Generation of structurally and functionally distinct factors from the basic helix-loop-helix gene Hes3 by alternative first exons.

The basic helix-loop-helix gene Hes3 is expressed by differentiating and mature Purkinje cells in the cerebellum and by neural precursor cells in the embryonal nervous system. We here found that the transcript of cerebellar Hes3, designated Hes3a, has a distinct 5'-terminal structure from that of embryonal Hes3, designated Hes3b, and that the two types of Hes3 transcripts are generated from different first exons. Hes3a lacks the amino-terminal half of the basic region and thus does not bind to the DNA, whereas Hes3b contains a complete basic region and binds to the N box sequence with a high affinity like Hes1, another member of the Hes family. Both types of Hes3 proteins functionally antagonize the neuronal determination factor Mash1, but only Hes3b represses transcription from the N box-containing promoter like Hes1. Furthermore, misexpression of Hes3b with retrovirus in neural precursor cells inhibits neuronal differentiation like Hes1, whereas Hes3a does not. Thus, alternative promoters and first exons that are differentially utilized during neural development generate structurally and functionally distinct proteins from a single Hes3 gene locus.

Mammalian neural development is controlled positively or negatively by multiple basic helix-loop-helix (bHLH) 1 transcription factors such as Mash1 and Hes1 (1)(2)(3)(4). Mash1 forms a heterodimer with the ubiquitously expressed bHLH factor E47, binds to the E box (CANNTG), and activates neuronalspecific gene expression (5). Forced expression of Mash1 promotes neuronal differentiation, whereas Mash1-null mutation causes defects of differentiation of subsets of neurons (6 -13). In contrast, Hes1, which is specifically expressed by neural precursor cells, not only inhibits the DNA binding activity of the Mash1-E47 heterodimer but also represses Mash1 expression (14,15). Forced expression of Hes1 inhibits neuronal differentiation, whereas in Hes1-deficient mice, Mash1 expression is up-regulated, and neurons differentiate prematurely without sufficient proliferation of neural precursor cells (16 -21). As a result, brain and eye morphogenesis is severely disturbed (17,18). Thus, Hes1 regulates neural development by preventing premature neuronal differentiation or maintaining neural precursor cells, and the balance between the positive and negative bHLH factors is important for the correct timing of neuronal differentiation (1,22). Another bHLH factor, Hes5, is also expressed by neural precursor cells and prevents premature neuronal differentiation (23)(24)(25). In Hes1-Hes5 double-deficient embryos, even more neurons differentiate prematurely than in Hes1-or Hes5-deficient embryos, indicating that Hes1 and Hes5 cooperatively regulate the maintenance of neural precursor cells (25). Nevertheless, still many neural precursor cells are present in the brain of Hes1-Hes5 double-deficient embryos, suggesting that there may be an additional gene that is involved in maintenance of neural precursor cells.
Hes3, another member of the Hes family, is initially identified as a cerebellar Purkinje cell-specific transcription factor (14). Purkinje cells, the only output neurons in the cerebellar cortex, originate from the ventricular zone of the neural tube, migrate radially into the cerebellar cortex, and form a monolayer called the Purkinje cell layer (26). This layer is formed within several days after birth and then Purkinje cells start extension and arborization of their dendrites. Hes3 expression starts in these differentiating Purkinje cells soon after birth and continues until adulthood (14). Thus, it is suggested that Hes3 may be involved in maturation and maintenance of Purkinje cells in contrast to Hes1, which inhibits neuronal differentiation. Furthermore, Hes3 lacks the amino-terminal half of the basic region and is therefore suggested to be unable to bind to the DNA unlike Hes1, which binds to the N box (CACNAG), further supporting the idea that Hes1 and Hes3 have distinct functions. However, recent studies revealed that Hes3 is also expressed by neural precursor cells in the ventricular zone of the developing nervous system like Hes1 (27, 28). Hes3 is initially expressed in the midbrain-hindbrain boundary and rhombomeres 2, 4, 6, and 7 at the early stage of neural development and subsequently in the ventricular zone of the most of the nervous system (27, 28). Thus, from the expression patterns it is speculated that Hes3 could have a similar function to Hes1 in embryos. To characterize the Hes3 functions in neural development, we cloned Hes3 cDNAs from mouse embryos and postnatal cerebellum and compared their structures and activities.
