Involvement of Hedgehog and FGF signalling in the lamprey telencephalon: evolution of regionalization and dorsoventral patterning of the vertebrate forebrain

The vertebrate telencephalon represents a highly sophisticated portion of the central nervous system (CNS). One component, the neocortex with its multilayered cortical neurons, can perform complicated processing of information, especially in mammals. Although the amphioxus FoxG1, considered a telencephalic marker in vertebrates, is expressed in a few cells in the cerebral vesicle (Toresson et al., 1998), detailed neuroanatomy supports the idea that a homologue of the vertebrate telencephalon is lacking in cephalochordates and urochordates (Wicht and Lacalli, 2005). How the telencephalon was obtained in evolution remains elusive, and this question is tightly linked with that of the origin of vertebrates per se. As a sister group of the jawed vertebrates (gnathostomes), lampreys serve as a valuable model system for studying the evolution of the telencephalon. This is because this animal lineage appears to have diverged in the earliest phase of vertebrate evolution and probably possesses ancient developmental features that have been modified or lost in gnathostomes (Kuratani and Ota, 2008; Murakami and Kuratani, 2008).

The vertebrate telencephalon is divided dorsoventrally into the pallium and subpallium. Of these, the subpallium is crucial because it not only differentiates into the basal ganglia but also serves as the source of various types of neurons (Moreno et al., 2009). This structure is further divided into lateral and medial ganglionic eminences (LGE and MGE, respectively). The MGE is specifically known to be the source of cortical interneurons, whereas the LGE gives rise to striatal projection neurons and olfactory bulb interneurons. The MGE produces GABAergic interneurons (Wichterle et al., 2001); curiously, precursors of these interneurons show active migration in gnathostome embryos, migrating along the tangential pathway to populate the cortex to yield most of the interneurons (Marin and Rubenstein, 2001). Although it is unknown whether the shark has an equivalent MGE region in the embryonic brain, migratory neuroblasts have been observed in this animal (Carrera et al., 2008). In the lamprey, the expression patterns of Pax6, Emx and Dlx homologues suggest that the embryonic brain also appears to possess a distinction equivalent to that found between the pallium and subpallium (Murakami et al., 2001) (see Fig. S4 in the supplementary material). Furthermore, based on the expression pattern of the Emx homologue, the pallium can be further divided into at least two domains that are specified dorsolaterally (Murakami et al., 2001). In addition, because the Nkx2.1 (also known as TTF-1) expression domain is absent in the subpallium (Ogasawara et al., 2001) and no Hh cognates are expressed in the telencephalon (Uchida et al., 2003; Osorio et al., 2005), the lampreys were assumed to lack a domain equivalent to the gnathostome MGE (Murakami et al., 2005). These traits pose a few questions regarding forebrain patterning in the lamprey. Does the lamprey really lack the MGE domain together with the Hh-Nkx2.1 expression domain (i.e. no other Hh genes expressed)? If so, which developmental and neurological functions does the lamprey subpallium possess? Finally, are there any differences between the lamprey and gnathostomes in the developmental mechanisms involved in the dorsoventral (DV) patterning of the telencephalic region?

In the present study, we aimed to isolate all the Hh paralogues from the lamprey, as well as those of Ptc, the genes encoding the suspected Hh protein receptors, to clarify whether the lamprey embryonic telencephalon exhibits functional Hh expression domains. We also characterized the gene expression profile of the forebrain to detect any LGE-associated properties in the lamprey subpallium. Finally, we performed inhibitory experiments for Hh and fibroblast growth factor (FGF) signalling using inhibitors to detect changes in developmental patterning in the lamprey telencephalon. We confirmed that there are no Hh expression domains at any of the observed stages, and the expression patterns of Gsh2, Isl1/2 and Sp8 suggest the presence of an LGE-like property of the lamprey subpallium. We also showed that although the lamprey subpallium does not possess any Hh expression domains, Hh signalling functions in the early DV patterning of the telencephalon at specific developmental stages, followed by involvement of FGF signalling. Thus, we have reconstructed an evolutionary scenario for the vertebrate telencephalon in terms of changes introduced to the developmental systems of ancestral animal lineages during evolution.

