Commissural neuron identity is specified by a homeodomain protein, Mbh1,that is directly downstream of Math1

An enormous variety of neuronal types are generated during vertebrate neurogenesis. Two critical steps in neurogenesis, generation of neural progenitor cells and their commitment to the neuronal fate, are controlled by proneural genes, which encode basic helix-loop-helix (bHLH) transcription factors, such as Mash1 (Ascl1 – Mouse Genome Informatics) and Math1(Atoh1 – Mouse Genome Informatics) (reviewed by Bertrand et al., 2002). Proneural genes are expressed in distinct domains or populations of cells and are thought to integrate positional information into the program of neuronal differentiation, resulting in the specification of neuronal identity. In sympathetic ganglia, Mash1 controls noradrenergic phenotypes through activation of a homeobox gene, Phox2a (reviewed by Goridis and Rohrer, 2002). Math1 controls the differentiation of cerebellar granule cells(Ben-Arie et al., 1997), hair cells in the inner ear (Bermingham et al.,1999) and commissural neurons in the spinal cord(Bermingham et al., 2001; Gowan et al., 2001). Many bHLH proteins activate transcription through binding to an E-box nucleotide sequence motif. A neuronal differentiation bHLH protein, NeuroM, has been shown to regulate Hb9, which is necessary for differentiation of motoneurons (Lee and Pfaff,2003). Less is known, however, about downstream targets of proneural bHLH proteins and their molecular mechanisms for specifying neuronal identity.

In the developing dorsal spinal cord, domains of progenitor cells are distinguished by the expression of bHLH genes, which are initially established by TGFβ-like signals (reviewed by Lee and Jessell, 1999; Caspary and Anderson, 2003; Helms and Johnson, 2003). Each domain produces a distinct set of neurons,which are marked by combinatorial expression of homeobox genes. Math1is expressed by dorsalmost cells adjacent to the roof plate, which give rise to dI1 cells positive for LIM-class homeodomain proteins, Lhx2 (LH2A) and Lhx9(LH2B). dI1 cells are lost in Math1 knockout mice(Bermingham et al., 2001; Gowan et al., 2001), whereas misexpression of Math1 increases the number of dI1 cells and commissural neurons (Gowan et al.,2001). A Bar-class homeobox gene, Mbh1 (Barhl2– Mouse Genome Informatics), is also expressed by dI1 cells, and Mbh1-positive cells give rise to commissural neurons(Saba et al., 2003). Dorsal cells that express Mbh1 ectopically are transfated to commissural neurons in the spinal cord, suggesting that Mbh1 is sufficient for the specification of commissural neuron identity(Saba et al., 2003).

In this study, we identified an enhancer directing Mbh1 expression in the spinal cord by analyzing transgenic mice that carried lacZwith Mbh1-flanking genome sequences. An E-box, which was conserved among mouse, rat and human sequences, was critical to drive lacZexpression in the dorsal spinal cord, suggesting that Mbh1 expression is regulated by a bHLH protein. Furthermore, chromatin immunoprecipitation(ChIP) experiments revealed that Math1 bound the Mbh1 enhancer containing the E-box in the spinal cord. Transfection assays in the mouse spinal cord indicated that expression of a reporter gene carrying the E-box was activated specifically by Math1. These results, taken together with Mbh1 expression in the gain- and loss-of-function experiments of Math1, indicate that Mbh1 is a downstream target gene of Math1. The function of Mbh1 was analyzed using chimeric proteins containing the homeodomain of Mbh1 and functional domains that can modulate transcription. A chimeric protein containing the Engrailed repressor domain generated commissural neurons, as Mbh1 and Math1 did. By contrast, a chimeric protein containing the VP16 activator domain inhibited generation of commissural neurons. These findings suggest that transcriptional repressor activity of Mbh1 is necessary and sufficient for the specification of commissural neurons. Thus, these studies revealed a cascade of events for specifying commissural neuron identity in the spinal cord.

