Expanded expression of Sonic Hedgehog in Astyanax cavefish:multiple consequences on forebrain development and evolution

The vertebrate forebrain is a sophisticated structure, which contains the neural regions controlling complex behaviours, such as sensorimotor, cognitive and limbic responses. The embryonic development of this structure,particularly the telencephalon and diencephalon, is orchestrated through a series of complex patterning, morphogenetic, migration and wiring events that are not completely understood. The genetic mechanisms controlling these developmental steps appear largely shared among vertebrates(Bachy et al., 2002; Murakami et al., 2005; Puelles et al., 2000; Rubenstein et al., 1994; Striedter, 1997; Wullimann and Mueller, 2004),even though the morphology and connectivity between forebrain areas vary considerably among different vertebrate species(Butler and Hodos, 1996).

Early in embryogenesis, the forebrain develops under the influence of signalling centres, called `secondary organisers' in mammals, which secrete morphogen molecules such as Wnt (Wingless-Int), Bmp (Bone morphogenetic protein), Fgf (Fibroblast growth factor), or Shh (Sonic hedgehog)(Marin and Rubenstein, 2001; Ohkubo et al., 2002; Rubenstein and Beachy, 1998; Rubenstein et al., 1998; Wilson and Rubenstein, 2000). These signalling centres are located at the dorsal (Wnt, Bmp), anterior (Fgf8)and ventral (Shh) midline of the telencephalon, respectively. Likewise,diencephalon development is largely controlled by the floor or basal plate and by the secondary organiser of the zona limitans intrathalamica (zli), which both produce Shh (Kiecker and Lumsden,2004; Vieira et al.,2005; Zeltser,2005). In amniotes, these midline signalling centres and interactions between them (Ohkubo et al.,2002; Shimogori et al.,2004; Tole et al.,2000) create a field of organisation in their zone of influence,and regulate the growth, patterning and regionalisation of forebrain areas. Thus, Shh midline signalling is well placed to have a major impact, not only on the development but also on the evolution of the forebrain, through subtle changes in signal intensity and/or position. For example, although the global developmental organisation of the forebrain is highly conserved in lampreys(jawless vertebrates) compared with gnathostomes(Murakami et al., 2001; Murakami et al., 2005), their ventral telencephalon `lacks' a pallidum(Weigle and Northcutt, 1999). Strikingly, there is no Shh signalling at the lamprey telencephalic ventral midline (Osorio et al., 2005),suggesting that Shh expression in the anterior ventral midline was responsible for the appearance of a novel forebrain subdivision, and further that midline signalling may be a powerful motor for forebrain evolution in jawed vertebrates.

The surface-dwelling and cave-living forms of Astyanax mexicanus have been used as an advantageous model system in evolutionary developmental biology (Jeffery, 2001). The two forms of this single species split from a common ancestor about 10,000 years ago, a relatively short period of time during which the cave animals have evolved both regressive and constructive features. These include the loss of eyes, pigmentation and aggressive behaviour, and the increase in feeding apparatus (jaws, teeth, taste buds) and body fat content. Among these evolutionary changes, the loss of eyes in cavefish has drawn much attention. Recently, Yamamoto et al. (Yamamoto et al., 2004) have shown that eye regression is a consequence of early overexpression of hedgehog genes (Hh, including Shh)at the ventral midline of cavefish embryos. This expansion of midline Hh signalling causes hyperactivation of downstream genes, lens apoptosis and eye degeneration. Thus, a gain of function in Hh signalling is at the origin of the regressive eye phenotype. In fact, eyes may have been lost by default as a consequence of natural selection for constructive traits, such as feeding structures, that are positively regulated by Hh signalling(Jeffery, 2005).

As shown in various model species, the Hedgehog family of secreted proteins regulates pattern formation, proliferation and differentiation. Here we have taken advantage of enlarged Shh expression in the embryonic midline of cavefish relative to surface fish embryos to test the idea of the impact of the ventral midline on the development and evolution of forebrain patterning,neuronal organisation and wiring. Through the analysis of Shh downstream genes of the Nkx and LIM-hd (LIM-homeodomain) families, cell proliferation patterns and GABAergic neuron differentiation, we show that Shh overexpression has multiple and region-specific effects on neuronal development in the hypothalamus, subpallium and olfactory bulbs. Moreover, comparison between cave and surface embryos allows the identification of new developmental mechanisms for neuronal specification in the fish telencephalon.

