fold and the lack of a distinct cerebellar domain at later stages. Our present results, combined with earlier observations (Reifers et al.,1998; Raible and Brand,2001), show that a molecular MHB is initially formed but is then lost during the early- to midsomitogenesis period in the mutants. The significant structural reorganization taking place at the embryonic neuraxis of the mutant is preceded by a marked alteration in the molecular identity of the cells of the isthmic and cerebellar primordia. The identity of the MHB domain is not maintained and, as a consequence, both the isthmic and cerebellar primordia acquire a more rostral, mesencephalic character in ace mutants. In fact, in ace mutants the dorsocaudal compartment (upper rhombic lip/dorsal r1), which normally gives rise to the cerebellum, develops as an enlargement of the midbrain tectum. Taken together,these results show that the agenesis of the isthmic constriction in ace embryos is not only a simple dysmorphology. The scheme in Fig. 8 summarizes the most important features of ace mutants in comparison with wild-type siblings.

The severe rostralization in gene expression and lack of cerebellar development suggests that the isthmic and cerebellar primordia in acemutants acquire a new identity. Mechanistically, however, the enlargement of a particular tissue domain could also be the result of excessive proliferation or a decrease in developmental cell death. Upon monitoring the cell proliferation characteristics of the mutants, a significant decrease in cell proliferation can be detected. ace mutants lose their caudal (upper rhombic lip) proliferation zone typical of wild-type brain, as revealed by anti-phospho histone H3 staining. However, this difference in the cell proliferation pattern is a relatively late phenomenon, typically seen from pharyngula stages onwards. During the segmentation period, the differences in cell proliferation are not evident. Therefore, we think that an increased proliferation is highly unlikely to contribute to the loss of the isthmic structures in ace mutants. Furthermore, the results obtained by analyzing the TUNEL and DIC images of ace mutant embryos show that a decrease in the rate of ontogenetic cell death cannot account for the expansion of the tectum and the lack of the isthmus. In ace mutants,apoptotic cell death is increased when compared with wild-type siblings. However, an increased number of dead cells can only be detected from mid-somitogenesis stages, when the rostralized gene expression profiles, at the relative positions where the isthmic and cerebellar primordia should form,are already evident in the mutant embryos. The decreased cell proliferation and increased apoptosis rate at later stages of development are consistent with the role of Fgfs, including Fgf8, as mitogen and survival factors for various neuronal and non-neuronal cell types(Lee et al., 1997;Hajihosseini et al., 1999; Trumpp et al.,1999; Chi et al.,2003). The sequence of mis-patterning and cell death seen in ace embryos is somewhat reminiscent of the kreisler phenotype in mice, where the cells in the developing hindbrain that would normally become specified as r5 and r6 adopt an r4 character instead, producing an excess of r4 cells that is disposed subsequently by apoptosis(McKay et al., 1994).

Our analysis suggests that quantitative processes, such as increased proliferation or a decreased apoptotic rate, are unlikely to play a major role in the restructuring process occuring within the mesencephalic-metencephalic region in ace mutants. By contrast, our cell lineage tracing experiments clearly indicate that the isthmic and cerebellar primordial cells become part of the tectum in ace embryos. Cells at the location of the wild-type MHB give rise to caudal tectum in ace embryos. Therefore, we think that the cells of the putative isthmic primordium in the mutants are qualitatively different to those of the wild type. The marked rostralization in gene expression is connected to the alteration of positional information and caudal-to-rostral transformation in the absence of the functional organizer signal Fgf8. Our results suggest that MHB and cerebellar primordial cells are transformed to tectal ones, or show a default tectal fate in ace.