Here, we found that embryonal Hes3, designated Hes3b, is encoded by a different first exon and has a complete basic region unlike cerebellar Hes3, now designated Hes3a. Furthermore, embryonal Hes3 represses transcription by directly binding to the N box and inhibits neuronal differentiation like Hes1. In contrast, cerebellar Hes3 does not have such functions. Therefore, alternative promoters and first exons that are differentially utilized during neural development generate structurally and functionally distinct proteins from a single Hes3 gene locus.

EXPERIMENTAL PROCEDURES
cDNA Library Screening-The cDNA libraries were constructed, as described previously (14). cDNA was synthesized by oligo(dT) priming of poly(A) ϩ RNA of mouse whole embryos of day 9.5 and cerebellum (three weeks old). Double-stranded cDNAs were then constructed, ligated to the EcoRI adapter (Stratagene), and cloned into the EcoRI site of the ZAP II vector (Stratagene). The rat Hes3 cDNA was used as a probe (14). Eight and five positive clones were obtained by screening 8 ϫ 10 5 plaques of cDNA library of the mouse cerebellum and 1 ϫ 10 6 plaques of cDNA library of the mouse embryo, respectively.
DNA Binding Analysis-The proteins were prepared as follows. The cDNA fragments of Hes3a (from the amino acid residue 2 to the carboxyl-terminal end), Hes3b (residue 2 to the end), Hes1 (residue 3 to the end), Mash1 (residue 77 to the end), and E47 (residue 473 to the end) were subcloned into pMNT T7 expression plasmid. JM109(DE3) cells transformed by expression plasmids were grown and treated with 1.6 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 h. The cells were collected and suspended in 0.02 volume of 30 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 20% sucrose. The proteins were purified from the SDSpolyacrylamide gel, incubated in 6 M guanidine-HCl for 20 min, and dialyzed against 0.1 M KCl/HM (20 mM HEPES (pH 7.9), 1 mM MgCl 2 , 2 mM dithiothreitol, and 17% glycerol) at 4°C for 12 h.
The probe DNAs were prepared as follows. For the N box and E box probes, the oligonucleotide (the top strand) containing either a putative N box (5Ј-CAGGCACCACGAGCTGG-3Ј) or a putative E box (5Ј-CAG-GCAGCAGGTGTTGG-3Ј) was labeled at the 5Ј-end and annealed with five molar excess of the bottom strand (N box: 5Ј-CCAGCTCGTGGT-GCCTG-3Ј; E box: 5Ј-CCAACACCTGCTGCCTG-3Ј), respectively. The probe DNAs were purified with MicroSpin G-25 columns (Amersham Pharmacia Biotech). The mutated N box probe (top strand, 5Ј-CAG-GCACCCCTAGCTGG-3Ј; bottom strand, 5Ј-CCAGCTAGGGGTGC-CTG-3Ј) and E box probe (top strand, 5Ј-CAGGCAGTGGGCATTGG-3Ј; bottom strand, CCAATGCCCACTGCCTG-3Ј) were used for DNA competitors. The electrophoresis mobility shift analysis (EMSA) with the N box or the E box probe was carried out as described previously (14).
Luciferase Assay-For the luciferase reporter plasmid, the doublestranded oligonucleotide fragment containing either six repeats of the N boxes (three repeats of CCACGAGCCACAAGG) or seven repeats of the E boxes (AGGCAGGTGGC) was cloned into the firefly luciferase reporter plasmid under the control of the ␤-actin promoter. For eukaryotic expression plasmids, full-length cDNA fragments of Hes1, Hes3a, and Hes3b were subcloned into the eukaryotic expression vector containing the cytomegalovirus promoter, pCI Vector (Promega). Expres- sion plasmids for Mash1 and E47 were described previously (14).
The luciferase reporter (0.1 g) and the eukaryotic expression plasmids (0.3-0.5 g each) were transfected with the FuGENE6 Transfection Reagent (Roche Molecular Biochemicals) into C3H10T1/2 cells, which were plated in 6-multiwell plates at the density of 1 ϫ 10 5 cells/well. The plasmid containing Renilla luciferase gene under the control of the SV40 promoter (2 ng) was also transfected as an internal standard to normalize the transfection efficiency. After 48 h, the cells were harvested, and luciferase activity was measured.