Laboratory stocks of A. mexicanus surface fish and cavefish(Pachòn population) were obtained from the Jeffery laboratory at the University of Maryland, College Park, MD. Fish were maintained at 23-26°C on a 12:12 hours light:dark cycle. Embryos were collected after spawning and fixed at various stages in 4% paraformaldehyde (PFA). After progressive dehydration in methanol, they were stored at -20°C. A. mexicanusdevelopment is almost identical to zebrafish in the first 48 hours postfertilisation (hpf), so the zebrafish staging table can be used(Kimmel et al., 1995).

Total RNA from surface fish brains was reverse transcribed with random primers using AMV reverse transcriptase (Promega). Partial cDNA sequences for Nkx2.1b (278 bp, GenBank accession no. DQ431669), Lhx6 (408 bp, DQ431667), Lhx9 (1.2 kb, EF175738) and GAD65 (445 bp,DQ431668) were amplified by PCR using degenerated primers (sequences on request) designed after alignments of several teleost sequences including zebrafish. PCR products were subcloned in TOPO-PCR II vector (Invitrogen) and sequenced. Nkx2.1a (AY661435), Shh (AY661431) and Pax6 (AY651762) cDNAs were previously isolated by the Jeffery laboratory. The Lhx7 and Dlx2 plasmids were kindly provided by David Stock.

Sequences of Lhx6 and Lhx7 genes (accession numbers on request) were aligned with Clustal W(Thompson et al., 1994). Phylogenetic tree (Neighbour-Joining) and bootstrap values were generated using the MEGA3.1 program (Kumar et al.,2004). During this work, an Astyanax Lhx6 sequence was isolated by David Stock (DQ822512). The two Lhx6 sequences shared 100% identity in the overlapping fragment.

cDNAs were amplified by PCR, and digoxygenin- or fluorescein-labeled riboprobes were synthesised from PCR templates(Nguyen et al., 2001). A protocol for automated whole-mount in situ hybridisation (Intavis) was performed (Deyts et al.,2005). For probe penetration, embryos were treated with proteinase K for 30 minutes at 37°C (10 μg/ml for 36 hpf, 20 μg/ml for 48 hpf,30 μg/ml for 60 hpf, 40 μg/ml for 72 hpf). Embryos were paraffinsectioned at 8 μm, or their brains were dissected out and mounted in glycerol. Depending on the gene marker and the stage considered, five to 20 embryos from each population were analysed (except for cyclopamine treatments,where n=4-5), and three to four were sectioned. Quantification of the expansion of Nkx2.1 and Dlx2 domains was performed using ImageJ, on in toto pictures of embryos at 24, 36 and 48 hpf coming from at least two independent experiments.

For Proliferating cell nuclear antigen (PCNA) immunohistochemistry embryos were fixed in Clark's solution (3:1 ethanol 100: acetic acid) overnight at 4°C. They were processed as described(Candal et al., 2005). Quantification was performed on the same serial section in cavefish and surface fish embryos. Classical morphometric measurement was done, using a grid superimposed on the sections, and counting the arbitrary surface covered by PCNA-positive cells.

For phospho-H3 immunohistochemistry, embryos were fixed in 4% PFA. A primary rabbit polyclonal anti-phosphoH3 antibody (1/100^e,Euromedex) was used. Quantification in the hypothalamus was performed by counting the total number of phospho-H3 positive cells in the hypothalamus of each embryo through all serial transverse sections obtained. For control quantification in the tectum, the first five sections at tectal level were counted. Results were expressed as phospho-H3 positive cell density after measurement of the surface area of the hypothalamus or the tectum for each section using ImageJ. Statistical analysis was carried out using Student's t-test.

At 15 hpf embryos were treated for 9 hours in 200 μmol/l or 20 μmol/l cyclopamine (Sigma) in water. Control embryos were exposed to 0.1% ethanol, as the cyclopamine stock solution was diluted in 100% ethanol. After treatment,they were washed twice, maintained in fresh water until 48 hpf and fixed.

Yamamoto et al. (Yamamoto et al.,2004) have recently shown that midline Shh expression is expanded at early embryonic stages (neural plate though 1-4 somites, or 10 hpf), and indirectly causes eye degeneration in cavefish. Here, we have investigated the impact of Shh modification later in forebrain embryogenesis, from 24 hpf to 1 week of development.

