Tracing of her5 progeny in zebrafish transgenics reveals the dynamics of midbrain-hindbrain neurogenesis and maintenance

Building of the vertebrate embryonic brain is a progressive process that involves a number of consecutive steps controlling patterning and neurogenesis events. Both processes respond to phases of induction and refinement, during which the positional identity and differentiation status of neural cells are specified, maintained or modified in a dynamically controlled manner. Unraveling the dynamics of neural patterning and neurogenesis are crucial steps in our understanding of brain development. Indeed, it will highlight the potentialities of given neural territories, thus revealing how their fate and differentiation are progressively restricted in vivo.

The midbrain-hindbrain (MH) domain of the embryonic neural tube displays extensive plasticity linked to specific ontogenic properties that make it an important model to study developmental dynamics(Martinez, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). The MH can be morphologically identified at early somitogenesis stages as comprising the mesencephalic vesicle and the first rhombencephalic vesicle (or metencephalon) (Fig. 1); the latter- also called `rhombomere A' in the chicken embryo(Vaage, 1969) - will later subdivide into rhombomeres (r) 1 and 2. Detailed fate map analyses in avian embryos demonstrated that the mesencephalon generates all midbrain structures,i.e. essentially an alar visual center, the tectum and a basal tegmentum,containing cranial motorneuron III (Marin and Puelles, 1994; Martinez and Alvarado-Mallart, 1989). In addition, the caudal third of the alar mesencephalic domain contributes to the dorsomedial part of the cerebellar plate (Hallonet and Le Douarin,1990; Hallonet et al.,1993; Martinez and Alvarado-Mallart, 1989), while the alar domain of r1 will give rise to remaining, lateral cerebellar structures(Wingate and Hatten, 1999)(Fig. 1). Finally, the basal r1 territory will generate the pons, of which a prominent output is cranial motorneuron IV. These distinct fates are prefigured by molecular gradients in the expression of MH genes such as engrailed-2/3 or ephrins(Martinez, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). MH structures, although physically and functionally distinct, develop in a concerted fashion. Their growth and patterning is dependent upon and coordinated by an organizing center [the `isthmic organizer' (IsO) or`isthmus'] located at the midbrain-hindbrain boundary (MHB)(Martinez, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001)(Fig. 1). Among the factors that likely mediate IsO activity are the secreted proteins Fgf8 and Wnt1,expressed on either side of the MHB. Accordingly, genetic analyses in the mouse, chicken and zebrafish demonstrate that a positive crossregulatory loop between the expression of IsO markers, of Pax2/5/8- and of Engrailed-family members is involved, at somitogenesis stages, in the stabilization of identities surrounding the MHB (Martinez,2001; Rhinn and Brand,2001; Wurst and Bally-Cuif,2001).

Transplantation studies further pointed to the remarkable plasticity of MH identities, the regionalisation of which becomes fixed only at a late stage. For example, in the avian embryo, midbrain AP polarity can be regulated until at least 12-13 somites: at that stage, it is corrected following an experimental rotation of the mesencephalic vesicle in ovo(Marin and Puelles, 1994), or is reorganized around ectopic transplants of MHB-containing tissue(Alvarado-Mallart et al., 1990; Gardner and Barald, 1991; Martinez et al., 1991; Nakamura et al., 1988) or around ectopic foci of Fgf8 expression(Crossley et al., 1996; Irving and Mason, 2000; Lee et al., 1997; Martinez et al., 1999). At the same stage, MH identity can also be changed into a diencephalic or more posterior hindbrain specification in misexpression experiments of diencephalic(Pax6) (Matsunaga et al.,2000a) or r2 (Hoxa2) (Irving and Mason, 2000) factors. Further insight into the potentialities of the MH domain will be provided by the analysis of mouse or zebrafish mutants deficient in IsO activity. The zebrafish mutants for Pax2.1(no-isthmus, noi) and Fgf8 (acerebellar, ace) functions are of particular interest, because in these backgrounds molecular MH markers are initially properly induced, but are not maintained(Brand et al., 1996; Lun and Brand, 1998; Reifers et al., 1998). During somitogenesis, strong noi alleles progressively loose the tectum,isthmus and cerebellum (Brand et al.,1996; Lun and Brand,1998). Similarly, ace mutants progressively lack the isthmus and cerebellum (Reifers et al.,1998); they only maintain tectal structures, which express low levels of Eng, ephrin-A5a and ephrin-A2, suggesting that they are of anterior identity (Brand et al., 1996; Picker et al.,1999). Currently, several (non-exclusive) interpretations can account for the loss of MH structures in noi and acemutants, among which are the death of MH precursors, their conversion to alternative fates (still to be determined) or their decreased proliferation. Understanding the fate of MH precursor cells in these backgrounds would reveal the initial potentialities of the MH anlage and would clarify the role of the IsO on cell fate.

The MH domain is also characterized by a striking profile of neurogenesis,where neuronal differentiation in the immediate vicinity of the MHB (the so-called `intervening zone', IZ) is much delayed compared to other domains of the neural tube (Bally-Cuif et al.,1993; Palmgren,1921; Vaage, 1969; Wullimann and Knipp, 2000)(Fig. 1). IZ formation is permitted by an active process of neurogenesis inhibition at the MHB(Geling et al., 2003) and, in zebrafish, the bHLH E(spl)-like factor Her5(Müller et al., 1996) was identified as the crucial element both necessary and sufficient for the formation of the basal IZ domain (Geling et al., 2003). The IZ plays a crucial role in controlling the extent of MH neurogenesis over time.

Understanding the dynamics of MH regional specification and neurogenesis are thus important issues as sustained MH plasticity correlates with the development of distinct and organized (1) MH derivatives and (2) neurogenesis domains. To approach this question, we chose to focus on the regulation of her5 expression. Two main reasons motivated our choice. First, her5 is the earliest known marker of the MH area(Bally-Cuif et al., 2000; Müller et al., 1996), and as such is the best candidate to label most MH precursors from the moment they are induced within the neural plate. If this proves true, tracing the descendants of cells expressing her5 at its onset thus should provide the best available means of assessing the fate of MH precursors in vivo. Second, because her5 expression within the IZ crucially controls the neurogenesis process, looking at the regulation of her5 expression should permit the appreciation of the dynamics of MH neurogenesis progression.

We report here the construction of zebrafish embryos where a stable reporter labels all descendents of her5-expressing cells. To maximize our chances of isolating all her5 regulatory elements, we used in vitro homologous recombination (ET-cloning)(Muyrers et al., 2000; Muyrers et al., 1999; Zhang et al., 1998) to introduce an egfp reporter cDNA at the her5 locus in a PAC containing more than 40 kb of her5 upstream sequence. We demonstrate in several independent lines that gfp expression in transgenic embryos carrying the recombined her5PAC:egfp construct faithfully reproduces her5 transcription at all stages, including the earliest step of her5 induction. Using the stability of GFP protein as a marker for the descendants of her5-expressing cells, we first demonstrate that the earliest her5-expression domain at gastrulation encompasses and thus is the first known marker of the whole MH anlage. By comparing the distribution of her5 RNA and GFP protein, we reveal a dynamic restriction of her5 expression to the MHB over time, and propose that this phenomenon permits the progression of neurogenesis in a converging manner towards the MHB during MH development. Finally, we use GFP to follow her5 progeny in the noi and acebackgrounds. We demonstrate that MH precursor cells are maintained but express alternative identities in noi and ace, albeit with striking differences between the two mutant contexts. Our results suggest a model for the progressive restriction of potentialities of MH precursors over time, and the respective roles of Pax2.1 and Fgf8 in this process.