Adult lampreys, Lethenteron japonicum, were collected from Miomote River, Niigata and Shiribetsu River, Hokkaido, Japan, during the breeding season (early June). Eggs were fertilized artificially and incubated in 10% Steinberg's solution (Steinberg, 1957) at 16-23°C. Lamprey embryos were staged according to Tahara's staging of Lethenteron reissneri (Tahara, 1988). For in situ hybridization, the embryos were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), then dehydrated in a methanol dilution series and stored in 100% methanol at −20°C.

Total RNA of L. japonicum was extracted from whole embryos of stages 25-27 using TRIZOL reagent (Invitrogen). Degenerate reverse transcription polymerase chain reaction (RT-PCR) was performed to amplify fragments of respective genes. For the LjHh genes, the degenerate primers 5′-GARGGNTGGGAYGAYGAYGGNCAYCA-3′ (forward) and 5′-ATGCGGTACAGCAGWCGCGAGTACCARTG-3′ (reverse) were designed based on the amino acid sequences EGWDEDGHH and HWYSRLLYRI, respectively. For LjIsl1/2 genes, the degenerate primers 5′-TTCAGCAAGACGGACTTCGTNATG-3′ (forward) and 5′-CGCGAACTCGCTGAGSGTYTTCCA-3′ (reverse) were designed based on the amino acid sequences FSKTDFVM and WKTLSEFA, respectively. For the LjGliA, LjSp8/9A and LjNkx2.2 genes, we used specific primers designed based on the nucleotide sequence of putative orthologue sequences found in the draft genome obtained by the Petromyzon marinus genome project (http://genome.wustl.edu/pub/organism/Other_Vertebrates/Petromyzon_marinus/). The LjPtcA gene was obtained using specific primers based on the nucleotide sequence of P. marinus Patched (Hammond and Whitfield, 2006). The PCR products were cloned into the pCRII-TOPO vector (Invitrogen). Amplified fragments were sequenced with the 3130 Sequence Analyzer (Applied Biosystems). The 5′ and 3′ ends were amplified with the GeneRacer Kit (Invitrogen). The cDNA sequences identified here have been deposited in GenBank under accession numbers AB583548-AB583554. Cloning of LjSproutyA (AB586026) and LjLhx6/7/8A (AB498801) will be published elsewhere.

Whole-mount in situ hybridization was performed as described previously (Takio et al., 2007) with minor modifications. Hybridization and post-hybridization washes were performed at 70°C to avoid non-specific cross-hybridization among paralogues. Some embryos were embedded in Tissue-Tek compound (Sakura, Japan) and sectioned using a cryostat microtome (HM505E; MICROM, Germany). Section in situ hybridization was performed in a Ventana automated machine (Roche).

Cyclopamine (Calbiochem) was dissolved in ethanol (EtOH) at 20 mM. SU5402 (Calbiochem) was dissolved in DMSO at 10 mM. Embryos were punctured with sharpened forceps and treated with 100 μM cyclopamine at 23°C or with 20 μM SU5402 at 16°C. Treatments were performed in 12-well plates, 50 embryos per well, in 1 ml of 10% Steinberg's solution. No effects were observed by exposure to DMSO or EtOH vehicle alone at the same concentrations used for the experimental treatments. Treated embryos were collected and fixed in 4% paraformaldehyde in PBS for in situ hybridization and sectioning.

We identified two Hh homologues from L. japonicum, designated LjHhB and LjHhC, in addition to LjHhA (Uchida et al., 2003). The genome of a closely related species, P. marinus, does not reveal any Hh homologues other than the orthologues of LjHhA, B and C. Two Hh homologues have been identified from another closely related species, Lampetra fluviatilis (Kano et al., 2010). Although the lack of sequence detail hinders confirmation of their orthology to any of the LjHh genes, LjHhA and LjHhB probably correspond to LfHha and LjHhb, respectively.