First, we asked whether enlarged Shh expression in cavefish was maintained throughout development. At 20, 36 and 48 hpf, Shh was still more largely expressed in cavefish compared with surface fish(Fig. 1). This expanded expression included all anterior Shh expression domains, including the floor plate, the zli, the hypothalamus(Fig. 1A-H) and the subpallium(Fig. 1G,H). These data indicate that Shh expansion was not transient in cavefish, but rather was maintained and even spread to new locations during late embryogenesis,opening the possibility of multiple consequences on forebrain development.

As a next step, we investigated the expression of patterning and regionalisation genes of the Nkx family. Nkx2.1 and Nkx2.2 homeodomain factors are downstream in the Shh signalling pathway(Briscoe et al., 1999; Rallu et al., 2002; Xu et al., 2005). In other well-studied teleost species (zebrafish, medaka), there are two Nkx2.1 members(Alunni et al., 2004; Rohr et al., 2001). In Astyanax, we also found that Nkx2.1a(Yamamoto et al., 2004) is accompanied by a duplicated gene, which we named Nkx2.1b. Consistent with expression patterns in other fishes, Nkx2.1a was found exclusively in the presumptive hypothalamus, encompassing almost the entire structure (Fig. 2I). By contrast, Nkx2.1b was expressed both in restricted nuclei of the hypothalamus, in the subpallial telencephalon (SP, subpallium), including the future basal ganglia and septum (see Wullimann and Mueller, 2004; Alunni et al., 2004), and in the preoptic region (po; Fig. 2E-G). Finally, Nkx2.2 showed a typical `alar-basal boundary' pattern, adjacent to and following the Shh pattern(Fig. 2C).

There was no significant difference in Nkx2.2 or Dlx2(subpallial marker) expression between surface and cave embryos at any stage(Fig. 2A-D). Quantification showed a non-significant +4.8% increase in Dlx2 territory in cavefish, suggesting that global pallial and subpallial patterning was identical in both populations. Nkx2.1b expression was similar in the presumptive hypothalamus of cavefish and surface fish embryos(Fig. 2E-H), but was significantly expanded in the cavefish SP and preoptic region relative to surface fish embryos (Fig. 2E-H; +35% after quantification). This result was found reproducibly from 24 to 72 hpf. To investigate whether the Nkx2.1bexpansion was at the expense of the rest of the subpallium, double labelling for Dlx2 and Nkx2.1b were performed. They showed that the domain expressing only Dlx2 was unchanged, and that Nkx2.1bterritory enlargement was due to an anteroventral expansion of the Nkx2.1b domain (see Fig. S1A,B in the supplementary material). Finally, between 24 and 72 hpf, Nkx2.1a expression was enlarged in the developing hypothalamus of cavefish compared with their surface counterparts (Fig. 2I,J; +38%),possibly defining a larger hypothalamic territory.

The above data, using regionalisation factors as comparative indexes of brain patterning between cave and surface embryos, suggested that specific presumptive regions are larger in cavefish embryos. To analyse the origin of these differences, particularly in po and hypothalamus, we compared cell proliferation in cavefish and surface fish embryos from 36 hpf to 5 days, by PCNA immunostaining (Fig. 3). Until 48 hpf, almost the entire brain of both populations was proliferating(not shown), preventing any comparison. Slightly later, when proliferation was less extensive, an increased number of PCNA-positive cells was observed in the cavefish basal diencephalon (Fig. 3A-H). This increase was maintained for up to 5 days of development and was specific to the developing hypothalamus and preoptic region. Quantification showed that the surface area covered by proliferating cells was 31 and 22% higher in the po of cavefish at 60 hpf and 5 days,respectively, and 75 and 60% higher in the hypothalamus of cavefish at the same stages. By contrast, the dorsal diencephalic and mesencephalic proliferative zones showed identical PCNA immunoreactivity in both populations(Fig. 3A-D).