Embryos were obtained from natural spawning of AB wild-type or transgenic fish, ace^ti282a or noi^tu29a adults(Brand et al., 1996); they were raised and staged according to Kimmel et al.(Kimmel et al., 1995).

Two independent PACs containing the genomic her5 locus(BUSMP706P0356Q2, BUSMP706H15152Q2) were isolated by PCR from pools of library 706 (RZPD, Berlin) using the following primers: her5 upstream 5′TAGTAGACCTAGCTGGTCTTTTCAGTCTTTGGAGAGC3′, her5 reverse 5′TAAAAAGGGCACGCACAGAGGAGAGTGATGAGGATGT3′, with a 59°C annealing temperature and 30 amplification cycles, producing a specific amplification product of 450 bp. PAC DNA was prepared according to the Qiagen Large Construct kit protocol. Genomic inserts are flanked by NotI sites; digestion with NotI followed by pulse field gel electrophoresis (PFGE) revealed that the inserts of both PACs were above 100 kb. Further restriction analyses and Southern blotting revealed that PAC BUSMP706H15152Q2 contained more than 40 kb of upstream her5 sequence;this PAC was chosen for further experiments. The genomic structure of her5 (Fig. 2A) was determined by PAC sequencing, and was verified on the endogenous her5locus by PCR amplification and sequencing of genomic DNA.

ET cloning was based on the protocol provided by Stewart, available on the ET cloning web page http://www.heidelberg.de/ExternalInfo/stewart/ETprotocols.html.

pEGFP-1 (Clontech); pSV40/Zeo (Invitrogen); pGlZl3_3 (modified pEGFP-1 with a loxP-flanked Zeo-cassette in AflII-site, see below); pGETrec, carrying arabinose-inducible recE gene(Narayanan et al., 1999); p705-Cre (Buchholz et al.,1996); and her5-containing PAC (pCYPAC2nbackbone) with a total of about 100 kb genomic insert and at least 40 kb upstream region of her5), further called her5PAC.

pEGFP-1 was digested with AflII, and an insert containing loxP and the restriction enzyme site NheI (produced by oligonucleotide annealing) was inserted at this site. Similarly, a loxP-NheI was introduced into the vector pSV40/Zeoafter restriction cutting with BamHI. pSV40/Zeo:loxP-NheI was further cut with NheI to release the NheI fragment containing full length of Zeo^R and loxP, which was inserted into pEGF:loxP-NheI open at NheI. This produced pEGFP:loxP-Zeo^R-loxP, further referred to as pGlZl3_3. All plasmids containing Zeo^R were grown in INFαF' cells.

The fragment for homologous recombination was prepared by PCR using the following primers. Primer ET2: 48 nucleotides specific to the 5′-sequence of her5 exon 2(Fig. 2A, fragment b) and 21 nucleotides specific to pGlZl3_3 (underlined) (sequence: 5′GTC CCC AAG CCT CTC ATG GAG AAA AGG AGG AGA GAT CGC ATT AAT CAA GTC GCC ACC ATG GTG AGC AAG3′). Primer ET1: 47 nucleotides specific to the 3′-sequence of her5 exon 2(Fig. 2A, fragment b) and 22 nucleotides specific to pGlZl3_3 (underlined) (sequence: 5′CTC ATT GTT TGT GTT CTC AAG TAA AAG CAT TCT CAA GGT TTC TAG GCT TAA CGC TTA CAA TTT ACG CCT3′).

Oligonucleotides ET1 and ET2 were resuspended in water and purified as follows. To 100 μl, 12 μl 3 M Sodium-Acetate (pH 7.5) and 120 μl phenol were added, vortexed and centrifuged for 3 minutes. Then 360 μl Ethanol was added, and the mix was placed 10 seconds at 80°C, washed once with 75% ethanol, dried and finally dissolved in 100 μl water.

PCR amplification of the fragment her5a-EGFP:loxP-Zeo^R-loxP-her5b: Template her5PACDNA was denatured for 2 minutes at 94°C, followed by two cycles of denaturation at 94°C for 40 seconds. A first, annealing was performed at 62°C for 30 seconds, with extension at 72°C for 2 minutes. This was followed by 35 amplification cycles with denaturation at 94°C for 40 seconds, annealing at 58°C for 30 seconds, extension at 72°C for 2 minutes. The reaction was stopped by a final extension at 72°C for 10 minutes and cooled at 4°C. The expected 2 kb amplification product was purified using the QIA gel extraction kit (Qiagen) as recommended, and eluted in 50 μl water.

The bacterial host cells DH10B containing her5PAC were transformed with pGETrec and prepared for the recombination with the linear her5a-EGFP:loxP-Zeo^R-loxP-her5b fragment as follows:starting from an overnight culture, the cells were grown at 37°C for 90 minutes (to OD₆₀₀=0.2-0.3) with shaking. L-arabinose was added to the culture to a final concentration of 0.2% and the culture was grown further until OD₆₀₀=0.5 was reached. The cells were then prepared as electro-competent as described in http://www.heidelberg.de/ExternalInfo/stewart/ETprotocols.html. Electroporation of 120 ng of her5a-EGFP:loxP-Zeo^R-loxP-her5b fragment was performed with 2.5 kV pulses and 25 μF in 100 μl, induced with 0.2% L-arabinose at 37°C for 90 minutes before harvesting and plating twice for selection.

Competent cells carrying the recombined her5PAC were transformed with p705-Cre using standard protocols. p705 is based on the pSC101 temperature-sensitive origin, which maintains a low copy number and replicates at 30°C but not at 40°C. Furthermore, Cre is expressed from the lambdaPR promoter weakly at 30°C and strongly at 37°C. Finally, these plasmids are lost from cells if incubated at temperatures above 37°C. Thus, after transformation the cells were incubated for 2 days at 30°C, followed by 1 day's incubation at 40°C to give a transient burst of Cre expression after which the plasmids will be eliminated from the cell. The cells were then further grown for day at 37°C, transferred once and finally tested by PCR for excision of the loxPZeo^R-loxP cassette, generating her5PAC:egpf. Because of the presence of a NotI site 3′ to the egfpgene, digestion of her5PAC:egfp with NotI generated two fragments of 45 and 60 kb in addition to the vector backbone. PFGE and Southern blotting with a her5 probe identified the 45 kb fragment as containing the coding her5 sequence, thus her5PAC:egfpcontains more than 40 kb upstream her5 sequence driving egfpexpression.

