Patterning the zebrafish diencephalon by the conserved zinc-finger protein Fezl

The vertebrate forebrain is a tremendously complex structure and carries out essential functions such as regulating emotions, learning and memory, and hormonal homeostasis. It arises from anterior neuroectoderm during gastrulation. By the end of somitogenesis, distinct subdivisions are morphologically visible. At least two sequential events are thought to occur that lead to the establishment of these distinct forebrain subdivisions(Kiecker and Lumsden, 2005; Rubenstein et al., 1998; Stern, 2001; Wilson and Houart, 2004): (1)the anterior neural tissue acquires a crude initial regional identity by avoiding exposure to caudalizing factors, such as Wnt and FGF proteins. This initial regionalization also establishes local organizing centers in the neural plate, such as the anterior neural border [ANB, also know as the anterior neural ridge (ANR)] and the midhindbrain boundary (MHB). (2) These local organizing centers further refine initial regional patterning and lead to the establishment of subdivisions that later give rise to various structures in the mature CNS.

Forebrain regionalization is best understood in the context of telencephalic development (Rallu et al.,2002; Wilson and Houart,2004; Wilson and Rubenstein,2000). It has been shown previously that the establishment of telencephalic identity requires local suppression of Wnt signaling - an evolutionarily conserved pathway that regulates diverse processes, including embryonic patterning, cell fate determination, cancer and synaptogenesis(Moon et al., 2002; Patapoutian and Reichardt,2000). In zebrafish, the proper development of the telencephalon requires the secreted Wnt antagonist Tlc(Houart et al., 2002), the transcriptional repressor Headless and/or Tcf3 (also known as Tcf7l1a -Zebrafish Information Network) (Kim et al., 2000) and the Wnt-pathway scaffolding-protein Masterblind and/or Axin1 (Heisenberg et al.,2001).

Compared with the telencephalon, much less is known as to how various subdivisions within the diencephalon are established. Although the embryonic diencephalon has been proposed to have a segmental organization(Figdor and Stern, 1993; Puelles and Rubenstein, 2003),cell-lineage restriction boundaries are not apparent between some of the segments (Larsen et al.,2001). During embryogenesis, a transient boundary-like structure called the zona limitans intrathalamica (ZLI) is present between the prethalamus and the thalamus. This region divides the diencephalon into the anterior (the prethalamus) versus the posterior, or caudal (the thalamus and pretectum) territories (Kiecker and Lumsden, 2005). In chick, the ZLI is predicted by the absence of lunatic fringe and the presence of Wnt8b expression, and, starting from the mid-somitogenesis stage, is demarcated by the expression of Shh, which has been shown to play a role in the proper maturation of the vertebrate diencephalon (Kiecker and Lumsden, 2004; Scholpp et al.,2006). High levels of Wnt signaling have also been suggested to promote posterior diencephalic fates (Braun et al., 2003; Kiecker and Niehrs, 2001; Masai et al.,1997; Nordstrom et al.,2002). However, the genes and pathways that establish the ZLI and other diencephalic subdivisions are elusive. Based on mis-expression studies(Kobayashi et al., 2002), it has been proposed that crossrepression between the Irx family of homeodomain proteins - which are expressed in the prospective caudal diencephalon(including the thalamus and pretectum) and midbrain(Lecaudey et al., 2005) - and Six3 homeodomain transcription factors - which are detected early in the entire anterior forebrain anlage and later mainly in the optic stalk and eye regions (Seo et al., 1998) -may contribute to the establishment of the ZLI and other diencephalic subdivisions. However, six3 (also known as six3a - Zebrafish Information Network) expression is dynamic and regresses rostrally as development progresses (Kobayashi et al.,2002; Seo et al.,1998), leaving the anterior diencephalic domain free of both irx gene family and six3 expression.