To further demonstrate the causal role of enlarged Shh signalling in the increased proliferation and expansion of the basal diencephalon in cavefish embryos, we performed cyclopamine treatments. Cavefish embryos were treated with cyclopamine between 15 and 24 hpf, were allowed to develop until 48 hpf and were assessed for proliferation pattern using phospho-histone H3(phosphoH3), which labels cells in M phase of the cell cycle and gives readily quantifiable number of immuno-positive cells. Cyclopamine treatment induced a reduction in cell proliferation both in the preoptic region and in the hypothalamus (see Fig. S2A-D in the supplementary material). Quantification of this effect in the hypothalamus showed a 38% reduction in the surface area of the hypothalamus, a 77% reduction in the number of phospho-H3 cells, and a significant decrease in the density of immuno-positive cells in this region in cyclopamine-treated embryos when compared with their controls(Fig. 3I). This suggested an Shh-dependent control of proliferation and size during early development of the basal diencephalon. The specificity of these effects were confirmed by parallel analysis in the optic tectum (Fig. 3J and see Fig. S2E,F in the supplementary material): there, a slight reduction of tectal surface (-11%) paralleled a slight reduction in the number of phospho-H3-positive cells (-18%), and resulted in a similar density of proliferating cells in control and treated embryos.

We next sought to estimate the impact of enlarged Nkx2.1a and Nkx2.1b expression on neuronal specification processes in the hypothalamus and basal telencephalon. In other vertebrates, Nkx2.1governs the expression of Lhx6 and Lhx7, neuronal specification factors of the LIM-hd family expressed in the basal forebrain(Sussel et al., 1999; Grigoriou et al., 1998). In mouse, Lhx6 governs the tangential migration of GABAergic neurons generated in the subpallium towards the pallium(Alifragis et al., 2004; Lavdas et al., 1999), whereas Lhx7 governs the specification of cholinergic neurons(Fragkouli et al., 2005; Mori et al., 2004; Zhao et al., 2003). The Astyanax Lhx6 and Lhx7 genes were identified as orthologs of Lhx6 and Lhx7 in other vertebrates, including zebrafish(Fig. 4A), and their expression was localised with respect to Shh expression using double-colour in situ hybridisation (Fig. 4B,C). In Astyanax, Lhx6 was expressed in the SP and preoptic region, and in a complex hypothalamic pattern partly similar to that of Nkx2.1b(Fig. 4B, see also Figs 6 and 9). Lhx7 transcripts were similarly detected in the SP and po(Fig. 4C, see also Figs 5 and 9), and also in the pituitary(see Fig. 4C-E and Fig. 5A-D). This pituitary expression has never been reported for Lhx7 in other species(Alunni et al., 2004; Bachy et al., 2001; Grigoriou et al., 1998), and might be Astyanax-specific. Double-colour in situ hybridisation further showed that Lhx6 and Lhx7 were expressed in the same subpallial domain but had different expression territories in the preoptic region, Lhx6 expression being restricted to the anterior po and included in a larger Lhx7 domain(Fig. 4D,E).

Lhx7 spatiotemporal expression patterns were compared between surface fish and cavefish embryos (Fig. 5). First, we observed a shift in the onset of Lhx7expression in the SP: whereas the cavefish basal telencephalon already contained numerous Lhx7-expressing cells at 24 hpf, only two or three cells were reproducibly detected in their surface fish counterparts at this stage (Fig. 5A,B). This heterochrony in the onset of Lhx7 in the telencephalon was rapidly compensated, as we could not detect significant differences in SP expression of Lhx7 in later embryos, ranging from stage 36 to 60 hpf(Fig. 5C-H). Of note, and with respect to the role of Lhx7 in cholinergic specification in mammals,the Astyanax ortholog was found in a few cells of the lateral subpallium at 48 and 60 hpf (Fig. 5G,H and inset). This region is called Vl in fish (for area Ventralis lateralis) and contains cholinergic neurons(Mueller et al., 2004). Lhx7 expression at this level also appears identical in cavefish and surface fish embryos (Fig. 5G,H). By contrast, Lhx7 was continuously expressed as a larger domain in the po region of cavefish relative to surface fish embryos(Fig. 5C-H). Finally,hypothalamic and pituitary expression were identical in both populations(Fig. 5C,D).

In sum, Lhx7 was expressed earlier in the SP and larger in the po of cavefish embryos, bringing up the possibility that Shh and Nkx2.1b expansion in these two regions could have effects on downstream neuronal specification events regulated by Lhx7.