Eleven overlapping lambda genomic DNA clones were obtained by screening a 129SvJ mouse genomic library (Stratagene) using Mbh1 cDNA (GenBank,AB004056) as a probe. These clones covered a 22.5 kb genome sequence including 4.5 kb 5′, 12.5 kb 3′ and introns. The transcription initiation site was determined using RNA from embryonic day (E) 12.5 mouse embryos with a 5′ RACE System for Rapid Amplification of cDNA Ends, Version 2.0 (Gibco BRL).

The lacZ-coding region with SV40 polyadenylation site derived from BGZA (Yee and Rigby, 1993; Helms and Johnson, 1998) was inserted downstream of the translation start site of Mbh1, which matched the translation start site of β-gal. In Tg5 and Tg6, the basalβ -globin promoter was used in place of the 1.0 kb 5′sequence. Tg18 was constructed by linking three tandem copies of the 123 bp fragment containing the E-box. The mutation was introduced into the E-box using an In Vitro Mutagenesis Kit (Takara).

Transgenic mice were generated and analyzed as described previously(Saba et al., 2003). Briefly,injected eggs were implanted into ICR female mice, and founders were collected and stained for β-gal activity. Stable transgenic lines were also generated for Tg2 and Tg4. Transgenes were detected by PCR for the lacZ gene using placenta DNA for embryos and tail DNA for pups.

Section and whole-mount in situ hybridization were performed as described in Saito et al. (Saito et al.,1996) and Wilkinson(Wilkinson, 1992),respectively. Antisense RNA probes were synthesized from plasmids carrying mouse cDNA clones: pMH4-1 and pNH10 for Mbh1; Lhx9 (gift of T. M. Jessell, Columbia University); Math1 (gift of R. Kageyama,Kyoto University). Immunofluorescent studies were performed as described(Saba et al., 2003). The following primary and secondary antibodies were used for visualizing the signals: rabbit anti-Math1 (Helms and Johnson, 1998); goat anti-β-gal (Biogenesis); mouse monoclonal anti-bromodeoxyuridine (BrdU) (Sigma); donkey anti-mouse and anti-rabbit IgGs conjugated with Cy3 (Jackson ImmunoResearch); donkey anti-goat IgG conjugated with Alexa Fluor 488 (Molecular Probes).

In vivo electroporation was performed as described before(Saba et al., 2003). Solution(1 μl) containing 140 nmol/l each plasmid in PBS was injected into the central canal of the spinal cord of E11.5 mouse embryos. Half-ring type electrodes were used to transfect DNA through the spinal cord. pEYFP, which carried EYFP downstream of a CAG promoter(Saito and Nakatsuji, 2001),was used as a control. pEYFP-Math1 and pEYFP-Mash1 were constructed by inserting the coding region of Math1 and Mash1 downstream of the second CAG promoter of pCAG-EYFP-CAG(Saito and Nakatsuji, 2001),respectively. The Math1-HA gene was constructed by inserting oligonucleotides encoding the HA tag immediately upstream of the translation termination codon of Math1. The En-Mbh1 and VP16-Mbh1chimeric genes were constructed by fusing the sequences encoding the Drosophila Engrailed repressor domain(Jaynes and O'Farrell, 1991)and the VP16 activator domain (Clontech) to the sequence encoding the C-terminal portion of the Mbh1 protein, respectively. These three genes were inserted downstream of the second CAG promoter of pCAG-EYFP-CAG to construct pEYFP-Math1-HA, pEYFP-En-Mbh1 and pEYFP-VP16-Mbh1. Each result of electroporation was confirmed by using at least two independently isolated clones with the same structure.