The zebrafish fezl gene was identified as a zinc-finger-containing gene induced by Dkk1 (Hashimoto et al.,2000), a secreted antagonist of Wnt signaling(Glinka et al., 1998). fezl has a paralog named fez, and both genes exhibit remarkable evolutionary conservation from flies to men(Matsuo-Takasaki et al.,2000). Little is known about the role of the DrosophilaFezl protein. In mammals, Fezl and Fez have been shown to regulate cortical neuronal differentiation(Chen et al., 2005a; Chen et al., 2005b; Hirata et al., 2004; Molyneaux et al., 2005) and olfactory-bulb development (Hirata et al.,2006b), respectively, but their early roles in brain regionalization are unclear, probably as a result of their extensively overlapping expression patterns. In zebrafish, a hypomorphic allele of fezl, too few (tof), reveals its role in monoaminergic neuron development (Guo et al.,1999; Jeong et al.,2006; Levkowitz et al.,2003). Moreover, the expression of zebrafish fezl (but not fez) is detected in distinct domains of forebrain primordia in early zebrafish embryos (Hashimoto et al.,2000) (Fig. 1, and data not shown). These observations intrigued us to investigate whether fezl has a role in early forebrain regionalization.

Here, we report that fezl is expressed exclusively in the presumptive telencephalon, diencephalon and hypothalamus shortly after gastrulation. Reduced activity of fezl results in a deficit of the prethalamus and a corresponding anterior expansion of the ZLI. Gal4-UAS-mediated fezl overexpression in late gastrula is capable of expanding the prethalamus, optic stalk, telencephalon and hypothalamus at the expense of the eyes, ZLI and posterior fore- and/or mid-brain regions. The enlargement of these forebrain subdivisions is preceded by an early downregulation of wnt expression in the prospective diencephalon. Finally, fezl overexpression is able to restore anterior forebrain and downregulate wnt1 expression in Headless- and/or Tcf3-deficient embryos. Our findings reveal a crucial role of Fezl in establishing regional subdivisions within the diencephalon, and also uncover the capability of Fezl in repressing Wnt proteins and in promoting the development of the telencephalon and hypothalamus.

The UAS-fezl transgene was cloned into the transposon vector pT2KXIG (Kawakami et al.,2004). The resulting plasmid DNA was injected, together with transposase RNA, into one-cell-stage embryos. Zebrafish embryos were raised at 28.5°C and staged according to Kimmel et al.(Kimmel et al., 1995). UAS-fezl-carrying transgenic founders were identified by PCR on pooled progeny and propagated by crossing with wild-type fish. UAS-fezl transgenic lines were subsequently crossed with hsp-gal4 containing transgenic fish to yield double-transgenic embryos.

The sequences for fezl splicing-blocking morpholinos (MOs) and the effectiveness of MOs to block fezl-RNA splicing was as previously described (Jeong et al.,2006). Embryos were injected with 3-4 nl of 0.1 mM fezlsMO at the one-to four-cell stages. The hdl/tcf3 MO was synthesized according to published information (Dorsky et al., 2003), and 3-4 nl of 0.3 mM was used per embryo.

fezl expression was first detected at ∼75% epiboly exclusively in anterior neuroectoderm (Fig. 1A); by the tailbud stage, it was confined to the prospective telencephalon, hypothalamus and diencephalon(Fig. 1B). Its expression in these regions was maintained throughout the segmentation stages(Fig. 1C,D). The expression of fezl overlapped with the telencephalic marker foxg1(Fig. 1E and see Fig. S1A in the supplementary material) and anterior forebrain marker six3(Fig. 1F and data not shown,and see Fig. S1B in the supplementary material) but not with the eye field marker rx3 (see Fig. S1C,D in the supplementary material). During early somitogenesis, whereas six3 and fezl expression overlapped in the prospective telencephalon, only fezl expression was detected in the prospective diencephalon(Fig. 1F and see Fig. S1E,F in the supplementary material). By late somitogenesis stages, fezl was strongly expressed in the dorsal telencephalon, prethalamus and hypothalamus(Fig. 1D), whereas six3 expression was largely confined to the optic stalk and eye region (see Fig. 1G,H in the supplementary material). Compared with posteriorly expressed genes, the fezl domain at the tailbud stage was separated from that of wnt1 and wnt8b by an approximately two- to three-cell diameter space; at least part of this area may represent the presumptive ZLI (Fig. 1G,H). By early somitogenesis, the fezl-expressing domain was separated by a gap (the presumptive ZLI) from wnt1-expressing cells in the roof plate and MHB (Fig. 1I,K), and from irx3a-expressing cells in the posterior diencephalon (Fig. 1L);moreover, fez1 expression abutted wnt8b expression in the prospective ZLI (Fig. 1J). These analyses indicate that fezl expression is initiated early and exclusively in the developing forebrain, and is subsequently maintained in discrete forebrain subdivisions throughout the somitogenesis stages. These observations led us to hypothesize that fezl might have a role in forebrain regionalization.