At all studied stages from 24 to 72 hpf, Lhx6 expression domains were significantly broader in cavefish embryos than in their surface fish counterparts. This included the SP and po regions (shown in Fig. 6A,B at 24 hpf), and the hypothalamus (shown in Fig. 6C,D at 36 hpf). Moreover, starting at 48 hpf, an additional area of Lhx6 expression was detected in the telencephalon of both Astyanax populations (Fig. 6E-H). At 48, 60 and 72 hpf, two bilateral columns of Lhx6-expressing cells extended from the dorsal aspect of the Lhx6-positive SP domain (dSP, Fig. 7A,B), ending as a pack of strong Lhx6-expressing cells located in the anterolateral part of the pallium(Fig. 6E-H and Fig. 7A,B). After analysis in sections, this Lhx6 expression pattern is suggestive of cells migrating out of the SP towards pallial areas, and may therefore represent the fish counterpart of the Lhx6-positive,ganglionic-eminence-originated, tangentially migrating GABAergic neurons in mammals. Importantly, Lhx6 expression was much more intense in the columns of cells emanating from the dorsal SP in cavefish than in surface fish embryos (Fig. 6E-H). Moreover,at 72 hpf, Lhx6 expression in the SP and associated bilateral columns of putative migratory cells continued to be stronger in cavefish embryos(Fig. 6G,H).

To further support the idea that Lhx6-positive cells move outside the subpallium towards the pallium, we analysed how Lhx6 expression relates to Pax6 and Lhx9. In fish, Pax6 marks the pallio-subpallial border (Wulliman and Mueller, 2004), and Lhx9 is a pallial marker (Alunni et al.,2004). We could confirm that Lhx6 cells cross the pallio-subpallial border, represented by the Pax6-expressing band in the telencephalon (Fig. 7J,K),and reach their target in a pallial region that also expresses weak levels of Pax6 (Fig. 7J-M). To further explore the characteristics of the columns of Lhx6-expressing cells, we showed that at 60 hpf, when the columns are well developed, they also expressed Nkx2.1b (Fig. 7E,F, see also Fig. 2G,H at 48 hpf). Moreover, the columns of Lhx6-positive cells expressed Astyanax GAD65 (Glutamic acid decarboxylase),demonstrating that they are GABAergic (Fig. 7C,D). Finally, we found that the columns ended in a pallial area corresponding to the developing olfactory bulbs, as demonstrated both by the positions of their termini in the forebrain and by fluorescent dye (DiI)labelling of the olfactory bulbs through the primary olfactory projection(Fig. 7G-I). As the Lhx6-positive columns are more important in cavefish than in surface fish, these data suggest that the expansion of the `ventralising' Shh morphogen has an impact on the establishment of neuronal circuits in various parts of the brain, including regions that are not directly in the range of Shh signalling activity.

Finally, we sought to demonstrate that the modifications of forebrain neuronal patterning in cavefish were indeed Shh-dependent. To this end, we performed cyclopamine treatments on cavefish embryos, trying to phenocopy the surface-type patterns by reduction of Shh signalling. Because we are studying late events in neurodevelopment, cavefish embryos were treated with cyclopamine between 15 and 24 hpf, and fixed at 48 hpf(Fig. 8). After cyclopamine treatment, Dlx2 expression was affected in the SP and thalamic prosomere p3 but not in the hypothalamus(Fig. 8A-D), and Nkx2.2 expression was strongly affected(Fig. 8E-H). The Nkx2.1a expression domain in the hypothalamus was also reduced by cyclopamine, phenocopying the difference between surface fish and cavefish embryos (Fig. 8I-L). Although Nkx2.1b expression was not strongly modified in the hypothalamus, it was reduced in size in the po and SP of treated cavefish(Fig. 8J-L). Finally, Lhx6 and Lhx7 expression were affected in the telencephalon(SP and po) of cyclopamine-treated embryos(Fig. 8Q-X), but not in the pituitary (Lhx7) nor in the hypothalamus (Lhx6). Of note,the columns of Lhx6-positive cells were affected similarly to the difference observed between cavefish and surface fish(Fig. 8U-X). In sum,cyclopamine treatments showed dose-dependent, gene-specific and region-specific effects, which demonstrate the Shh-dependence of the above described modifications in regional and neuronal patterning in cavefish(Fig. 9).