Our molecular phylogenetic analysis supported the orthology of LjHhA to gnathostome Shh (see Fig. S1 in the supplementary material). By contrast, LjHhB and LjHhC appeared to be products of a gene duplication unique to the lamprey (or cyclostome) lineage (see Fig. S1 in the supplementary material). Although not strongly supported, these two genes showed a closer relationship to desert hedgehog (Dhh) (see Fig. S1 in the supplementary material). Overall, this phylogenetic analysis resulted in many tree topologies supported with similarly high likelihood.

We performed expression analyses of the Hh genes LjHhB and LjHhC together with the previously identified LjHhA. LjHhA was expressed in the midline mesoderm from stage 18 (neural fold stage) onwards (Fig. 1A,A′). At this stage, LjPtcA, encoding the Hh receptor, was expressed only in the anterior boundary between neural and non-neural ectoderm (see Fig. S5E,F in the supplementary material). However, at stage 19, expression of LjPtcA appeared in the ventral neural ectoderm adjacent to the Hh-expressing mesoderm (see Fig. S5G,H in the supplementary material). At stage 21, LjHhA was expressed in the notochord, the floor plate and particularly strongly in the prechordal plate (Fig. 1B,B′). In a stage 23 embryo, the notochordal expression of LjHhA started to disappear in a rostral to caudal direction, while it started to be expressed in the pharyngeal endoderm (Fig. 1C). By stage 24, LjHhA transcripts had not been detected in the neural tube rostral to the zona limitans intrathalamica (zli), but they began to be detected in the hypothalamus from stage 26, which was discontinuous from the more caudal expression domain (Fig. 1D,E) (Osorio et al., 2005). Otherwise, transcripts were also distributed in the endostyle (Fig. 1E).

LjHhB was not expressed in the notochord, but it started to be expressed in the floor plate and the prechordal plate at stage 21 (Fig. 1G,H,H′). At stage 23, it was also expressed in the ventral part of the forebrain. Unlike the report on LfHh expression by Osorio et al. (Osorio et al., 2005), this expression domain of LjHhB was apparently continuous posteriorly with that in the floor plate (Fig. 1I), reminiscent of a gnathostome embryo pattern. Its expression in the pharyngeal endoderm peaked adjacent to the stomodeal ectoderm (Fig. 1J,K). Expression of LjHhC could not be detected by in situ hybridization, so its expression level must be extremely low. From these results, LjHhA and LjHhB exhibited expression patterns similar to those seen in gnathostomes for diencephalic expression but showed slight differences in some minor domains. By observing the expression patterns of Pax6 and Dlx cognates, the lamprey telencephalon is probably regionalized into a pallium and a subpallium (Murakami et al., 2001). Histologically, LjPax6 and LjDlxA expression domains in the stage 26 telencephalon were complementary in a DV fashion (see Fig. S4A,B in the supplementary material) (Kuraku et al., 2010; Murakami et al., 2001).

In the lamprey, the boundary between the telencephalon and the diencephalon can be found in the anterior intraencephalic sulcus (sa in Fig. 2A) (von Kupffer, 1906; Murakami et al., 2001). In the lamprey forebrain before embryonic stage 26, although the LjHh genes were expressed in a similar manner to those in gnathostomes, especially in the floor plate and zli, they were not expressed in the ventral telencephalon (Fig. 2A,B). To rule out possible heterochrony specifically associated with the telencephalic upregulation of these genes, we performed expression analyses on sections of a stage 30 ammocoete larva, which also gave no signals (data not shown). In gnathostome embryos, Nkx2.1 expression is maintained by Shh signals to specify the MGE (Machold et al., 2003). LjNkx2.1, the lamprey homologue of Nkx2.1, is also expressed in the hypothalamus but not in the telencephalon (Murakami et al., 2001; Ogasawara et al., 2001; Uchida et al., 2003) (Fig. 2F). To exclude the possible involvement of unknown LjHh paralogues, the expression of LjPtcA, the lamprey homologue of Ptc and Nkx2.2, was studied. In gnathostomes, Ptc is both a receptor for Hh signalling and a transcriptional target of the Hh pathway; Nkx2.2 is induced by Hh signals expressed caudally to the diencephalon and is required for specification of the p3 domain in the spinal cord (Dessaud et al., 2008). For both genes, expression patterns were similar to those in gnathostomes in the diencephalon and more caudal domains but were undetectable in the telencephalon (Fig. 2D,E).