The fragment containing 3650bp of her5 upstream sequence was obtained by digestion of her5PAC:egfp with NotI +BglII followed by pulse field gel electrophoresis, identification by Southern blotting with a probe covering the her5 5′ region, and gel purification (Qiagen Gel extraction kit). The fragment was subcloned into pBS(SK) for amplification, and was repurified by digestion and gel extraction before injection. All other constructs were prepared as PCR fragments from her5PAC:egfp and purified using the Qiagen PCR purification kit. All fragments were eluted in H₂O (Ambion).

her5PAC:eGFP DNA was isolated using the Qiagen Large Construct Kit, eluted in H₂O and injected (in circular form) into fertilized eggs at the one-cell stage at a concentration of 50 ng/μl. All other constructs were injected as linear fragments at the same concentration. Injected embryos were raised to adulthood and mated to wild-type adults. F1 embryos expressing eGFP were then sorted-out, raised and crossed to wild-type fish to establish the lines. We obtained integration and expression in three from 600 injected fish for her5PAC:egfp and in average three from 50 injected fish for the other fragments. All results presented in this work were verified over at least three generations.

In situ hybridisation and immunocytochemistry were carried out according to standard protocols (Hauptmann and Gerster,1994). The following in situ antisense RNA probes were used: her5 (Müller,1996; Thisse et al.,1993); egfp (Clontech); pax6(Krauss et al., 1991); fgfr3 (Sleptsova-Friedrich et al., 2002); otx2 (Li et al., 1994b); hoxa2(Prince et al., 1998); and krx20 (Oxtoby and Jowett,1993).

For immunocytochemistry, the following antibodies were used: mouse anti-GFP`JL-8' (Chemicon) used at a dilution of 1/100; mouse anti-invected 4D9 (DHSB),which recognises all zebrafish Eng proteins, used at a dilution of 1/8; and rabbit anti-phosphohistone H3 (Upstate Biotechnology, no.06-570) used at a dilution of 1/200. They were revealed using goat-anti-mouse-HRP or goat-anti-rabbit-HRP (Chemicon) (dilution 1/200) followed by DAB/H₂O₂ staining, or goat-anti-mouse-FITC (Dianova)(dilution 1/200), as appropriate. Double in situ hybridisation and immunocytochemistry staining on transgenic embryos were performed as follows:whole-mount embryos were first processed for in situ hybridisation, then cryostat-sectioned at 8 μm thickness and the sections were subjected to immunocytochemistry following standard protocols. In Fig. 7K-M, immunocytochemical detection was performed after in situ hybridisation on whole-mount specimen. Embryos were scored and photographed under a Zeiss SV11 stereomicroscope or a Zeiss Axioplan photomicroscope.

her5PAC:egfp transgenic embryos were injected at the one-cell stage with 7 mg/ml DMNB-caged fluorescein (10 kDa, Molecular Probes), and were left to develop in the dark. When GFP protein first became visible (at 95%epiboly), small groups of four or five cells located within the most anterior or most posterior rows of GFP-positive cells (see yellow and red dots on Fig. 5A) were UV-irradiated for 2 minutes using DAPI illumination and a 0.1 mm pinhole under a 63× water objective, according to Kozlowski and Weinberg(Kozlowski and Weinberg,2000). Embryos were fixed at 24 hpf and uncaged fluorescein was detected by immunocytochemistry as described by Dickmeis et al.(Dickmeis et al., 2001).

For characterization of cell death, embryos were stained according to Williams and Holder (Williams and Holder,2000), with minor modifications. Briefly, embryos were incubated for 20 minutes in 5 μg/ml Acridine Orange (Sigma) in embryo medium, washed three times for 5 minutes in embryo medium and observed under fluorescence microscopy with FITC filter.

Because gene regulatory elements might be located at a distance from the transcriptional start site, we chose to search for her5 enhancers using a homologous recombination approach in large genomic fragments. Two PACs were isolated that contained the genomic her5 locus, and the PAC insert containing the longest 5′ sequence (over 40 kb, as determined from pulse field gel electrophoresis and Southern blotting) was selected. We found that the complete her5 coding sequence overlaps three exons,where exon 1 contains the transcription start site and encodes the 17 N-terminal Her5 amino acids (Geling et al., 2003). Exon 2 codes for the 32 following amino acids,comprising the basic domain, helix 1 and part of the loop domain of Her5(Fig. 2A), and exon 3 for the last 165 amino acids. We used the ET-cloning technology(Muyrers et al., 2000; Muyrers et al., 1999) to recombine the egfp cDNA in frame after amino acid 33 of Her5 (end of the basic domain) (Fig. 2A). The egfp cDNA was terminated with a stop codon and polyadenylation signal, thus translation of the recombined mRNA is stopped after a fusion protein that does not comprise the protein interaction motifs of Her5 (HLH and more C-terminal domains). We expected that this fusion protein would not interfere with the activity of other bHLH factors. In line with this prediction, we did not detect any morphological or molecular phenotype in all our transient or stable expression assays (see below, and data not shown). Three independent transgenic lines were established that carried the recombined her5 PAC (her5PAC:egfp lines). All showed an identical gfp RNA expression profile at all embryonic stages examined(data not shown). These lines will be used indiscriminately below.

At early gastrulation, her5 is transcribed in a subset of anterior endodermal precursors ('e' in Fig. 2C) (Bally-Cuif et al.,2000). Accordingly, we detected GFP expression in the pharynx at 24 hours post-fertilization (hpf) (Fig. 2D, Fig. 3) in all her5PAC:egfp embryos. These results make of her5 the earliest selective pharyngeal marker known to date, and are in line with the proposed role of endodermal Her5 activity in attributing pharyngeal fate(Bally-Cuif et al., 2000). In addition, wild-type her5 expression is initiated at the 70% epiboly stage in a V-shaped neuroectodermal domain ('MH' in Fig. 2C) that was fate-mapped to the midbrain at 90% epiboly(Müller et al., 1996). Accordingly, strong GFP expression was found in the MH domain (Figs 2D and 3).

The early control of MH her5 expression involves two distinct phases: expression is initiated at 70% epiboly by currently unknown regulators, and is maintained and refined after the five-somite stage by the MH regulatory loop (Lun and Brand,1998; Reifers et al.,1998). To determine whether egfp transcription was a faithful reporter of her5 expression, we performed double in situ hybridisation experiments with gfp and her5 probes on her5PAC:egfp embryos between 60% epiboly and 24 hpf. her5PAC-driven gfp transcription faithfully reproduced expression of endogenous her5 at all embryonic stages tested, both in its onset and spatial extent (Fig. 3A,C,D, and data not shown). In particular, gfpexpression was initiated at 70% epiboly within the neural plate and maintained in the MH domain thereafter, demonstrating that both the initiation and maintenance phases of her5 transcription are recapitulated by expression of the transgene. Together, these observations demonstrate that the her5PAC construct comprises all the regulatory elements that control endogenous her5 expression at embryonic stages.

