Emx2 directs the development of diencephalon in cooperation with Otx2

Initial regionalization of the neural plate along its anterior-posterior axis occurs concurrent with, or shortly after, neural induction. Recent studies have suggested that anterior neuroectoderm is induced by permissive and instructive signals from anterior visceral endoderm, anterior definitive mesendoderm and others (Ang et al., 1996; Thomas and Beddington, 1996; Beddington and Robertson, 1998; Beddington and Robertson, 1999; Shawlot et al., 1999; Tam and Steiner, 1999; Kimura et al., 2000). Otx2, Rpx/Hesx1 and Six3 are the genes expressed in the early anterior neuroectoderm after induction (Suda et al., 1999; Martinez-Barbera et al., 2000a; Martinez-Barbera et al., 2000b; Wallis and Muenke, 2000). The initial morphological landmark of regionalization in the anterior neuroectoderm is the preotic sulcus, which corresponds to the future boundary between rhombomeres r2 and r3. This landmark becomes apparent around the three-somite stage. No molecular devices are known in the establishment of the boundary or initial development of r1 and r2. Rostral to the preotic sulcus, the midbrain/hindbrain junction (or isthmus) establishes molecularly around the six-somite stage (Crossley and Martin, 1995; Millet et al., 1999; Suda et al., 1999) and morphologically around 9.0 days post coitus (dpc). A growing number of molecular studies have been conducted on developmental patterning of the midbrain and the establishment of the isthmus. Otx2 and Otx1, Pax2 and Pax5, and En1 and En2 cooperate in midbrain development (Hanks et al., 1995; Acampora et al., 1997; Acampora et al., 1998; Acampora et al., 1999; Suda et al., 1996; Suda et al., 1997; Suda et al., 1999; Urbanek et al., 1997; Schwarz et al., 1999). Wnt1 functions to maintain the midbrain, possibly via the regulation of En expression (McMahon et al., 1992; Danielian and McMahon, 1996). The isthmus is established where Otx2 and Gbx2 meet (Wassarman et al., 1997; Broccoli et al., 1999; Hidalgo-Sanchez et al., 1999; Millet et al., 1999; Simeone, 2000), and plays an essential role in the rostrocaudal patterning of the midbrain. A loop of mutual interactions between Fgf8, En1, En2, Pax2, Pax5 and Wnt1 genes has been suggested in the isthmus (Balley-Cuif and Wassef, 1995; Joyner, 1996). Fibroblast growth factor (FGF8) is currently the sole factor demonstrating midbrain patterning activity of the isthmus (Crossley et al., 1996; Lee et al., 1997; Martinez et al., 1999).

In contrast, the mechanisms that delineate the telencephalon and diencephalon are poorly understood. The forebrain or primary procencephalon segregates into the diencephalon and secondary procencephalon, from which alar plates protrude to generate the telencephalon. The boundary between the mesencephalon and diencephalon has been proposed to occur around the ten-somite stage, at the point where Pax6 expression segregates from that of Pax2 and En1 (Araki and Nakamura, 1999; Schwarz et al., 1999; Matsunaga et al., 2000). The posterior commissure, which is formed near 10.5 dpc, is the morphological landmark of its dorsal boundary. Zona limitans interthalamica (zlth), which divides the ventral and dorsal thalamus, is formed earlier. It occurs molecularly near the five-somite stage and morphologically at 10.0 dpc, approximately where axial mesoderm is subdivided into the prechordal plate and notochord (Figdor and Stern, 1993; Shimamura et al., 1995; Inoue et al., 2000). The diencephalic region caudal to zlth is compatible with transformation into a mesencephalic phenotype by ectopic transplantation of isthmus/FGF8-soaked beads and by ectopic En expression (Crossley et al., 1996; Araki and Nakamura, 1999; Martinez et al., 1999). Sonic hedgehog (SHH) is expressed by zlth (Erickson et al., 1995), and it has been suggested that zlth functions not only as a barrier to restrict cell mixing and the spread of pattern information, but also as a source of morphogenetic information or as a local organizing center (Balley-Cuif and Wassef, 1995).

A number of genes are now known to be expressed in specific domains of the forebrain, and a neuromeric organization of the forebrain has been proposed (Bulfone et al., 1993; Shimamura et al., 1995; Shimamura et al., 1997). Otx2, Otx1, Emx2 and Emx1, mouse cognates of Drosophila head gap genes, otd and ems, are among these genes. Otx2 expression occurs throughout the anterior neuroectoderm during the initial phase of its induction. At 10.5-12.5 dpc, Otx2 expression regresses in the dorsal telencephalon corresponding to the presumptive cerebral cortex (Simeone et al., 1993; Mallmaci et al., 1996). Otx1 expression is evident around the 1 somite stage. At 10.25-12.5 dpc, Otx1 expression covers the region from the cerebral cortex to the midbrain (Simeone et al., 1993). Emx2 expression is evident around the three-somite stage. At 10.5-12.5 dpc, Emx2 expression ranges from a portion of the subcortical domain in the telencephalon to the diencephalon rostral to zlth (Simeone et al., 1992a; Shimamura et al., 1997; Mallamaci et al., 1998). Emx1 expression occurs in the cortical region around 9.5 dpc. As a result, Otx2, Otx1, Emx2 and Emx1 genes constitute a ‘nested’ expression, and the roles of these genes in brain regionalization have been suggested (Simeone et al., 1992b).

However, the phenotype of each single mutant of these genes has not been informative with respect to their roles in rostral brain regionalization. Otx2^−/− mutants fail to develop a rostral head by the loss of earlier Otx2 functions in visceral endoderm (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996; Rhinn et al., 1998; Kimura et al., 2000). Defects are apparent exclusively in the archipallium in Emx2^−/− mutants (Pellegrini et al., 1996; Yoshida et al., 1997). Otx1^−/− and Emx1^−/− single mutants exhibit only subtle defects (Suda et al., 1996; Acampora et al., 1996; Qiu et al., 1996; Yoshida et al., 1997). In contrast, data from Otx1 and Otx2 double heterozygous mutants (Otx1^+/−Otx2^+/−) indicate that Otx2 and Otx1 cooperate in developing mesencephalon and posterior diencephalon at the time of brain regionalization (Acampora et al., 1997; Suda et al., 1997).