In gnathostomes, LIM-homeodomain protein-encoding genes Lhx6 and Lhx7 (also called Lhx8) play a role in the differentiation of interneurons derived from the MGE; Lhx6 contributes to GABAergic interneurons, whereas Lhx7 is involved in the formation of the striatal cholinergic interneuron (Marin and Rubenstein, 2001). Therefore, we isolated LjLhx6/7/8A, a putative homologue of gnathostome Lhx6/7/8 (see Fig. S3C in the supplementary material). Although LjLhx6/7/8A was expressed in the oral mesenchyme as in gnathostomes (Grigoriou et al., 1998), it could not be observed in the telencephalon (Fig. 2G,H). From these results, we conclude that the lamprey telencephalon lacks the Hh, Nkx2.1 and Lhx6/7/8 expression domains, together with a region corresponding to the gnathostome MGE as suggested by Murakami et al. (Murakami et al., 2001).

Targeted disruption of Nkx2.1 in the mouse leads to a smaller MGE and a concomitant expansion of the LGE (Sussel et al., 1999). Thus, the entire lamprey subpallium might be equivalent to the LGE because this animal lacks expression domains for Hh and Nkx2.1 (Murakami et al., 2005; Osorio et al., 2005). However, it is unknown whether this region possesses the expected developmental traits as in the LGE.

Gsh2 (Gsx2 – Mouse Genome Informatics) is known to be necessary for regional specification of the LGE in the mouse (Hebert and Fishell, 2008). The orthology of LjGshA to gnathostome Gsh2 was confirmed (Kuraku et al., 2009). When we observed the embryonic expression of this gene in the lamprey, it was expressed in the subpallium in a pattern similar to that observed for LjDlxA (Fig. 3B,C).

In the mouse, a zinc finger transcription factor-encoding gene, Sp8, is expressed in the dorsal LGE, which might be a prerequisite for the differentiation of olfactory bulb interneurons (Waclaw et al., 2006). Expression of Isl1 is also necessary for the differentiation of striatal projection neurons in the LGE (Stenman et al., 2003). Therefore, we isolated Sp8 and Isl1 cognates in the lamprey to observe their embryonic expression patterns. We identified LjSp8/9A as a member of the Sp8/9 subfamily, being clustered with the gnathostome Sp8 (see Fig. S3B in the supplementary material). It was detected slightly dorsally in the subpallium at stage 26 (Fig. 3D). Transcripts of LjIsl1/2B, a member of the Isl1/2 subfamily, were also present in the lamprey subpallium (Fig. 3E; see Fig. S3A in the supplementary material). Thus, the lamprey embryonic subpallium exhibits a similar gene expression profile to that observed in the gnathostome LGE.

Hh signalling is crucial for the DV patterning of the neural tube. The signal is induced by the notochord and emanates ventrally from the floor plate, counteracting the dorsally originated factor Gli3 (Lupo et al., 2006). Shh restricts the dorsalizing function of Gli3 and controls the positioning of the DV boundary (Hebert and Fishell, 2008). As with gnathostomes, Hh expression patterns in the lamprey CNS are distributed in the diencephalon as well as in the ventral moiety of the more posterior neural tube. Curiously, no transcripts were observed in the telencephalon, unlike in gnathostomes. This suggests that in the lamprey, Hh signalling might not act in the DV patterning of the telencephalon. To test this, we isolated a Gli3 homologue in the lamprey.