To narrow down the sequences directing MH and/or endodermal expression of her5, we performed a deletion analysis series of the her5PAC:egfp transgene. A comprehensive series of reporter constructs of varying length encoding the Her5-eGFP fusion protein and comprising between 60 and 3650 bp upstream of the her5 transcriptional start site(Geling et al., 2003) were amplified by PCR from her5PAC:egfp and tested in transient or transgenic assays (black or red lines in Fig. 2B, respectively). In the latter case, at least two independent lines were established for each construct. Transient assays generally produced ectopic expression sites when compared with transgenic analyses of the same fragments; however, comparison of a sufficient number of injected embryos (n>30) allowed to reliably predict the reporter expression profile (not shown). All results are summarized in Fig. 2B,D. In summary, we observed that all fragments containing 240 bp or more of upstream sequence lead to non-neural expression(Fig. 2B). Transgenic lines established with 770 bp upstream region (-0.7her5:egfp) faithfully recapitulated her5 endodermal expression, with similar onset and anteroposterior extent (Fig. 2D-H and data not shown). These results locate the her5endodermal enhancer to the first upstream 770 bp, the first her5intron (contained in all constructs) or a combination of both.

We next examined the regulatory elements controlling neural expression of her5. We found that all constructs containing more than 770 bp of upstream sequence directed, in addition to endodermal expression, GFP fluorescence within the neural tube (Fig. 2D-G). However, MH selectivity in stable assays was only achieved with upstream sequences of 3.4 kb or more (-3.4her5:egfp lines)(Fig. 2E), whereas shorter elements triggered GFP expression over the MH as well as fore- and hindbrain(e.g. -1.7her5:egfp lines, Fig. 2F,G). Double in situ hybridisation experiments with gfpand her5 probes demonstrated that gfp transcription in -3.4her5:egfp transgenics faithfully reproduces expression of endogenous her5, including its induction and maintenance phases(Fig. 3B,E,F, and data not shown). Thus, all regulatory elements driving correct MH her5 both in time and space appear contained within the -3.4her5:egfp construct. Together, our analysis of the her5 enhancer demonstrates that spatially distinct and dissociable elements drive endodermal and MH expression of her5 during embryogenesis.

her5 expression in endodermal precursors is initiated at 30%epiboly, and switched off at 90% epiboly(Bally-Cuif et al., 2000). We could detect GFP protein in the pharynx until 26-30 hpf (e.g. see Fig. 2D,F,H), thus GFP protein is stable for ∼18-20 hours in this tissue in our lines. We reached a similar conclusion for GFP stability in the neural tube, where posterior her5-positive cells at 75% epiboly rapidly switch off her5expression and give rise to metencephalic derivatives that loose GFP protein around 24 hpf (see below). Thus, the GFP protein profile observed at a given time corresponds to all descendants of the cells that expressed gfpunder her5 regulatory elements between 18-20 and a few hours before the moment of analysis. The stability of the GFP protein in her5PAC:egfp embryos thus offers the unique opportunity of following the fate of her5-expressing cells, from the onset of endogenous her5 expression and throughout embryogenesis.

The MH anlage is composed of precursors for the midbrain, isthmus, r1 and r2 (Fig. 1). These domains are together characterized by the expression of Eng2 proteins at somitogenesis stages, but an early molecular marker of the entire presumptive MH remains to be identified. The onset of her5 expression within the neural plate is at 70% epiboly, and GFP protein becomes visible in this location around 90%epiboly (not shown). To determine the fate of these early her5-expressing cells, we performed a detailed spatiotemporal analysis of GFP distribution by fluorescence microscopy on live embryos and immunocytochemistry on whole-mount or sectioned specimen(Fig. 4A-J). When necessary,GFP protein distribution was compared with the expression of diagnostic molecular markers for diencephalic (see Fig. 8A,E,G,I) or hindbrain domains (Fig. 4K-Q).

The morphological constriction marking the midbrain-hindbrain boundary(MHB) becomes visible from the 10-12-somite stage onwards, and is prominent by 20 somites (Fig. 4E, arrow). At the 12- and 20-somite stages, GFP protein clearly distributes over the entire midbrain as well as posterior to the MHB(Fig. 4A,C,E), and a cross-section at the MHB level demonstrates that all neural tube cells are stained (Fig. 4B, top). Whole-mount analyses and lateral sections further reveal intense GFP staining in neural crests streams that exit the midbrain area towards anterior and ventral (Fig. 4C,D). At 25 somites and later, the isthmic fold has formed and the cerebellar anlage is discernible. GFP protein is detected in the midbrain, isthmus, cerebellar fold and pons (Fig. 4F-H, see also Fig. 6A,B). The intensity of GFP staining in the metencephalon is, however, weak compared with midbrain expression, and becomes undetectable after 26 hpf(Fig. 4I). GFP expression at 26 hpf remains prominent in the midbrain, albeit with a clear caudorostral decreasing gradient. After 30 hpf, GFP protein is maintained only at the MHB(Fig. 4J), in a profile reminiscent of late her5 RNA transcription (see Fig. 6C).

To position precisely the spatial limits of GFP protein distribution, we compared its anterior and posterior borders with the expression of diagnostic markers. pax6.1, the zebrafish ortholog of murine and chicken Pax6, is expressed within the anterior alar plate with a posterior limit at the dimesencephalic boundary (Li et al., 1994a; Macdonald et al., 1995). From the onset of pax6.1 expression (12 somites) until at least the 30-somite stage, we found that GFP- and pax6.1-positive cells precisely abut each other at the di-mesencephalic border (Fig. 8A,C,E,G). The posterior extent of GFP protein distribution was determined by comparison with the expression of hoxa2 from 10 somites onwards, when hoxa2 exhibits a sharp anterior limit of expression at the r1/r2 boundary (Prince et al.,1998), and with the expression of krox20 that marks r3 and r5 (Oxtoby and Jowett,1993). At 10 somites and subsequent stages until at least 30 somites, GFP distribution overlaps r2 (Fig. 4K,M,O,Q). A few GFP-positive cells can also transiently be found within r3 and r4 at 10 somites (Fig. 4L). At this stage, cells in r3 co-express GFP protein and krox20 (Fig. 4N,yellow arrows). However the contribution to r3 and r4 is marginal and no longer detectable at 20 somites (Fig. 4P).

To ascertain that her5 expression at its onset within the neural plate comprises all MH precursors, we determined whether the spatial organization of the earliest her5-expressing cells prefigures the later distribution of MH cells along the AP axis. To this aim, we fate mapped the anterior and posterior extremities of the her5 domain at 70%epiboly. To reliably identify this domain, we relied on its giving rise to the earliest detectable GFP expression using fluorescence microscopy in her5PAC:egfp embryos, at 90-95% epiboly. Thus we activated caged-fluorescein in small groups of four or five GFP-positive cells located at the edges of the GFP domain in transgenic embryos at 95% epiboly(Fig. 5A), and followed these cells at 24 hpf (Fig. 5B-E). We found that anterior activated cells always gave rise to cell clones distributing within the anterior midbrain(Fig. 5B,C) (n=4),while posterior activated cells populate r2(Fig. 5D,E) (n=5). Thus, the anterior and posterior extremities of the earliest her5-expressing domain at 70% epiboly prefigure the corresponding extremities of the later MH.