Chromatin was prepared as described in Forsberg et al.(Forsberg et al., 2000) with minor modifications. To examine binding of endogenous Math1, the spinal cord from ∼28 mouse embryos at E10.5 was used for one assay. To analyze binding of misexpressed Math1, EYFP⁺ sides of the spinal cord were dissected out 24 hours after electroporation at E11.5, and ∼10 electroporated embryos were used for one assay. The dissected spinal cord was fixed with 1% formaldehyde in PBS for 3 hours on ice. After cell lysis and sonication, ChIP was performed following the manufacture's protocol in the ChIP Assay Kit (Upstate). The following antibodies and IgG (3 μg) were used: rabbit anti-Math1; rabbit anti-neurofilament 200 (Sigma); rat anti-HA(3F10, Roche); rat IgG (Immunotech). Immunocomplexes were pulled down using protein A and protein G-agarose beads for rabbit and rat IgGs, respectively. A 502 bp fragment spanning nucleotides +5507 to +6009 containing the E-box of the Mbh1 enhancer was amplified by semi-quantitative PCR using the following primers: sense, TTCCAGGTGCCCGCCTCTTCTGA; antisense,TTCGCGGATCCAAGCACAACTCATT. As a negative control, a 546 bp DNA fragment spanning nucleotides –6 to +540 of the Mbh1 gene was amplified using the following primers: sense, GTAGAAATGACAGCAATGGAAGG; antisense,CCTGAAGCTCTCGTGTGC. The intensity of PCR bands was analyzed using a Typhoon 9410 fluorescence imager (Amersham Bioscience). For all experiments,immunoprecipitated DNA templates were well under the saturation level.

Forty-eight hours after electroporation, EYFP⁺ regions in one side of the spinal cord were dissected out in cold PBS under a fluorescent stereomicroscope and suspended with lysis buffer from the High Sensitivityβ-galactosidase Assay Kit (Stratagene). Approximately 150 μg of protein was obtained from the EYFP⁺ region of one embryonic spinal cord. β-gal activity was measured using the Assay Kit according to the manufacturer's protocol. The efficiency of transfection was normalized with the intensity of EYFP fluorescence, which was measured using the Typhoon 9410 fluorescence imager.

Mbh1 was expressed in several areas of the developing nervous system, including the ventral telencephalon, diencephalon, mesencephalon,hindbrain and spinal cord (Fig. 1). In the E10.5 spinal cord, the expression pattern of Mbh1 was similar to that of Math1, as previously analyzed by sectioning (Saba et al.,2003). At E11.5, Mbh1 expression extended ventrally to the deep dorsal horn in the spinal cord, reflecting ventral migration of Mbh1⁺ cells (Saba et al., 2003).

The mouse Mbh1 gene spanned 5.5 kb and was composed of three exons. The 5′ end of the Mbh1 transcript was mapped to an A residue 641 bases upstream of the translation start site(Fig. 2A). The transcription initiation site was located in a GC-rich region, which contained a CAAT box,and binding sites for Sp1, ETS and CREB/ATF, but not a discernible TATA box(the nucleotide sequence was deposited as AB063281 in GenBank).

To search for an enhancer responsible for Mbh1 expression, we constructed several transgenes carrying parts of the Mbh1 genome encompassing from –4.5 to +11 kb(Fig. 2A). The lacZreporter gene was inserted immediately downstream of the translation start site of the Mbh1 open reading frame. Each transgene was injected into fertilized eggs, and lacZ expression was analyzed in transgenic embryos developed from the eggs at E11.5, at which stage Mbh1expression was pronounced in the spinal cord. We had previously reported the result of Transgene (Tg)4 (Saba et al.,2003) and extended our analysis to other Tgs. Tg1 did not demonstrate lacZ expression, whereas transgenic embryos harboring Tg2 and Tg4, which contained 1.0 kb 5′ and 2.5 kb 3′ sequences,expressed lacZ in the dorsal spinal cord(Fig. 2B-D). The expression of lacZ closely resembled that of endogenous Mbh1 (see Fig. 1 for whole embryos and Saba et al., 2003 for sections). Moreover, β-galactosidase (β-gal)⁺ cells were labeled with anti-Mbh1 antibody (Saba et al., 2003), indicating that the expression of lacZrecapitulated endogenous Mbh1 expression. Sections of the transgenic embryos showed β-gal⁺ axons projecting to the floor plate and ventral funiculi, confirming that β-gal⁺ cells gave rise to commissural neurons (Fig. 2F,G). The temporal expression patterns of lacZ were examined using transgenic mice bearing Tg2 and Tg4. The onset of lacZexpression matched that of endogenous Mbh1 expression, which was first detected at E10.5 (Saba et al.,2003). Preceding Mbh1 expression, Math1expression was detected at E9.5 (Helms and Johnson, 1998) (data not shown). At E10.5, when Math1⁺cells started to migrate ventrally, they expressed lacZ at the lateral border of the Math1⁺ domain(Fig. 2H). Whereas Math1 is expressed in both proliferating precursors and postmitotic cells (Helms and Johnson,1998), only postmitotic cells expressed lacZ (see Fig. S1 in the supplementary material).