We have taken advantage of subtle, viable and probably adaptive expansion of Shh expression at the ventral midline of the cave-living form of Astyanax mexicanus to investigate the impact of midline signalling on forebrain neuronal patterning during development and evolution. The main results are as follows: (1) Shh expansion in cavefish is maintained throughout development; (2) transcription factors of the Nkx2.1 (but not Nkx2.2 or Dlx2)families are expressed in larger domains in the basal forebrain of cavefish embryos; (3) the LIM-hd cellspecification genes Lhx6 and Lhx7 also show broader expression in a region-specific manner in the subpallium and hypothalamus; (4) GABAergic interneuron migration towards the olfactory bulbs is increased in cavefish; (5) the hypothalamus of cavefish is larger, probably due to increased cell proliferation; and (6) all these effects are Shh-dependent. These findings indicate that a slight modification of midline Shh signalling can affect the specification, size and presumably the migration and connectivity of specific neuronal populations, resulting in a global increase of the olfactory and neuroendocrine components of the brain in cave-living animals.

We have shown that expanded Shh expression is maintained throughout in every region of the anterior central nervous system (ventral midline, zli, subpallium) up to 72 hpf in cavefish embryos. Such a pleiotropic`upregulation' of Shh strongly suggests that a quantitative element, as opposed to spatio-temporal regulatory elements(Jeong et al., 2006), in the transcriptional regulation of Shh has been changed in cavefish. Alternatively,a global increase of an upstream regulator that would quantitatively control Shh expression independently of spatial regulatory elements is possible. In any case, as increased Shh signalling has been found in several independently evolved Astyanax cavefish populations(Yamamoto et al., 2004), it will be extremely interesting to identify this viable regulatory mutation.

Continuous expanded Shh signalling would be expected to induce major perturbations in brain development. However, the global patterning of forebrain development is not modified in cavefish relative to surface fish embryos, and only some of the specific gene cascades downstream of Shh signalling are increased in these embryos (summarised in Fig. 9). Hence, the expression of Nkx2.2 is identical in surface and cavefish, and so is the expression of Dlx2 (see also Xu et al., 2005), reinforcing the specificity of the effects discussed below, and suggesting that at least some natural compensatory mechanisms must be at work in cavefish to avoid a massive perturbation in brain development after Shh expansion. In fact, when cavefish are treated with cyclopamine, Dlx2 and Nkx2.2 expressions are diminished, demonstrating the functional regulatory link between Shh and Dlx2 or Nkx2.2, and unmasking a probable compensatory mechanism that has evolved in cavefish. Moreover and globally, the cyclopamine treatments also show that the observed effects on regionalisation and neuronal specification genes are indeed due to increased Shh signalling.

As schematised in Fig. 9, Nkx2.1a and Nkx2.1b expressions are enlarged in the cavefish hypothalamus and SP or po, respectively. However, Nkx2.1b is unchanged in the hypothalamus. This differential regulation can be viewed in terms of subfunctionalisation of the two Nkx2.1 genes, which may constitute a feature characteristic of teleosts. Concerning LIM-hd genes, the enlarged expression of Lhx6 in the hypothalamus can be interpreted as a consequence of Nkx2.1a expansion, and the enlarged expression of Lhx7 and Lhx6 in the SP or po as a consequence of Nkx2.1b expansion. Thus, we uncover specificity of the increased`cascades', where each Nkx2.1 paralog has an effect on the expression of one or the two LIM-hd paralogs, and where Shh signalling eventually regulates the neuronal composition of the forebrain. These results fit nicely with recent findings in mouse showing that Shh signalling during neurogenesis maintains cortical interneuron identity through regulation of Nkx2.1(Xu et al., 2005).

Our findings suggest that the expanded Nkx2.1a and Nkx2.1b,Lhx6 and Lhx7 expression domains in the hypothalamus and po of cavefish embryos reflects a higher level of cell proliferation in these regions relative to surface fish embryos, eventually resulting in enlargement of this part of the basal forebrain. Consistent with this possibility, the mature hypothalamus is larger in cavefish than in surface fish adults (D. Soares and W. R. Jeffery, unpublished). The implication of Shh in the control of proliferation in the brain (and other organs) is documented(Britto et al., 2000; Fu et al., 2004; Ishibashi and McMahon, 2002; Lewis et al., 2004; Moshiri et al., 2005; Wechsler-Reya and Scott,1999). However, despite the fact that Shh controls the proliferation of both ventral and dorsal brain structures(Dahmane et al., 2001), we observed a specific effect only in the hypothalamus and preoptic region. By contrast to the case of the telencephalon, where other organising centres secreting Bmps or Fgfs might naturally compensate for Shhoverexpression (Ohkubo et al.,2002; Shimogori et al.,2004; Tole et al.,2000), the hypothalamus develops under the direct influence of the ventral midline and alar-basal boundary and in the vicinity of the zli diencephalic organiser, all of which secrete Shh. Thus, there is probably no possible compensation by neighbouring centres in this part of the forebrain,and this might explain why Shh expansion results in a bigger hypothalamus in cavefish. Does this have any physiological and/or adaptive consequences? Cave animals are fat and have lost aggressive behaviours compared with surface fish (Jeffery,2001). These traits might be related to the neuroendocrine controls exerted by the hypothalamus. However, several nuclei of the hypothalamus and preoptic region serve as tertiary gustatory centres(Folgueira et al., 2003; Lamb and Finger, 1996), and the increase in these nuclei might correspond to the increase in taste bud number in cavefish (Jeffery et al.,2000). Further investigations are clearly needed to functionally interpret this hypothalamic difference between cavefish and surface fish.