In our bead implantation experiments the molecular and morphological features of the MHB could be restored in ace mutant embryos,indicating that Fgf8 is either directly or indirectly necessary to execute the proper morphogenetic program at the mesencephalic and hindbrain alar plate. Moreover, when Fgf8 is provided ectopically in wild-type embryos it is able to restructure even the caudal parts of the forebrain, as was known from previous work in chick (Martinez et al.,1999; Shamim et al.,1999). Fgf8, besides being necessary and sufficient to reprogram and restructure the surrounding tissues, seems to act in a dose-dependent manner. Several pieces of evidence support the idea that different doses of Fgf8 may be responsible for the specification of distinct structures of the MHB region, and that the activity of Fgf8 is directly coupled to the dose of Fgf8 protein acting on the target tissue(Xu et al., 2000; Sato et al., 2001). Gain-of-function experiments performed with two different Fgf8 isoforms(MacArthur et al., 1995; Liu et al., 1999; Sato et al., 2001) show that the type-difference in the activity of the a and b isoforms can be attributed to the signal intensity, as electroporation of 100-fold less Fgf8b expression vector exerts similar effects to Fgf8a in chick embryos(Sato et al., 2001). To achieve cerebellar differentiation a high Fgf8 concentration is required, as reflected by the transformation of the tectal structures into cerebellum upon ectopic overexpression of the stronger Fgf8b isoform. During normal development, the stronger Fgf8b is the predominant isoform of Fgf8 in the isthmic region (Sato et al.,2001). The region that is exposed to strong Fgf8 signal (e.g. to the more abundant and physiologically more active Fgf8b), which expresses Gbx2 and Irx2, may then acquire r1 characteristics, from which the dorsal rhombic lip and, later, the cerebellum will differentiate(Sato et al., 2001). Regions in which the tectal differentiation program will be executed receive only weak Fgf8 signaling, as inferred from the results of diencephalic overexpression of the weakly active Fgf8a isoform (Sato et al., 2001). In addition, different doses of Fgf8 might influence the fine patterning of midbrain regions(Lee et al., 1997; Picker et al., 1999). Our bead implantation experiments support dose dependency of Fgf8 action, as reflected by the graded induction of tectal ephrin A expression at ectopic locations in wild-type embryos, or in ace embryos at positions where the isthmic organizer normally forms. Close to the Fgf8 source, this protein is normally expressed at high levels, and its expression gradually decreases as a function of the distance from the source. In ace embryos the gradient is absent, and the mutants show no or only low levels of expression,characteristic of the anterior tectum in wild types(Picker et al., 1999) (present study).

The next question is how Fgf8 triggers and orchestrates the reorganization of the isthmic region when applied in the form of coated beads. We have suggested previously that the implantation of Fgf8 beads into the mesencephalic alar plate of the mutants may re-activate a set of regulatory genes that are acting during an earlier phase of the normal MHB development(Reifers et al., 1998). During normal development, these genes are initially independent of Fgf8, but at the time the bead implantation experiments were performed many of these key regulators (her5, wnt1, pax2a, pax8, spa1a, spry4) are directly or indirectly dependent on Fgf8 function, as is evident from the fact that they disappear in the absence of functional Fgf8 in ace mutants. Indeed,bead implantation experiments done on chick and fish embryos support this notion. Fgf8- or Fgf4-soaked beads implanted into the posterior forebrain or hindbrain alar plate are capable of inducing isthmic, cerebellar or midbrain structures, and the key regulatory genes (Wnt1, En2 and Fgf8itself) are induced around the bead in the implanted embryos(Crossley et al., 1996; Martinez et al., 1999; Shamim et al., 1999). Similarly in zebrafish, we found that implanted Fgf8-coated beads in wild-type embryos are able to induce fgf17, en2, sprouty4 and other targets of the MAPK pathway (Reifers et al.,2000b; Fürthauer et al.,2001; Raible and Brand,2001). Similar to the results of the above-mentioned in vivo bead implantation experiments, Fgf8b-coated beads are able to trigger expression of En1, En2, Pax5, Wnt1 and Gbx2 in mouse embryonic midbrain and diencephalic explants, and repress Otx2 in mesencephalic explants(Liu et al., 1999). This proposed action of Fgf8 triggering key regulators of MHB development, which then create ectopic MHB identity, is presumably different from the normal steps of MHB development because early expression of these markers is independent of Fgf8. These events can be considered as a shortcut in the developmental program, leading to recapitulation of the MHB cascade. Upon re-initiation of key regulators of the MHB cascade, it is likely that the mutual effects of regulatory genes create secondary genetic events that then stabilize the identity of the tissue, as is presumed to happen during the maintenance phase of MHB development.