Embryos harboring Tg5 and Tg6, in which the 1.0 kb 5′ fragment of Tg4 was replaced with the β-globin basal promoter, expressed lacZ in the dorsal spinal cord(Fig. 2E), suggesting that the 2.5 kb 3′ fragment is sufficient to drive lacZ expression in an Mbh1-specific manner. The 3′ fragment functioned irrespective of its orientation as well as shorter fragments (see below).

To identify a cis-regulatory element directing Mbh1expression in the dorsal spinal cord, we made a series of deletions of the 2.5 kb 3′ fragment (Fig. 3A). As demonstrated by Tg12, a 517 bp fragment, which was 5.7 kb downstream of the translation start site, was sufficient to drive lacZ expression in the dorsal spinal cord of some transgenic embryos. Further deletions suggested that the 5′ portion of the 517 bp fragment contained a critical site for Mbh1 expression (see also Fig. S2C-F in the supplementary material). It should be noted that β-gal activity became weaker, as the genome sequences were progressively deleted. This finding suggests that several sites of the 3′ fragment are involved in upregulating Mbh1expression.

The 5′ portion of this 517 bp fragment, a 123-base sequence(Fig. 3E), was relatively well conserved among mouse, rat and human (95.6% identity between mouse and rat,and 76.7% identity between mouse and human). Conversely, the sequence encompassing from +4.5 to +11.0 kb downstream of the translation start site showed 79.0% identity between mouse and rat, and 60.0% identity between mouse and human. We searched the 123-base sequence for binding sites of transcription factors using the MatInspector professional(http://www.genomatix.de/index.html). Although there were some potential transcription-factor-binding sites, only one E-box (CAGCTG) was conserved among the three species.

To examine if the E-box was essential for Mbh1 expression, we introduced a mutation into the E-box (ATTCTG) of the 517 bp and 308 bp fragments, thereby constructing Tg16 and Tg17. This mutation is known to disrupt the binding activity of Math1 to DNA(Helms et al., 2000). None of the transgenic embryos harboring these two mutant Tgs expressed lacZin the dorsal spinal cord (Fig. 3A and see Fig. S2I,J in the supplementary material), suggesting that this E-box is required for Mbh1 expression in the dorsal spinal cord.

The involvement of the E-box in Mbh1 expression was further supported by Tg18. Transgenic embryos harboring Tg18, which carried a trimer of the 123 bp fragment, expressed lacZ in the dorsal spinal cord(Fig. 3D and see Fig. S2G,H in the supplementary material). However, lacZ expression from Tg18 was also detected at ectopic sites, the midbrain and somites, in most of theβ-gal⁺ embryos, suggesting that the sequence outside these 123 bp is involved in restricting Mbh1 expression in the dorsal spinal cord.

The E-box necessary for Mbh1 expression completely matched the site to which the Math1 protein could bind efficiently in vitro(Akazawa et al., 1995; Helms et al., 2000). Mbh1 is expressed by a lineage of cells that have expressed Math1(Saba et al., 2003). Moreover, Mbh1 expression started in cells expressing Math1(Fig. 2H). To clarify a genetic relationship between Mbh1 and Math1, Mbh1 expression was examined in Math1 knockout embryos(Fig. 4 and see Figs S3, S4 in the supplementary material). Mbh1 expression was lost in the spinal cord of Math1^–/– embryos, indicating that Math1 was necessary for Mbh1 expression. Expression of Lhx9,which resembled that of Mbh1, was also lost in the Math1^–/– spinal cord(Fig. 4G). By contrast, in the developing dorsal diencephalon, where Math1 was not expressed, Mbh1 expression was not perturbed by the Math1 null mutation(Fig. 4H).