In the present study, we have generated Emx2 homozygous and Otx2 heterozygous double mutants (Emx2^−/−Otx2^+/−), and Emx2 knock-in mutants into the Otx2 locus (Otx2^+/Emx2), in order to genetically determine the interaction between Emx2 and Otx2 genes in forebrain development. Otx2^−/− homozygous mutants exhibit defects caused by the lack of early Otx2 function, as described above. Consequently, interactions between the two genes in brain regionalization could not be evaluated in the Emx2^−/−Otx2^−/− double homozygous mutant state. Nevertheless, Emx2^−/−Otx2^+/− double mutants demonstrate that these two genes indeed play crucial roles in the development of the ventral and dorsal thalamus, and anterior pretectum (precommissural region of pretectum) (Martinez and Puelles, 2000). Emx2 ectopic expression by its knock-in into the Otx2 locus results in a phenotype complementary to the double mutants; Emx2 expression must be suppressed for the development of the posterior pretectum (commissural region of the pretectum).

Emx2 (762 bp) cDNA lacking 3′UTR was isolated from a 11.5 dpc mouse cDNA library. The neomycin resistance gene (neo^r), which has no polyadenylation signal and is driven by the Pgk1 promoter, was flanked by loxP sequences (see Fig. 6A). These sequences (Emx2-loxp/neo/loxp) were inserted into the NlaIII site at the initial codon of the Otx2 gene. The in-frame ligation was confirmed by nucleotide sequencing. Lengths of the homologous regions were 6.9 kb and 4.1 kb at the 5′ and 3′ sides of the insert (Emx2-loxp/neo/loxp), respectively, in the targeting vector. The diphtheria toxin-A fragment (DT-A) gene, driven by the MC1 promoter, was used for negative selection of homologous recombinants as described (Yagi et al., 1993b). (The details of vector construction can be provided upon request.)

TT2 ES cells (Yagi et al., 1993a) were cultured, electroporated with SalI linearized targeting vector and selected against G418 as described (Nada et al., 1993; Yagi et al., 1993b). Homologous recombinants were assessed by long-PCR with a sense primer, p1 (5′-ATCGCCTTCTTGACGAGTTCTTCTG), in the neo^r gene and an antisense primer, p2 (5′-CTTATAATCCAAGCAATCAGTGGTTGAG), in the Otx2 genome located downstream of the 3′ end of the homologous region in the targeting vector (see Fig. 6A). Homologous recombinants were obtained at a frequency of 115 of 120 G418 resistant clones. Recombinants were confirmed by Southern blot hybridization using probes in the neo^r gene and in the third exon of the Otx2 genome (SphI/HindIII fragment in Fig. 6A). Chimeric mice were obtained through injection of targeted TT2 cells into eight-cell stage embryos, as described (Yagi et al., 1993b). Chimeras were crossed with Cre females, which expressed Cre in their zygotes (Sakai and Miyazaki, 1997), in order to generate F₁ heterozygotes. Otx2 and Emx2 heterozygotes were obtained as described (Matsuo et al., 1995; Suda et al., 1996; Yoshida et al., 1997). The genetic background of animals used in the present study is noted in Results. Mice were housed in environmentally controlled rooms of the Laboratory Animal Research Center of Kumamoto University under University guidelines for animal and recombinant DNA experiments.

Genotypes of newborn mice and embryos were routinely determined by PCR analyses. Confirmation, when necessary, was by Southern blot analyses. Genomic DNAs used in the analyses were prepared from tails or yolk sacs. In PCR analysis, knock-in alleles, which retained or lost the neo^r gene, were detected as the 2.1 and 0.7 kb PCR products, respectively. A sense primer, p3 (5′-CCGAGAGTTTCCTTTTGCACAACGC), was placed in the Emx2 cDNA and an antisense primer, p4 (5′-TGTGGCACTCGGCAGTTTGGTAGC), was in the first exon of the Otx2 genome (see Fig. 6A). Genotypes of Otx2 and Emx2 knockout alleles were identified as described previously (Matsuo et al., 1995; Suda et al., 1996; Yoshida et al., 1997).

Mouse embryos were fixed with Bouin’s fixative solution at room temperature for 18-24 hours. Specimens were subsequently dehydrated and embedded in paraplast. Serial sections (8 μm) were prepared and stained with Hematoxylin and Eosin or with 0.1% Cresyl Violet (Sigma).

Embryos were dissected in PBS and fixed overnight at 4°C in 4% paraformaldehyde (PFA) in PBS. Specimens were gradually dehydrated in methanol/PBT (PBS containing 0.1% Tween-20) up to 100% methanol and stored at −20°C. The protocol for in situ hybridization of embryos was as described previously(Wilkinson,1993). Single-stranded digoxigenin-UTP-labeled (Boehringer Mannheim) RNA probes were used. The probes were as described for Wnt1 and Wnt7b (McMahon and Bradley,1990), BF1 (Foxg1 – Mouse Genome Informatics) (Tao and Lai, 1992), En1 and En2 (Davis and Joyner,1988), Pax2 (Dressler et al., 1990), Pax5 (Asano and Gruss, 1992), Pax6 (Walther and Gruss, 1991), Dlx1 (Bulfone et al., 1993), Fgf8 (Crossley and Martin,1995), Otx1 and Otx2 (Matsuo et al., 1995), Gbx2 (Bulfone et al., 1993), Six3 (Oliver et al., 1995), Tcf4 (Korinek et al.,1998), ephrin-A2 (Efna1 – Mouse Genome Informatics) (Flenniken et al., 1996), Mek4 (Cheng and Flanagan, 1994), Lim1 (Lhx – Mouse Genome Informatics) (Fujii et al., 1994), Ebf1 and Ebf3 (Garel et al., 1997), and COUP-TFI (Nr2f1 – Mouse Genome Informatics) (Qiu et al., 1994). A 3′UTR probe was used as described (Yoshida et al., 1997) in order to detect endogenous Emx2 expression in knock-in mutants. A cDNA probe was used for the detection of both endogenous and ectopic Emx2 expression.

