Zinc-finger genes Fez and Fez-like function in the establishment of diencephalon subdivisions

The forebrain of adult mammals is one of the most complicated biological structures; it is essential for higher neural functions, such as memory,emotion, reasoning and the planning of coordinated movements. Various models for forebrain subdivisions (neuromeres) have been developed over the past 10 years (reviewed by Kiecker and Lumsden,2005). Among them, Puelles and Rubenstein and colleagues proposed a neuromeric organization of the forebrain on the basis of the differential expression of neural marker genes and morphological considerations (the so called `prosomeric model') (Puelles and Rubenstein, 1993; Puelles and Rubenstein, 2003; Rubenstein et al., 1994; Rubenstein et al., 1998). In the most recent model(Puelles and Rubenstein,2003), the forebrain can be divided into rostral and caudal regions. The rostral part of the forebrain is the telencephalon and can be divided into several anatomical and functional territories, including the pallium, subpallium, preoptic area and hypothalamus. The caudal part of the forebrain is the diencephalon and can be divided into three transverse subdivisions known as prosomeres: the prethalamus (also called the ventral thalamus, p3), thalamus (also known as the dorsal thalamus, p2), and pretectum(p1). The eminentia thalamic, habenula/epithalamus and posterior commissures are located dorsally to the prethalamus, thalamus and pretectum,respectively.

Although the structures of the diencephalic subdivisions become obvious in mice late in gestation, individual subdivisions can be distinguished by their expression of genetic markers at the beginning of forebrain patterning. Members of the Dlx family of genes are expressed in the prethalamus, Gbx2 is expressed in the thalamus(Bulfone et al., 1993) and Ebf1 and Lhx1 are expressed in the anterior and posterior pretectum, respectively (Barnes et al.,1994; Garel et al.,1997). Studies of genetically modified mice have revealed several genes involved in the patterning and/or development of the diencephalon. These include Pax6, Otx2, Emx2 and Six3(Kimura et al., 2005; Kurokawa et al., 2004a; Kurokawa et al., 2004b; Lagutin et al., 2003; Stoykova et al., 1996; Suda et al., 2001). However,it is largely unknown how these genes function in the formation of the diencephalon subdivisions and what other genes are involved in diencephalon patterning.

The zona limitans intrathalamica (ZLI) is located on the boundary between the prethalamus and thalamus (Larsen et al., 2001; Shimamura et al.,1995). The ZLI marks the interface between regions of different ability to respond to inductive signals such as Fgf8 and Shh; for instance, in neural tissue rostral to the ZLI, Fgf8 induces the expression of Foxg1 (also called BF-1), but caudal to the ZLI it induces the expression of En2 (Shimamura and Rubenstein, 1997). The ZLI also has inductive influences on the adjacent subdivisions (prethalamus and thalamus). Shh expressed in the ZLI is involved in development of the prethalamus and thalamus in chick and zebrafish embryos; inhibition of Shh signaling represses the expression of the prethalmus and thalamus markers(Kiecker and Lumsden, 2004; Scholpp et al., 2006; Vieira et al., 2005). In chick embryos, the future position of the ZLI is at the abutting expression domains of two homeobox genes, Six3 rostrally and Irx3 caudally(Kobayashi et al., 2002). Six3 and Irx3 regulate expression of each other, and the misexpression of Six3 or Irx3 affects the formation of the prethalamus and thalamus (Braun et al.,2003; Kobayashi et al.,2002). Irx3 expression confers competence to respond to the Fgf8 and Shh inductive signals by expressing thalamus-specific genes(Kiecker and Lumsden, 2004; Kobayashi et al., 2002). The expression of Six3 and Irx3 is regulated by Wnt signaling(Braun et al., 2003); Wnt1,Wnt3 and Wnt3a are expressed in the dorsal neural tissue caudal to the prospective ZLI when the expression of Six3 and Irx3begins (Braun et al., 2003; Lagutin et al., 2003; Roelink and Nusse, 1991; Salinas and Nusse, 1992). These reports suggest that Wnt signaling controls the rostro-caudal polarity of the forebrain through the regulation of Six3 and Irx3,and the mutually repressive interaction of Six3 and Irx3controls the position of the ZLI and thus confers differential competence to respond to the inductive signals to form the prethalamus and thalamus. This hypothesis is based mainly on the results of misexpression studies and explant assays in the chicken embryo, but there is little genetic evidence to support it. Six3-deficient mice display a strong reduction of the neural tissue rostral to the ZLI, but still express Shh in the ZLI and rudimentary rostral tissue (Lagutin et al., 2003), suggesting that other genes may cooperate with or function parallel to Six3 to determine the position of the ZLI or control the formation of the rostral diencephalon.