LjGliA clustered with gnathostome Gli3 (see Fig. S2A in the supplementary material). Its transcripts were detected in the dorsal neural fold (see Fig. S5I-L in the supplementary material), the dorsal neural tube and in the dorsal brain (Fig. 4C,C′) including the telencephalon (Fig. 4C″). Therefore, it seems that the isolated LjGliAcorresponds functionally to the gnathostome Gli3 that represses Hh signalling. Thus, LjGliA is expressed in a gnathostome-like dorsally restricted pattern when the ventral Hh signal is absent. What could be the ventral signal for the lamprey telencephalon?

In some gnathostomes, Shh involved in DV patterning of the telencephalon can be found in the prechordal mesoderm, and its inductive function appears only in a small window of development. In the chicken, it can be detected before Hamburger-Hamilton (HH) stage 6 (Gunhaga et al., 2000), in the mouse before embryonic day (E) 9 (Fuccillo et al., 2004) and in the zebrafish before the eight-somite stage (Danesin et al., 2009). Curiously, during these stages, Shh expression in the telencephalon is not detected in any of these animals, possibly implying that the telencephalic Hh expression is not necessarily involved in the initial DV patterning of the gnathostome telencephalon.

In stage 18 and 19 lamprey embryos, LjHhA was already upregulated in the midline mesoderm corresponding to the notochord and prechordal plate (Figs 1A,A′; see Fig. S5A-D in the supplementary material), whereas at these stages, LjHhA, LjNkx2.1 and LjNkx2.2 could not be detected in the neural ectoderm (see Fig. S5A-D,O-R in the supplementary material). Hh expression in the prechordal plate persisted up to stage 21 (Fig. 1B′,H′). To detect whether these expressions are responsible for DV patterning of the telencephalon, we administered cyclopamine, an Hh signal inhibitor (Chen et al., 2002; Taipale et al., 2000), to early lamprey embryos. Because LjHhA starts to be expressed from stage 18, we studied stage 17 embryos corresponding to the neural plate stage of the lamprey. As negative controls, we employed stage 20 embryos just after neurulation.

In the embryos treated from stage 17, the ventral part of the telencephalon was smaller, whereas the dorsal half was enlarged (Fig. 6A,B). The morphology of the ventral domain was also altered in the diencephalic region (Fig. 6C,D). In these embryos, expression of putative dorsal specifiers such as LjPax6 and LjGliA expanded ventrally at stage 26, covering almost the entire telencephalon (Fig. 5A,B,E,F). On the other hand, the ventrally upregulated LjDlxA and LjSp8/9A had lost their transcripts in the experimental embryos (Fig. 5C,D,G,H). However, embryos treated from stage 20 did not show any difference in the developmental patterns from non-treated embryos (Fig. 5I-L). Therefore, we conclude that the prechordal mesoderm-derived Hh signal is responsible for DV patterning of the lamprey telencephalon by stage 20 (before neurulation), and its inhibition results specifically in the ventral patterning of the telencephalon, possibly via dorsalization of the telencephalon. In the Shh^−/− mutant mouse, there is no Fgf expression because of the repressive action of Gli3 (Aoto et al., 2002). Curiously, in the cyclopamine-treated lamprey embryos, LjFgf8/17 and LjSproutyA, a putative downstream target of the FGF signalling, were still expressed in the ventral telencephalon, as in controls (Fig. 8).

Shh signalling in the telencephalon patterning might suppress dorsalization by repressing GLI3 and might be insufficient for subpallial patterning (Rallu et al., 2002). Thus, in the Shh^−/−Gli3^−/− double knockout (KO) mouse, DV patterning of the telencephalon was restored and expression of subpallial genes was maintained. Therefore, the required inductive signal for the subpallium could be the FGF signal (Gutin et al., 2006).

In the lamprey telencephalon, LjFgf8/17 were expressed from stage 24, and by stage 26, they were active in the anteroventral telencephalon (Fig. 4E,E′) (Guerin et al., 2009; Uchida et al., 2003). To test whether FGF signalling is responsible for the lamprey subpallium patterning, we applied SU5402, a potent FGF receptor (FGFR) inhibitor (Mohammadi et al., 1997), onto embryos from stage 24 to stage 26. The experimental embryos had lost the expression of LjSproutyA, a downstream gene of FGF and a putative homologue of the gnathostome Sprouty (Komisarczuk et al., 2008) (Fig. 7E,J), suggesting that SU5402 is also capable of inhibiting lamprey FGF signalling. At stage 26, the expression of LjPax6 expanded ventrally, whereas that of LjDlxA, LjGshA and LjSp8/9A was lost in embryos treated with SU5402 between stages 24 and 26 (Fig. 7A-D,F-J). Thus, besides the early Hh signalling shown previously, FGF signalling is responsible for DV patterning of the telencephalon in the lamprey.

Our degenerate RT-PCR amplified two novel Hh homologues. In general, the landscape of regulatory gene repertoires in the lamprey has been obscured by insufficient gene sampling and by ambiguous signatures in lamprey sequences when assigning their orthology (Kuraku, 2008). Nonetheless, our analysis supported the orthology of LjHhA to gnathostome Shh genes (see Fig. S1 in the supplementary material), together with Gli, Ptc, Isl1/2 and Sp8/9 (see Figs S2 and S3 in the supplementary material), for which we also observed gene duplications in early vertebrate evolution. Thus, the last common ancestor of all extant vertebrates had already experienced gene duplications resulting in multiple subtypes with differential expressions (Kuraku et al., 2009). This needs to be confirmed with reinforced data sets after thorough gene sampling in genome-wide resources of hagfishes and lampreys.

The ventral telencephalon of the lamprey does not express LjNkx2.1 or LjHhA, leading to the hypothesis that this animal does not possess a mechanism to specify the MGE (Murakami et al., 2001; Osorio et al., 2005). Here, we isolated two more LjHh genes, Nkx2.2 and Ptc homologues of the lamprey. Their expression patterns indicated the absence of the Hh expression domain in the ventral telencephalon.

In gnathostomes, the MGE gives rise to GABAergic interneurons in a Hh-dependent manner, which later migrate cortically (Marin and Rubenstein, 2001) and play important roles in ensuring the function of the gnathostome cerebral cortex. This migration has been observed in the chicken (Cobos et al., 2001), turtle (Métin et al., 2007), Xenopus (Moreno et al., 2008) and shark (Carrera et al., 2008) embryos. Thus, this pattern of telencephalic development is widespread among gnathostomes. In gnathostomes, cholinergic interneurons are also generated in the MGE and migrate to the striatum (Marin and Rubenstein, 2001). However, no migratory interneurons have been identified in the lamprey. Lhx6 is required for the migratory GABAergic interneurons, whereas Lhx7/8 regulates development of the migratory striatal cholinergic interneurons (Marin and Rubenstein, 2001). A lack of expression of at least one of the Lhx6/7/8 homologue genes, LjLhx6/7/8A, was observed in the telencephalon (Fig. 2G,H). Although GABA-immunoreactive cells and cholinergic neurons were found in the pallium (Melendez-Ferro et al., 2002; Robertson et al., 2007) and in the striatum (Pombal et al., 2001) in the developing and adult lamprey, respectively, it is unclear whether these cells are migrated interneurons or projection neurons. Based on the expression of LjLhx6/7/8A, we speculate that these cells are pallial projection neurons or migrated interneurons, which are generated through a different specification mechanism from that of gnathostomes. Moreover, they lack the pallidum, the MGE derivative identified in the telencephalon (Nieuwenhuys and Nicholson, 1998). Altogether, it appears most likely that MGE and migrating interneurons derived from MGE were acquired secondarily in the gnathostome lineage after divergence from the lamprey lineage (Fig. 9), because the key regulatory genes for MGE development (Hh and Nkx2.1) and for migrating interneurons (Lhx6/7/8) are absent in the lamprey telencephalon.