The peptide of 50 amino acid residues (EMX2-C50) from the C-terminal of mouse EMX2 protein was chemically synthesized. The rabbit anti-EMX2-C50 antiserum was obtained with the peptide as described (Tanaka et al., 1991). The antiserum was absorbed with mouse liver powder and used at the dilution of 1:2000 in PBS containing 1% goat serum. Paraffin sections of embryos were prepared as described (Gurdon et al., 1976; Mallamaci et al., 1996). Sections were deparafinized in xylene. Subsequently, specimens were rehydrated and incubated with 4% blocking serum for 1 hour, followed by incubation with antibodies. EMX2 expression was detected with rabbit anti-mouse EMX2-C50 antiserum and the secondary antibody (biotinylated goat anti-rabbit IgG) at 1:200 dilution. Posterior commissure neurons were detected with anti-mouse GAP43 monoclonal antibody (Sigma) at a dilution of 1:2000 and horseradish peroxidase-labeled secondary antibody (goat anti-mouse IgG, ZYMED, USA) at a dilution of 1:500. Chromogenic staining was effected according to the protocol supplied by the manufacturer (VECTASTAIN Elite ABC kit, Vector Laboratories).

Genetic background influences the phenotype of Otx2 heterozygous mutants (Matsuo et al., 1995). Our ES cells, TT2, were derived from an F₁ embryo between C57BL/6 and CBA mice. Otx2 is expressed in cephalic neural crest cells (Kimura et al., 1997), and F₁Otx2 heterozygotes obtained from crosses of chimeras with C57BL/6 females exhibit craniofacial defects, yielding few fertile heterozygotes. F₁ heterozygotes that resulted from mating chimeras with CBA females did not possess the defects. Consequently, Otx2 heterozygotes have been maintained by crosses with CBA mice. Accordingly, Emx2 heterozygotes employed in this study were obtained via crosses of chimeras with CBA females and have been maintained by crosses with CBA mice (Yoshida et al., 1997; Miyamoto et al., 1997). The F₅ and F₃ generations of Otx2 and Emx2 heterozygotes, respectively, were the source for the present analysis.

Double heterozygotes (Emx2^+/−Otx2^+/−) were obtained at Mendelian ratio by crosses between Emx2^+/− and Otx2^+/− single heterozygotes. These double heterozygotes were subsequently mated with Emx2^+/− single heterozygotes so as to generate Emx2^−/− homozygous and Otx2^+/− heterozygous (Emx2^−/−Otx2^+/−) double mutants (hereafter referred to as double mutants). The double mutants were obtained at Mendelian ratio at 15.5 dpc; however, none survived beyond 16.5 dpc. Morphologically, defects were not apparent in Otx2^+/− mutants of this pedigree (Matsuo et al., 1995; Fig. 1). Emx2^−/− mutants exhibited defects solely in the medial pallium, as previously reported (Yoshida et al., 1997; Fig. 1). Emx2^+/−Otx2^+/− double heterozygotes displayed defects in forebrain structures similar to, but significantly milder than, Emx2^−/−Otx2^+/− double mutants.

The 15.5 dpc Emx2^−/−Otx2^+/− double mutant brain demonstrated severe defects in the dorsal forebrain (Fig. 1A-H). In the telencephalon, cerebral hemispheres were present, but greatly diminished. The medial region of the cerebral cortex, which juxtaposes to the choroidal roof, is normally the territory of archicortex structures, i.e. the hippocampus (the dentate gyrus and CA fields) and para-hippocampal subicular cortex. The choroidal roof was expanded in the double mutants. Furthermore, no histological signs denoting archipallium structures could be identified; the defects were substantially more severe than those in Emx2^−/− single mutants (Yoshida et al., 1997; Fig. 1). The choroid plexus did not develop in the third ventricle. In the cerebral cortex that was present laterally, no axonal fasciculation of the corpus callosum, anterior commissure, fornix or fimbria was in evidence. Cortical layers were poorly differentiated and laminar organization was not apparent; that is, the cortical plate was thin and sparse, and boundaries between the cortical plate and other layers could not be distinguished (data not shown). Defects were not observed in anterior structures, including the septum, but lateral and medial ganglionic eminences were hyperplastic.

The dorsal and ventral thalamus and pretectum were not apparent in Emx2^−/−Otx2^+/− double mutants. The hypothalamus, mammillary region and hypophysis were relatively normal. The neurohypophysis was normal, but the adenohypophysis was irregularly shaped (data not shown). Zlth, which divides the ventral and dorsal thalamus, or the habenulopeduncular tract separating the pretectum and dorsal thalamus, were not observed in the double mutants. The posterior commissure typically occurs between the pretectum and tectum. The commissure was located proximal to the sulcus telodiencephalicus in the double mutants (Fig. 1D); moreover, only traces of structures that may have corresponded to the posterior pretectum were evident. The posterior commissure was reduced in size and poorly fasciculated. Additionally, the anterior boundary of the mesencephalon was poorly differentiated. The tectum was greatly enlarged occupying the original epithalamus and pretectum regions; in this midbrain, the darkly stained ventricular proliferating field was expanded, whereas the differentiating field was narrowed. The tegmentum developed normally.

Histological analysis was subsequently conducted at 12.5 dpc upon near completion of rapid cell proliferation in the telencephalon. In Emx2^−/−Otx2^+/− double mutants, the telencephalon was small and diminished, particularly in the dorsomedial aspect (Fig. 1I-L). Consequently, the telencephalic roof was enlarged and exposed. Evagination of the medial telencephalic pallium beyond the sulcus telodiencephalicus was poor and the hippocampal region did not develop in the double mutants (Fig. 1M-P). Ganglionic eminence and the mesencephalon were not hyperplastic at this stage. Neither the ventral nor dorsal thalamus was apparent; however, a commissural structure was present that may have corresponded to the posterior pretectum. Ventral structures, possibly corresponding to the hypothalamus, mammillary region and tegmentum, were evident. Isthmic constriction was observed, however, it was shifted rostrally (Fig. 1P).

Wild-type olfactory bulb primordium exhibits mitral and tufted cell layers at 12.5 dpc. The double mutants displayed a very small olfactory bulb that lacked the layered structure. Histologically tufted cells were present, but mitral cell layers were not (data not shown). Olfactory neurons did not project to the olfactory bulb; rather, the nerve fivers were tangled outside the bulb. Several defects also occurred in the eyes. Lenses were irregularly shaped and the outer/inner layers of the retina were hyperplastic (data not shown). Consequently, Otx2 and Emx2 appear to cooperate in several steps of forebrain development. These genes interact in corticogenesis, archipallium/roof development and the formation of sensory organs, the details of which will be reported elsewhere. This study focuses on the roles of Emx2 and Otx2 in diencephalon development.

