Functional Properties of the Neuronal Nicotinic Acetylcholine Receptor β3 Promoter in the Developing Central Nervous System*

Within the chick central nervous system, expression of the β3 nicotinic acetylcholine receptor gene is restricted to a subset of retinal neurons, the majority of which are ganglion cells. Transient transfection in retinal neurons and in neural and non-neural cells from other regions of the chick embryo allowed the identification of the cis-regulatory domain of the β3 gene. Within this domain, a 75-base pair fragment located immediately upstream of the transcription start site suffices to reproduce the neuron-specific expression pattern of β3. This fragment encompasses an E-box and a CAAT box, both of which are shown to be key positive regulatory elements of the β3 promoter. Co-transfection experiments into retinal, telencephalic, and tectal neurons with plasmid reporters of β3 promoter activity and a number of vectors expressing different neuronal (ASH-1, NeuroM, NeuroD, CTF-4) and non-neuronal (MyoD) basic helix-loop-helix transcription factors indicate that the cis-regulatory domain of β3 has the remarkable property of discriminating accurately between related members of the basic helix-loop-helix protein family. The sequence located immediately 3′ of the E-box participates in this selection, and the E-box acts in concert with the nearby CAAT box.

In vertebrates, both negative and positive regulations play an important role in determining neuronal gene expression. Negative regulation (reviewed in Refs. 1 and 2) is best exemplified by REST/NRSF, a factor that represses transcription of the SCG-10 and type II Na ϩ channel genes in non-neuronal cell-types, whereas most neurons lack REST/NRSF and thus express these two genes (3,4). Conversely, the Olf-1 transcription factor positively regulates several genes (e.g. the olfactory neuron-specific G protein, the type III adenylyl cyclase) specifically expressed in olfactory neurons (5). Vertebrate homologs of the basic helix-loop-helix (bHLH) 1 factors involved in Drosophila neurogenesis (6) act as positive or negative regulators in the acquisition of neuronal identity. For instance, the atonaland achaete-scute-related activators are transiently expressed in parts of the central and peripheral nervous system during early development, and their null mutation or ectopic expres-sion profoundly influences neurogenesis (reviewed by Lee et al. (7)). However, the direct regulation by bHLH proteins of genes that define neuronal identity has never been documented.
Several neuronal nicotinic acetylcholine receptor (nAChR) genes are expressed early in neural development (8 -12), and, since they encode transmembrane sensors capable of fluxing Ca 2ϩ and other cations upon stimulation (reviewed in Ref. 13), an understanding of their regulation should help explain how the genetic program puts together the mechanisms needed for epigenetic environmental cues to participate in development. This is illustrated in a recent report by Shatz and associates (14) showing that, in the developing retina, cholinergic synaptic transmission between newly generated amacrine and ganglion cells is responsible for the propagation of spontaneous waves of action potentials that may be critical for the establishment of visual system circuitry.
Several nAChR subunit genes, including ␣4, ␤2, ␤3, and ␤4, are expressed in the chick retina (10), 2 and expression of ␤3 is confined to ganglion cells and amacrine neurons (15). Forsayeth and Kobrin (16) have shown that the ␤3 subunit co-assembles in vivo with the ␣4, ␤2, and ␤4 subunits to form a functional nicotinic receptor endowed with distinctive properties. We have isolated a short 5Ј-sequence of the ␤3 gene containing promoter elements that are sufficient to target reporter gene expression to those retinal neurons that normally express ␤3 in vivo (11,15). The stringent neuronal specificity of the ␤3 promoter and its activation during the period of neuronal fate determination make it an attractive system in which to study the functional interactions between transcription factors and cis-acting regulatory elements that help establish the diverse neuronal phenotypes.
In this report, we carry out a functional analysis of the cis-regulatory domain, establishing that transcription of the ␤3 gene is under the direct control of bHLH factors. Moreover, we show that the ␤3 promoter is able to discriminate accurately between related members of the bHLH family, thereby effecting the stringent neuron-specific regulation of the gene.

EXPERIMENTAL PROCEDURES
DNA Constructions-Standard molecular biology techniques were used (17) unless otherwise stated. Construction of the reporter plasmids ␤3RS-CAT and ␤3RS-lacZ and of the reference plasmids SV-CAT and SV-lacZ was described by Hernandez et al. (15). Point mutations in the ␤3RS sequence (Fig. 1A) were introduced by polymerase chain reaction and checked by sequencing. The mutant DNA fragments were cloned at the SmaI site of the pCAT00 plasmid (18). SF-E and SF-3E were obtained by ligation of PvuII linkers (CCAGCTGG; New England Biolabs) to the 5Ј-end of fragment SF ( Fig. 2A). ␣1KK encompasses nucleotides 151-334 of the chick ␣1 nAChR promoter (19,20) (GenBank TM accession number M15307), flanked by KpnI restriction sites that were used for subcloning into pCAT00. The ␣1/␤3 hybrid promoter was obtained by ligation of the 5Ј-end of ␣1KK with the 3Ј-end of ␤3RS at their shared PvuII restriction site (Fig. 7A) Roztocil et al. (21).
Expression and Purification of the CTF-4 Protein-A HindIII cDNA fragment encoding the bulk of the CTF-4 protein (22) was cloned in phase in the appropriate pDS bacterial expression vector (23). High level accumulation of His-tagged fusion protein was achieved by incubating Escherichia coli transformants at 37°C for 2 h with 2 mM isopropylthiogalactoside in rich broth. Purification of the fusion protein was performed on Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) in denaturing conditions. The eluted protein (48.5 kDa) was renatured from 8 M urea by serial dialysis and stored at Ϫ20°C in 10 mM Tris-Cl, pH 7.6, 1 mM EDTA, 1 mM ␤-mercaptoethanol.
Isolation of Nuclear Proteins and Gel Mobility Shift Analysis-Retinae and optic tecta were dissected in cold phosphate-buffered saline, and nuclei were isolated as described by Matter-Sadzinski et al. (24). Nuclear proteins were obtained by adding NaCl to the suspension of nuclei to a concentration of 1 M, in the presence of the protease inhibitors leupeptin, phenylmethylsulfonyl fluoride (both at 0.7 g/ml), and Trasylol (1%), incubating for 15 min on ice, and centrifuging for 45 min at 50,000 rpm (Beckman TLA 100). The proteins in the supernatant were quantified (Bio-Rad protein assay) and stored at Ϫ70°C. The probes were 35-bp double-stranded fragments (underlined in Fig. 1A) end-labeled by fill-in of 5Ј-overhangs with Klenow enzyme in presence of [␣-32 P]dATP. For band shifts, bovine serum albumin was added to nuclear proteins to keep the total protein load at 6 g/lane. Poly(dI-dC) (2.5 g), used as nonspecific competitor DNA, was mixed with the probe (20,000 cpm) in FT buffer (25 mM HEPES, pH 7.6, 5 mM MgCl 2 , 40 mM KCl), proteins were added, and the mixture (20 l) was incubated for 15 min on ice. The samples were then loaded on a 7.5% polyacrylamide gel containing 2.5% glycerol. After a 2.5-h run at 200 V, the gel was dried and exposed overnight at Ϫ70°C with an intensifying screen.
Cell Cultures, Transfection, and CAT and ␤-Galactosidase Assays-All tissue culture reagents except as indicated were purchased from Life Technologies, Inc. Plasticware was from Nunc. Chick embryos were staged according to Hamburger and Hamilton (25). Cells from different regions of the CNS were prepared as described previously (11,24). Neuroretina, pigment epithelium, optic tectum, telencephalon, and cerebellum were dissected from E4 -E13 embryos, collected in Ca 2ϩ -and Mg 2ϩ -free Hanks' balanced salt solution and incubated with 0.05% trypsin for 10 min (E4 -E8) or with 0.1% trypsin for 20 min (E9 -E13). Desoxyribonuclease I (Boehringer Mannheim) was added to 30 g/ml for 5 min, and then trypsin was inactivated by adding fetal calf serum to 5%. Cells were pelleted, rinsed in Opti-MEM medium, resuspended in Opti-MEM at densities as indicated below, and subjected to the transfection procedure. Plasmid DNA (5 g in 100 l of Opti-MEM) was mixed with Lipofectin reagent (20 g diluted in 200 l of Opti-MEM), and the transfection solution was added to 200 l of cell suspension containing 4 -6 ϫ 10 6 cells. In co-transfection experiments with two constructs, 5 g of reporter plasmid was mixed with 3.5 g of expression vector. In all instances, the ratio of DNA to Lipofectin was 4/1. Glial cells and primary or secondary cultures of CEFs were prepared and transfected as described by Matter-Sadzinski et al. (24). 24 or 48 h after transfection, cells were collected and processed for CAT assay. The amount of acetylated 14 C-labeled chloramphenicol was determined by scintillation counting and/or by scanning the chromatogram with a Linear Analyzer LB 284/285 (Berthold). In each experiment, an aliquot of cells was transfected with pSV-CAT, and the resulting CAT activity was arbitrarily set at 100. The activities obtained in parallel with other constructs were calculated relative to this value. 100 -150 g of cytosolic proteins were used in CAT assays such that the proportion of acetylated [ 14 C]chloramphenicol in cells transfected with pSV-CAT did not exceed 70%. The activity of the SV40 promoter was similar in different neuronal and non-neuronal cell-types at different embryonic stages (24). The means and S.D. values were calculated with data obtained in at least five independent experiments.
Cells transfected with ␤-galactosidase reporter plasmids were plated into the chambers of a poly-DL-ornithine-coated plastic chamber slide (Lab-Tek). 24 or 48 h after transfection, X-gal staining was performed as described by Hernandez et al. (15). Blue cells were counted in 20 -30 grid areas that each contained about 10 3 positive cells upon transfection with pSV-lacZ.

Identification of Regulatory Elements in the ␤3
Promoter-In the chick CNS, the ␤3 nAChR gene is selectively expressed in the neuroretina. We have previously shown that essential neuron-specific promoter elements are located in a short EcoRI-SphI DNA fragment, 143 bp in length and located just upstream of the transcription start site (11,15). This ␤3RS fragment contains several putative binding sites for transcription factors: CACCC and CAGCTG (E-box) motifs, a CAAT box, and two TATA-like motifs located at Ϫ56 bp and Ϫ30 bp relative to the transcription initiation site (Fig. 1A). To investigate how these different sites contribute to promoter activity, point mutations were introduced by polymerase chain reaction into each site, and the mutated fragments were fused to the CAT reporter gene. The constructs were transfected into neurons The ␤3RS promoter (143 bp) extends between the indicated EcoRI and SphI sites. B, the wildtype ␤3RS and mutant sequences (*) were linked to the chloramphenicol acetyltransferase (CAT) gene. Cells isolated from E5 neuroretina were transfected with the constructs (5 g) and assayed for CAT activity 48 h after transfection. The CAT activity obtained with the wild-type ␤3RS fragment is arbitrarily set at 100, and activities of the mutated promoters are given relative to this value. C, nuclear protein extracts prepared from E5 neuroretina and E9 optic tectum were used for gel mobility shift assays. The DNA binding affinity of these extracts (1, 3, or 5 g of protein/assay) was tested on the wild-type (WT) or E-box mutant (E-box*) double-stranded 35-mer underlined in A. F, free probe; B, bound probe. freshly dissociated from E5 chick neuroretina (24), and 48 h later the transfected cells were processed for CAT activity. Mutations in the E-box and in the CAAT box produced spectacular effects, completely abolishing promoter activity in retinal cells. In contrast, mutations in the other motifs had no significant impact on promoter activity (Fig. 1B).
The binding of nuclear proteins to a very short fragment (35 bp, underlined in Fig. 1A) encompassing both the E-box and the CAAT box was tested in gel mobility shift experiments. Extracts from E5 retina bound strongly to this region, whereas binding was much weaker with extracts prepared from cells that do not express ␤3, such as tectal neurons (Fig. 1C) or glial cells (data not shown). Consistent with the transfection experiments, a DNA fragment bearing a mutant E-box (TAGCTA) did not bind nuclear proteins.
The E-box Is a Key Element of the ␤3 Promoter-Deleting the 5Ј-end of ␤3RS, which contains the CACCC and E-box motifs (SF; Fig. 2A), resulted in the complete inactivation of the promoter in retinal cells (Fig. 2B). Since mutating the E-box in the full-length ␤3RS fragment (E-box*, Fig. 2A) was sufficient to obtain a similar effect, we tested whether it is possible to reactivate the truncated SF fragment by adding one or several E-boxes at its 5Ј-end (constructs SF-E and SF-3E; Fig. 2A). The addition of one E-box restored promoter activity in retinal cells to a level comparable with that of the complete ␤3RS fragment, and three E-boxes allowed activity levels consistently higher than ␤3RS (Fig. 2, B and C). The tissue specificity of the reactivated promoters was examined by transfection into cells in which the ␤3 promoter is normally silent, namely cells from the optic tectum and telencephalon, CEFs or glial cells (Fig.  2C). No promoter activity was detected in any of them, indicating that SF-E and SF-3E are regulated as specifically as the ␤3RS fragment. Thus, the addition of one E-box is sufficient to reconstitute a promoter with the same specificity and activity as the wild-type fragment, demonstrating that the E-box is a key regulator of the ␤3 nAChR gene.
Precise Positioning of the E-box and CAAT Box Is Required for Promoter Activity-Mutation of the CAAT box, which is located 9 bp downstream of the E-box, abolishes promoter activity in retinal cells (Fig. 1B), suggesting that direct interactions between proteins bound to these two elements may take place. Due to the helicity of DNA, the addition or deletion of base pairs in the intervening sequence should disrupt the alignment of bound factors, thereby decreasing promoter activity. Several mutants were constructed by the addition or deletion of nucleotides between the E-box and the CAAT box, and they were tested for promoter activity in E5 retinal cells (Fig.  3). Reducing the distance by 1, 2, and 3 bp was sufficient to decrease promoter activity 5, 8, and 10-fold, respectively. In contrast, the addition of 1-4 bp had either a modest effect (Fig.  3) or no effect at all (SF-E, Fig. 2), depending on the particular additional base pairs. This suggests that steric hindrance between bound factors contributes to the decrease in promoter activity when the distance between the two motifs is reduced.
Functional Analyses of Neuronal bHLH Proteins-Since the E-box is a binding site for transcription factors of the bHLH family, we postulated that expression of the ␤3 gene is under the control of such a factor in the neuroretina. Several bHLH genes are sequentially expressed in the developing chick CNS. While CASH-1 is expressed in proliferating cells (26), NeuroM is transiently expressed in cells that have withdrawn from the FIG. 2. Analysis of the E-box in the ␤3 promoter. A, ␤3RS is the wild-type promoter. In the mutated E-box (E-box*), the wild-type sequence was mutated to TAGCTA. In SF, the E-box was truncated, and the 5Ј-flanking sequence was deleted. In SF-E and SF-3E, oligonucleotide linkers containing, respectively, one (CCAGCTGG) and three Eboxes (CCAGCTGG) 3 were added at the 5Ј-end of SF. B, the DNA fragments described in A were fused to the CAT gene and transfected into cells isolated from E5 neuroretina. Cells were assayed for CAT activity 48 h after transfection. C, the constructs were also transfected into neurons from the optic tectum and telencephalon, in glial cells, and in CEFs. The CAT activity obtained upon transfection of each cell type by SV-CAT is arbitrarily set at 100, and the promoter activities of the different DNA fragments are given relative to this value.

FIG. 3. Analysis of E-box/CAAT box spacing mutants. A, brackets
indicate positions of the 1-, 2-, or 3-bp deletions in the intervening sequence between the E-box and the CAAT box. The arrow marks the position where 1 bp was added. B, ␤3RS sequences bearing the different mutations were linked to the CAT gene, and the constructs were transfected in cells isolated from E5 neuroretina. Cells were assayed for CAT activity 48 h after transfection. The CAT activity obtained upon transfection with the wild-type ␤3RS fragment is arbitrarily set at 100, and activities of the mutant promoters are given relative to this value. mitotic cycle but have not yet migrated in the outer layers, whereas NeuroD labels neurons that are migrating and differentiating (21). CTF-4 (22) is widely distributed in the nervous system, but its expression is significantly enhanced in the retina. 3 Since ␤3 expression at early stages of retina development coincides with expression of these different bHLH proteins, we tested if any of them was able to trans-activate the ␤3RS promoter. Co-transfections with expression vectors encoding CASH-1, NeuroM, NeuroD, or CTF-4 into E8 tectal and E9 telencephalic neurons revealed that none of these proteins was able to trans-activate the ␤3RS promoter ectopically, although all four were functional transcription factors capable of strongly enhancing activity of the E-box-driven promoter of the muscle nAChR ␣1 subunit (␣1KK fragment, Fig. 4A, Table I).
Next, we tested whether overexpression of these neuronal transcription factors in retinal cells increased the activity of the ␤3 promoter. Cells isolated from E5 retinae were co-transfected with the ␤3RS-lacZ reporter and each of the CASH-1, NeuroM, NeuroD, or CTF-4 expression vectors. On its own, the ␤3RS promoter drives the expression of ␤-galactosidase in 15 Ϯ 2% of E5 retinal cells, and this proportion remained unchanged in cells that overexpressed CASH-1, NeuroM, NeuroD, or CTF-4 (Table I). Moreover, no influence of these factors on ␤3 promoter activity has been detected at later stages of retina development (data not shown). The ␣1 nAChR gene is not expressed in retina, but the transfected ␣1KK fragment displayed a significant promoter activity in retinal cells (most likely due to transactivation by endogenous bHLH factors), and co-transfection of the ␣1KK-lacZ reporter with the different neuronal bHLH proteins strongly increased the proportion of X-gal-positive retinal cells (Table I). Thus, although CASH-1, NeuroM, NeuroD and CTF-4 are functional transcription factors and instances of overlapping expression between these bHLH genes and ␤3 in the retina have been detected, 4 none of these factors appears to be directly involved in the regulation of ␤3. Gel mobility shift analyses reveal that CTF-4 binds to the ␤3 E-box in vitro (Fig. 4B) and that NeuroD binds to the CAGCTG motif (27). If such interactions take place in vivo, they are probably not sufficient for promoter activation. Weintraub et al. (28) have shown that the base pairs flanking one of the IgH E-boxes constitute part of a negative cis-acting element enabling the IgH gene to discriminate between bHLH proteins. In the case of ␤3, mutation of the base pairs flanking the E-box did not modify the promoter specificity toward neuronal bHLH proteins (Fig. 4C). Ectopic Activation of the ␤3 Promoter by a bHLH Protein-Because functional domains are well conserved among members of the bHLH protein family, we asked if MyoD, a transcriptional activator of muscle-specific genes that binds to the CAGCTG motif (29), was capable of replacing endogenous neuronal bHLH proteins. Co-transfections with a MyoD expression vector revealed that ectopic expression of MyoD was sufficient to trans-activate the ␤3RS promoter in neurons from the optic tectum, telencephalon, and cerebellum that do not normally express ␤3 (Fig. 5, A and B). In parallel experiments, telencephalic or tectal cells were co-transfected with MyoD and ␤3RS-lacZ constructs, and trans-activation was observed by X-gal staining in about 20% of transfected cells (Table I). This activation is mediated by the E-box, since MyoD failed to activate the promoter bearing a mutant E-box (Fig. 5A). Moreover, as determined by Northern blot analysis with RNA obtained from transfected tectal cells, MyoD was also able to trans-activate the endogenous ␤3 gene (Fig. 5C), clearly indicating that the transfected ␤3RS promoter behaves in neurons much as the native promoter does.
We also tested the ability of MyoD to trans-activate the ␤3RS promoter in non-neuronal cells. In cells from the pigment epithelium of the retina, MyoD was unable to trans-activate the ␤3RS promoter or the endogenous ␤3 gene (Fig. 5, A and C). We used the ␣1KK promoter to ascertain that functional MyoD was indeed synthesized in transfected pigmented retinal cells. Although the ