The same enhancer regulates the earliest Emx2 expression in caudal forebrain primordium, subsequent expression in dorsal telencephalon and later expression in the cortical ventricular zone

In the mouse brain, Emx2 is first expressed in caudal forebrain primordium at embryonic day (E) 8.5 (3- to 8-somite stage) and after E9.5 in dorsal telencephalon, forming a gradient along the anterior-posterior and dorsomedial-lateral axes, with the highest expression in the caudal-dorsomedial domain (Simeone et al., 1992; Gulisano et al., 1996; Mallamaci et al., 1998; Suda et al., 2001). At E12.5, the cells in prethalamus, thalamus and non-commissure region of the pretectum are descendants of the cells that once expressed Emx2 (Kimura et al., 2005). However, after E9.5, Emx2 expression does not occur in the majority of the diencephalon. After E12.5, during cortical lamination, Emx2 is expressed in proliferating cells in the ventricular zone and Cajal-Retzius cells in the marginal zone; it is not expressed in differentiated neurons in cortical plate or the intermediate zone (Gulisano et al., 1996; Mallamaci et al., 1998; Shinozaki et al., 2002). Emx2 has thus been presumed to play an essential role in each step and site of forebrain development; this has indeed been demonstrated in mouse mutants.

Emx2;Otx2 and Emx2;Pax6 double mutants displayed Emx2 functions in diencephalon development (Suda et al., 2001; Kimura et al., 2005). Emx1;Emx2 double mutants indicated Emx functions in the development of medial pallium (Pellegrini et al., 1996; Yoshida et al., 1997; Shinozaki et al., 2004); they also showed their roles in cortical lamination (Shinozaki et al., 2002). It is interesting to know how these Emx2 functions are differentially regulated. A previous study by Theil et al. (Theil et al., 2002) identified an Emx2 enhancer immediately upstream of the Emx2 translational start site that directs the Emx2 expression in dorsal telencephalon but not in caudal forebrain primordium at the 3- to 6-somite stage. This enhancer was proposed to comprise two elements, 450 base pair (bp) DT1 and 180 bp DT2, 1.0 kb apart, both of which were essential and sufficient to direct the expression in dorsal telencephalon, but had no activity in caudal forebrain primordium at E8.5. However, we have realized that the promoter region adjacent to the Emx2 translational start site does not harbor any enhancer activities. With an aim to correctly map the enhancer and to identify the caudal forebrain primordium enhancer, we have examined the enhancer activities of 28 non-coding domains that are conserved among mouse, human, chick and Xenopus at the mouse Emx2 locus.

Here we report that DT2 exists immediately upstream of the Emx2 translational start site, but DT1 exists immediately 3′ downstream of the last coding exon. Among mouse, chick and Xenopus Emx2 loci, DT1 is conserved at a domain we named θ but DT2 is not. DT2 is unnecessary to the enhancer activity of the DT1-θ domain. In addition, the θ domain has the enhancer activity not only in dorsal telencephalon but also in caudal forebrain primordium at the 3- to 5-somite stage. After E11.5, it has the activity in the cortical ventricular zone but not in cortical hem or Cajal-Retzius cells in the marginal zone. The enhancer is regulated not only by Tcf and Smad, but also by Otx. Two other sites for unknown transcriptional factors, TGTTTTTTGCATGCTTCATTTGCTT and GTGCAAATCAGTTTAAGCAATTATC, were also demonstrated to be essential for the enhancer activity. Of note is that not only the Emx2 expression in dorsal telencephalon after E9.5, but also the expression in caudal forebrain primordium at the 3- to 5-somite stage and in the ventricular zone at E15.5 was regulated by all these factors. Mutant mice that lack this enhancer indicated that the enhancer is indeed essential to Emx2 expression during forebrain development. However, the mutants also indicated the presence of another enhancer for the Emx2 expression in telencephalon unique to mammals.

Comparison of mouse Emx2 genomic sequences with other species was conducted with the Berkeley Genome Pipeline and Genomic VISTA (Couronne et al., 2003) programs. Additionally, sequence alignment was confirmed with the BLAST program (Mayor et al., 2000). Putative transcription binding sites were predicted with the TFSEARCH program (Heinemeyer et al., 1998). The source of finch data is Ensemble Genome Data Resources, Sanger Center Institute (http://www.ensembl.org/index.html).