At 11.5 dpc, several genes are expressed region specifically in the diencephalon. Affected structures were confirmed with these molecular markers; expression of each marker used is schematically summarized in Fig. 9B. Tcf4 is strongly expressed in the dorsal thalamus and pretectum, whereas expression in the ventral thalamus is weak (Fig. 2A; Cho and Dressler, 1998; Korinek et al., 1998). The double mutants lost the Tcf4-weak ventral thalamus and the majority of Tcf4-intense structures (Fig. 2B); however, traces of structures displaying intense Tcf4 expression were in evidence. The posterior commissure, which originates from the posterior pretectum (Mastick et al., 1997), developed in the double mutants. Consequently, this Tcf4-positive structure in the double mutants most probably corresponds to the posterior pretectum. The posterior pretectum normally expresses Lim1 (Barnes et al., 1994; Fujii et al., 1994; Mastick et al., 1997). Indeed, Lim1 expression was typically present in the double mutants (Fig. 2C,D). Wild-type embryos display Dlx1 expression in the ventral thalamus, entopedunucular area, hypothalamic cell cord and ganglionic eminence (Fig. 2E; Bulfone et al., 1993; Eisenstat et al., 1999). Dlx1-positive ganglionic eminence, hypothalamic cell cord and entopedunucular area were present in the double mutants; however, the ventral thalamus was absent (Fig. 2F), as indicated by Tcf4 expression. Gbx2 is expressed in the dorsal thalamus and ganglionic eminence (Fig. 2G; Bulfone et al., 1993). Expression was present in the ganglionic eminence; however, Gbx2-positive dorsal thalamus was absent in the double mutants (Fig. 2H). Ebf1 is expressed in the anterior pretectum, anterior tectum and tegmentum (Fig. 2I; Garel et al., 1997). Double mutants exhibited the expression in the anterior tectum and tegmentum, but these mutants lacked Ebf1-positive anterior pretectum (Fig. 2J).

Pax6 is expressed in the telencephalon and diencephalon. At 11.5 dpc, it is strong in the ventral thalamus, posterior pretectum and tegmentum (Fig. 2K; Walther and Gruss, 1991). The strong expression in the basal posterior pretectum and midbrain tegmentum was retained, but all other Pax6 expression was lost in the double mutants (Fig. 2L). Otx1 is expressed spanning the telencephalon, diencephalon and mesencephalon. The most intense Otx1 expression is observed in the anterior pretectum, tegmentum and a boundary region along zlth (Fig. 2M; Simeone et al., 1993; Boncinelli et al., 1993). Double mutants lacked the Otx1-rich anterior pretectum and zlth region (Fig. 2N).

Morphologically, the mesencephalon was normal in the 12.5 dpc double mutants. It is possible, however, that the diencephalic defects affected rostrocaudal regionalization in the mesencephalon. Mek4 is expressed in a gradient in wild-type embryos. High levels of Mek4 occur in the anterior tectum, whereas low levels occur caudally (Fig. 2O; Cheng and Flanagan, 1994). In contrast, ephrin-A2 expression is high in caudal mesencephalon and low in anterior mesencephalon (Fig. 2Q; Flenniken et al., 1996; Feldheim et al., 1998; Feldheim et al., 2000). The double mutants exhibited normal patterns for Mek4 and ephrin-A2 expression (Fig. 2P,R).

Marker analyses, in concert with morphological features, indicate that the precommissural region of the pretectum and dorsal and ventral thalamus do not develop in double mutants. The commissural region of the pretectum and the ventral structures were present, including the hypothalamus, mammillary region and tegmentum.

No diencephalic structures are subdivided morphologically at 9.5 dpc; nor are any region-specific molecular markers expressed in the diencephalon. At this stage, telencephalic vesicle formation begins and BF1 is expressed throughout the entire telencephalic neuroepithelium with the exceptions of the most medial and caudal regions adjacent to the choroidal roof and diencephalon (Tao and Lai, 1992). In the 9.5 dpc double mutants, the BF1-positive region was largely normal (Fig. 3A,B). Normally Otx2 is expressed in the forebrain and midbrain with the caudal limit at the mid/hindbrain junction (Simeone et al., 1993; Matsuo et al., 1995). In comparison with Otx2^+/− single heterozygotes, Otx2 expression from the remaining allele was greatly reduced in Emx2^−/−Otx2^+/− double mutants, particularly in the region caudal to the sulcus telodiencephalicus (Fig. 3C,D). The caudal limit shifted rostrally and expression was restricted to the more dorsal region.

Pax6 expression is found over the telencephalon and diencephalon at 9.5 dpc in wild-type embryos (Fig. 3E; Walther and Gruss, 1991; Stoykova et al., 1996). Pax6 expression was scarcely observed in double mutants, however, its expression in the eyes was retained (Fig. 3F). Wnt7b expression was present in the double mutant telencephalon, but the mutants lost a Wnt7b-positive diencephalon (data not shown, compare Fig. 3F with Fig. 7G). Fgf8 expression in the anterior neural ridge was normally present, but the roof at the position of the telodiencephalicus or the roof of ventral thalamus was absent (Fig. 3M,N).

Wnt1 is expressed in the dorsal midline over the caudal diencephalon and mesencephalon, and in a characteristic lateral stripe at the isthmus in 9.5 dpc wild-type embryos (Fig. 3G; Parr et al., 1993). Double mutants demonstrated a reduced Wnt1-positive region in the roof as well as a decrease in the level of Wnt1 expression laterally. The isthmic stripe was formed, but it was shifted rostrally. In wild-type embryos, caudally to the isthmus, the anterior metencephalon, r1 and r2, does not express Wnt1 (Fig. 3G; McMahon et al., 1992; Echelard et al., 1994). This Wnt1-negative region was expanded in the double mutants (Fig. 3H).

En2 is expressed in a gradient over the caudal mesencephalon and r1 at 9.5 dpc in wild-type embryos (Fig. 3I; Davis and Joyner, 1988; Davis et al., 1988). The double mutants exhibited expanded En2 expression rostral to the isthmic constriction, and the gradient is less distinct (Fig. 3J). Concomitantly, the region between the anterior terminus of En2 expression and the sulcus telodiencephalicus was greatly diminished. Moreover, the En2-positive region caudal to the isthmic constriction and the region between the posterior terminus of En2 expression and the otic vesicle appeared enlarged. Normally Pax5 is also expressed in the mesencephalon and r1 (Fig. 3K; Asano and Gruss, 1992; Urbanek et al., 1994). The Pax5-positive region both rostral and caudal to the isthmic constriction was enlarged, while the negative region extending up to the sulcus telodiencephalicus was greatly reduced in double mutants (Fig. 3L). The region between the caudal end of Pax5 expression and the otic vesicle appeared enlarged. Isthmus also expresses Fgf8 and Gbx2 (Fig. 3M,O; Bulfone et al., 1993; Crossley and Martin, 1995; Mahmood et al., 1995). These expressions were shifted anteriorly in the double mutants (Fig. 3N,P). In addition, the distance between the Fgf8- and Gbx2-positive isthmic stripe and otic vesicles appeared increased by the double mutation.