Fez (Fezf1 - Mouse Genome Informatics) and Fez-like (Fezl, Fezf2 - Mouse Genome Informatics) are closely related genes that encode transcriptional repressors containing six C2H2-type zinc fingers and an Eh1 (Engrailed homology 1) repressor motif, and were originally isolated as anterior neuroectoderm-specific genes in Xenopus and zebrafish (Hashimoto et al., 2000; Matsuo-Takasaki et al., 2000). Orthologs of these two genes exist in the puffer fish, zebrafish, mouse and human (Hirata et al., 2006). In mouse and zebrafish, Fez and Fezl are expressed in overlapping domains in the forebrain during development (Hashimoto et al.,2000; Hirata et al.,2006; Hirata et al.,2004; Matsuo-Takasaki et al.,2000). A number of studies have previously investigated the specific roles of Fez and Fezl in neural development in both zebrafish and mice. The zebrafish mutant too few (tfu) has a fezl gene mutation and displays a loss or reduction of dopaminergic neurons in the hypothalamus (Levkowitz et al., 2003). Fezl-deficient mice show no abnormalities in the dopaminergic neurons, but do show abnormal formation of the subplate neurons and thalamocortical axons, and loss of the fornix/fimbria system(Hirata et al., 2004). Fezl has also been shown to be required for the development of subcerebral projection neurons in layer V of the neocortex(Chen et al., 2005a; Chen et al., 2005b; Molyneaux et al., 2005). Fez-deficient mice display abnormal olfactory sensory neuronal projections and olfactory bulb formation(Hirata et al., 2006). The relatively weak forebrain phenotypes of Fez and Fezl-deficient mice suggest that Fez and Fezlfunction redundantly in the patterning and development of the forebrain.

Here, we generated mice deficient in both the Fez and Fezl genes. The Fez;Fezl-deficient mice showed defects in the rostro-caudal patterning of the diencephalon. We found that Fezand Fezl redundantly control the rostro-caudal patterning of the diencephalon by repressing the caudal diencephalon fate in the prospective prethalamic region, and that ZLI formation depends on Fez/Fezl-mediated formation of diencephalon subdivisions.

A mouse Fezl cDNA fragment (GenBank Accession Number AI325906) was originally reported as the mouse ortholog of Xenopus Fez(Matsuo-Takasaki et al.,2000). However, comparing the sequences of zebrafish fez(AB207804) and fezl (AB041824), mouse Fez homolog(AK014242), and the full-length clone of mouse Fezl (AB042399)(Hashimoto et al., 2000; Hirata et al., 2006) revealed that the mouse gene reported by Matsuo-Takasaki et al. is more similar to zebrafish fezl than to Xenopus Fez(Hashimoto et al., 2000; Matsuo-Takasaki et al., 2000). Therefore, we refer to AB042399 as mouse Fezl(Hashimoto et al., 2000; Hirata et al., 2004), and to AK014242, which is more similar to Xenopus Fez, as mouse Fez(Hirata et al., 2006). Fezl^+/- and Fez^+/- mice were previously described (Hirata et al.,2006; Hirata et al.,2004). Both the Fez^+/- and Fezl^+/- heterozygous mice were originally established in a 129sv genetic background and backcrossed to the C57BL/6 background for several generations. Fez^+/-Fezl^+/- double heterozygous mice were generated by crossing Fez^+/- and Fezl^+/- heterozygotes, and were used to generate Fez^-/-Fezl^+/-,Fez^+/-Fezl^-/- and Fez^-/-Fezl^-/- embryos. Mice were housed in an environmentally controlled room at the Animal Facility of the Center for Developmental Biology (CDB), RIKEN, under the guidelines for animal experiments from RIKEN CDB. The genotypes of newborn mice and embryos were determined by PCR analysis (Hirata et al.,2006; Hirata et al.,2004). The primer sequences and PCR conditions are available on request. Noon of the day on which the vaginal plug was detected was designed as embryonic day (E) 0.5.

Brains or embryos were fixed with Carnoy's solution at room temperature overnight. Specimens were dehydrated and embedded in paraffin. Serial sections were prepared and stained with 0.1% Cresyl Violet (MERCK).