Each mouse Emx2 genomic domain and Xenopus Emx2 θ domain was obtained by PCR with primers located more than 100 bp apart from the 5′ and 3′ ends of the domains conserved between mouse and Xenopus, except for the (ζ–η–θ) domain. A 6.8 kb fragment containing mouse Emx2 exon 1-3 (mEmx2GA2) was isolated from a BAC clone. The (ζ–η–θ) domain was obtained as a 4.9 kb fragment by NruI and NotI digestion of the mEmx2GA2 plasmid. The mouse θ domain (∼0.7 kb) was obtained by PCR with the following primers: 5′-TCTGAGAGATCTTCCACTCT-3′ and 5′-GTACCCACAATACAAACTCC-3′. The Xenopus θ domain was obtained as an 809 bp fragment from Xenopus tropicalis genome DNAs by PCR with the following primers: 5′-ACTCAACAAACCGATTGCGAATGC-3′ and 5′-ATCGTAATAATCTGGGGTTGAGTC-3′. The DNAs were provided by the National Bio-Resource Project (NBRP) of the MEXT, Japan. The PCR products were confirmed by sequencing. Tcf and Smad binding sites were transversely mutated and Otx binding sites (TAATCC) were mutated to GGCGCC, utilizing the PCR-based overlap extension method (Kucharczuk et al., 1999). Each 25 bp of the mouse N-fragment was replaced with a luciferase sequence (CTGGAGCCTGAGGAGTTCGCTGCCT) by the same method. All genomic DNA isolated or mutagenized by PCR was sequenced to verify the absence of spurious mutations.

P19 cells expressing FLAG-HA-tagged Otx2 were established by transfecting the Otx2 gene under the E1a promoter. The cells were neuralized by treating them with 1 nM retinoic acid from 24 hours after plating for a 48-hour period as previously described (Pachernik et al., 2005); the cells were further cultured for 48 hours in the absence of retinoic acid and nuclear extracts were prepared as previously described (Dignam et al., 1983). Under these conditions, the cells expressed early neuroectoderm marker Sox1, rostral brain markers Otx1 and Otx2, forebrain markers Emx2 and Pax6, and midbrain marker Dmbx1, but not hindbrain marker Gbx2, mesoderm marker Bra nor endoderm marker Gata4. Otx2 protein was isolated with FLAG-M2 beads (Sigma) as per manufacturer's instructions. The electrophoretic mobility shift assay was conducted as previously described (Takasaki et al., 2007).

The assay with the neuralized P19 cells was conducted as previously described (Agata et al., 2001). Otx2 antibodies were purchased from Chemicon (#1 in Fig. 6C) and R & D (#2 in Fig. 6B,C). The assay with E11.5 telencephalon was performed as previously described (Visel et al., 2009). Antibodies used were rabbit anti-Otx2 (ab21990, Abcam), rabbit anti-Tcf1 (C63D9, Cell Signaling), rabbit anti-Tcf4 (C48H11, Cell Signaling), rabbit anti-Smad1/5/8 (N-18, Santa Cruz Biotechnology) and rabbit control IgG (ab46540, Abcam) coupled to Dynabeads Protein G (Invitrogen). Primers for the PCR were p1 (5′-ACCAGGAATATGAAGGGAAAAGAGG-3′) and p2 (3′-GCATTGTAACTTGGCCTTATCACAG-5′), of which locations are indicated in Fig. 4B.

The isolation of the recombinant clones and production of the FB enhancer mutant mice (Accession No. CDB0073K) were performed as previously described (http://www.cdb.riken.go.jp/arg/download_file/vector_09.pdf). The primers used to identify the wild-type allele were p1 (5′-ATCAGTTTAAGCAATTATCATACCAG-3′) in the θ domain and p2 (5′-TGTAAGGAGACTATACTCTACTCTAT-3′) in its 3′ downstream domain (Fig. 8A); those used to identify the FB neo allele were p3 (5′-GTACTCGGATGGAAGCCGGTCTTGTC-3′) in the neo cassette and p2; those used to identify the ΔFB allele were p4 (5′-TCAATATGCCCTCAGAGTAGTTGTCT-3′) in front of the insertion site of the neo cassette and p2 (Fig. 8A). Otx2-null (Accession No. CDB0010K), Emx1-null (Accession No. CDB0021K) and Emx2-null (Accession No. CDB0018K) mutants and Cre knock-in mutant mice at the Emx2 locus (Accession No. CDB0020K) were established previously and genotyped as described (http://www.cdb.riken.jp/arg/mutant%20mice%20list.html) (Matsuo et al., 1995; Yoshida et al., 1997; Kimura et al., 2005). ROSA26R mice were obtained from Jackson Laboratory (http://jaxmice.jax.org/strain/003474.html). Mice were housed in environmentally controlled rooms under the RIKEN Center for Developmental Biology (CDB) guidelines for animal and recombinant DNA experiments.

RNA isolation and semi-quantitative reverse transcription (RT) PCR were performed according to Kimura et al. (Kimura et al., 2001). Primer sets used were as follows: Emx2, 5′-CCGAGAGTTTCCTTTTGCACAACGC-3′ and 5′-GCCTGCTTGGTAGCAATTCTCCACC-3′; HPRT, 5′-GAAATGTCAGTTGCTGCGTC-3′ and 5′-GCCAACACTGCTGAAACATG-3′. Quantitative RT-PCR was carried out as previously described (Shibata et al., 2008). For all primer sets tested, correlation (R2) was higher than 0.98 and the slope was –3.1 to –3.6 in each standard curve. Primers to detect the expression of each gene were designed in a single exon encoding a 3′UTR: forward, 5′-CTGCACACACATCCACCGAG-3′, and reverse, 5′-GCGTCACTGCTCTGATTCCC-3′ for Emx2; forward, 5′-GTGATGTGAAGTTCCCCATAAGG-3′, and reverse, 5′-CTACTGAACTGCTGGTGGGTCA-3′ for Tbp (Svingen et al., 2009).

As true of mouse Emx2, chick and Xenopus Emx2 genes are expressed in caudal forebrain primordium at neural plate stage, in dorsal telencephalon after neural tube closure and in ventricular zone of the developing cortex (Bell et al., 2001; Pannese et al., 1998). Therefore, we rationalized that the enhancers directing these expressions are conserved among tetrapods. We thus compared 556-kb non-coding genomic sequences between the genes adjacent to Emx2 at the 5′ and 3′ ends (PDZK8 and Rab11FIP2) among mouse, human, chick and Xenopus (Fig. 1A). We chose 5 (α–ϵ) and 23 (ζ-ω5) domains at the 5′ and 3′ ends of the translational start site, respectively, as the domains conserved among tetrapods. The DT1 domain from Theil et al. (Theil et al., 2002) corresponds to the θ domain, whereas the DT2 domain is not conserved, being located between the γ and δ domains (Fig. 1B). A 2.1 kb region proximal to the translational start site (Fig. 1B) that covers the δ and ϵ domains did not express any apparent β-gal expression in E8.0-E15.5 transgenic embryos when fused with a lacZ reporter gene (2.1-lacZ). This 2.1 kb fragment was then used as the promoter in the enhancer assay; each domain was combined with 2.1-lacZ and the enhancer activity was examined by generating transgenic embryos (Fig. 1C). It was only the (ζ–η–θ) domain that exhibited enhancer activities in anterior neuroectoderm and forebrain during E8.0-E15.5; this domain exhibited early activity in caudal forebrain primordium at the 3- to 6-somite stage.