Marker analyses at 9.5 dpc are consistent with the defects at 11.5 dpc. Telencephalon and mesencephalon formation is fairly normal; however, the major region of the diencephalon does not develop and Pax6 expression in the forebrain is lost in Emx2^−/−Otx2^+/− mutants. Thus, the onset of the defects precedes this stage. The expansion of the anterior hindbrain was notable.

Marker analyses were next performed at the six-somite stage, the point at which the initial brain regionalization is completed. The caudal limit of Otx2 expression becomes distinct at the isthmus and the territory of the future forebrain and midbrain is established by this stage (Fig. 4A; Acampora et al., 1995; Rubenstein and Shimamura, 1998). Double mutants displayed a greatly reduced Otx2-positive region (Fig. 4B). Pax6 expression normally occurs in the forebrain, as well as in the laterocaudal component of the incipient optic cup, and the hindbrain and spinal cord (Fig. 4C; Walther and Gruss, 1991). Pax6 expression was unchanged in the hindbrain and spinal cord, and remained evident in the optic cup in double mutants (Fig. 4D); however, it was scarcely observed in the forebrain.

The Wnt1 expression was already reduced and shifted anteriorly in the double mutants at the six-somite stage (Fig. 4E,F; McMahon et al., 1992; Bally-Cuif et al., 1995). In addition, the Wnt1-negative anterior hindbrain, corresponding to future r1 and r2, exhibited expansion. In wild-type brain, the En1- and Pax5-positive regions, which initially cover the entire prospective mesencephalon (compare with Fig. 8G,I), have completed regression to the caudal portion by this stage (Fig. 4G,I; Joyner, 1996). Double mutants demonstrated somewhat expansion of the positive domains accompanied by an anterior shift (Fig. 4H,J). Fgf8 expression in the midbrain-hindbrain boundary in wild-type embryos begins to be confined to the future isthmus at the six-somite stage (Fig. 4K; Suda et al., 1997). However, this event was less distinct in the double mutants (Fig. 4L). Six3 expression is observed in the most rostral neuroectoderm at the six-somite stage in wild-type embryos (Fig. 4M; Oliver et al., 1995). Double mutants exhibited normal Six3 expression (Fig. 4M,N). Consequently, the diencephalon corresponding to the future ventral thalamus, dorsal thalamus and anterior pretectum is probably already lost at this point. Additionally, anterior hindbrain expansion, the loss of Pax6 expression, reductions in Otx2 and Wnt1 expression, and the elevation of En1 and Pax5 expression were also evident at the six-somite stage in Emx2^−/−Otx2^+/− mutants.

Otx2 is expressed in the anterior neuroectoderm from the earliest stage of its induction (Fig. 5A-D; Acampora et al., 1995). As a result, the onset of the double mutant defects should closely correlate with the onset of Emx2 expression. Emx2 was not detectable at the two-somite stage, but at the three-somite stage, it occurred in the laterocaudal forebrain primordia (Fig. 5E,F). This Emx2 expression increases and expands rostrally to the future dorsal aspect by the five-somite stage (Fig. 5G,H; Shimamura et al., 1997). Pax2 expression, which corresponds to the future midbrain, also occurs around the three-somite stage in the caudal portion of the Otx2-positive region (Fig. 5I,J; Rowitch and McMahon, 1995). At this stage, the anterior limit of Pax2 expression appears to overlap with the posterior limit of Emx2 expression; moreover, its caudal limit extends beyond the caudal limit of Otx2 expression. At the four-somite stage, Pax2 expression extends caudally into the preotic sulcus at the most lateral (future dorsal) edge; however, in the medial (future lateral and ventral) region, expression terminates more rostrally (Fig. 5K,L). This laterally Pax2-positive and medially Pax2-negative region corresponds to future r1 and/or r2. Gbx2 is expressed in this preotic as well as the otic region at the 3-4 somite stage (Fig. 5W). Pax6 expression occurs at the four-somite stage, and its caudal limit nearly coincides with the caudal limit of Emx2 expression (Shimamura et al., 1997; Inoue et al., 2000).

The defect was evident at the three-somite stage upon evaluation of the double mutant phenotype via Otx2 expression (Fig. 5M,N). The Otx2-positive domain was reduced both rostrocaudally and lateromedially (future dorsoventrally). The rostrocaudal reduction of the Otx2-positive domain was more apparent at the four-somite stage (Fig. 5O,P). Pax2 expression was more diffuse and expanded rostrally in double mutants at the three-somite stage (Fig. 5Q,R). At the four-somite stage, the caudal limit of Pax2 expression in the medial region had shifted rostrally. Furthermore, the laterally Pax2-positive and medially Pax2-negative region had expanded (Fig. 5S,T), and the anterior Pax2-negative domain was greatly reduced (Fig. 5U,V). Concomitantly, the anterior limit of the Gbx2 expression in the preotic region shifted rostrally (Fig. 5X). Thus, it appears most likely that the Emx2-positive domain failed to develop, the Pax2-positive domain underwent a rostral shift and the Gbx2-positive region anterior to preotic sulcus expanded by the four-somite stage in the Emx2^−/−Otx2^+/− double mutants (see Fig. 9A). Wnt1 expression was faint, whereas Pax6 expression was not detected in double mutants at the four-somite stage as in later stages (data not shown).

At 10.5 dpc, Emx2 is typically expressed rostrally to zlth. Emx2 expression is not observed in the dorsal thalamus, pretectum or tectum (Simeone et al., 1992a; Shimamura, et al., 1997; Yoshida et al., 1997). Emx2 functions in the development of diencephalic structures were also examined by its ectopic expression in these endogenously Emx2-negative regions under the Otx2 heterozygous state. For this purpose the Otx2 gene was replaced with Emx2 cDNA by homologous recombination in TT2 ES cells, as described in Fig. 6A. The neo-resistant gene directed by the phosphoglycerate kinase 1 (Pgk1) promoter for the selection of the recombinants was flanked with loxP sequences. To exclude possible effects of the Pgk1 promoter on the Otx2 transcriptional machinery, the gene was deleted by crossing chimeras derived from the homologous recombinants with transgenic females expressing Cre in their zygotes (Fig. 6B; Sakai and Miyazaki, 1997). F₂ embryos derived from these F₁ heterozygotes (Otx2^+/Emx2) expressed Emx2 in the dorsal thalamus, pretectum and tectum, as expected (Fig. 6C-H).