To determine the role of Fezl in forebrain development, we knockeddown fezl activity with two specific splicing morpholinos (MOs)(Jeong et al., 2006): both MOs gave similar results, whereas the control MOs had no effect. fezlMO-injected embryos (hereafter referred to as morphants) were examined at∼28 hours post fertilization (hpf), when brain subdivisions are distinct by morphology and gene expression; they were also examined at earlier stages(neural-plate and mid-somitogenesis) in order to better define the timing of fezl action.

In ∼28 hpf fezl morphants, we found that dlx2a(Akimenko et al., 1994)expression in the prethalamus was significantly reduced in a dose-dependent fashion (Fig. 2B, 97%, n=65; and data not shown), while remaining unaffected in the pharyngeal arch progenitors (insets of Fig. 2B). In addition to dlx2a, the expression of lhx5 (Peng and Westerfield,2006; Scholpp et al.,2006; Toyama et al.,1995) in the prethalamus region was significantly reduced, but its expression in the posterior tubercular area remained largely normal(Fig. 2D, 90%, n=58). Despite the strong expression of fezl in the telencephalon and hypothalamus (Fig. 1B,D),expression of the telencephalic foxg1(Fig. 2F, 98%, n=60)and the hypothalamic nk2.1a (also known as titf1a -Zebrafish Information Network) (Fig. 2H, 98%, n=81) appeared largely normal (perhaps slightly reduced); the expression of six3b also appeared largely normal(Fig. 2J, 88%, n=33).

Analyses of early-stage fezl morphants revealed largely normal patterning at the neural-plate stage (data not shown, and see Fig. 2 in the supplementary material). However, by mid-somitogenesis, the expression of both dlx2a (Fig. 2L, 90%, n=119) and lhx5 (Fig. 2N, 85%, n=44) in the prospective anterior diencephalon was significantly reduced, and dlx2a expression in the prospective telencephalon was also reduced at this early stage(Fig. 2L, 81%, n=119),suggesting a transient and later-recoverable telencephalic defect in the fezl morphants, possibly due to genetic compensation by other anterior forebrain factors (see Discussion). Together, these results indicate that fezl is essential for the development of anterior diencephalon(the prethalamus), starting at around mid-somitogenesis.

Since proliferation and cell death were not significantly affected in the fezl morphants (data not shown), we next asked whether the prethalamic defects might be a result of altered regionalization within the diencephalon. We examined the expression of shha (also known as shh - Zebrafish Information Network, and hereafter referred to as shh) and foxa2, two markers of the ZLI, which is located immediately posterior to the prethalamus. In control embryos, a gap occupied by the prethalamus existed between shh-expressing cells and the telencephalic boundary (Fig. 3A,C,E), but in ∼28 hpf fezl morphants, this gap was filled in by expanded shh expression(Fig. 3B,D, 99%, n=83; Fig. 3F, 98%, n=46). The expression of foxa2 was also expanded to abut the telencephalon(Fig. 3H, 95%, n=86),but foxa2 expansion was noticeably most pronounced in the dorsal region, suggesting perhaps an easier transformation of the dorsal area as compared with the more ventral territory of the prethalamus. In addition to shh and foxa2, the expression of dbx1a - which is detected in the anterior diencephalon adjacent to or overlapping with the ZLI,and also in the posterior diencephalon (the thalamus)(Scholpp et al., 2006) - was significantly expanded in the anterior diencephalon(Fig. 3J, arrow, 90%, n=40), and moderately increased in the thalamic region(Fig. 3J, arrowhead, 90%, n=40). These analyses suggest that fezl activity is required to repress the fate of ZLI and possibly also the posterior diencephalon.