Embryos were fixed overnight at 4°C in 4% paraformaldehyde (PFA) in PBS. Specimens were gradually dehydrated in ethanol/H₂O and stored in ethanol at -20°C. The protocol for in situ hybridization was described previously (Hirata et al.,2006; Hirata et al.,2004). Single-stranded digoxigenin-UTP-labeled (Roche) RNAs were used. In situ signals were detected with an anti-digoxigenin antibody and BM Purple (Roche). For two-color staining of the histological sections in Fig. 6D, Fez and Irx1 probes labeled with FITC-UTP and digoxigenin-UTP (Roche), were used respectively; BM Purple and Fast Red (Roche) were used for staining. For Fig. 6H-M, Fez and Fezl probes were labeled with FITC-UTP and Irx1 probe with digoxigenin-UTP. The hybridized signals for Fez and Fezlwere detected by an alkaline phosphatase-conjugated anti-Fluorescein antibody(Roche) and BM Purple, and those for Irx1 were detected by a peroxidase-conjugated anti-digoxigenin antibody (Roche) and tyramide signal amplification system (TSA-Plus Fluorescein System, PerkinElmer). The precise protocols for in situ hybridization are available on request. The probes were: Dlx1 (Bulfone et al.,1993), Gbx2 (Bulfone et al., 1993), Lhx1(Fujii et al., 1994), Ebf1 (Garel et al.,1997), Sox14(Hashimoto-Torii et al.,2003), Emx2 (Yoshida et al., 1997), Pax6(Walther and Gruss, 1991), Shh (Echelard et al.,1993), Tcf4(Ishibashi and McMahon, 2002), Lhx5 (Nakagawa and O'Leary,2001), Fgf8 (Crossley and Martin, 1995), Irx1 (a gift of T. Ogura), Wnt3a (Takada et al.,1994), En2 (Joyner and Martin, 1987), Foxg1(Tao and Lai, 1992), Fez (Hirata et al.,2006) and Fezl(Hirata et al., 2004). In situ hybridization images were taken using an AxioPlan2 microscope or a SteREO Lumar V12, equipped with an AxioCam CCD camera (Zeiss). Figures were assembled using AxioVision version 4.3 and Adobe Photoshop CS2.

We isolated an approximately 8.2 kbp enhancer/promoter region of Fezl from the bacterial artificial chromosome clone containing the coding and non-coding regions of Fezl(Hirata et al., 2004). To make the Fezl enhancer/promoter-driven lacZ(β-galactosidase) transgene constructs, the 8.2 or 2.7 kbp enhancer/promoter region of Fezl was connected to lacZ cDNA at the position of the translational initiation site of Fezl(Fezl8.2p-lacZ, Fezl2.7p-lacZ). For misexpression of Fezl,its 2.7 kbp enhancer/promoter was connected to Fezl or FezcDNA, followed by internal ribosomal entry site (IRES)-Gap43-Venus(Fezl2.7p-Fezl-IRES-Venus and Fezl2.7p-Fez-IRES-Venus)(IRES-Gap43-Venus, a gift of Y. Yoshihara). Similarly, Otx2 FM enhancer (1.4 kbp) (Kurokawa et al.,2004a) and the promoter of mouse heat shock protein 68 (pHsp68)(Sasaki and Hogan, 1996) were connected to Fezl cDNA and IRES-Gap43-Venus(OtxFM-Hsp68-FeZLIRES-Venus). The transgene-derived Fezlexpression was monitored as Venus expression with an epifluorescence microscope AxioPlan2. Genotypes of the transgenic mice were analyzed by PCR with 5′-AAACCCTGGCGTTACCCAACT-3′ and 5′-ACGACAGTATCGGCCTCAGGA-3′ for the lacZ reporter lines,5′-TGTGTCTGCAGAGAGTGCTGGCCTG-3′ and 5′-CTGGCTGCTGCTCACCCCAAGCTTT-3′ for the Fezl2.7p-Fezl-IRES-Venus and OtxFM-Hsp68-Fezl-IRES-Venuslines and 5′-AAAACGTATTTAGCCGAAAGGAAT-3′ and 5′-ACTTTACACACGAAGGGTCTGG-3′ for the Fezl2.7p-Fez-IRES-Venus lines. Transient transgenic embryos were generated and β-galactosidase staining was performed as described previously (Kimura et al.,1997; Kimura et al.,2000).

To reveal whether Fez and Fezl have redundant roles in forebrain formation, we crossed Fez- and Fezl-deficient heterozygous mice (Hirata et al.,2006; Hirata et al.,2004) and generated mice deficient homozygously or heterozygously in the Fez and/or Fezl genes(Fig. 1). Fez^-/-Fezl^+/- embryos showed a small olfactory bulb at E15.5 (Fig. 1F), as Fez^-/- embryos do(Hirata et al., 2006). Fez^+/-Fezl^-/- embryos showed defects in the formation of the dentate gyrus at E17.5(Fig. 1K). Fez^-/-Fezl^-/- embryos showed more severe phenotypes in the olfactory bulb and hippocampus: nearly complete loss of the olfactory bulb, stronger reduction of the CA3 region and loss of dentate gyrus in the hippocampus. Fez^-/-Fezl^-/- embryos also showed impaired neocortex formation. In addition, Fez^-/-Fezl^-/- embryos were found to have abnormalities in the formation of diencephalon, including loss of prethalamus and a strong reduction of the thalamus in size(Fig. 1H, arrowheads). These diencephalon defects were not observed in the Fez^-/-Fezl^+/- and Fez^+/-Fezl^-/- embryos(Fig. 1F,G), indicating a redundant role of Fez and Fezl in the formation of diencephalon. Although it is intriguing to study the phenotypes of the olfactory bulb, hippocampus and neocortex in these mice, we restricted the focus of our study to the role of Fez and Fezl in diencephalon formation.