The ventral and dorsal thalamus, tectum, hypothalamus, mammillary region and tegmentum were morphologically normal in these 12.5 dpc Otx2^+/Emx2 mutants (Fig. 7A,B). A structure similar to the precommissural region of the pretectum was also evident, but the commissural region of the pretectum was not apparent.

Endogenous Emx2 and Emx1 were expressed typically in the 10.5 dpc telencephalon of knock-in mutants (Fig. 7C-F; Simeone et al., 1992a; Yoshida et al., 1997). Wnt7b is normally expressed in the 9.5 dpc forebrain, however, expression does not occur in its most posterior aspect (Fig. 7G; Parr et al., 1993). This Wnt7b-negative caudal diencephalon was nearly lost in knock-in mutants (Fig. 7H). Tcf4-intense diencephalon existed in the 10.5 dpc mutants; however, the domain was narrowed and its posterior aspect was not distinctly delineated, as in wild-type embryos (Fig. 7I,J). Gbx2-positive dorsal thalamus and Ebf1-positive anterior pretectum and anterior tectum were present despite the lack of a Ebf1-negative posterior pretectum in the 12.5 dpc knock-in mutants (Fig. 7K,L; data not shown; compare with Fig. 2I). Pax6 expression in the 10.5 dpc forebrain was largely normal, although its caudal-most expression, typically corresponding to the posterior pretectum, was not sharply delineated (Fig. 7M,N). At 12.5 dpc, a Pax6-intense ventral thalamus was present, but the posterior pretectum was not observed (Fig. 7O,P). Ventral thalamus development was also confirmed with Dlx1 (data not shown, compare with Fig. 2E). The Lim1-positive tegmentum was present, whereas the Lim1-positive posterior pretectum was not found in the 10.5 dpc knock-in mutants (Fig. 7Q,R). Anti-GAP43 antibody specifically detects the posterior commissure (Fig. 7S; Matsunaga et al., 2000). A GAP43-positive structure was not identified in 12.5 dpc Otx2^+/Emx2 mutants (Fig. 7T).

Among gene activity dorsally spanning the caudal diencephalon and mesencepalon,Wnt1 and Ebf3 (Garel et al., 1997) expression were decreased at 10.5 dpc (Fig. 8A-D). COUP-TFI expression (Qiu et al., 1994) was unchanged (Fig. 8E-F), whereas En1 and Pax5 expression was strongly enhanced (Fig. 8G-J). Knock-in mutants, however, displayed normal patterns of Mek4 and ephrin-A2 expression (data not shown; compare with Fig. 2O,Q). Isthmus formation typically occurs with normal Fgf8 expression. Fgf8 expression in the anterior neural ridge and the roof at the position of the telodiencephalicus was also normally present (Fig. 8K,L).

Marker analyses, in conjunction with morphological observations, indicate that the telencephalon, ventral and dorsal thalamus, anterior pretectum and mesencephalon developed in the knock-in mutants. However, the commissural region of the pretectum specifically failed to develop as a consequence of ectopic Emx2 expression.

In contrast to the hindbrain, the forebrain has long been thought not to possess neuromeric organization. However, a prosomeric model that postulates forebrain neuromeric organization has been presented on the basis of region-specific gene expression and other evidence. It is now generally believed to divide the caudal forebrain along the rostrocaudal axis into p1 pretectum, p2 dorsal thalamus/epithalamus and p3 ventral thalamus (Bulfone et al., 1993; Figdor and Stern, 1993; Puelles and Rubenstein, 1993; Rubenstein et al., 1994; Shimamura et al., 1995). Adjacent to the boundary between the tectum and p1 pretectum lies the posterior commissure; between p1 and p2 dorsal thalamus, the habenulopeduncular tract is found; between p2 and p3 ventral thalamus, the zona limitans interthalamica develops, respectively. The ventral and dorsal thalamus, epithalamus and anterior pretectum were lost in Emx2^−/−Otx2^+/− mutants (Fig. 9B). Consequently, Emx2 and Otx2 play crucial roles in diencephalon development. No other such genes are known.

Otx2 and Emx2 are co-expressed in a variety of stages and sites during brain development. The double mutant defects at 15.5 dpc may reflect each of these Otx2 and Emx2 functions. Defects in the formation of the diencephalon, however, appear to reside within their functions at the three-somite stage when Emx2 expression first occurs (Fig. 9). Otx2 and Gbx2 are among the genes known to be first expressed in the neuroectoderm. Otx2 expression hallmarks the anterior neuroectoderm which corresponds to the future forebrain and midbrain. In contrast, Gbx2 marks the posterior neuroectoderm (Wassarman et al., 1997); at the three- to four-somite stage its expression is found in preotic and otic regions. Pax2 and En1 expression occurs nearly simultaneously at the three-somite stage in the entire prospective midbrain (Rowitch and McMahon, 1995), probably as a result of signals such as Fgf4 from the anterior notochord (Shamim et al., 1999). Their rostral limits likely overlap with Emx2 expression. Wnt1 expression begins at this stage in a similar manner with respect to En1 (Bally-Cuif et al., 1995). Onset of Pax6 expression is somewhat late around the four-somite stage. Moreover, the caudal limits of Pax6 and Emx2 expression appear to coincide with one another at this stage (Shimamura et al., 1997; Inoue et al., 2000).

At the three-somite stage, double mutants displayed an Otx2-positive region which was reduced rostrocaudally. Additionally, the Pax2-positive domain was diffuse and extended in a rostral manner. By the four-somite stage, the anterior Pax2-negative region was reduced, whereas the caudal limit of Pax2 and the anterior limit of the Gbx2 expression shifted rostrally; the Gbx2-positive region rostral to preotic sulcus expanded (Fig. 9A). It is quite simple to interpret the phenotype as a consequence of the loss of the Emx2-positive domain in double mutants and the shift of the Pax2-positive domain to the anterior Otx2-positive/Emx2-negative domain.