It has been previously reported that a slight dorsal extension of shh expression at early- to mid-somitogenesis stages probably demarcates the location of the future ZLI, and dbx1a expression is detected ventrally adjacent to the ZLI at this early stage(Scholpp et al., 2006). In∼16-somite stage fezl morphants, dbx1a expression adjacent to the ZLI appeared slightly increased(Fig. 3L, 85%, n=42);moreover, an increased and anteriorly expanded shh expression in the prospective ZLI was readily discernible, which was accompanied by the loss of the gap between the hypothalamic- and ZLI-expression, whereas its expression in the anterior hypothalamus was moderately decreased(Fig. 3N, 88%, n=50). irx3a expression, which is in the prospective caudal diencephalon and midbrain and in close proximity to shh-expressing prospective ZLI cells (Lecaudey et al., 2005),was moderately expanded anteriorly; enhanced irx3a expression was observed in a ventral location that might be overlapping with the shh-expressing prospective ZLI cells; in addition, a small cluster of irx3a⁺ cells was detected in the anterior forebrain,possibly in the prospective telencephalon or in the dorsal prethalamus(Fig. 3P, 99%, n=40). This anterior forebrain expression appeared to be transient and might represent an incomplete fate switch because it was not observed in ∼28 hpf fezl morphants (data not shown), and because other thalamic markers,such as dbx1a, were not detected in the corresponding location (data not shown). These results suggest that fezl plays a role in repressing the early ZLI and posterior diencephalic fate, beginning around the mid-somitogenesis stage.

To further elucidate the role of fezl in forebrain regionalization, we investigated the consequence of overexpressing fezl. As the delivery of fezl mRNA into one- to four-cell-stage embryos led to severe embryonic deformity (data not shown)(Levkowitz et al., 2003; Yang et al., 2001), we employed the heat inducible Gal4-UAS system(Scheer and Camnos-Ortega,1999) to achieve a temporal control of fezloverexpression. We established transgenic lines carrying the UAS-fezltransgene and crossed them with another line carrying the hsp-gal4transgene. The embryos derived from this cross were subjected to heat shock for 15, 30 or 45 minutes at ∼75% epiboly, analyzed for gene expression at∼28 hpf, and subsequently genotyped to determine the presence or absence of the transgenes. Heat shock at later stages (e.g. 10-somite and 24 hpf) had little or no effect on brain regionalization (data not shown). Because the earliest time when the endogenous fezl expression was observed is at∼75% epiboly, our analyses were focused on the embryos that were heat shocked at this stage.

Ubiquitous fezl expression was detected ∼2 hours post heat-shock through to 6 hours (Fig. S3 in the supplementary material). In the heat-shocked double-transgenic embryos, prethalamic and hypothalamic dlx2a expression was expanded in proportion to the duration of heat shock, whereas pharyngeal arch dlx2a expression was relatively normal(Fig. 4A-D, n=62). To determine whether the expansion of dlx2a-expressing domains might be at the expense of more posterior brain tissues, we examined the expression of shh. Remarkably, shh-expressing ZLI was shifted posteriorly,reduced or eliminated in the double-transgenic embryos in proportion to the duration of heat shock, whereas the floor plate expression of shh was unaffected (Fig. 4E-H, n=34). These analyses indicate that fezl overexpression is capable of expanding the prethalamus at the expense of the ZLI.