In mammals, migrating interneurons are also generated in the caudal ganglionic eminence (CGE) of the basal telencephalon (Nery et al., 2002). In the mouse, ~30% of all cortical interneurons are CGE derived (Miyoshi et al., 2010). However, it is unknown whether other gnathostomes possess a CGE. We could not identify a CGE in lampreys in this study.

Expression of Gsh2 in the gnathostome LGE is necessary for its development (Toresson and Campbell, 2001; Yun et al., 2003). Here, the expression of LjGshA, the Gsh2 orthologue in the lamprey, colocalized with the region of the subpallium that also expresses LjDlxA (Fig. 3B,C), suggesting that the LGE is present in the lamprey telencephalon.

The gnathostome LGE is also known as the developmental source for striatal projection neurons and olfactory bulb interneurons (Wichterle et al., 2001). Normal expression levels of Isl1 and Sp8 are required for the differentiation of these neurons, respectively (Stenman et al., 2003; Waclaw et al., 2006). Lamprey cognates of these genes (LjIsl1/2B, LjSp8/9A) are also expressed in the ventral telencephalon (Fig. 3D,E), and this animal also possesses olfactory bulb interneurons and striatal cholinergic neurons, as found previously (Nieuwenhuys, 1967; Pombal et al., 2001). Although Isl1 is also expressed in MGE, Isl1-expressing cells require Nkx2.1 and Lhx6/7 (Fragkouli et al., 2009), and these genes are absent in the lamprey telencephalon. Therefore, it appears that the expression of LjIslet1/2B corresponds to that of LGE and not to that of MGE in gnathostomes.

Altogether, the subpallium domain of the lamprey appears to possess properties similar to the gnathostome LGE, which suggests that this domain was established in the last common ancestor of cyclostomes and gnathostomes.

Inhibiting Hh signalling with cyclopamine from embryonic stage 17 led to a change in the developmental pattern of the ventral telencephalon. This was associated with the ventral expansion of the LjPax6/LjGliA domain that is normally restricted to the dorsal telencephalon, as well as the loss of LjDlxA/LjSp8/9A expression normally observed in the ventral telencephalon (Fig. 5). This treatment also led to the anatomical pattern of the ventral forebrain (Fig. 6). These effects were only observed when the treatment was applied from stage 17, and were not observed when cyclopamine was administrated after stage 20 (Fig. 5). Moreover, LjPtcA was expressed from stage 19 in the ventral forebrain (see Fig. S5H in the supplementary material), whereas expression of the Hh genes was observed only in the mesoderm from stage 18 (Fig. 1A,A′; see Fig. S5A-D in the supplementary material), suggesting that the forebrain might receive Hh signalling from the prechordal mesoderm at this stage. Therefore, it appears to be most likely that Hh signalling is functional in DV patterning of the telencephalon at early stages before the completion of neural tube closure in this animal, as in other gnathostomes (Danesin et al., 2009; Fuccillo et al., 2004; Gunhaga et al., 2000).

In gnathostomes, there appear to be two phases in telencephalon patterning in terms of Shh expression patterns. The first is the inductive activity derived from the prechordal plate, leading to the DV patterning of the telencephalon before E9 of mouse development. The second depends on expression in the MGE anlage to function in the maintenance of Nkx2.1 expression, active after E9 (Fuccillo et al., 2006). In lampreys, because there are no Shh expression domains in the telencephalon anlage, the second phase of Shh signalling might represent a function that was new to gnathostomes in evolution. In other words, the first phase of action of Shh would be primary and more ancestral than the second function, most probably possessed by the common ancestor of vertebrates. The secondary acquisition of the telencephalon-derived Shh signalling appears to be a key innovation that would have permitted further evolution of the MGE and cortical interneurons specific for gnathostomes (Fig. 9).

Recent studies have demonstrated that not only Hh but also FGF signalling is involved in DV patterning of the telencephalon in gnathostomes. In the Shh^−/−Gli3^−/− double KO mouse, ventral patterning is restored greatly over that observed in single Shh mutants, suggesting that Hh-independent signals are involved in the DV patterning of the telencephalon (Rallu et al., 2002). FGF signalling has important roles in this process (Gutin et al., 2006).