Pax6 expression was not induced in double mutants. Apparently, Emx2 and Otx2 are located upstream of Pax6 in development of the diencephalon. Sey mutants, which possess a null mutation in the Pax6 gene, suggested that Pax6 is important for correct dorsoventral and anteroposterior patterning in the diencephalon (Grindley et al., 1997; Stoykova et al., 1997; Warren and Price, 1997). Each region of the diencephalon, ventral thalamus, dorsal thalamus and pretectum, however, emerges in sey/sey mutants. Consequently, Emx2 and Otx2 regulate other genes with respect to diencephalon development. Wnt1 expression also decreased at the initial stage in double mutants. Emx-binding sites exist in the cis regulatory region of the Wnt1 gene (Iler et al., 1995). Wingless and engrailed are hypothesized to be targets of the head gap genes otd and ems in Drosophila (Royet and Finkelstein, 1995). However, Wnt1 mutants as well as En1 and En2 mutants develop diencephalic structures (Joyner et al., 1991; McMahon et al., 1992; Millen et al., 1994; Wurst et al., 1994).

The 15.5 dpc defects in the neocortex are apparently due to later Emx2 and Otx2 functions in corticogenesis, as suggested by BF1 expression at 9.5 dpc; however failure of archipallium formation in the double mutants may also reside within the Emx2 functions at the three-somite stage (Fig. 9). In the prosomeric model, the archipallium belongs to the p4 structures anterior to p3 ventral thalamus (Rubenstein et al., 1998). Wild-type embryos demonstrate a rostral extension of the lateral (future dorsal) portion of the anterior-most Emx2-positive region (Fig. 5C,D; Fig. 9A). This extension may include the presumptive dorsomedial region of the forebrain or archipallium (Rubenstein and Shimamura, 1998). The Otx2-positive region was reduced not only rostrocaudally but also lateromedially (future dorsoventrally) in double mutants at the three-somite stage (Fig. 5G,H; Fig. 9B). The detailed analysis of archipallium defects in double mutants will be reported elsewhere.

The expansion of the anterior hindbrain in 9.5 dpc Emx2^−/−Otx2^+/− double mutants is puzzling. The region between the otic vesicle and isthmus, indicated by Fgf8 and Gbx2 expression, appeared expanded. The region between the otic vesicle and the caudal terminus of En2 and Pax5 expression was enlarged; their expression in r1 was also expanded. The Wnt1-negative region in the anterior hindbrain increased at 9.5 dpc and the six-somite stage. Consequently the expanded region may correspond to r1 and r2. The r1/2 also enlarges in Otx1^+/−Otx2^+/− double heterozygotes that fail to develop a midbrain (Suda et al., 1997). We speculated that suppressive signals originating within the midbrain might be necessary for correct r1/2 development; however, such signals from the diencephalon are very unlikely.

Of note is the expansion of the laterally Pax2-positive (but medially Pax2-negative region) or the Gbx2-positive region anterior to the preotic sulcus in double mutants at the four-somite stage (Fig. 5S,T,X; Fig. 9). This event is most probably brought about by the loss of the Emx2-positive region and anterior shift of Pax2-positive region. Morphologically, the anterior neuroectoderm is initially demarcated by the preotic sulcus, which is apparent around the three-somite stage and includes future r1 and r2. Unfortunately, no molecular markers are available specific to the prospective r1/2 territory at early stages. As a result, more detailed analyses are necessary in future studies in order to determine the direct correlation of the defect at the four-somite stage with r1/2 expansion at 9.5 dpc.

The loss of anterior hindbrain in Gbx2-null mutants accompanies a posterior expansion of the Otx2-positive midbrain (Wassarman et al., 1997; Miller et al., 1999). En1, En2, Pax2, Pax5 or Wnt1 mutants fail to develop midbrain without expansion of r1/2. These midbrain genes, however, may function in midbrain development after the establishment of its territory (McMahon et al., 1992; Schwarz et al., 1999; Liu and Joyner, 2001). It is tempting to speculate that the region anterior to the preotic sulcus initially develops as a unit and that the anterior hindbrain expands to compensate when more anterior structures are lost before their subdivision into forebrain, midbrain and the anterior hindbrain. Clarification of the development of the preotic sulcus and r1/r2 might be important regarding investigations of rostral brain regionalization.

The posterior limit of Pax6 expression and the anterior limit of En1 and Pax2 expression are known to overlap at the six- to eight-somite stages in chicken. However, the expression of these genes segregates by the 10 somite stage (Matsunaga et al., 2000). The interaction between Pax6, En1 and Pax2 at this stage is believed to establish the boundary between the diencephalon and mesencephalon (Araki and Nakamura, 1999; Schwarz et al., 1999; Matsunaga et al., 2000). In the absence of Pax6 expression, however, the posterior pretectum and tectum were regionalized in Emx2^−/−Otx2^+/− double mutants; the present study does not exclude the role of Pax6 in the fine tuning of the boundary. After segregation from the Pax6-positive region, En1, En2 and Pax2 expression regress caudally (Dressler et al., 1990). At 10.5 dpc, the caudal limit of Emx2 expression is in zlth (Simeone et al., 1992a; Shimamura et al., 1997), whereas that of Pax6 occurs in the boundary between the diencephalon and mesencephalon (Walther and Gruss, 1995). En1 and En2, and Pax2 and Pax5 are expressed in the caudal mesencephalon (Davis and Joyner, 1988; Dressler et al., 1990; Asano and Gruss, 1992; Millet et al., 1999). These observations raise the question of why the Pax6-positive, Emx2-negative anterior pretectum and dorsal thalamus are lost in Emx2^−/−Otx2^+/− double mutants.

A cell lineage analysis has revealed that restriction in cell movement between regions anterior and posterior to the presumptive zlth occurs immediately after the onset of Pax6 expression at the four-somite stage (Inoue et al., 2000). At this juncture, the prospective pretectum and dorsal thalamus regions are diminutive. Consequently, two refined analyses are necessary to assess the origin of dorsal thalamus and pretectum cells. The caudal limit of Emx2 expression must first be determined by expression analysis at the single cell level. This process will ascertain the distinction between its coincidence with the caudal limit of Pax6 expression or the presence of Emx2-negative and Pax6-positive cells in this region already at the four-somite stage. The analysis should demonstrate the sequence of growth of the Emx2-negative, Pax6-positive diencephalon. Secondly, cell lineage analysis by genetic approach is necessary in order to determine the fate of Emx2-positive cells at the three- to four-somite stage. These studies are in progress; however, we speculate that the Emx2-positive domain at the three-somite stage includes cells of the future dorsal thalamus and anterior pretectum, where Emx2 expression is subsequently lost. Their development is compatible with Emx2 expression as demonstrated by the Otx2^+/Emx2 knock-in mutation. Of course, it can not be excluded that a population of Emx2-negative and Pax6-positive cells at the four-somite stage generates the dorsal thalamus and anterior pretectum, depending on signals from Emx2-positive ventral thalamus.