To further characterize the fezl gain-of-function (GOF) phenotype,we assessed the extent to which fezl overexpression could alter brain regionalization by examining additional region-specific genes: the expression of foxg1 (Fig. 5B, n=24) in the telencephalon and nk2.1a in the hypothalamus(Fig. 5D, n=42), as well as pax2a in the optic stalk(Fig. 5F, n=10), were all expanded; by contrast, the expression of pax2a(Fig. 5F, n=10) and engrailed 2 (also known as engrailed 2a - Zebrafish Information Network) (Fig. 5J, n=22) in the MHB, and irx3a in the caudal diencephalon and midbrain (Fig. 5H, n=10) were lost, and the eyes were significantly reduced(Fig. 5F, n=10). However, krox-20 (also known as egr2a - Zebrafish Information Network) expression in the hindbrain rhombomeres r3 and r5 remained(Fig. 5J, n=22). Therefore, it appears that ectopic expression of fezl at late gastrulation can expand the telencephalon, hypothalamus and prethalamus at the expense of the eyes, ZLI, caudal diencephalon, midbrain and MHB.

To understand better the underlying cause of the fezl GOF phenotype, we examined fezl overexpressing embryos at early stages. At the tailbud stage, the brain subdivisions are not yet morphologically discernible, but the Wnt antagonist Tlc and transcription factors Emx1 and Foxg1 are expressed at the anterior margin of the neural plate, overlapping with the expression of Fezl (Fig. 1E and Fig. 6A,C;and data not shown). Both wnt1 and wnt8b are expressed in the presumptive caudal diencephalon (Fig. 1G,H). In tailbud-stage fezl-overexpressing embryos, the expression of emx1, tlc, foxg1, six3b, wnt8b, irx3a and fgf8was unaffected (Fig. 6A-H,O-R;and data not shown). However, the expression of wnt1(Fig. 6J,N, n=34) and pax2a (Fig. 6L,N,P, n=38) was severely deficient. By the mid-somitogenesis stage, the expression of wnt8b in the prospective ZLI and MHB(Fig. 6V, n=19) and fgf8 in the MHB (Fig. 6X, n=16) was found reduced, and the expression of telencephalic fgf8, six3b, emx1 and foxg1 was found extended caudally (Fig. 6T,X; and data not shown). These observations reveal that the earliest effect of fezl overexpression is an inhibition of wnt1 and pax2a at the neural-plate stage, followed by deficits of two important organizers - the ZLI and MHB - by the mid-somitogenesis stage. These effects could underlie the GOF phenotypes observed at ∼28 hpf.

headless and/or tcf3 (hdl/tcf3), which encode a transcriptional repressor of Wnt target genes (headless being the mutant form of tcf3), has been shown previously to be crucial for the development of the forebrain: its inactivation leads to a headless phenotype and a dramatic expansion of wnt1 expression into anterior forebrain(Kim et al., 2000). Injection of six3 mRNA, a known repressor of wnt1 transcription, into Hdl/Tcf3-deficient embryos has been shown to rescue the eyes, as well as the head, based on gross morphology, but the extent of head rescue by six3 has not been examined at the molecular level(Lagutin et al., 2003).

We reasoned that, if Fezl were a crucial player in establishing forebrain subdivisions, overexpression of fezl should restore these subdivisions to Hdl/Tcf3-deficient embryos. Because the maternal zygotic hdl/tcf3 mutant had a rather variable severity of head loss (our unpublished observations), we delivered the hdl/tcf3 MO, which was previously shown to be effective and specific in knocking down hdl/tcf3 activity (Dorsky et al.,2003; Dorsky et al.,2002), to embryos derived from the hsp-gal4/+×uas-fezl cross. These hdl/tcf3morphants were subjected to heat shock at ∼75% epiboly, analyzed for forebrain gene expression at ∼28 hpf, and subsequently genotyped for the presence or absence of the transgenes. At the 3-somite stage, fezlexpression was dramatically reduced and wnt1 expression was significantly increased in the hdl/tcf3 morphants(Fig. 7A,B, n=43). fezl overexpression was sufficient to repress wnt1expression in the hdl/tcf3 morphant(Fig. 7C, 81%, n=36). Similarly, whereas the lack of hdl/tcf3 activity led to severe deficits in the expression of foxg1(Fig. 7D, 62%, n=52), dlx2a (Fig. 7G, 67%, n=42) and nk2.1a (Fig. 7J, 45%, n=29) in anterior brain regions, fezloverexpression was not only able to restore, but was also able to expand, the telencephalic foxg1 expression(Fig. 7F, 80%, n=35),prethalamic dlx2a expression (Fig. 7I, 91%, n=34) and hypothalamic nk2.1aexpression (Fig. 7L, 84%, n=13) in the hdl/tcf3 morphants. These analyses demonstrate that Fezl overexpression is sufficient to repress wnt1 and to promote telencephalic hypothalamic and prethalamic identity even in the absence of the wnt signaling repressor Hdl/Tcf3.

The establishment of subdivisions is a crucial early step towards the formation of stereotyped neuronal patterns, intricate neural connectivity and subsequent functioning of the vertebrate forebrain. Through loss-of-function(LOF) studies, we have revealed an essential role of fezl in establishing proper subdivisions within the diencephalon: fezl is required to specify the prethalamus and to repress the fate of the ZLI starting at around the mid-somitogenesis stage. GOF analyses show that Fezl is capable of expanding the prethalamus at the expense of ZLI. These LOF and GOF data nicely complement each other and demonstrate that fezl has a crucial role in diencephalic regionalization.

In addition to affecting the diencephalon, Fezl GOF is also able to expand the telencephalon and hypothalamus at the expense of eyes and other fore-and/or mid-brain regions. However, the telencephalon and hypothalamus remain largely normal in fezl morphants. Although we cannot fully rule out the possibility that the GOF phenotypes observed for the telencephalon and hypothalamus are `artifacts' due to overexpression of fez1, several lines of evidence suggest that it is not the case. First, such GOF phenotypes were not observed in embryos that were heat-shocked at later stages,suggesting that these phenotypes are not simply due to an overproduction of Fezl. Second, whereas anterior-posterior patterning is disturbed in fezl GOF embryos, other axial development (e.g. dorsoventral patterning) appears largely normal. Finally, endogenous fezlexpression was detected strongly in both the telencephalon and hypothalamus during the early development of these embryos. Therefore, one plausible explanation for the strong GOF but weak LOF phenotypes in the telencephalon and hypothalamus is that other forebrain-expressed factors may compensate for the loss of Fezl in these brain regions. One such factor might be Six3, which is essential for murine forebrain development(Lagutin et al., 2003) and medaka fish eye formation (Carl et al.,2002; Del Bene et al.,2004), and has also been shown to regulate forebrain development in zebrafish (Ando et al.,2005; Kobayashi et al.,1998). Another compensating factor may be lhx5, which was recently shown to promote forebrain development via the transcriptional activation of secreted Wnt antagonists(Peng and Westerfield, 2006). Thus, it is possible that fezl, six3 and lhx5 together may ensure the proper establishment of multiple anterior neural subdivisions that include the telencephalon, eyes, prethalamus and hypothalamus.

Because fezl expression is not detectable in dopaminergic neurons(Levkowitz et al., 2003) a cell non-autonomous function of fezl has been previously proposed for their development. However, recent evidence favors the hypothesis that fezl is expressed in the dopaminergic progenitor cells to cell autonomously control their development(Jeong et al., 2006). Several lines of evidence suggest that fezl may also act cell autonomously in the specification of the prethalamus. First, fezl is expressed in the prethalamus. Second, fezl LOF leads to a deficit of prethalamus,whereas fezl GOF leads to an expansion of prethalamus. Third, fezl has previously been shown to cell autonomously induce the expression of dlx2a, a prethalamic marker(Yang et al., 2001). Whereas fezl is likely to act cell autonomously in the specification of the prethalamus, its requirement for repressing ZLI may be cell non-autonomous because fezl appears not detected in this region.