We then generated permanent transgenic lines that harbored (ζ–η–θ)-2.1-lacZ to compare temporal changes of the activities with the endogenous Emx2 expression in detail (Fig. 2). The β-gal expression was first observed at the 3- to 6-somite stage in caudal forebrain primordium, as was endogenous Emx2 expression (Fig. 2C,G). At E9.5, endogenous Emx2 expression was found in telencephalon but not in diencephalic region (Fig. 2D,E). The β-gal expression, however, remained in the diencephalic region and was rarely found in the telencephalic region (Fig. 2H). At E10.5, β-gal expression was still retained in the diencephalon, but also occurred in a dorsocaudal part of the telencephalon (Fig. 2I,J); β-gal expression covered the entire diencephalon except for the commissure region of pretectum. At E11.5, the expression faded in the diencephalon and covered the entire dorsal telencephalon (Fig. 2K,L). Therefore, the enhancer activity of the (ζ–η–θ) domain shifted from the diencephalic region to the telencephalic region, as endogenous Emx2 expression does, but the shift was delayed. Moreover, endogenous Emx2 expression includes the cortical hem, although not the choroid plexus or choroidal roof, and is graded high dorsomedially in the cortex (Fig. 2F,M,M′,O). The activity of the (ζ–η–θ) domain did not include the cortical hem and was not graded in the cortex (Fig. 2L,N,N′,P). The endogenous Emx2 expression was weakly found in the ventricular zone of lateral ganglionic eminences and a part of medial ganglionic eminences (Fig. 2M). The (ζ–η–θ) domain exhibited significant activities not only in the ventricular zone but also in the differentiating field of lateral ganglionic eminences; it had no activity in medial ganglionic eminences (Fig. 2N). In E15.5 cortex, endogenous Emx2 expression is found in the ventricular zone and in Cajal-Retzius cells in the marginal zone, but not in the cortical plate or intermediate zone (Fig. 2Q,Q′) (Shinozaki et al., 2002). The majority of Cajal-Retzius cells originate from the Emx2-positive cortical hem (Takiguchi-Hayashi et al., 2004; Yoshida et al., 2006). The (ζ–η–θ) domain was active in the ventricular zone, but it had no activity in the marginal zone (Fig. 2R,R′), coincident with the lack of the activity in the cortical hem (Fig. 2N,N′). These differences in the enhancer activities of the (ζ–η–θ) domain from the endogenous Emx2 expression were commonly observed in three transgenic lines independently established. β-gal protein is relatively stable, and differences might exist between lacZ mRNA expression and β-gal protein expression. However, being directed by the (ζ–η–θ) domain, the expression pattern of the lacZ mRNA was almost the same as that of the β-gal expression.

At E11.5, Emx2 is also expressed in part of the pretectum and tegmentum, and weakly in the thalamus, prethalamus and thalamic eminence (Fig. 2O). The (ζ–η–θ) domain had significant activities at these sites (Fig. 2P). Endogenously, Emx2 is also expressed in the urogenital ridge (data not shown) and nasal epithelia (Fig. 2O). The (ζ–η–θ) domain did not have activities at these sites (Fig. 2P; data not shown).

The enhancer activities of the (ζ–η–θ) domain were also determined by injecting the (ζ–η–θ)-2.1-Cre transgene, in which the domain was combined with the 2.1 kb promoter and Cre gene (2.1-Cre), into zygotes from ICR females mated with homozygous ROSA26R mice; the β-gal expression in the embryos developed from the zygotes represents all the cells in which the (ζ–η–θ) domain was at one time active. The expression was found in caudal forebrain primordium at E8.5 and in telencephalon and diencephalon at E10.5 (Fig. 2S-U). This is as observed in ROSA26R embryos obtained by cross with Emx2^+/Cre mice in which the Cre recombinase gene was knocked-in into the Emx2 gene locus to be expressed in place of Emx2 (Fig. 2V-X). Of note is that the (ζ–η–θ) domain exhibited activity in telencephalon at E9.5 in this assay (Fig. 2T).

In dissecting the enhancer activities of a 4.9 kb A fragment that covers the (ζ–η–θ) region, we took advantage of the Cre-ROSA26R system by which enhancer activities both at earlier stages in caudal forebrain primordium and at subsequent stages in telencephalon can be determined by a single assay at E9.5. The C (3.1 kb), F (1.8 kb), G (1.0 kb) and Q (0.7 kb) fragments retained enhancer activities the same as the 4.9 kb A fragment; the activity was finally confined to the 400 bp BglI-HindIII fragment (N) (Fig. 3A,Ba-d). Further dissection of the N fragment into three fragments (R, S and T in Fig. 3A) abolished activity both in caudal forebrain primordium and telencephalon; the middle T fragment exhibited aberrant activity (Fig. 3Be) and the 5′ R or 3′ S fragment exhibited no activity (Fig. 3A). The activity of the N fragment was also examined by combining it with 2.1-lacZ (Fig. 3C). It exhibited the same activity as the (ζ–η–θ) domain (Fig. 2). However, the shift of its activity from diencephalon to telencephalon took place earlier than the (ζ–η–θ) domain, although still later than endogenous Emx2 expression. The N fragment coincides with the DT1 element from Theil et al. (Theil et al., 2002) and the θ domain conserved among tetrapod Emx2 loci (Fig. 1A,B); we refer here to the enhancer existing in the N fragment or θ domain as the FB (forebrain) enhancer.