Together, our findings demonstrate that the early neural expression of her5 is a marker of the entire MH anlage, and it appears as the earliest MH marker known to date. Furthermore, this early domain displays some degree of ordered cell distribution, such that its anterior and posterior limits contain precursors for the anterior and posterior extremities of the later MH. Specifically, at 70% epiboly, anterior her5-positive cells abut and exclude the diencephalon anlage, while posterior cells comprise precursors for r1 and r2.

Her5 crucially controls MH neurogenesis(Geling et al., 2003), making it important to analyze the regulation of its expression. In 30-somite her5PAC:egfp embryos, we observed a dramatic difference in the AP extent of her5 transcription and GFP protein distribution(Fig. 6A-C). This observation suggests that MH precursors loose her5 expression upon division, such that the her5-positive territory shrinks from a domain covering the entire MH anlage at early gastrulation, to be maintained at 30 somites at the MHB only. To confirm this hypothesis, and assess the progression of this phenomenon in time and space, we conducted a precise comparison of her5 RNA and GFP protein distributions between 90% epiboly (first stage where GFP protein becomes detectable in the MH domain) and 24 hpf. To this aim, double in situ hybridisation and immunocytochemical detection was performed on whole-mount embryos or serial sagittal sections (minimum three embryos per stage). At 90% epiboly and until the one- to two-somite stage, the anterior borders of her5 RNA and GFP protein expression were coincident (Fig. 6D, and data not shown). However their posterior limits differed by approximately one or two cell rows (Fig. 6D, and data no shown). Thus, between the onset of her5 expression in the neural plate (70% epiboly) and 90% epiboly, her5 transcription becomes restricted of a few cell rows posteriorly, although it is maintained in all its progeny cells anteriorly (Fig. 6P, parts a,b). At three somites, her5 transcripts distribute over approximately eight cell rows along the AP axis, while GFP protein covers 15-18 rows (Fig. 6E,F). From this stage onwards, prominent differences in the AP extent of her5 RNA and GFP protein are detectable posteriorly but also anteriorly, on the lateral and basal domains of the midbrain(Fig. 6E, black arrows). By contrast, her5 expression still mostly matches GFP staining along the dorsal midline of the neural tube (Fig. 6E, blue arrow). Similar observations can be made until the 12- to 14-somite stage (Fig. 6G,H). At 16 somites, the dorsal expression of her5 dramatically regresses and her5 expression is restricted to a band of 4-6 cell rows across the entire DV extent of the neural tube (Fig. 6I). At this stage, MH cells have further divided as GFP protein extent now covers approximately 27-30 rows along AP(Fig. 6J). This progression is ongoing at least until the 30-somite stage, when GFP protein extends over 45-50 rows, against 3-5 rows for her5 RNA(Fig. 4H, Fig. 6K,L).

To ascertain the directionality of the progressive restriction of her5 expression in MH precursors, we revealed her5 RNA and GFP protein on single sagittal sections in double fluorescence experiments(Fig. 6M-O). Such stainings unambiguously located the final her5 expression domain to the center of the GFP-positive domain, confirming that her5 expression is lost both anteriorly and posteriorly upon cell divisions. Several other MH markers,e.g. pax2.1, eng1, wnt1 and fgf8, display an expression profile that globally compares in extent with her5 at early and late stages, and GFP distribution in pax2.1:gfp transgenics(Picker et al., 2002) and wnt1:gfp-injected embryos (Lekven et al., 2003) suggests that the expression of these genes follow a restriction similar to her5 over time.

We conclude from these observations that (1) her5 expression within the MH domain is subject to a highly dynamic regulation and is progressively lost upon cell divisions between 70% epiboly and 24 hpf(Fig. 6P), (2) the restriction of her5 expression occurs in a centripetal manner towards the MHB,and (3) it follows a precise spatial sequence: it is initiated posteriorly (in the future metencephalon) before affecting the basolateral and finally the dorsal mesencephalic areas. her5 expression, at least in the basal plate, is always adjacent to neurogenesis sites(Geling et al., 2003). Thus,our observations imply that neurogenesis within the MH domain is also a spatially dynamic process, and converges towards the MHB over time (red arrows in Fig. 6P, part d).

We next used the stability of the GFP protein to study the potentialities of MH precursors in terms of their spatial identity. MH precursors remain plastic until late stages, and the choice and reinforcement of their specification are incompletely understood. We addressed the role of Pax2.1 and Fgf8 in these processes, by studying GFP distribution in noi and ace mutants, where the fate of the presumptive MH anlage is unknown.

We first ascertained that GFP protein could be used as a reliable marker of MH fate in noi and ace. To this aim, we verified that gfp transcription faithfully recapitulated her5 expression in these mutant contexts. Double in situ hybridisations with the her5and gfp probes were performed on transgenic mutant embryos, and demonstrated an identical initiation (not shown) and later downregulation of her5 and gfp transcription in these backgrounds(Fig. 7A-D). Near-complete downregulation of gfp expression was observable at 24 hpf in her5PAC:egfp;ace embryos (Fig. 7B) and at the 10-somite stage in her5PAC:egfp;noi(Fig. 7D), like expression of endogenous her5. We conclude that the distribution of GFP protein can be used as a faithful tracer of MH precursors in the ace and noi contexts.

Live observation of 24 hour-old transgenic mutant embryos first revealed that a significant number of GFP-positive fluorescent cells was maintained at that stage in both the ace and noi backgrounds(Fig. 7E-G). These cells distribute over an AP territory that approaches wild-type size (compare Fig. 7F,G with 7E), and throughout the entire DV extent of the neural tube. No signs of aberrant cell migration were apparent at any stage, i.e. no patches of unstained cells were observed within the GFP-positive domain, and conversely, no patches of positive cells were found outside the main GFP-positive domain. In addition,at 15 somites, no difference was observed in the rate of cell death (as revealed with Acridine Orange) (25±12 cells in wild-type, 26±9 cells in ace, 30±5 cells in noi; n=20)(Fig. 7H-J) and cell proliferation (anti-phosphohistone H3) (61±9 cells in wild-type,55±6 cells in ace; 59±8 cells in noi; n=10) (Fig. 7K-M) in that area between wild-type, ace and noi embryos. Together,these observations suggest that the normal complement of MH precursors is present in the mutants at least until the 15-somite stage. The expression of MHB markers (Brand et al.,1996; Lun and Brand,1998; Reifers et al.,1998), however, and of basal MH derivatives such as the III and IV cranial nerves (Fig. 7E-G,insets), is absent. Together, these results suggest that MH precursors remained in place but, at least in part, display alternative identities in the mutants. We used the co-detection of GFP protein and diagnostic molecular markers expression on single sections to verify this hypothesis. We demonstrate below that MH progeny cells display aberrant specification in noi and ace as early as at 15 somites, when, as described above, the survival, proliferation and migration of MH cells do not show signs of perturbation.