Here, LjFgf8/17 was expressed from embryonic stage 24 in the ventral forebrain (Fig. 4D) and upregulated in the anterior part of the ventral telencephalon at stage 26 (Fig. 4E,E′). To determine whether FGF signalling is responsible for this DV patterning, we applied an inhibitor against FGF signalling from stages 24 to 26 (Fig. 7). The expression pattern of LjPax6, a dorsal marker, expanded ventrally, whereas the ventral telencephalon markers LjDlxA, LjGshA and LjSp8/9A disappeared. Thus, FGF signalling is likely to be responsible for DV patterning of the telencephalon, in addition to early Hh signalling in the lamprey. We also estimated the effective time window of this signalling during lamprey development. Hh signalling during neurulation (stages 18-20) is essential for DV patterning in the forebrain, and FGF signalling is likely to act between stage 24 and 26. Actual changes in DV patterning were observed at stage 26.

Generally, FGF signalling also controls cell proliferation (Mason, 2007). However, the lamprey telencephalon did not seem to be reduced in size by inhibiting FGF signalling (Fig. 7). Furthermore, inhibition experiments of FGF signalling using zebrafish embryos resulted in no significant changes in cell proliferation in the dorsal or ventral telencephalon (Shinya et al., 2001). Therefore, inhibiting FGF signalling might not affect cell proliferation in the lamprey telencephalon at these stages.

Although the Shh^−/− mutant mouse shows loss of Fgf expression, the Shh^−/−Gli3^−/− double mutant restores Fgf expression (Aoto et al., 2002). It is thus conceivable that Gli3 represses FGF signalling (Gutin et al., 2006). By contrast, FGF signalling did not seem to be repressed in the lamprey by blocking Hh signalling (Fig. 8). It seems likely that these phenomena represent differences in developmental mechanisms for the telencephalon between gnathostomes and lampreys.

We propose an evolutionary scenario for the vertebrate telencephalon (Fig. 9). This domain is most likely to have been established in the vertebrate ancestor before the dichotomy between the gnathostome and cyclostome lineages. There would have already been a distinction between the pallium and the subpallium. It is likely that the ventral telencephalon of the ancestor would have possessed a function similar to that of the gnathostome LGE. Even in the absence of the Hh expression domain in that region, Hh signals emanating from the prechordal plate and FGF signals from the ventral telencephalon would have functioned in the DV patterning of the telencephalon. After the divergence of the cyclostome and gnathostome lineages, an Hh-Nkx2.1 expression domain arose in the gnathostome ventral telencephalon de novo, which led to an apomorphic specification of the MGE in this animal lineage independently. Acquisition of an MGE provided migratory neuroblasts for cortical interneurons, leading to the sophisticated function of the gnathostome pallium.

Finally, how could the MGE region have been established in the gnathostome lineage? In this regard, both in gnathostomes and in lampreys, transcripts of Nkx2.1 and Hh colocalize in the hypothalamus in the ventral diencephalic domain, which appears as a serial homologue of the pallidum. This suggests that co-option of a gene network to pattern the hypothalamus into a more rostral domain (the ventral telencephalon) might have generated the MGE in the gnathostomes de novo. To test this hypothesis, we need to perform further comparisons of gene expression profiles between the hypothalamus and the pallidum, and to investigate the regulatory mechanisms of these genes in the ventral forebrain. Further studies on regulatory genes in the lamprey will provide valuable insights into the evolution of the vertebrate forebrain.

We are grateful to Tsukasa Shimojo, a member of the Fishery Association of Shiribetsu River, for collecting lampreys. We also thank Kinya G. Ota, Masaki Takechi, Yasuhiro Oisi, Noritaka Adachi and Hiroki Higashiyama for maintaining the aquarium facilities. We thank Rie Kusakabe, Motoki Tada and Hiroshi Nagashima for technical advice. The sea lamprey P. marinus genomic data were produced by the Genome Center at the Washington University School of Medicine in St Louis.