We also performed gain-of-function analysis using mouse in vivo electroporation (Fig. 5). On Math1-transfected sides of the developing spinal cord, Mbh1expression was induced ectopically, indicating that Math1 was sufficient for Mbh1 expression in the spinal cord(Fig. 5B). Math1 also induced ectopic expression of Lhx9 (Fig. 5C). These results indicate that Math1 is both necessary and sufficient for Mbh1 expression in the spinal cord.

To determine whether Math1 bound the E-box in vivo, we performed ChIP experiments. Chromatin was prepared from the spinal cord at E10.5, at which stage endogenous Math1 expression peaked. An anti-Math1 antibody specifically immunoprecipitated Mbh1 enhancer DNA fragments, which contained the E-box (Fig. 6A). Mbh1 genomic fragments that were 5.8 kb far from the E-box were not immunoprecipitated. To examine binding of misexpressed Math1 to the E-box, Math1-HA was transfected to the spinal cord. Math1-HA exhibited the same activity as Math1 in the spinal cord (data not shown). A monoclonal anti-HA-antibody specifically immunoprecipitated the DNA fragments containing the E-box (Fig. 6B). The monoclonal antibody did not co-precipitate the DNA fragments after transfection of Math1, which did not contain the HA tag (data not shown). These findings indicate that Math1-DNA complexes were specifically immunoprecipitated by these antibodies, and that both endogenous and misexpressed Math1 bound the Mbh1 enhancer in the spinal cord.

To examine if Math1 specifically regulated expression of a transgene, we performed transient transfection assays in the developing spinal cord. Tg12 was transfected into the E11.5 mouse spinal cord with an expression vector,pEYFP-Math1, using in vivo electroporation. E10.5 embryos could not survive after electroporation into the spinal cord. EYFP and Math1were co-expressed in the same cells by a double promoter vector, which carried the two genes under two separate promoters on the same plasmid(Saito and Nakatsuji, 2001; Saba et al., 2003). Two days after electroporation, fluorescence of EYFP was detected through the spinal cord (Fig. 7A,E). Strong lacZ expression was detected, when Tg12 was co-transfected with Math1 (Fig. 7F). By contrast, transfection of Tg12 with EYFP alone generated only a fewβ-gal⁺ cells close to the roof plate (arrow in Fig. 7D), reflecting that endogenous Math1 expression is limited to a smaller number of dorsal cells after E11.5. Since most dorsal commissural neurons had already migrated away from the ventricle at E11.5 (Saba et al., 2003), these genes were mostly transfected into cells that had not been fated to commissural neurons. Therefore, transfection of EYFP alone labeled less commissural neurons(Fig. 7C and also see Fig. 8A). By contrast,transfection of Math1 generated more commissural neurons(Fig. 7G and also see Fig. 8C), as dorsal neurons were transfated into commissural neurons by Math1(Saba et al., 2003). Many of the transfated neurons were β-gal⁺(Fig. 7H).

To measure the transcriptional activity of Math1, EYFP⁺ regions of the electroporated spinal cord were dissected out, and cell extracts were analyzed (Fig. 7I). lacZ expression from Tg12 was activated ∼5-fold by Math1 but not by Mash1. This result was consistent with the data of transgenic embryos,where lacZ was not expressed from Tg12 in Mash1⁺ domains,such as the dorsal half of the spinal cord and autonomic nervous system. These results indicate that the expression of the transgene is specifically activated by Math1.

We analyzed the function of Mbh1 using chimeric proteins, which were expected to exert opposite functions. As the C-terminal portion, which included the homeodomain, was well conserved among Bar-class homeodomain proteins, the N-terminal portion of the Mbh1 protein was replaced with a functional domain: the repressor domain of Drosophila Engrailed(Jaynes and O'Farrell, 1991)for En-Mbh1, or the activation domain of herpes simplex virus VP16(Triezenberg et al., 1988) for VP16-Mbh1. Their genes were transfected into the E11.5 spinal cord by in vivo electroporation and co-expressed with EYFP in the same cells using the double promoter vector. More commissural neurons were generated by transfection of Mbh1 as well as Math1(Fig. 8B,C), as described previously (Saba et al.,2003). Similarly, EYFP⁺ commissural neurons were generated by misexpression of En-Mbh1(Fig. 8D) but not VP16-Mbh1 (Fig. 8E),suggesting that Mbh1 functions as a transcriptional repressor.