The anterior limit of GFP protein abuts at all stages the caudal border of pax6.1 expression (Fig. 8A,C,E,G), a marker for the posterior diencephalic alar plate. Strikingly, however, ace mutants showed a significant overlap between these two patterns at the 30-somite stage(Fig. 8B,B′), where a large number of cells in the anterior part of the GFP-positive territory co-expressed pax6.1. A transient overlap in the expression of Pax6 and En has been documented in chicken(Matsunaga et al., 2000a),suggesting that the co-expression GFP and pax6.1 in acemight result from a failure to downregulate pax6.1 in anterior MH precursors. However, in a precise comparison of pax6.1 and GFP, as well as of pax6.1 and Eng proteins expression in zebrafish, we failed to observe an overlap of these markers at any stage(Fig. 8E,G and data not shown). Thus, the co-expression of GFP and pax6.1 in ace rather reflects aberrant pax6.1 transcription in MH precursors. A time-course experiment further revealed that GFP-positive cells in ace express a pax6.1-positive identity at least as early as the 15-somite stages (Fig. 8,compare F,F′ with E and H,H′ with G). In striking contrast to these findings, a distinct pax6.1/GFP border was maintained in noi, although pax6.1 expression appeared extended posteriorly compared with its wild-type pattern (compare Fig. 8C with 8D).

Diencephalic cells are also characterized by the expression of fgfr3 (Fig. 8I). In wild-type transgenic embryos, the GFP-positive territory abuts the caudal border of fgfr3 expression (Fig. 8I, green arrowheads), which thus shares a common posterior limit with pax6.1. As reported previously, we found that fgfr3expression extends ectopically towards caudal in ace and noi(Sleptsova-Friedrich et al.,2002). Double labeling of transgenic mutants reveals, in addition,that GFP and fgfr3 expression overlap extensively in noi,where all GFP-positive cells co-express fgfr3(Fig. 8K), at least from the 15-somite stage onwards (Fig. 8L). By contrast, the fgfr3/GFP border is maintained in the ace alar plate. Both markers overlap in the ace basal plate (Fig. 8J), however,further documenting the differential plasticity of basal and alar MH precursors (see Lun and Brand,1998; Reifers et al.,1998; Sleptsova-Friedrich et al., 2002).

Metencephalic derivatives such as the cerebellum fail to develop in both ace and noi, but the fate of metencephalic progenitors is unknown. To address this question, we relied on the expression of otx2, a marker of the fore- and midbrain, but not hindbrain territories. In ace mutants, we found that the posterior limit of otx2 expression precisely coincided with the posterior border of GFP protein distribution (Fig. 8N). Because no extensive cell death was observed in the mutants(Reifers et al., 1998)(Fig. 7I and data not shown),this result highlights that metencephalic precursors display an otx2-positive identity in the absence of Fgf8 function. By contrast,in noi mutants, the caudal border of otx2 expression appeared to be located half way through the GFP-positive domain, in a manner reminiscent of the wild-type situation(Fig. 8M,O). Thus, some AP distinctions related to ante- and post-MHB differences are maintained by the descendents of MH progenitors in noi. Posterior GFP-positive, otx2-negative cells also express fgfr3 at high levels(Fig. 8K), suggesting that they are of posterior r1 or r2 identity. However, because of the dynamic posterior limit of GFP protein distribution in the hindbrain(Fig. 4K-Q), it was not possible to follow these cells.

Together, our findings demonstrate that MH precursors display aberrant spatial identities in ace and noi, in a manner that strikingly depends on the mutant context. An interpretative summary of our results is presented in Fig. 9.

In this article, we construct transgenic tools to trace precisely the progeny of her5-expressing cells during zebrafish embryogenesis, and we use these tools in a detailed analysis of the dynamics of MH development. Our tracing of her5 progeny in wild-type and mutants leads to three important conclusions. First, we demonstrate that her5 expression at its onset defines the MH anlage, making her5 the first marker of the MH territory. Second, we show that her5 expression is progressively lost upon cell division in a spatially controlled manner towards the MHB. Because Her5 activity negatively defines neurogenesis sites(Geling et al., 2003), this result implies that MH neurogenesis is dynamically regulated and progresses towards the MHB over time. Finally, we demonstrate that MH precursors are mostly maintained but harbor alternative identities in noi and ace, and we show that these identities depend on the mutant context. Together, our findings provide models for the dynamics of MH neurogenesis and maintenance, and directly determine pax2.1 and fgf8 mutant gene activities for their effect on cell identity choices.

During embryogenesis, her5 expression follows at least three distinct phases: it is first transcribed in a subset of endodermal precursors,then induced and maintained within the presumptive MH. In addition, each phase is subject to dynamic regulation, as endodermal expression is transient(Bally-Cuif et al., 2000) and MH expression is drastically downregulated over time (this paper). Because the her5 enhancer had not been characterized and her5 expression is complex, we chose the ET-cloning in vitro recombination technology(Muyrers et al., 2000; Muyrers et al., 1999; Zhang et al., 1998) to build transgenic lines where gfp expression is driven by the complete set of her5 regulatory elements. Precise analysis of her5PAC:egfp embryos reveals that our lines indeed fully recapitulate the phases and dynamics of in vivo her5 expression. Our results confirm the power of artificial chromosome transgenesis in zebrafish to decipher the complexity of developmental gene regulation in vivo.

All early MH markers studied to date, including zebrafish her5, pax2.1,eng2, fgf8 and wnt1, follow a bi-phasic mode of regulation:their expression is induced at late gastrulation, probably by independent pathways, and maintained after the five-somite stage in a mutually interdependent process (Lun and Brand,1998; Reifers et al.,1998; Scholpp and Brand,2001). These phases correspond to distinct regulatory elements on the promoters of zebrafish pax2.1(Picker et al., 2002), mouse Pax2 (Pfeffer et al.,2002; Rowitch et al.,1999) and mouse En2(Li Song and Joyner, 2000; Song et al., 1996). Our deletion analysis (Fig. 2 and data not shown) failed to dissociate initiation and maintenance elements within the her5 enhancer, suggesting that they are closely linked and/or overlapping at the her5 locus. The `maintenance' elements of mouse En2 depend upon Pax2/5/8 binding sites(Li Song and Joyner, 2000; Song et al., 1996); those of mouse Pax2 are at least targets for auto- or crossregulation by Pax2/5/8 proteins (Pfeffer et al.,2002). her5 expression is dependent upon the presence of Pax2.1 at somitogenesis (Lun and Brand,1998; Reifers et al.,1998); however, analysis of the her5 enhancer sequence failed to reveal binding sites for this maintenance factor (A.T. and L.B.-C.,unpublished). In addition, we showed previously that her5 expression was not subject to autoregulation (Geling et al., 2003). Maintenance of her5 expression at somitogenesis thus likely involves relay factors that have yet to be identified.