A series of sagittal sections revealed that the Fez^-/-Fezl^-/- embryos had posterior commissures and habenulo-interpeduncular tracts at E18.5(Fig. 1P,Q), located on the dorsal side of the pretectum and on the boundary between the thalamus and pretectum, respectively (Fig. 1M,N). However, the habenulo-interpeduncular tracts in the Fez^-/-Fezl^-/- embryos tracts were small and abnormally located (Fig. 1Q). This is probably due to the strong reduction of the thalamus. The Fez^-/-Fezl^-/- embryos did not have a mammilothalamic tract, which is located in between the prethalamus and thalamus (Fig. 1O). These data indicate that the Fez^-/-Fezl^-/- embryos had no prethalamus and displayed a strong reduction of thalamus, while having a relatively normal development of the pretectum.

We next examined the Fez;Fezl-deficient mice with various genetic markers at E12.5. Dlx1 is expressed in the ventral telencephalon(including the medial and lateral ganglionic eminences), hypothalamus and prethalamus, but not the thalamus (Fig. 2A) (Bulfone et al.,1993; Stuhmer et al.,2002). In the Fez^-/-Fezl^-/-embryos, Dlx1 expression was maintained in the ganglionic eminences and hypothalamus, but was completely absent in the diencephalon(Fig. 2D). Gbx2 is expressed strongly in the thalamus and weakly in the ganglionic eminences(Fig. 2E)(Bulfone et al., 1993). In Fez^-/-Fezl^-/- embryos, Gbx2 expression in the thalamus was markedly reduced and shifted rostrally, but not abrogated(Fig. 2H). Lhx1 is expressed in the posterior pretectum and ZLI(Fig. 2I)(Barnes et al., 1994; Fujii et al., 1994; Mastick et al., 1997; Suda et al., 2001). Although its expression in the posterior pretectum was not affected significantly, we found that it was not expressed in the ZLI in the Fez^-/-Fezl^-/- embryos(Fig. 2L). Ebf1 is expressed in the anterior pretectum (Garel et al., 1997; Suda et al.,2001) (Fig. 2M),and its expression was expanded rostrally in the Fez^-/-Fezl^-/- embryos(Fig. 2P). Sox14 is expressed in the rostral part of the thalamus(Hashimoto-Torii et al., 2003)(Fig. 2Q), and its expression was absent in the Fez^-/-Fezl^-/- embryos(Fig. 2T). Neither the Fez^-/-Fezl^+/- nor the Fez^+/-Fezl^-/- embryos showed abnormal expression of Dlx1, Gbx2, Lim1, Ebf1 or Sox14 at E12.5(Fig. 2B,C,F,G,J,K,N,O,R,S),further confirming the strictly redundant function of Fez and Fezl in the diencephalon patterning. The data indicate that, in the Fez^-/-Fezl^-/- embryos at E12.5, the prethalamus did not form, the thalamus formed at a reduced size and the anterior pretectum expanded rostrally, suggesting that Fez and Fezl play an important role in rostro-caudal patterning of the diencephalon. Although we detected Tuj-1-postive postmitotic neurons in the thalamus at E11.5 in the Fez^-/-Fezl^-/- embryos (data not shown), the Fez^-/-Fezl^-/- embryos failed to form the rostral part of the thalamus (Sox14-positive region), the formation of which is known to depend on Shh from the ZLI(Hashimoto-Torii et al.,2003). The data indicate that the redundant function of Fez and Fezl also controls the rostro-caudal patterning of the thalamus directly or indirectly.