To clarify the role of Mbh1 in the differentiation of commissural neurons,these chimeric genes were cotransfected with Math1. Whereas co-transfection of En-Mbh1 with either Math1 or Mbh1 generated commissural neurons in a manner similar to the transfection of each gene (data not shown), the generation of commissural neurons by Math1 and Mbh1 was inhibited by VP16-Mbh1 (Fig. 8F;data not shown). These results suggest that VP16-Mbh1 functions as a dominant negative form of Mbh1, and that transcriptional repressor activity of Mbh1 is required for the differentiation of commissural neurons downstream of Math1.

Both gain- and loss-of-function experiments of Math1, taken together with Mbh1 expression in a lineage of cells that have expressed Math1, indicate that Mbh1 is expressed downstream of Math1 in the spinal cord. Moreover, ChIP assays showed in vivo binding of Math1 to the enhancer containing the E-box, which was crucial for Mbh1 expression in the spinal cord. These results indicate that Mbh1 is a direct downstream target of a proneural protein, Math1. Previous analyses indicate that Tag1 (Cntn2 – Mouse Genome Informatics) and Dcc are induced by misexpression of Mbh1 (Saba et al.,2003). As Mbh1 is a potential transcriptional repressor, Mbh1 may de-repress the expression of these two genes, by downregulating expression of a repressor. These findings delineate a cascade of genes in the differentiation of commissural neurons in the spinal cord(Fig. 9).

Many homeobox genes are expressed downstream of proneural bHLH genes. Some homeodomain proteins are directly involved in the specification of neuronal identity, for example, Phox2a regulates the expression of the noradrenaline-synthesizing enzyme, dopamine-β-hydroxylase (reviewed by Goridis and Rohrer, 2002). However, it remained to be determined what genes are direct downstream targets of proneural bHLH proteins. In this study, we demonstrated that Math1 directly activated expression of a homeobox gene, Mbh1, which is necessary and sufficient for the specification of commissural neuron identity. This finding indicates that some aspects of neuronal identity are determined immediately downstream of proneural bHLH proteins.

Enhancer analyses using transgenic mice showed the critical role of the E-box 3′ of the Mbh1 gene for its expression. Math1 is suggested to autoregulate its expression through an E-box 3′ of the Math1 gene (Helms et al.,2000). The nucleotide sequences of the two E-boxes were identical. There is another Bar-class homeobox gene, Mbh2 (Barhl1– MGI), of which expression is lost in Math1^–/– mice(Bermingham et al., 2001). We could not find a sequence similar to the 123 bp fragment containing the E-box 3′ of the Mbh1 gene in the Mbh2-flanking genome sequences. Expression patterns of Mbh1 and Mbh2 are similar but not identical (T.S., T. Hama and R.S., unpublished), as is the case for their Xenopus orthologs, Xbh1 and Xbh2(Patterson et al., 2000). Regulatory mechanisms of Mbh1 and Mbh2 may be different.

The sequence outside the 123 bp was required for efficient and restricted expression of lacZ in transgenic embryos, suggesting that Math1 regulates Mbh1 by collaborating with another factor that may bind a site other than the E-box. Mash1 could not activate lacZ expression from Tg12 in reporter assays using the spinal cord. However, this finding may not simply imply that Mash1 cannot bind the E-box, because misexpression of Mash1 efficiently activated lacZ expression from Tg18, which carried three copies of the E-box, in 10T1/2 cells (data not shown). These findings also suggest that there is a factor that specifically interacts with Math1 to activate gene expression in the spinal cord. Misexpression of Math1 induced ectopic expression of Mbh1 in more ventral regions, suggesting that the factor may not be restricted to the dorsalmost area of the spinal cord, where endogenous Mbh1 expression starts. A bHLH protein, NeuroM, has been shown to interact with LIM-type homeodomain proteins and regulate Hb9 (Lee and Pfaff, 2003). In the dorsal spinal cord, Lhx9 is expressed downstream of Math1 (Figs 4, 5), but misexpression of Lhx9 did not activate either endogenous Mbh1 expression(Saba et al., 2003) or lacZ expression from Tg12 in reporter assays (data not shown).