Generation of Recombinant Retrovirus-The CLIG retrovirus vector contains the cytomegalovirus enhancer in the 5Ј-long terminal repeat, internal ribosomal entry site of the encephalomyocarditis, enhanced GFP gene (CLONTECH), and 3Ј-long terminal repeat (31). For retrovirus expression plasmids, full-coding cDNA fragments of Hes1, Hes3a, and Hes3b were subcloned into the upstream of the internal ribosomal entry site of CLIG. Retrovirus was produced by transfecting the retroviral DNA constructs into the packaging cell line 2mp34 (30). Retrovirus solution was passed through a 0.45-m filter and concentrated, as described previously (30).
Neural Precursor Cell Culture-The primary culture of neural precursor cells was performed as described previously (16). The uterus containing mouse whole embryos of day 11.5 was excised and incubated in minimal essential medium-␣ for 45 min. Then the embryos were isolated, and their heads were excised. With fine forceps, the epidermis, mesenchyme, and meninges were removed. The neuroepithelium of the forebrain was transferred into fresh phosphate-buffered saline and partially dissociated by pipetting. The cells were resuspended in the Dulbecco's modified Eagle's medium/Ham's F-12 medium (Life Technologies, Inc.) containing 2% fetal bovine serum, 25 g/ml of transferrin, 15 g/ml of insulin, 20 nM progesterone, 30 nM sodium selenite, 1.7 ng/ml basic fibroblast growth factor, 3.3 ng/ml of epidermal growth factor, 10 ng/ml nerve growth factor, and 10 ng/ml of cholera toxin. This cell suspension was plated onto a tissue culture dish, and on the next day nonadhesive cells were collected and replated onto polyethylenimine and fibronectin-coated chamber slides (Lab-Tek) at the density of 3 ϫ 10 4 . After the cells were attached to the slides, virus solution with 8 g/ml of polybrene was added to the culture. 3 h later, the solution was aspirated, and fresh medium was added. After 2 weeks of culture, the cells were fixed and examined by immunohistochemistry as follows.

Structural Analysis of Embryonal and Cerebellar
Hes3-To determine the structure of embryonal and postnatal cerebellar Hes3, we screened mouse cDNA libraries made from mouse whole embryos of day 9.5 and postnatal cerebellum. This screening yielded five embryonal (out of 1 ϫ 10 6 independent plaques) and eight cerebellar cDNA clones (out of 8 ϫ 10 5 ). Sequencing analysis revealed that all the cDNA clones isolated from the cerebellar cDNA library (Hes3a) had essentially the same structure as the previously published rat Hes3 (14). Hes3a had the first methionine codon in the middle of the basic region, thus lacking its amino-terminal half (Fig. 1C). This methionine codon was located in exon 2, and therefore the first exon only had a noncoding sequence (Fig. 1B) (32). In contrast, the embryonal type cDNA clones (Hes3b) had a different structure in the 5Ј region, suggesting that the 5Ј region is derived from a different exon. The other regions corresponding to the second, third, and fourth (last) exons were identical between the two types of Hes3 cDNAs. Sequence comparison between the 5Ј region of Hes3b and the Hes3 gene (32) demonstrated that a new first exon, designated exon 1b, was located in the first intron (Fig. 1, A and B). The previously identified first exon of Hes3a is now designated exon 1a. Interestingly, Hes3b transcript had the first methionine codon at a further upstream position in exon 1b and encoded a complete basic region (Fig. 1,  B and C). Hes3a and Hes3b consisted of 175 and 200 amino acid residues, respectively. In the basic region of Hes3b, there was a proline residue that is conserved by all Hes factors except for Hes3a (Fig. 1C), and the bHLH domain of Hes3b had 60% identity to that of mouse Hes1. Both Hes3a and Hes3b contained the carboxyl-terminal Trp-Arg-Pro-Trp sequence (the WRPW domain), which recruits the corepressor Groucho (33)(34)(35).
Determination of Transcriptional Initiation Sites-To determine the transcription initiation sites of Hes3 gene, primer extension analysis was performed. The labeled antisense primer corresponding to the region from nucleotide residues 211 to 240 in exon 1a (primer 5, Fig. 2A) was hybridized to embryonal and postnatal cerebellar RNA and subjected to reverse transcription. This experiment generated one specific band with the size of 240 nucleotides when the primer was hybridized to cerebellar RNA (Fig. 2B, lane 3) but not when the same primer was hybridized to embryonal RNA (lane 4). This confirmed that exon 1a is transcribed only in postnatal cerebellum and that transcription initiates at the nucleotide position 1 (Fig. 1B). When the antisense primer corresponding to the region from 831 to 855 in the second exon (primer 6, Fig.  2A) was hybridized to postnatal cerebellar RNA, a 418-nucleotide band was detected (Fig. 2B, lane 1). This product was likely to be derived from the transcript of exon 1a. In contrast, when the same primer (primer 6) was hybridized to embryonal RNA, a band with the size of 115 nucleotides was detected (lane 2). This product was likely to be derived from the transcript of exon 1b, and the size of the band suggested that the transcription initiation site of exon 1b was located at 621 (Fig. 1B).