During normal development, repressive genetic interactions between Otx2 and Gbx2 (Gbx1 in zebrafish) seem to determine the caudal limit of the tectum and the position of the molecular MHB (Broccoli et al., 1999; Millet et al.,1999; Katahira et al.,2000; Martinez-Barbera et al.,2001; Rhinn et al.,2003) (M. Rhinn and M.B., unpublished). In vertebrates, both Fgf8 and Gbx2 have a repressive activity on Otx2expression in the metencephalic alar plate (Liu and Joyner, 1999; Martinez et al., 1999; Millet et al., 1999; Tour et al., 2002). In addition, zebrafish gbx2 is fully dependent on Fgf8 during the maintenance period (Rhinn et al.,2003). In ace mutants, one of the key determinants of the MHB position, otx2, expands at the expense of the fgf8 and gbx1 (Gbx2 in amniotes) domain. It is therefore conceivable that upon displaying functional Fgf8 to otx2-expressing acemutant tectal cells, the repressive genetic interactions(Martinez-Barbera et al.,2001) playing a role during normal MHB maintenance are re-established, accounting for the reversal of the tectal morphology and identity on the operated side in ace embryos. Conversely, in ace mutants where functional Fgf8 signaling is absent(Reifers et al., 1998; Araki and Brand, 2001), otx2 transcripts are not repressed, causing a strong similarity to rostral mesencephalic regions experiencing putatively low or no Fgf8 signal as a function of the distance from the emitting source. Consequently, the mesencephalic/tectal differentiation program is executed in place of the metencephalic alar plate in ace mutants. Our observations are in good agreement with the results of Otx2 misexpression studies, where the ectopic appearance of Otx2 changes the fate of the metencephalic alar plate to a more rostral, tectal fate(Broccoli et al., 1999; Katahira et al., 2000). Taken together, the transformation process that takes place at the expense of the isthmic and cerebellar primordia can be viewed as an imbalance between mutual repressive effects of the MHB cascade genes, caused by lack of the orchestrating Fgf8 signal. As a consequence of the lack of Fgf8 function, the self-maintenance of isthmic and cerebellar primordial cells is impaired resulting in a rostralization of gene expression profiles, and transformation of the isthmic and cerebellar primordium in ace mutants. In agreement with recent observations (Tallafußand Bally-Cuif, 2003), our results suggest a role for Fgf8 in protecting the isthmic and metencephalic precursors from acquiring more rostral identities, and in self-maintaining the MHB identity.

The embryonic brain seems to be composed of sequential modules that represent histogenetic fields specified by position-dependent expression of patterning genes (reviewed by Redies and Puelles, 2001). This early, embryonic modularity of the nervous system is later on translated into functional modularity. Furthermore, the suggested modular organization of the early brain allows for adaptive modification during evolution (Redies and Puelles, 2001). The modularity of the developing brainstem structures allows for spatial and temporal changes at early stages of development, for example, transformation of one particular structure to another one. During early stages of ontogenesis, such plasticity of the developing CNS makes escape from an immediate developmental arrest possible in CNS mutant embryos. Cases of transformation and their mechanisms are well documented during insect development (Sato and Denell, 1985; Klingensmith et al., 1989), and during development of the vertebrate caudal hindbrain, where the modular organization is evident owing to the rhombomeres (McKay et al., 1994; Gavalas et al., 1998; Popperl et al., 2000; Wiellette and Sive, 2003)(reviewed by Lumsden and Krumlauf,1996). The rostralization, transformation and reversal of the abnormalities by bead implantation in ace mutants reveal the plasticity of the developing upper brainstem. An interesting question is therefore whether the observed transformation of the isthmic region hints at a modular organization of this region as well. Anatomically, a modular organization of the midbrain and the isthmic region has not yet been revealed(Redies and Puelles, 2001). Experiments performed in chick reveal the unique developmental fate of r1,being the only rhombomere in which no Hox genes are expressed as consequence of Fgf8 action in the anterior hindbrain. Fgf8 acts to set aside the territory from which the cerebellum will eventually develop through restriction of Hoxa2 expression in r1 (Irving and Mason, 2000). Visualizing the formation of a new otx2/epha4a interface in ace mutants reveals a bipartite organization of r1, where the rostral part of r1 acquires a new,tectal identity. The modular plasticity also applies to more anterior parts of the mesencephalon and the diencephalon as seen, supported by bead implantation and electroporation experiments, both in chick(Martinez et al., 1999; Sato et al., 2001) and fish(present study). We demonstrated that functional Fgf8 is able to both revert the transformation of the isthmic and cerebellar primordia in mutant embryos and induce the diencephalon and mesencephalon to undergo transformation in wild-type embryos. Taken together, this reveals an extensive plasticity of the isthmic and r1 region, suggesting that transformation of cell fates seems to be the most economic way of restructuring the MHB territory in the absence of functional organizer activity in ace mutants.

The authors thank Florian Raible, Gerlinde Reim and Lilla Farkas for critical reading of the manuscript, and Naomi Foster for linguistic suggestions. This work was supported by the European Union(QLG3-CT-2000-02310), the Deutsche Forschungsgemeinschaft, and the Max-Planck Society.