We analyzed the Fez;Fezl-deficient mice with genetic markers at E10.5, when forebrain patterning becomes apparent. Emx2 is expressed in the dorsal telencephalon (pallium) (Fig. 3A) (Yoshida et al.,1997), and its expression domain was reduced in the Fez^-/-Fezl^-/-, but not in the Fez^-/-Fezl^+/- or Fez^+/-Fezl^-/- embryos(Fig. 3B-D). This is consistent with the reduction of the neocortex and loss of the hippocampus, as Emx2 (and Emx1) is required for neocortical and hippocampal formation (Pellegrini et al.,1996; Yoshida et al.,1997). Pax6 is initially expressed in the dorsal telencephalon and diencephalon, but its expression in the thalamus is reduced by E10.5 (Fig. 3E)(Mastick et al., 1997; Stoykova et al., 1996; Stoykova and Gruss, 1994; Warren and Price, 1997). The rostral and caudal limits of Pax6 expression in the diencephalon were the same in Fez^-/-Fezl^-/- embryos as in wild type, Fez^-/-Fezl^+/- and Fez^+/-Fezl^-/- embryos(Fig. 3E-H). However, the Pax6 expression in the thalamus was not reduced in the Fez^-/-Fezl^-/- embryos(Fig. 3H), suggesting that the rostro-caudal patterning, but not the formation of the diencephalon, was affected in the Fez^-/-Fezl^-/- embryos. The Dlx1 expression in the prethalamus was already absent at E10.5 in Fez^-/-Fezl^-/- embryos(Fig. 3L), indicating that the prethalamus was not established. Shh is expressed in the anterior entopeduncular area (AEP) of the ventral telencephalon, the basal plate of the entire neuroectoderm (including the hypothalamus) and the ZLI(Fig. 3M)(Echelard et al., 1993; Ericson et al., 1995) in the wild-type, Fez^-/-Fezl^+/- and Fez^+/-Fezl^-/- embryos, but its expression in the ZLI was absent in the Fez^-/-Fezl^-/- embryos(Fig. 3M-P), suggesting that the ZLI was not established in the absence of Fez and Fezl. Tcf4 is expressed at a high level in the thalamus and pretectum(Cho and Dressler, 1998; Korinek et al., 1998) in the wild type, Fez^-/-Fezl^+/- and Fez^+/-Fezl^-/- embryos, whereas the Tcf4-high domain expanded rostrally in the Fez^-/-Fezl^-/- embryos(Fig. 3Q-T). Lhx5 is expressed in the prethalamus in the wild type, Fez^-/-Fezl^+/- and Fez^+/-Fezl^-/- embryos(Fig. 3U-W)(Nakagawa and O'Leary, 2001),but was not detected in the Fez^-/-Fezl^-/-embryos (Fig. 3X). Taking these data together, loss of the prethalamus and expansion of the caudal diencephalon took place before E10.5, and the ZLI, the prethalamus-thalamus boundary, was not established in the absence of Fez and Fezl.

Formation and patterning of the forebrain are regulated by transcription factors and inductive signals expressed in specific rostro-caudal positions in the neuroectoderm at the beginning of neurogenesis and neural patterning. We examined gene expression in wild type and Fez^-/-Fezl^-/- embryos at around E9.5. The expression domain of Emx2 and Pax6 in the dorsal telencephalon was already reduced in the Fez^-/-Fezl^-/- embryos at E9.5(Fig. 4A-D). Fgf8 was expressed in the commissural plate, the infundibulum of the hypothalamus, the dorsal part of the prethalamus (eminentia thalami) and the mid-hindbrain boundary (isthmus) in wild-type embryos(Fig. 4E)(Crossley and Martin, 1995),whereas its expression in the prospective prethalamus was specifically absent in the Fez^-/-Fezl^-/- embryos(Fig. 4F). Wnt3a is expressed in the dorsal neural tissue caudal to the prospective ZLI(Fig. 4G)(Roelink and Nusse, 1991; Salinas and Nusse, 1992), and the rostral limit of its expression was shifted rostrally in the Fez^-/-Fezl^-/- embryos(Fig. 4H). En2expression in the midbrain was not affected in the Fez^-/-Fezl^-/- embryos(Fig. 4I,J)(Joyner and Martin, 1987). All these data indicate that the diencephalon patterning was already affected when neural patterning began. In chick, ZLI is positioned on the boundary of the expression domains of Six3 and Irx3, and Six3 and Irx3 can repress the expression of each other, possibly determining the position of the ZLI (Kobayashi et al.,2002). In mouse, Irx1 and Irx3 display a similar expression profile in the neuroectoderm, but the rostral limit of the Irx1 expression is more rostral than that of Irx3 and is positioned on the ZLI (Bosse et al.,1997; Cohen et al.,2000). We examined the expression of Irx1 in wild-type and double-mutant embryos (Fig. 5). The rostral limit of Irx1 expression had shifted rostrally at E9.0, suggesting that Fez and Fezl are required to repress the Irx1 expression in the rostral diencephalon. We found that the expression of Six3 was not affected at E8.5 and 9.5(Fig. 5I,J, data not shown for E8.5), suggesting that the expansion of the caudal diencephalon in the Fez^-/-Fezl^-/- embryos was independent of Six3.