To confirm the positions of the two transcriptional initiation sites, we next carried out RT-PCR experiments with different sets of primers (the positions of the primers are shown in Fig.  2A). Whereas primers 1 (corresponding to the region from Ϫ25 to Ϫ1) and 6 yielded no band (Fig. 2C, lanes 1 and 5), primers 2 (1-25) and 6 yielded a 418-bp band only when postnatal cerebellar RNA was used as a template (lane 2), agreeing with the above result that transcription starts at 1 in postnatal cerebellum. In contrast, whereas primers 3 (596 -620) and 6 yielded no band (lanes 3 and 7), primers 4 (621-650) and 6 yielded a 115-bp band only when embryonal RNA was used as a template (lane 8), agreeing with the result that transcription starts at 621 in embryos. These data indicated that exon 1a is transcribed only in postnatal cerebellum, whereas exon 1b is only in embryos. In the upstream region of each transcription initiation site, there is a TATA-like sequence, TATATA (35 nucleotides upstream of exon 1a) and TATAAA (27 nucleotides upstream of exon 1b) (Fig. 1B, underlined).
DNA Binding Analysis-Hes3a has a helix-loop-helix domain but lacks the amino-terminal half of the basic region, whereas Hes3b has a complete bHLH domain, raising the possibility that these two proteins have distinct DNA binding properties. To determine the DNA binding activity, each pro-tein was expressed in Escherichia coli and applied to the EMSA. Whereas Hes3a did not bind to the N box (Fig. 3A, lane  2), Hes3b bound to the N box very efficiently (lane 5). These results suggested that the complete basic region is essential for DNA binding. The binding intensity of Hes3b was comparable with that of Hes1 (lane 8), which binds to the N box with a high affinity (14). The binding was abrogated when the excess amount of the N box-specific competitor was mixed (lanes 6 and 9) but not when a mutated competitor was mixed (lanes 7 and 10), indicating that the N box binding was specific. These results demonstrated that Hes3b is functionally different from Hes3a but more similar to Hes1.
Because Hes1 is known to inhibit the Mash1-E47 heterodimer from binding to the E box (14), we next examined whether Hes3a and Hes3b have the similar activity. Mash1 and E47 bound to the E box with a high affinity (Fig. 3B, lane  3). This binding was abrogated by the E box-specific (lane 4) but not by a mutated competitor (lane 5). Coexpression of Hes3a or Hes3b with the Mash1-E47 complex completely inhibited the E box binding, like Hes1 (lanes 6 -8), indicating that both Hes3a and Hes3b had a similar inhibitory activity on Mash1 and E47. Thus, the basic region is not required for the inhibition of the DNA binding of Mash1 and E47.
Transcriptional Analysis-To determine the transcriptional activity, we next performed a transient transfection assay with the luciferase reporter genes. The reporter gene under the control of the N box-containing promoter was coexpressed with Hes3a or Hes3b. We previously showed that Hes1 represses transcription by binding to the N box (14) (Fig. 4A, row 4). Similarly, Hes3b efficiently repressed transcription (row 3), whereas Hes3a did not (row 2), agreeing well with the EMSA data that Hes3b and Hes1 bound to the N box with a high affinity, whereas Hes3a did not. These results again indicated that Hes3b is functionally different from Hes3a but more similar to Hes1.
When the reporter gene under the control of the E boxcontaining promoter was used, coexpression of Mash1 and E47 significantly induced the expression (Fig. 4B, row 4), as described previously (5). Both Hes3a and Hes3b efficiently inhibited Mash1 and E47-dependent transcription like Hes1 (rows 5-7). These data agreed with the above EMSA results, which showed that Hes3a and Hes3b inhibited the DNA binding of Mash1 and E47.