The nucleotide sequence of the θ domain was deeply conserved not only among amniotes but also in Xenopus (Fig. 4A,B). Smad and Tcf binding sites are present in the mouse θ domain as reported by Theil et al. (Theil et al., 2002), and they are conserved among tetrapod Emx2 θ domains. In addition, two potential Otx binding sites exist in Emx2 θ domains of these animals (Fig. 4B). To examine whether these sites are indeed essential to the FB enhancer activities in caudal forebrain primordium and telencephalon, the enhancer activities of the N fragments in which mutation or deletion was introduced into these sites were examined by the Cre-ROSA26R system (Fig. 5). The transverse mutation of the Smad binding site reduced β-gal expression at E9.5 (Fig. 5Bb), and β-gal expression was residual when the Tcf binding site was transversely mutated (Fig. 5Bc). Mutation in both Tcf and Smad sites abolished β-gal expression completely (Fig. 5Bd). In addition, both deletion and transverse mutation of the two Otx binding sites abolished the activities of the N fragment in both diencephalon and telencephalon (Fig. 5Be,f). Tcf, Smad and Otx binding sites were also essential to the activity in the E15 cortical ventricular zone (Fig. 5Bg-i).

Smad and Tcf were reported to bind to the Smad and Tcf binding sites, respectively, by Theil et al. (Theil et al., 2002). Electrophoretic mobility shift analysis was conducted to demonstrate that Otx protein also binds to the potential binding sites in the 0.4 kb FB enhancer. The Otx2 protein, isolated from neuralized P19 cells expressing FLAG-HA-tagged Otx2 under the EF1a promoter (see Materials and methods), yielded a uniquely shifted band (Fig. 6B, lanes 1 and 2). This band was almost lost in competition with a wild-type competitor (Fig. 6A, WT; Fig. 6B, lanes 3 and 4) or mutant competitors in which only one of the two Otx sites was mutated (Fig. 6A, 1M and 2M; Fig. 6B, lanes 5-8). However, the shifted band was not lost in competition with a mutant competitor in which both Otx sites were mutated (Fig. 6A, 1-2M; Fig. 6B, lanes 9 and 10). Furthermore, anti-Otx2 antibody super-shifted the band (compare lanes 11 and 12 in Fig. 6B), suggesting that the complex represented by the shifted band indeed contains Otx2 protein.

We next performed a chromatin immunoprecipitation (ChIP) assay to confirm that endogenous Otx binds to these Otx sites. The ChIP assay with the neuralized P19 cells was conducted with two different antibodies (Fig. 6C) and PCR primers were set up so that a 144 bp sequence, including the two Otx sites, was amplified (Fig. 4B). The ChIP assay with Otx2 antibodies indeed yielded the 144-bp products (Fig. 6C); they were confirmed by sequencing. The assay was also performed with E11.5 telencephalon. Otx2, Tcf4 and Tcf1 (Hnf1a – Mouse Genome Informatics) antibodies precipitated the enhancer (Fig. 6D), although we were unable to confirm Smad1/5/8 binding to the Smad site with the antibody used.

Otx, Tcf and Smad sites do not explain the deletion analysis of the N fragment into R, S and T subfragments or the requirement of the 5′ R subfragment (Fig. 3). It is also possible that sequences other than the Tcf, Smad and Otx sites are essential in the N fragment. A linker scanner assay was then conducted by replacing each 25 bp sequence of the N fragment with a 25 bp unrelated luciferase sequence (CTGGAGCCTGAGGAGTTCGCTGCCT); the sequence has no obvious consensus binding sites for known transcription factors by TRANSFAC database search. The mutant N fragments were numbered from #1, in which the first 25 bp was replaced with the luciferase sequence, to #17, in which the last 25 bp was replaced with the luciferase sequence. Mutant N fragments #1-4, #6, #7, #11, #12 and #14-17 exhibited almost the same activities as the wild-type N fragment (Fig. 7A). By contrast, #5, #9, #10 or #13 mutant N fragments exhibited no activity and the #8 mutant fragment exhibited only a weak activity in caudal telencephalon (Fig. 7A,Bb-f). The #8 sequence contains a Smad binding site, #9 a Tcf binding site and #13 two Otx binding sites. Of particular interest were the #5 and #10 sites that do not have a consensus sequence for the binding of known transcriptional factors; the #5 site explained the requirement of the R subfragment. Both were also essential to the activity in the E15.5 ventricular zone (Fig. 7Bg-i) and are well conserved among tetrapods (Fig. 4B). Despite intensive efforts, however, we have so far been unsuccessful in identifying the factors that bind to sequences #5 and #10.

To confirm the roles for the FB enhancer in Emx2 expression, we next generated mouse mutants that lack this enhancer (Fig. 8A,B); the G domain (Fig. 3A) was first replaced with the neo cassette flanked by loxP sequences (FB-neo or neo allele) and the cassette was then deleted (ΔFB allele). Semi-quantitative RT-PCR and RNA in situ hybridization demonstrated that the Emx2 expression in caudal forebrain primordium at E8.5 was lost in the homozygous FB enhancer mutants (Emx2^neo/neo and Emx2^ΔFB/ΔFB mutants; Fig. 8Bc,Ca,e). However, at E9.5 and E10.5, Emx2 expression decreased to about half, but was present in both Emx2^neo/neo and Emx2^ΔFB/ΔFB mutants (Fig. 8Bc,Cb-d,f-h). The loss of the Emx2 expression in caudal forebrain primordium at E8.5 and the reduction in E10.5 telencephalon was further confirmed by quantitative RT-PCR (Fig. 8D). The level of Emx2 expression in E8.5 mutant head and in E10.5 diencephalon was less than 10% of that in wild-type counterparts, whereas the level of Emx2 expression in E10.5 mutant telencephalon was about 40% of that in wild-type telencephalon. Sequencing of the amplified products with E10.5 mutant telencephalon indicated that the products are indeed Emx2 transcripts. Therefore, we conclude that the FB enhancer is responsible for Emx2 expression in caudal forebrain primordium at E8.5, but another enhancer must exist for Emx2 expression in telencephalon.

Emx2 function in the development of caudal forebrain primordium is represented in the defects of Emx2 and Otx2 double mutants (Emx2^–/–;Otx2^+/–), in which the thalamic eminence, ventral thalamus (prethalamus), dorsal thalamus (thalamus) and non-commissure regions of the pretectum are lost (Fig. 9Aa,b) (Suda et al., 2001; Kimura et al., 2005). Emx2^neo/neo;Otx2^+/– and Emx2^ΔFB/ΔFB;Otx2^+/– mutants lost these diencephalic structures as did Emx2^–/–;Otx2^+/– (Fig. 9Ac,d). This is consistent with the loss of Emx2 expression in the E8.5 Emx2^neo/neo and Emx2^ΔFB/ΔFB mutants, confirming that the FB enhancer is solely responsible for Emx2 expression in E8.5 caudal forebrain primordium.

Emx2 function in early telencephalon or medial pallium development is represented in Emx1^–/–;Emx2^–/– double mutants (Shinozaki et al., 2004); the loss of medial pallium is significant in Emx2^–/– mutants and remarkable in Emx1^–/–;Emx2^–/– mutants (Fig. 9Ba-c). The loss was moderate in Emx2^ΔFB/ΔFB, Emx1^–/–;Emx2^neo/neo and Emx1^–/–;Emx2^ΔFB/ΔFB mutants (Fig. 9Bd-f). The cortical hem, where the FB enhancer is inactive, develops into fimbria. At E18.5, histologically dentate gyrus is lost in Emx2^–/– mutants, and fimbria, dentate gyrus and hippocampus are lost in Emx1^–/–;Emx2^–/– mutants (Fig. 9Ca-c) (Shinozaki et al., 2004). Dentate gyrus was lost, but fimbria developed, and hippocampus was moderately reduced in Emx1^–/–;Emx2^neo/neo and Emx1^–/–;Emx2^ΔFB/ΔFB mutants (Fig. 9Cd-f).

Subsequent Emx2 functions in Cajal-Retzius cell development and cortical lamination are also demonstrated in Emx1;Emx2 double mutant cortex (Fig. 9Da-f) (Shinozaki et al., 2002). Histologically, defects in cortical laminar structure were moderate in the Emx1^–/–;Emx2^neo/neo and Emx1^–/–;Emx2^ΔFB/ΔFB mutant cortex (Fig. 9Dg-l) when compared with the Emx1^–/–;Emx2^–/– defect. The cortical hem is the major source of Cajal-Retzius cells. Reelin-positive cells are lost in Emx1^–/–;Emx2^+/– mutant cortex (Fig. 9Dc,f) (Shinozaki et al., 2002), but were present in Emx1^–/–;Emx2^neo/neo and Emx1^–/–;Emx2^ΔFB/ΔFB mutants (Fig. 9Di,l). Furthermore, interneurons are scarce in Emx1^–/–;Emx2^+/– mutant cortex owing to the failure of their tangential migration, causing the hyperplasia of ganglionic eminences (Fig. 9Ca,c) (Shinozaki et al., 2002). However, ganglionic eminences were apparently normal in Emx1^–/–;Emx2^neo/neo and Emx1^–/–;Emx2^ΔFB/ΔFB mutants (Fig. 9Ce,f). Therefore, the telencephalic phenotypes of the double mutants with Emx1^–/– indicated the hypomorphic nature of Emx2^neo/neo and Emx2^ΔFB/ΔFB mutations and the presence of the second enhancer for Emx2 expression in telencephalon. No difference was apparent in any of these double mutant phenotypes between Emx2^neo/neo and Emx2^ΔFB/ΔFB mutants.

The θ domain is conserved in the Xenopus Emx2 locus. Otx, Tcf and Smad binding sites are perfectly conserved, the #10 site has only one base change among its 25 bases, and 20 bases out of the 25 are the same in the #5 site (Fig. 4B). The enhancer activity of the Xenopus θ domain was then examined in mouse forebrain. It indeed exhibited activities in caudal forebrain primordium at E8.5, in dorsal telencephalon at E9.5 and in the ventricular, but not marginal, zone at E15.5 (Fig. 4C). These activities of the Xenopus θ domain were almost the same as those of the mouse; however, the shift of its activity from diencephalon to telencephalon took place at E9.5. Thus, the FB enhancer is most probably responsible for the basal Emx2 functions in tetrapod forebrain, but not for the functions unique to mammals; the FB enhancer lacked the activities in the cortical hem and Cajal-Retzius cells and its activity was not graded in the cortex.

We expected that different enhancers direct each Emx2 expression: in caudal forebrain primordium at the 3- to 6-somite stage, in dorsal telencephalon during its initial development and in the cortical ventricular zone and Cajal-Retzius cells during cortical lamination (Kurokawa et al., 2004a; Kurokawa et al., 2004b). Unexpectedly, however, the FB enhancer identified in this study had activities in caudal forebrain primordium, in dorsal telencephalon and in the ventricular zone. Not only Tcf and Smad sites, but also Otx sites and two other sites for unknown factors – sequences #5 (TGTTTTTTGCATGCTTCATTTGCTT) and #10 (GTGCAAATCAGTTTAAGCAATTATC) – were essential to all these activities. The FB enhancer mutants indicated that the enhancer is solely responsible for the Emx2 expression in caudal forebrain primordium but that there is another enhancer for the Emx2 expression unique to mammalian telencephalon.

The FB enhancer is conserved among tetrapod Emx2 loci; the Xenopus θ domain exhibited enhancer activities in mouse caudal forebrain primordium at E8.5, in dorsal telencephalon at E9.5 and in the ventricular zone at E15.5. These activities of the Xenopus θ domain are almost the same as those of the mouse. Moreover, Otx, Tcf, Smad, #5 and #10 sites are well conserved among tetrapod θ domains, suggesting that the upstream mechanisms that control the θ domain enhancer activity at each site must also be conserved among tetrapods. The θ domain enhancer and its upstream factors thus would have been established in ancestor tetrapods to shoulder the Emx2 functions in forebrain development common to tetrapod Emx2 genes.

The change in the active sites of the (ζ–η–θ) domain solved a question about early Emx2 expression. Previous cell lineage analysis with Cre knock-in into the mouse Emx2 locus indicated that although most of diencephalic cells do not express Emx2 at E9.5, they are descendants of Emx2-positive caudal forebrain primordium at E8.5 (Kimura et al., 2005). They precisely coincide with the cells in which the (ζ–η–θ) domain is active at E9.5 (Fig. 2J). The cell lineage analysis, however, could not tell whether Emx2-positive telencephalic cells at E9.5 are descendants of Emx2-positive caudal forebrain primordium at E8.5 or if the Emx2 expression newly begins in the telencephalic cells that have not expressed Emx2 at E8.5. The activity shift of the (ζ–η–θ) domain apparently suggested that the FB enhancer initially does not have an activity in telencephalic primordium when it is active in diencephalic precursor cells. The profile of the (ζ–η–θ) domain activities thus suggests that the endogenous Emx2 expression first takes place in diencephalic precursor cells at the 3- to 4-somite stage but becomes suppressed by E9.5 in their descendant cells. Coincidentally, the endogenous Emx2 expression takes place in dorsal telencephalon by E9.5. In Xenopus at early neurula, it was reported that Emx2 expression is first found in an area that generates diencephalon, being distinct from Emx1-positive future telencephalon; soon after neural tube closure, Emx2 is transcribed in almost the same domain as Emx1 at the level of the dorsal telencephalon (Pannese et al., 1998). This is also likely to be the case in chick forebrain development (Bell et al., 2001).

This study opens the question how the temporal changes in the forebrain activities are regulated. Otx is essential but cannot explain the forebrain activity specific to the caudal forebrain primordium at E8.5; at this stage, Otx2 is expressed in the entire anterior neuroectoderm and Otx1 expression also starts in the ectoderm (Simeone et al., 1992). Wnt and Bmp proteins are expressed in anterior neuroectoderm and/or surface ectoderm adjacent to the neuroectoderm; they are essential to but cannot explain the Emx2 expression specific to the caudal forebrain primordium. After closure of the neural tube, Wnt and Bmp proteins continue to be expressed at the dorsal midline of diencephalon. Otx2 and Otx1 also continue to be expressed in diencephalon. Thus these cannot explain why the FB enhancer activity ceases in diencephalon later than E9.5. When the neural plate is closed at the telencephalic level, Bmp proteins are expressed in the choroidal roof, choroid plexus and cortical hem (Furuta et al., 1997). A series of Wnt proteins are also expressed in medial pallium in a nested pattern (Grove et al., 1998; Lee et al., 2000). Otx2 is also expressed in these sites (Kurokawa et al., 2004a; Kurokawa et al., 2004b). The forebrain activity appears to take place from the caudodorsomedial telencephalon (Fig. 2H,I; Fig. 3Cc,d), and Wnt, Bmp and Otx would explain the forebrain activity in telencephalon. However, it is a question why the FB enhancer, which is regulated by Otx, Tcf and Smad, does not have activity in the cortical hem, choroid plexus or roof.

At subsequent stages Wnt, Bmp and Otx2 expression persists in the dorsomedial telencephalon; after E9.5, Otx2 expression is lost in neopallium, ganglionic eminences and the hypothalamus (Simeone et al., 1992; Kurokawa et al., 2004b), but Otx1 expression continues to be expressed in the entire cortex, including the cortical hem. At E15.5, Otx2 is not expressed in the ventricular zone of the neopallium, whereas Otx1 is expressed in the entire cortical ventricular zone (Simeone et al., 1993; Frantz et al., 1994). Otx1 would thus explain the forebrain activity in the ventricular zone, but Emx2 is expressed in the ventricular zone of Otx1 mutants (Suda et al., 1997) (our unpublished data). The Tcf and Smad sites in the FB enhancer were essential to its activity even in the lateral and ventral pallium; however, the source of signalings for Tcf and Smad expression in the pallium is not clear. Identification of the factors that bind to #5 and #10 sites is the first step to address these questions.

The FB enhancer mutants indicated the presence of the second enhancer for the Emx2 expression in telencephalon. In Emx2^ΔFB/ΔFB mutant telencephalon at E9.5 and E10.5, Emx2 expression was reduced but present at about 40% strength; Emx1^–/–;Emx2^ΔFB/ΔFB mutant phenotypes in medial pallium development and cortical lamination were much milder than Emx1^–/–;Emx2^–/– phenotypes. In contrast to the FB enhancer, the second enhancer must be responsible for the graded Emx2 expression, exhibiting activity in the cortical hem and Cajal-Retzius cells. Hippocampal structures have remarkably evolved in amniotes (Butler and Hodos, 2005) and the cortical hem is thought to be the organizing center of the medial pallium development (Mangale et al., 2008). Graded Emx2 expression in the cortex, together with Pax6 expression in the reciprocal gradient, has been suggested to regulate area patterning in mammalian telencephalon (O'Leary and Nakagawa, 2002). Laminar structure and area patterning are characteristic to mammalian telencephalon. The second enhancer must thus be unique to mammals, and its identification, together with the determination of its upstream factors, will bring us valuable information on the evolution of mammalian telencephalon.

We are indebted to the Laboratory for Animal Resources and Genetic Engineering of RIKEN CDB for animal housing and mutant mouse production. This work was supported by a Grant-in-Aid for Creative Scientific Research from the Japan Society for the Promotion of Science.