A restricted subset of players involved in MH induction has been identified: the Oct-like transcription factor Spiel-ohne-Grenzen (Spg)/Pou2(Belting et al., 2001; Burgess et al., 2002) and the Btd/Sp1-like zinc finger protein Bts1(Tallafuss et al., 2001). Accordingly, Oct- and Sp1-binding sites are found on the early-acting enhancer of mouse Pax2, and at least the Oct sites are required for enhancer activity (Pfeffer et al.,2002). Similarly, we found that several Oct and Sp sites are present on the her5 MH enhancer (A.T. and L.B.-C., unpublished). The requirement for these specific sites for her5 induction remains to be directly demonstrated; suggestively, however, endogenous her5expression (Reim and Brand,2002) and her5PAC-driven gfp expression (A.T. and L.B.-C., not shown) followed the same decreased and delayed induction in spg/pou2 mutants compared with wild-type siblings. Factors restricting her5 expression to the MH anlage also remain crucial components of the MH induction process to be identified. Some of these likely bind the distal region of the her5 enhancer, as proximal domains drove unrestricted reporter expression to the anterior brain in our transgenic assays (Fig. 2F,G).

GFP protein distribution (Fig. 4) and direct mapping of the anterior and posterior extremities of the earliest her5-positive domain within the neural plate(Fig. 5) position the early anterior her5 expression border to the di-mesencephalic boundary,while the posterior border of her5 expression is more dynamic and expands, at early stages, a minor contribution into r3 and r4. GFP-positive cells found in r3 and r4 might be accounted for by a transient overlap of her5 expression with the r3/r4 anlage at gastrulation. However, at this stage, her5 is not co-expressed with hoxa-1 (later renamed hoxb1b) (A.T. and L.B.-C., unpublished), interpreted to extend to the r3/r4 boundary (Koshida et al., 1998). Thus, alternatively, the contribution of GFP-positive cells to r3 and r4 might result from the migration of metencephalic cells towards caudal, followed by an acquisition of posterior identities (as revealed, for example, by their expression of krox20)(Fig. 4N). Such migration has been documented in the chicken embryo at a later stage(Marin and Puelles, 1995). We cannot formerly exclude either possibility at this point.

Outside this marginal contribution to posterior rhombomeres, the large majority of GFP-positive cells is confined to mesencephalic (midbrain,isthmus) and metencephalic (r1, r2) derivatives. GFP expression encompasses the entire extent of the MH domain, and displays a ubiquitous distribution within this domain. Thus, our results identify her5 expression at its onset as a comprehensive marker of MH precursors. her5 expression at 90% epiboly was fate mapped to the midbrain only(Müller et al., 1996), an observation in agreement with the immediate restriction of her5expression from posterior cells between 70% and 90% epiboly(Fig. 6D), and with the identification of these posterior cells as metencephalic precursors(Fig. 5D,E). The MH domain is generally considered as an entity because its different sub-territories develop in a concerted fashion (in direct or indirect response to IsO activity), and because it is globally characterized by the expression of molecular markers (such as En2) at somitogenesis stages(Martinez, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). Our results add support to these ideas by providing the first direct molecular evidence for the definition of a MH prodomain ('pro-MH') at early developmental stages. Furthermore, they show that the AP distribution of precursors within this pro-domain displays some degree of spatial coherence as it prefigures the organization of the later MH.

The earliest her5 expression domain defines pro-MH cells although Her5 function itself does not control the acquisition or maintenance of MH identity (Geling et al.,2003). It is thus likely that (as yet unidentified) MH identity factors display an expression profile similar to her5 at gastrulation. These factors might be rapidly relayed in time by Pax2.1 and/or Eng2/3.

An important demonstration of our study is the highly dynamic regulation of her5 expression over time. Indeed her5 expression restricts from a domain covering the entire MH anlage at 70% epiboly to a few cell rows at the MHB at late somitogenesis (Fig. 6). We believe that this restriction is functionally relevant, as the spatiotemporal distribution of the Her5 protein is likely to follow very closely that of her5 mRNA. Indeed, her5 mRNA is always found directly adjacent to sites undergoing neurogenesis(Geling et al., 2003)(Fig. 6P), and Her5 protein potently inhibits neurogenesis (Geling et al., 2003).

Between 70% epiboly and late somitogenesis, the total number of her5-expressing cells remains roughly unchanged; by contrast, the number of MH cells greatly increases. This observation demonstrates that her5 expression is progressively lost upon cell divisions in a converging manner from anterior and posterior towards the MHB. Whether this progressive downregulation follows an asymmetrical mode of cell division,where her5 expression is maintained in every other progeny cell at each cellular generation, or rather results from the progression of a maturation gradient within the MH in a manner unrelated to cell cycle events,remains to be determined. This will require the tracing of single GFP-positive cells.

Our results correlatively demonstrate that primary neurogenesis converges from anterior and posterior towards the MHB over time(Fig. 5P), and suggest that neurogenesis progression is permitted by the dynamic downregulation of her5 expression (Geling et al.,2003). Along the DV axis of the neural tube, the combinatorial differentiation-promoting and differentiation-inhibiting activities of Shh and Wnt signaling, respectively, has been proposed to account for the global ventral-to-dorsal progression of neuronal maturation(Megason and McMahon, 2002). Her5 might be regarded as a counterpart to Shh and Wnt along DV, which controls the spatial order of neurogenesis progression along AP within the MH domain.

Within the MH basal plate, neuronal identity varies according to, and has been postulated to depend on, the position of the population considered relative to the MHB (Agarwala and Ragsdale,2002; Broccoli et al.,1999; Wassarman et al.,1997). For example, nMLF reticulospinal neurons lie at the anterior border of the mesencephalon, while motoneurons (of cranial nerves III and IV) are found adjacent to the MHB. Our results on her5 and neurogenesis dynamics also imply that these neurons are generated at different times, the former being an early and the latter a late neuronal type. Along this line, the combined action of the two E(spl)-like factors Hes1 and Hes3 is required for IZ maintenance in the E10.5 mouse embryo(Hirata et al., 2001), and premature neurogenesis at the MHB in Hes1^-/-;Hes3^-/- embryos is correlated with the loss of some but not all neuronal identities that normally develop around the MHB after E10.5 (Hirata et al.,2001). Whether the primary determinant of neuronal identity is the AP location of the different populations, or rather is the timing of their engagement into the differentiation process, primarily controlled by her5 restriction, becomes an important aspect of MH development to address in future studies.

A major interest of our lines is to permit the direct tracing of MH precursors in mutant or manipulated contexts. We focused here on the noi and ace mutants, where the fate of pro-MH cells is unknown (Lun and Brand, 1998; Reifers et al., 1998). Our tracings first demonstrate that, in these mutants, a large proportion of these cells are maintained but partially acquire alternative AP identities. Second,they reveal dramatic differences in the final identities of MH precursors between the noi and ace contexts. These findings, discussed below, suggest models for the acquisition of MH fate in vivo, and clarify the respective roles of Pax2.1 and Fgf8 in this process.