Functional Analysis in Neuronal Differentiation-The DNA binding and transcriptional analyses indicated that Hes3a and Hes3b have different activities. To define their functions in neural development, each factor was misexpressed with retrovirus in neural precursor cells. For misexpression, we used the replication-incompetent retrovirus CLIG (31), which directs expression of a test gene and GFP as a marker (Fig. 5A). Neural precursor cells were prepared from mouse embryos at day 11.5, and virus was applied on the next day. After 2 weeks, by which time many cells differentiated into neurons, the fate of the virus-infected cells was determined by monitoring GFP ϩ cells. When the control CLIG virus was infected, approximately 40% of the virus-infected cells differentiated into neurons, which extended multiple neurites and expressed the neuron-specific microtubule associated protein MAP2 (Figs. 5, B-D, and 6A). Approximately 20% of the CLIG-infected cells remained as neural precursor cells, which expressed the progenitor-specific intermediate filament nestin (Figs. 5, E-G, and 6B). When CLIG-Hes1 was infected, about 90% of the virus-infected cells remained as nestin-positive precursor cells (Figs. 5, W-Y, and  6B). These virus-infected cells were clearly demarcated from the surrounding MAP2-positive neurons (Fig. 5, T-V), and only 4% of them differentiated as neurons (Fig. 6A), agreeing with the previous results that Hes1 inhibits neuronal differentiation (16,17). Similarly, when CLIG-Hes3b was infected, about 85% of the virus-infected cells remained as nestin-positive precursor cells (Figs. 5, Q-S, and 6B), and they were segregated from MAP2-positive neurons (Fig. 5, N-P). Only 5% of the virusinfected cells differentiated into MAP2-positive neurons (Fig.  6A). These results indicated that Hes3b inhibited neuronal differentiation and maintained neural precursor cells like Hes1. In contrast, when CLIG-Hes3a was infected, about 35% of the virus-infected cells differentiated into MAP2-positive neurons (Figs. 5, H-J, and 6A), and only 30% of the cells remained as nestin-positive neural precursor cells (Figs. 5, K-M, and 6B). These ratios were very similar to those of the control CLIG virus infection, suggesting that Hes3a has no inhibitory or promoting activity to neuronal differentiation. These results demonstrated that Hes3a and Hes3b have different functions in neural development.

Two Types of Hes3
Proteins-Hes3 is expressed by neural precursor cells in the embryonal nervous system and by differentiating and mature Purkinje cells in the postnatal cerebellum. From these expression patterns, it has been suggested that Hes3 has different functions in embryos and postnatal cerebellum, but the molecular nature for the possible functional difference was not known. We here showed that embryonal and cerebellar Hes3 transcripts are structurally different and generated from alternative first exons. The cerebellar type Hes3a transcript has the first methionine codon in the middle of the basic region, and therefore Hes3a lacks the amino-terminal half of the basic region. In contrast, the embryonal type Hes3b transcript has the first methionine codon at the further upstream position, and thus Hes3b has a complete basic region. Interestingly, these two types of Hes3 proteins are different in function as well.
Hes3a and Hes3b Have Different Functions in Neuronal Differentiation-Hes3a does not bind to the DNA, whereas Hes3b binds to the N box and represses transcription from the N box-containing promoter like Hes1 and Hes5. Furthermore, Hes3b inhibits neuronal differentiation and maintains neural precursor cells, whereas Hes3a does not. Thus, Hes3b is functionally different from Hes3a but very similar to Hes1 and Hes5, which are expressed by neural precursor cells and play an important role in maintenance of neural precursor cells. In Hes1-Hes5 mutant brains, neural precursor cells are not prop-erly maintained and differentiate prematurely (17,18,25). As a result, these embryos exhibit severe defects of brain and eye morphogenesis (17,18). Nevertheless, these mutant embryos still have many neural precursor cells, suggesting that an additional factor may be involved in maintenance of neural precursor cells. Our present study suggests that Hes3b may be responsible for this activity.