In mammals, the function of Lhx7 is well studied: Lhx7 specifies the cholinergic phenotype in the basal forebrain(Mori et al., 2004; Zhao et al., 2003), and mice lacking Lhx7 show impaired spatial learning and memory(Fragkouli et al., 2005). More recently, Lhx7 was also involved in the control of GABAergic fate in the basal forebrain (Manabe et al.,2005; Bachy and Rétaux, 2006). There, Lhx7inhibits GABAergic differentiation at early stages of embryogenesis to preserve a pool of future cholinergic neurons, the number of which is tightly controlled and which will undergo terminal differentiation much later in development (Bachy and Rétaux, 2006). Thus, our observation of earlier Lhx7 expression in cavefish embryos might simply reflect the fact that upstream Shh and Nkx2.1b are increased and reach a`threshold' level earlier than in surface fish to induce Lhx7expression. Then, when the correct pool of cholinergic neurons is obtained,progenitors stop dividing, and the expression of Lhx7 in surface fish and cavefish becomes similar after 36 hpf. In this case, and contrary to the case of Lhx6 regulation, the (unknown) mechanisms that control the number of cholinergic neurons in the forebrain would naturally compensate for the primary effects of Shh overexpression in cavefish, and would not be expected to have any obvious consequence on SP cholinergic neuronal patterning.

By striking contrast to the case of the SP, Lhx7 expansion is maintained in the po of cavefish relative to surface fish. The po is another cholinergic region in mammals and also contains cholinergic neurons in fish(Mueller et al., 2004) and other nonmammalian vertebrates (Marin et al., 1997; Pombal et al.,2001; Rodriguez-Moldes et al.,2002). Moreover, in fish, the po projects to the olfactory bulbs(Folgueira et al., 2004), and these projections might represent the fish counterpart of the pathway from basal forebrain cholinergic nuclei of the diagonal band onto the olfactory bulbs of mammals (Shipley,1985). Importantly, this cholinergic innervation of the bulbs actively participates in olfactory functions, such as early olfactory learning and discrimination (Fletcher and Wilson,2002; Linster and Cleland,2002; Wilson et al.,2004). Further analysis will be needed to investigate whether cavefish indeed have enhanced cholinergic projections to the olfactory bulbs,and whether this has functional implications (see also below).