The expression of Fez and Fezl in the forebrain is initiated at E8.0 and 8.5, respectively(Hirata et al., 2006; Hirata et al., 2004)(Fig. 6B). At E12.5, Fez and Fezl are expressed in the pallium, septum,hypothalamus and prethalamus, which are located rostral to the ZLI(Hirata et al., 2004; Hirata et al., 2006). We examined the early expression domains of Fez and Fezl, and compared them with those of Irx1. At E8.5, the expression domain of Fez was slightly wider than that of Fezl(Fig. 6A,B), and the caudal limit of Fez abutted the rostral limit of Irx1(Fig. 6D). Two-color staining for Fez or Fezl with Irx1 revealed that the expression of Fez and Fezl was strictly rostral to that of Irx1 at E9.5 (Fig. 6H-M).

The expression of Fez and Fezl suggests that Fezand Fezl function in the region rostral to the ZLI and repress the caudal diencephalon fate. To address this issue, we misexpressed Fezor Fezl caudal to the ZLI using the Fezl gene enhancer/promoter or the Otx2 forebrain-midbrain (FM) enhancer. We constructed β-galactosidase (lacZ) reporter genes in which the lacZ gene was connected to the 8.2 kbp or 2.7 kbp enhancer/promoter region upstream of the translational initiation site of the mouse Fezl gene (Fig. 7A),and examined the lacZ expression in the resulting transgenics(Fig. 7B-D). The 8.2 kbp Fezl enhancer/promoter recapitulated the expression in the forebrain at E8.5 (Fig. 7B). By contrast,the 2.7 kbp Fezl enhancer/promoter (Fezl2.7p) drove the lacZ expression in a wider region at E8.5 than the 8.2 kbp promoter did (Fig. 7C). At E9.5 in transgenic mice with the 2.7 kbp Fezl promoter-lacZ, lacZactivity was detected in a wider region than the endogenous Fezlexpression (Fig. 7D),indicating that the enhancer/promoter could drive expression caudal to the ZLI. Using Fezl2.7p, we expressed Fezl or Fez cDNA ectopically in the caudal diencephalon. We monitored the exogenous Fezl or Fez expression with a green fluorescence protein variant, Venus (IRES-Venus) (Nagai et al.,2002). In embryos with the Fezl2.7p-Fezl-IRES-Venus or Fezl2.7p-Fez-IRES-Venus transgene, the expression of Foxg1in the telencephalon (Tao and Lai,1992) was not affected, but the expression of Irx1 was reduced and its rostral limit was shifted caudally at E9.5(Fig. 8B,C,S). The rostral limit of Irx1 corresponded to the caudal limit of the exogenous Fezl-IRES-Venus expression in these transgenic embryos(Fig. 8A,R). Similarly, in the Fezl2.7p-Fezl-IRES-Venus and Fezl2.7p-Fez-IRES-Venustransgenic embryos, the Tcf4-high domains, corresponding to the thalamus and pretectum, were strongly reduced at E10.5(Fig. 8E,F,U). In transgenic embryos strongly expressing Fezl, Dlx1 expression expanded caudally(n=1/2), with Gbx2 expression being prominently reduced(n=2/3) at E12.5 (Fig. 8,M,N,P,Q). Furthermore, the misexpression of Fezl by Fezl2.7p abolished Shh expression in the ZLI and ventral diencephalon (n=2/4) or shifted the ZLI-specific Shhexpression caudally (n=2/4) at E10.5(Fig. 8H,I,K). The misexpression of Fez by Fezl2.7p abolished Shhexpression in the ZLI (data not shown, n=1/1). We also used an Otx2FM enhancer, which can drive a transgene in the midbrain,diencephalon and archicortex (Kurokawa et al., 2004a), to misexpress Fezl. Fezl misexpression under this enhancer also suppressed Irx1 expression without affecting Foxg1 expression (Fig. 8W). These data indicate that Fez and Fezl can suppress the caudal diencephalon fate.