The functional difference in regulation of neuronal differentiation between Hes3a and Hes3b is likely to reside in the basic region. Whereas Hes3b represses transcription by directly binding to the N box like Hes1 and Hes5, Hes3a cannot because of its incomplete basic region. It seems that transcriptional repression by directly binding to the promoter, which is exhibited by Hes1, Hes5, and Hes3b but not by Hes3a, is important to block neuronal differentiation. For this transcriptional repression, the carboxyl-terminal WRPW domain is essential (33)(34)(35). This domain interacts with the corepressor Groucho, which recruits a histone deacetylase protein and actively represses transcription (36). The target genes of Hes3b are not known, but Hes1 is known to repress Mash1 expression by directly binding to its promoter (15). In the absence of Hes1, Mash1 expression is up-regulated, and neurons differentiate prematurely (17,18), suggesting that transcriptional repression of neuronal genes such as Mash1 may be important to block neuronal differentiation. Because Hes3b and Hes1 have very similar functions and expression patterns, it is possible that Hes3b may also repress Mash1 expression by directly binding to its promoter. Another possible target gene is Krox20, a zinc finger gene expressed in rhombomeres 3 and 5 (37). Because Hes3b is expressed in rhombomeres 2, 4, 6, and 7, where Krox20 is not expressed, it is possible that Hes3b may be involved in repression of Krox20 expression in such rhombomeres.
Although Hes3a can inhibit Mash1-dependent transcription in a dominant-negative manner, it does not inhibit neuronal differentiation. The apparent requirement for the DNA binding activity to inhibit neuronal differentiation is intriguing because the Drosophila counterparts, E(spl) proteins, can suppress neuronal differentiation even if the bHLH domain is deleted (38), suggesting that the DNA binding activity is not essential. In this case, the carboxyl-terminal region including the WRPW domain is essential (38), suggesting that E(spl) proteins may inhibit differentiation by direct interaction with other proteins but not by DNA binding. Thus, although we do not have any evidence, it is still possible that Hes3a might repress differentiation of some types of neurons, particularly ones that depend on Mash1 for differentiation. Nevertheless, because Hes3a is expressed only by differentiating and mature Purkinje cells, it is more likely that Hes3a may be involved in their maturation, although the precise mechanism remains to be determined. One possible function of Hes3a is sequestration of Groucho because Hes3a has a WRPW domain and therefore interacts with Groucho but cannot bind to the DNA. Thus, Hes3a may inhibit the activity of negative bHLH factors such as Hes1 by sequestration of the corepressor and thereby promote maturation of neurons. However, misexpression of Hes3a in neural precursor cells did not promote neuronal differentiation (Figs. 5 and 6), and therefore the exact functions of Hes3a in Purkinje cells remain to be determined.
Differential Expression of Hes3a and Hes3b during Neural Development-Exon 1a is transcribed only by postnatal Purkinje cells, whereas exon 1b is transcribed only in embryos. The mechanism of how these two promoters and exons are differentially selected remains to be analyzed. There are only 248 nucleotide residues between the 3Ј-end of exon 1a and the 5Ј-end of exon 1b, and it is possible that embryo-specific enhancer sequences are present within this region. Sequence examination of this region indicated a possible binding site of RBP-J. RBP-J is a DNA-binding protein and mediates the activity of the transmembrane protein Notch, which plays an important role in maintenance of precursor cells and generation of cell type diversity (39 -45). Upon activation by its ligands, Notch is cleaved and its intracellular domain (ICD) is released from the membrane region. Because ICD has a nuclear localization signal, it moves into the nucleus, where it forms a complex with the DNA-binding protein RBP-J. The ICD and RBP-J complex binds to the promoter and induces gene expression. The consensus sequence of the RBP-J-binding sites is (C/T)GTG(G/A)GAA(A/C), and it has been shown that Hes1 and Hes5 are the target genes of ICD-RBP-J complex (39,40,45). Thus, Notch regulates cell differentiation by inducing Hes1 and Hes5 expression (25). In Hes3 gene, a sequence similar to the RBP-J-binding site, TGTGTGAAC (614 -622), is present upstream of exon 1b. In addition, there is another possible RBP-J-binding site, AGTGGGAAC (731-739) downstream of exon 1b. They matched eight and seven nucleotides, respectively, out of nine (underlined). However, the active form of Notch failed to up-regulate the Hes3b promoter activity in transient transfection assay. 2 Similarly, Hes3a promoter activity is not induced either by the Notch pathway (40), suggesting that Hes3 expression is not regulated by Notch signaling. It is known that basic fibroblast growth factor and epidermal growth factor play an important role in the maintenance of neural precursor cells (46), and thus it is possible that expression of Hes3b as well as of Hes1 and Hes5 could be regulated by these growth factors. However, comparison of the upstream region of exon 1b with Hes1 and Hes5 promoters failed to reveal specific elements conserved among the three genes. Thus, it remains to be determined whether or not the three genes are controlled by common regulatory factors in neural precursor cells.