This study distinguished the developmental nature of the anterior and posterior pretectum. The anterior pretectum was lost in the double mutants, while the Lim1-positive commissural region of the pretectum developed (Fig. 9C). In contrast, formation of not only the entire endogenously Emx2-positive region, but also of the Emx2-negative dorsal thalamus, anterior pretectum and tectum, was compatible with ectopic Emx2 expression by the Otx2^+/Emx2 knock-in mutation. Specifically, the development of the commissural region of the pretectum was not allowed by ectopic Emx2 expression. Unfortunately, molecular markers for posterior pretectum were unavailable at earlier stages. As a result, analyses of defect onset is left to future investigations. The cellular origin of the posterior pretectum is the most significant aspect to be determined.

The tract of the posterior commissure forms just rostral to the caudal border of the Pax6-positive domain. Although few posterior commissure neurons express Pax6 at 10.5 dpc, the point at which the posterior commissure can be identified, the neurons are thought to originate from Pax6-positive pretectum cells (Mastick et al., 1997). Schwartz et al. argue that Pax6 is sufficient and necessary for the development of the commissure on the basis of its loss in their sey/sey mutants and its subsequent recovery by Pax6-transgenesis under a Pax2 enhancer (Schwartz et al., 1999). Pax6 expression was lost in Emx2^−/−Otx2^+/− double mutants, however, the posterior commissure was present despite blurring. In the sey/sey mutants reported, the commissure was either completely absent (Stoykova et al., 1996; Warren and Price, 1997), lacking in dorsal axons (Mastick et al., 1997) or present but smaller than normal (Grindley et al., 1997).

Curiously, Wnt1 expression decreased and En1, En2, Pax2 and Pax5 expression were enhanced by both double and knock-in mutations. However, the midbrain was normal in size at 12.5 dpc and displayed the normal rostrocaudal pattern of Mek4 and ephrin-A2 expression in both mutants. The anterior tectum developed as indicated by Ebf1, as well as Mek4 expression. The causes and significance of changes in the expression of several midbrain genes, in conjunction with possible roles of the commissural region of the pretectum or diencephalon with respect to later midbrain development, including its hyperplasia at 15.5 dpc (Fig. 1), require further study.

This study examined genetically the existence of interaction between Emx2 and Otx2 genes. It indicated that these two genes indeed interact in diencephalon development. Otx1 also participates in diencephalon formation, as demonstrated by Otx1/Otx2 double mutation (Acampora et al., 1997; Suda et al., 1997). Consequently Otx2 interacts with both Emx2 and Otx1 in the development of the diencephalon. In addition, both Emx2/Otx2 and Otx1/Otx2 double mutants indicated the gene dose-dependent nature of the interactions between these genes. Each head subdomain appears to require differing otd and ems levels in Drosophila (Royet and Finkelstein, 1995). However, the question regarding the specifics of how Emx2 interacts with Otx2 remains unanswered. Both Otx1 and Otx2 gene products (OTX2 and OTX1) exhibit paired-like homeodomains; interaction between OTX2 and OTX1 through these homeodomains is probable. Furthermore, the homeodomains of Emx1 and Emx2 gene products (EMX1 and EMX2) belong to HEX class. The HEX homeodomain is too distant to allow homeodomain-homeodomain interactions with the paired-like homeodomains, however, our preliminary studies have suggested a direct protein-protein interaction between OTX2 and EMX2 proteins, possibly through homeodomain and non-homeodomain regions (Nakano et al., 2000). Alternatively, these genes might bind to each recognition site. Synergistic binding would activate the expression of common target genes.

The role of EMX2 in diencephalon development indicated by Emx2^−/−Otx2^+/− double mutants, however, must reconcile with the fact that the diencephalon develops normally in Emx2^−/− single mutants. Otx2 expression is not restricted to the diencephalon. Several explanations are possible; however, it is quite simple to hypothesize that there exists a gene that complements Emx2 and regulates diencephalon development through gene dose-dependent interactions with Emx2, Otx2 and Otx1. Emx1 would be a primary candidate for this unidentified gene. However, Emx1 becomes expressed later at 9.5 dpc and not in the early diencephalon. Two genes, Vax1 and Vax2, are also known that locate very closely to Emx2 and Emx1 genes in chromosomes 19 and 6, respectively. They possess similar homeodomains (Hallonet et al., 1998; Ohsaki et al., 1999), but their expression does not overlap with Emx2 expression in the forebrain. Neither the Emx1, Vax1 or Vax2 gene was ectopically induced in the diencephalon of Emx2 single mutants at the four-somite stage (data not shown). The Not class of genes also exhibits homeodomains similar to Emx2 (Stein and Kessel, 1995). Chicken Cnot1 is expressed in prospective forebrain at the early neurula stage and later in the diencephalon. A Not2 cognate, floating head, is essential for the development of the epiphysial region in zebrafish (Masai et al., 1997). However, no mouse Not homologues have been identified. The identification of a gene that complements Emx2 using Emx2 single mutants at the three- to four-somite stage is eagerly awaited, in addition to the examination of other possibilities.

We thank Mr Naoki Takeda (Division of Transgenic Technology, Center for Animal Resources and Development, Kumamoto University) for production of chimeric mice. We are indebted to Drs A. P. McMahon, G. Martin, A. Joyner, P. Gruss, E. Lai, J. L. R. Rubenstein, H. Clevers, M. Taira, J. G. Flanagan, D. G. Wilkinson, P. Charnay and M.-J. Tsai for in situ hybridization probes. We thank Dr J. Miyazaki for providing the Cre mice. We are also grateful to the Laboratory Animal Research Center of Kumamoto University for housing of the mice. This work was supported in part by grants-in-aid from the Ministry of Education, Science, Sports and Culture of Japan (Specially Promoted Research); the Science and Technology Agency, CREST, JST; and the Ministry of Public Welfare, Japan.