What are the mechanisms by which fezl regulates diencephalic regionalization? One possible avenue is through repressing the target genes of the wnt and/or β-catenin signaling pathway, which has been shown to repress the anterior, and promote the posterior, neural fate. Our LOF analysis shows an expansion of irx3a, a gene induced by wnt signaling(Braun et al., 2003). Our GOF analysis demonstrates the capability of fezl to repress wnt1and pax2a, which is followed by the loss of two important organizers- the ZLI and MHB - which may be the cause of the GOF phenotypes. Moreover,the rescue of the headless/tcf3-deficient embryos by fezloverexpression further substantiates the ability of Fezl to repress wnt signaling. The largely unaffected wnt expression in the fezl morphants (see Fig. S2E-J in the supplementary material) may be due to genetic compensation, possibly provided by six3 or lhx5. Alternatively, the interaction between fezl- and wnt-signaling may not be at the level of direct transcriptional regulation of the wnt family of genes. Taken together and consistent with a suggested requirement of inhibiting wnt activity for the development of the prethalamus (Braun et al.,2003; Kiecker and Niehrs,2001; Lagutin et al.,2003), our results provide strong evidence that such repression of wnt activity can be achieved by fezl. Finally, it is worth pointing out that our data are consistent with the possibility that fezl may also play a role in directly promoting the prethalamic fate independent from repressing wnt activity.

A clear understanding of the biochemical mechanisms underlying fezl function requires the future identification of its direct target genes. Moreover, it is also of great interest to know mechanistically how fezl expression is restricted to distinct anterior forebrain subdivisions. Consistent with the Wnt gradient hypothesis in forebrain regionalization (Wilson and Houart,2004), fezl is inducible by Dkk1(Hashimoto et al., 2000),again, making the Wnt pathway an attractive candidate in regulating fezl expression. Future studies will provide crucial insights into possible cross regulations between fezl and wntsignaling.

Our study uncovers an essential role of fezl in diencephalic patterning, prior to its function in neuronal subtype differentiation(Chen et al., 2005a; Chen et al., 2005b; Guo et al., 1999; Hirata et al., 2004; Jeong et al., 2006; Levkowitz et al., 2003; Molyneaux et al., 2005). Given its restricted expression in the vertebrate forebrain and a demonstrated role in regulating reward-associated behaviors(Lau et al., 2006), it would be interesting to test in the future whether the deregulation of Fezl may be involved in human neurological disorders that have a developmental origin.

While our manuscript was under review, it was reported that the mouse Fez and Fezl proteins together also play a crucial role in the establishment of the diencephalon divisions (Hirata et al.,2006a), suggesting an evolutionarily conserved role of the Fez-and Fezlgene families. Interestingly, a notable difference exists between zebrafish and mice: whereas our LOF and GOF analyses indicate a clear role of zebrafish fezl in promoting the prethalamus and in repressing the ZLI, fez^-fezl^- double-mutant mouse embryos had a loss of both prethalamus and ZLI, but an expansion of caudal diencephalon(the thalamus and pretectum). However, similar to our GOF data in zebrafish,overexpression of mouse fezl or fez abolished shh-expressing ZLI. This difference in LOF phenotypes could be species-dependent. Alternatively, the zebrafish fezl morphant perhaps represents a weaker LOF of fezl (also with fez being intact), whereas the fez^-fezl^- double-mutant mice are null conditions. Therefore, it is attractive to hypothesize that the levels of fezl and/or fez activity may determine distinct subdivisions along the rostrocaudal axis. Future experiments are necessary to test this hypothesis.

We thank many colleagues for providing reagents and probes, especially Chongmin Wang, Moloy Goswami and Ajay Chitnis for the headless/tcf3mutant fish; and Sara Mercurio, Marina Mione, Stephen Wilson, Gilbert Weidinger, Randy Moon, Kiyoshi Kawakami, Masahiko Hibi, Gang Peng and Monte Westerfield for plasmids. We thank Ajay Chitnis, Samuel Pleasure and John Rubenstein for valuable discussions and helpful comments on the manuscript. The study was supported by the NIH.