We interpret the Lhx6 expression domain in Astyanaxtelencephalon as a stream of migrating cells originating from the SP and populating the pallial region forming the olfactory bulbs. These cells fulfil criteria for migrating cells: (1) they express Nkx2.1b, a subpallial marker, expressed in the ventricular zone of the SP; (2) their pallial region of destination never expresses Nkx2.1 or Lhx6 in the ventricular or subventricular zone, suggesting that they are not derived from the pallial progenitor zone; (3) they follow a tangential trajectory in the neuroepithelium, where they end up in a subpial, marginal zone location; and(4) they cross the pallio-subpallial border defined by Pax6 domain. These cells are GABAergic, as they express GAD65, and thus probably constitute a population of interneurons. We therefore propose that Lhx6 labels the equivalent of the mammalian rostral migratory stream(RMS), which contributes granule and periglomerular GABAergic interneurons to the olfactory bulbs (Lois and Alvarez-Buylla, 1994; Luskin,1993; Wichterle et al.,2001). In mammals, the RMS begins to migrate during the late embryonic and perinatal periods and continues throughout adulthood(Wichterle et al., 2001). The timecourse observed here in fish is similar, as the peak of Lhx6expression in the putative migratory cells is observed in 48 hpf embryos and continues at later stages. We did not observe at any stage Lhx6-positive cells that could constitute the fish counterpart of tangentially migrating GABAergic cortical interneurons(Anderson et al., 1999; Anderson et al., 1997; Anderson et al., 2001; Marin and Rubenstein, 2001; Marin and Rubenstein, 2003). In fish, these GABAergic pallial interneurons constitute a relatively scarce neuronal population (Anglade et al.,1999), and the absence of pallial Lhx6 expression,together with the absence of Dlx gene expression in the fish pallium(Akimenko et al., 1994; Alunni et al., 2004; Zerucha and Ekker, 2000)(present data), suggest that these interneurons do not share common genetic specification mechanisms with their mammalian counterparts. In fact, an Lhx6.1 isoform of Lhx6 in mouse was reported to be expressed both in the cortex and the olfactory bulbs (Kimura et al., 1999). Altogether, this may suggest that an Lhx6-positive migratory pathway from the SP to the olfactory bulbs is ancestral in vertebrates, and that the dorsal pallium-directed Lhx6migrations are an innovation of mammals.

Strikingly, lampreys do not express Shh and Nkx2.1 in their embryonic ventral forebrain(Murakami et al., 2001; Osorio et al., 2005), they do not have a pallidal division in their adult subpallium(Weigle and Northcutt, 1999),and they do not possess GABAergic interneurons in their pallium(Melendez-Ferro et al., 2002). Thus, the presence of regulatory cascades controlling cell migrations and generating diversity in neuronal patterning in the forebrain seems to be a novelty that emerged at the transition from agnathans to fish, and for which the impact of Shh midline signalling is crucial.

We have observed an increased Lhx6-positive migratory stream in the forebrain of cavefish relative to surface fish. From an adaptive point of view, what kind of advantage does having more olfactory bulb interneurons confer to cavefish for their life in perpetual darkness? In the mammalian bulbs, periglomerular GABA interneurons increase the contrast between activation levels in different glomeruli, and augment odour discrimination(Aungst et al., 2003; Lledo et al., 2004), whereas granular cells increase contrast and specificity of olfactory information by synchronising the discharges of mitral projection neurons(Lagier et al., 2004; Laurent, 2002; MacLeod and Laurent, 1996; Saghatelyan et al., 2005). In sum, bulbar interneurons are crucial in olfactory circuitry, and a positive correlation between the abundance of newly generated bulbar interneurons and olfactory performances was reported(Scotto-Lomassese et al.,2003). An increase in olfactory modality through an augmentation or more rapid turnover of the RMS might therefore be advantageous for cavefish. Indeed, Shh was shown to be required for the production of new olfactory neurons from the stem cell niche of the forebrain subventricular zone in vivo (Palma et al.,2005).

Using the two forms of Astyanax as an evolutionary developmental model, we have shown that changes in Shh midline signalling can have various effects on brain development and evolution. Because Shh expansion in cavefish is physiological and probably adaptive, it could be compensated by other signalling pathways to prevent deleterious effects on brain development. Thus, Shh expansion in cavefish has highly specific consequences on the expression of downstream genes of the Nkx2.1 and LIM-hd families,resulting either in expression heterochrony of Lhx7, which is rapidly compensated and has no consequences on telencephalic patterning, or in an increase of Lhx6 expression, which might confer a selective advantage to the cavefish, or in an increase of the size of the hypothalamus, which is more difficult to interpret in terms of adaptation. The finding that cavefish have enhanced olfactory and neuroendocrine structures when compared to surface fish illustrates how midline signalling centres can have a major impact on forebrain development, evolution and diversification.

We sincerely thank Laurent Legendre and Matthieu Simion for help in fish care, Carole Deyts and Aurélie Heuzé for managing the robot, Amy Parkhurst for collecting embryos, David Stock for sharing sequence data, and Franck Bourrat, Eva Candal and Mario Wullimann for constructive discussions. Work supported by CNRS and GIS Génomique Marine to S.R., EEC grant QLK3-CT-2001-01890 to J.-S.J., and NSF (IBN-0110275) and NIH (EY01-46190)grants to W.R.J. A.M. was a CNRS postdoctoral fellow, and A.A. is an INRA postdoctoral fellow.