The Fez^-/-Fezl^-/- mutant embryos, but not the Fez^-/-Fezl^+/- or Fez^+/-Fezl^-/- embryos, showed defects in patterning of the diencephalon (Fig. 1), indicating a strictly redundant role for Fez and Fezl in diencephalon development. This is consistent with the overlapping expression of Fez and Fezl in the prethalamus at later stages (Hirata et al.,2006; Hirata et al.,2004). Fez and Fezl, respectively, begin expression at E8 and 8.5 (Fig. 6) (Hirata et al.,2006; Hirata et al.,2004). We detected a defect in the rostro-caudal polarity of the diencephalon: rostral expansion of the Irx1 expression at E9.0,indicating that the Fez/Fezl-mediated diencephalon patterning starts soon after the onset of Fez and Fezl expression. Marker analyses showed that the prethalamic region was not established; instead, the caudal diencephalon, which includes the thalamus and the anterior pretectum,expanded rostrally in the double-mutant embryos (Figs 3, 4). There are two possible explanations for the phenotype of Fez;Fezl-deficient embryos: (1)transformation of the prethalamus into the thalamus; and (2) truncation of the prethalamus and the rostral shift of the caudal neural tissue. We found that the size of the diencephalon was not significantly different in the double-mutant embryos, compared with the wild-type embryos, when the rostro-caudal patterning in the diencephalon became abnormal (Figs 3, 4). Our data strongly suggest that the transformation of the prethalamus into the caudal diencephalon takes place during early neural patterning in the absence of Fez and Fezl (Fig. 9). Future studies including the Cre-loxP-mediated cell-fate mapping of Fez and Fezl-expressing cells will definitely clarify this issue. The misexpression of Fez or Fezl suppressed the caudal diencephalon fate and induced the expression of Dlx1, which normally marks the prethalamus, in the region caudal to the ZLI(Fig. 8). All of these data indicate that Fez and Fezl function to repress the caudal diencephalon fate and establish the prethalamus fate(Fig. 9). Although the caudal diencephalon was initially expanded, the thalamus eventually became smaller and the rostral part of the thalamus was missing in the Fez;Fezl-deficient embryos at E12.5 (Figs 1, 2). The later development of the thalamus is known to be dependent on inductive signals from the ZLI(discussed below). The loss of the ZLI in the Fez;Fezl-deficient embryos secondarily affects the development of the thalamus in these embryos(Fig. 9).

Both FEZ and FEZL contain an Eh1 repressor motif, which interacts with the Groucho/TLE family of transcriptional co-repressors(Bae et al., 2003; Kobayashi et al., 2001; Muhr et al., 2001; Shimizu et al., 2002). The Eh1 repressor motif of zebrafish Fezl is required for at least part of the Fezl function in this animal (Levkowitz et al.,2003), suggesting that FEZ and FEZL function as transcriptional repressors to regulate patterning of the diencephalon. In the Fez;Fezl double-mutant embryos, expression domains of the caudal diencephalic genes, such as Irx1, Wnt3a and Tcf4(high-expression domain), expanded rostrally, and misexpression of Fez or Fezl caudal to the ZLI inhibited the expression of Irx1 and Tcf4 in the caudal diencephalon (Figs 4, 5, 8). Fez and Fezl are expressed in the region rostral to the rostral limit of Irx1 expression, which marks the prospective position of the ZLI. Taking these findings together, we conclude that FEZ and FEZL directly or indirectly repress the caudal diencephalon genes in the rostral diencephalon(Fig. 9). Identification of target genes for FEZ and FEZL and/or chromatin immunoprecipitation assay of FEZ/FEZL-binding genomic fragments will help to clarify the precise mechanism by which Fez and Fezl control the rostro-caudal polarity of diencephalon.

SIX3 negatively regulates the caudally expressed Wnt1(Lagutin et al., 2003) and functions as a transcriptional repressor(Kobayashi et al., 2001; Lopez-Rios et al., 2003; Zhu et al., 2002). Six3-deficient mice show strong reduction of the neural tissue rostral to the ZLI (Lagutin et al.,2003). Six3 negatively controls the expression of Irx3 in chick (Kobayashi et al.,2002). These reports suggest that the role of Fez and Fezl in diencephalon patterning is similar to that of Six3,at least in part. However, Six3-deficient mice have rudimentary tissue rostral to the ZLI, and express Shh in the ZLI(Lagutin et al., 2003), and Fez;Fezl-double mutants display complete loss of the prethalamus and ZLI, implying there is a difference between the functions of Six3 and Fez/Fezl. We examined the expression of Six3 in the Fez^-/-Fezl^-/- embryos, but did not observe any significant alteration in the Six3 expression at E8.5 and 9.5(Fig. 5, data not shown for E8.5), suggesting that Six3 does not function downstream of Fez and Fezl. Rather, Six3 may function upstream of, or in parallel with, Fez and Fezl. Future studies examining Fez and Fezl expression in Six3-deficient embryos and combinatory gene disruption of Six3 and Fezand/or Fezl will clarify this issue.

How Fez and Fezl expression is regulated remains unclear. There are several genes with expression domains that overlap with those of Fez and Fezl. They include Pax6, Emx1/2, Dlx1/2/5/6and Otx1/2 in addition to Six3(Bulfone et al., 1993; Oliver et al., 1995; Simeone et al., 1992a; Simeone et al., 1993; Simeone et al., 1992b; Stuhmer et al., 2002; Walther and Gruss, 1991). These genes might be involved in the regulation of Fez and Fezl expression. The expression of zebrafish fezl and Xenopus Fez is negatively regulated by Wnt signaling(Hashimoto et al., 2000) (M. Matsuo-Takasaki, personal communication). Thus, the initial expression of Fez and Fezl may be controlled by the rostro-caudal polarity information, in which Wnt signaling is strongly involved(Niehrs, 2004). In this context, Fez and Fezl may link the rostro-caudal polarity information to the subdivision formation in the diencephalon.

The ZLI position is predicted as abutting the expression domains of rostral Six3 and caudal Irx3 in chick embryos(Kobayashi et al., 2002). We found that it is predicted by the expression of rostrally expressed Fez and Fezl and caudally expressed Irx1 in mouse embryos (Fig. 6). Furthermore,the deficiency of both Fez and Fezl led to loss of the ZLI,and misexpression of Fez or Fezl could inhibit ZLI formation or shift its position (Figs 3, 8). These data indicate that Fez and Fezl are involved in the formation and position of the ZLI. The ZLI initially forms as a wedge-shaped structure on the boundary between the prethalamus and thalamus, which is characterized by a gap in Lfrg expression; subsequently, it collapses to a narrow band(Zeltser et al., 2001). It is not clear whether Fez and Fezl are expressed in the prospective ZLI domain, although they are not expressed in the ZLI at E12.5(Hirata et al., 2006; Hirata et al., 2004). Fez and Fezl are expressed in the prethalamus, but they do not induce Shh expression there. Thus, Fez and Fezlare not likely to be instructive factors, but rather to function as permissive factors. Alternatively, Fez and Fezl may regulate ZLI formation indirectly and non-cell-autonomously by controlling formation of the prethalamus. Grafting experiments in chick embryos show that Shhexpression is induced in the interface between the prechordal (rostral to ZLI)and epichordal plate neuroepithelia (caudal to ZLI)(Vieira et al., 2005),suggesting that an interaction between prethalamus and thalamus is involved in the induction of ZLI. The complete loss of the prethalamus in the Fez^-/-Fezl^-/- embryos might lead to the loss of non-cell-autonomous signals (secreted or membrane-associated molecules) from the prethalamus and subsequently result in loss of the ZLI. In this scenario,formation of the ZLI may be controlled by signals from both the prethalamus and caudal diencephalon. Zli formation is also dependent on Shh emanating from the basal plate (Zeltser,2005). Thus, Fez and Fezl cooperate with Shh from the basal plate to determine the position of the ZLI.

Inductive signals from the ZLI are required for the formation of the prethalamus and thalamus (Kiecker and Lumsden, 2004; Scholpp et al.,2006; Vieira et al.,2005). Shh expressed in the ZLI is involved in the formation and patterning of the thalamus (Kiecker and Lumsden, 2004; Scholpp et al.,2006; Vieira et al.,2005). In the Fez^-/-Fezl^-/-embryos, the Shh-expressing ZLI was not established, and the thalamus became small, although the caudal diencephalon initially expanded rostrally. Therefore, the reduced thalamus is likely to be a secondary consequence of the loss of the ZLI in the Fez^-/-Fezl^-/- embryos(Fig. 9). Consistent with this,the Fez^-/-Fezl^-/- embryos showed complete loss of the Sox14 expression (Fig. 2), which is dependent on Shh from the ZLI(Hashimoto-Torii et al.,2003). Our findings support the inductive role of the ZLI in thalamus development.

In summary, Fez and Fezl are essential factors for development of the forebrain, playing an important role in rostro-caudal patterning of the diencephalon and in ZLI formation. The involvement of repressor-type zinc-finger proteins in forebrain formation provides a new mechanism for the formation and patterning of the forebrain subdivisions.

We thank A. P. McMahon, G. Martin, A. Joyner, P. Gruss, J. L. R. Rubenstein, M. Ishibashi, T. Ogura, M. Frohman, P. Charnay, M. Taira, Y. Yoshihara, K. Shimamura, Y. Nakagawa, H. Sasaki, D. Kurokawa and S. Aizawa for providing the DNA samples; C. Hirata-Fukae, K. Bando and A. Katsuyama for their technical assistance with the histological analysis; Y. Katsuyama and H. Enomoto for their comments on the manuscript; and M. Matsuo-Takasaki for her comments and providing unpublished information about Xenopus Fez. We are grateful to the Laboratory for Animal Resources and Genetic Engineering for the generation of transgenic embryos and for the housing of mice. This work was supported by Grants-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science, Sports and Technology, and by a grant from RIKEN (to M.H.).