Consistent with the cascade from Math1 to Mbh1, Mbh1 was expressed in many domains that expressed Math1, including the dorsal spinal cord and hindbrain. But Mbh1 was not expressed in all places that expressed Math1, such as the inner ear. This finding suggests that Math1 is not sufficient and requires an additional factor for Mbh1 expression. β-gal activity in the transgenic mice bearing Tg4 started to be detected at the same time as endogenous Mbh1expression. The β-gal activity faded after E13.5, whereas endogenous Mbh1 expression persisted to at least E18.5. This finding indicates that the 2.5 kb 3′ fragment containing the E-box was sufficient for initiation, but not for maintenance, of Mbh1 expression. In agreement with this, Math1 is abundant at the onset of Mbh1 expression.

Mbh1 was also expressed in domains that did not express Math1, such as the dorsal diencephalon(Saito et al., 1998). lacZ expression was not detected in those areas even in transgenic mice carrying DNA fragments encompassing from –4.5 kb to +11 kb of the Mbh1 genome, suggesting that Mbh1 expression and its maintenance in various domains is controlled by many cis-regulatory elements dispersed throughout the Mbh1 locus.

Analysis using chimeric proteins, En-Mbh1 and VP16-Mbh1, indicated that Mbh1 is a potential transcriptional repressor. Bar-class homeodomain proteins contain a short sequence motif (FxIxxIL), called FIL, in their N-terminal regions (Saito et al., 1998). This motif closely resembles some examples of the eh1 motif(Smith and Jaynes, 1996),which mediates transcriptional repression by interacting with Groucho-family co-repressors. Many of the transcription factors that are expressed in progenitor domains of the ventral spinal cord function as transcriptional repressors (Muhr et al., 2001; Novitch et al., 2001; William et al., 2003). Transcriptional repression may be a general feature in regulating the neuronal fate. Interestingly, a number of genes expressed by postmitotic neurons are silenced by Nrsf/Rest in non-neuronal cells (reviewed by Schoenherr and Anderson,1995).

One of the genes, Scg10 (Stmn2 – Mouse Genome Informatics), which is a pan-neuronal marker, is de-repressed downstream of bHLH proteins. As Mbh1 is expressed by postmitotic neurons in the spinal cord,Mbh1 might also be implicated in pan-neuronal differentiation through the repression of Nrsf (Rest – Mouse Genome Informatics). However, misexpression of Mbh1 in NIH3T3 cells could not activate Scg10 expression (data not shown), suggesting that Mbh1 is involved only in the specification of commissural neuron identity. These findings indicate that Mbh1 functions in a cascade controlling specific differentiation to commissural neurons. In parallel with this cascade, another cascade controlling pan-neuronal differentiation may be activated by Math1.

Quantitative analysis of enhancers and promoters has been mostly performed using cell lines. Cell lines that are suitable for the analysis of a particular gene, however, are not always available. We have established a quick and efficient method to introduce DNA into the developing nervous system using in vivo electroporation, even if the size of DNA is larger than 12 kb(Saito and Nakatsuji, 2001; Saba et al., 2003). Our present work indicates that embryonic tissues can be a good source for transcriptional analysis using in vivo electroporation, because EYFP⁺ portions of the embryonic spinal cord provided enough protein for the analysis. This approach will be very powerful for examination of gene function, where suitable cell lines are not available.

We gratefully acknowledge H. Y. Zoghbi for providing the Math1 mutant strain and N. Nakatsuji for support. We thank M. Tanaka fortechnical assistance, P. Parab for preparation of Math1 knockout embryos and H. Kimura for advice on ChIP assays. We are grateful to T. M. Jessell, R. Kageyama and J. B. Jaynes for plasmids. This work was supported in part by Grants-in-Aids for Scientific Research on Priority Areas-Neural Net Project and Advanced Brain Science Project from Ministry of Education, Culture,Sports, Science and Technology, Japan (to T.S.).