As previously discussed, the expression of her5 reveals that a pro-MH domain is identified at gastrulation stages within the neural plate. Several studies demonstrate that, at somitogenesis stages, the IsO is then necessary for the development of structures surrounding the MHB (such as the posterior tectum, isthmus and cerebellum). Thus, one likely function of the IsO is to permit or reinforce the diversification of MH identities at the center of the MH pro-domain. In addition, we directly demonstrate here that,in noi and ace, at least anterior and posterior MH precursors acquire characteristics of non-MH neighboring territories. Anteriorly, MH precursors express diencephalic markers (pax6.1 in ace, fgfr3 in noi). Posteriorly, in noi,otx2-negative MH precursors express pax6.1 and fgfr3(Fig. 7), suggesting that r1 precursors express r2 characteristics. Thus, another function of IsO factors such as Pax2.1 and Fgf8 is to stabilize MH identity at the extremities of the MH pro-domain. Several mechanisms could account for this function. For example, the activity of IsO factors could (directly or not) act on cell movements to retain MH precursors away from more anterior or posterior patterning sources. Alternatively, IsO factors could bias or stabilize anterior and posterior MH precursors in their choice(s) of cell identity,favoring the reinforcement of MH values. We favor this interpretation, because we did not detect obvious signs of ectopic cell migrations in noi and ace, where GFP-positive cells remained in a compact and homogeneous domain. IsO factors could either be in themselves instructive to impart or reinforce MH identities at the boundaries, or render pro-MH cells responsive to instructive cues at the proper time.

The stage at which this activity takes place cannot directly be inferred from our data. Gain- and loss-of-function experiments in the mouse, chick and zebrafish demonstrated an antagonism between the expression of Pax6 and En factors to delimit the di-mesencephalic border(Araki and Nakamura, 1999; Liu and Joyner, 2001; Mastick et al., 1997; Matsunaga et al., 2000b). Our time-course expression studies in ace however suggest that aberrant cell identity choice occurs at least as early as the 12-somite stage (see pax6.1 at 15 somites on Fig. 8F, and data not shown). Furthermore, in noi, MH precursors acquire fgfr3 expression but the pax6.1/GFP boundary is maintained. Therefore, in anterior cells of the MH pro-domain, we favor a model where IsO factors influence cell identity choices independently of Pax6 action, most probably at an earlier stage than the Pax6/En interplay. Because most MH markers display normal expression profiles in both mutants until the five-somite stage, impaired choices of identity in MH precursors in noi and ace might occur after that stage, in relation with a deficient MH maintenance loop. Alternatively, they might occur before the maintenance phase, when Pax2.1 and Fgf8 are broadly expressed within the MH anlage (Lun and Brand, 1998; Reifers et al., 1998). These choices might take place progressively, perhaps in a manner starting at the extremities of the pro-MH domain and converging towards the MHB, as suggested by the progressive restriction of MH markers (directly demonstrated here for her5) and the progression of maturation events such as neurogenesis(this paper).

In noi, pax6.1 expression is extended posteriorly (this paper Fig. 6); however, this diencephalic expansion does not occur by recruiting mesencephalic precursors. It is possible that cell death (Brand et al., 1996) or lower proliferation rate at a late stage, or altered influences of midbrain cells on diencephalic development account for the observed posterior expansion of pax6.1 expression. These results stress the importance of direct lineage tracing in the interpretation of patterning phenotypes.

In the light of the model proposed above, our results highlight strikingly different impacts of the noi and ace backgrounds on the orientation of identity choices of MH precursors. Major differences are (1)the anterior expression of pax6.1 in alar MH precursors in ace but not noi; (2) the acquisition of fgfr3expression by all alar cells in noi, while no alar cells express fgfr3 in ace; and (3) the expression of otx2 by posterior MH precursors in ace. Several (non-exclusive)interpretations can account for the differential plasticity of pro-MH cells in ace versus noi. Timing might be involved: the downregulation of MHB markers occurs generally later in ace (completed around the 20-somite stage) than in noi (completed around the 10-12-somite stage), making it possible that partial IsO activity is maintained until a later stage in ace and prevents, for example, the turning-on of fgfr3 expression by most alar MH precursors. More likely, Pax2.1 and Fgf8 exert distinct functions in the orientation of identity choices of pro-MH cells. First, fgf8 and pax2.1 are expressed in overlapping but non-identical domains, thus their primary and secondary target cells are probably distinct. In addition, they probably control different cellular processes. Pax2.1 appears generally required to prevent the pro-MH territory as a whole from acquiring an fgfr3-positive fate. In noi,because otx2-positive and -negative domains are maintained, the easiest interpretation of the fgfr3 phenotype is an anteriorization of mesencephalic precursors and a posteriorization of metencephalic precursors. Thus, we propose that Pax2.1 activity in vivo prevents mes- and metencephalic precursors from choosing immediately neighboring, non-MH fates. These results extend previous findings in the mouse that implied Pax2(together with Pax5) in the maintenance of MH identity or the IsO as a whole(Schwarz et al., 1997; Urbanek et al., 1997). Some MH characters are however retained in MH precursors in noi, like the non-expression of pax6.1. Because of the antagonistic effects of noi and ace on pax6.1 and fgfr3expression, this is possibly due to the maintenance of early Fgf8 activity in noi.

By contrast, our results suggest distinct functions for Fgf8. First, Fgf8 expression prevents only the most anterior alar mesencephalic precursors from acquiring a partial diencephalic identity. Thus, we propose that Fgf8 is involved, at a distance, in the choice or reinforcement of an anterior tectal fate. Furthermore, we report that all GFP-positive cells in ace are otx2 positive. As no cell death was observed, this strongly suggests that metencephalic precursors mostly choose an anterior identity in the absence of Fgf8 function. These results validate earlier interpretations of the Fgf8 mutant or gain-of-function phenotypes(Brand et al., 1996; Reifers et al., 1998; Liu et al., 1999; Martinez et al., 1999; Sato et al., 2001). Thus,another function of Fgf8 is to maintain AP distinctions within the MH pro-domain itself and permit the individualization of metencephalic versus mesencephalic identities. Ectopic expression experiments in chicken at somitogenesis stages demonstrated an antagonism between Fgf8 and Hoxa1 expression to delimit the r1/r2 boundary and determine r1 versus r2 identities (Irving and Mason,2000). These results, together with ours, further suggest distinct functions of Fgf8 over time: at an early stage, Fgf8 would orient the choice of a metencephalic versus mesencephalic identity within the MH pro-domain;later, within the metencephalic anlage, it would reinforce an r1 versus an r2 character. In the mouse and chicken, Fgf8 has also been proposed to control proliferation (Lee et al.,1997). However, we could not detect gross alterations in the number of her5 progeny cells between ace mutants and wild-type siblings at the stage of our analysis, suggesting that Fgf8 alone,in the zebrafish, does not initially play a major role in MH growth.

We are grateful to J. Sitz for advice on the ET-cloning strategy, to N. David for isolating the her5 genomic PACs, and to A. Folchert and B. Tannhäuser for expert technical assistance and laboratory management. We acknowledge J. A. Campos-Ortega, V. Prince, T. Jovett, H. Takeda, S. Wilson and V. Korzh for gifts of probes, and the RZPD (Berlin) for providing the genomic PAC library. We thank all members of the L.B.-C.'s and K. Imai's laboratories for support and constructive input, J. A. Campos-Ortega for discussions, and P. Chapouton, F. Rosa and M. Wassef for critical reading of the manuscript. Work in L.B.-C.'s laboratory is supported by a VolkswagenStiftung `junior research group' grant and DFG grant BA2024/2-1.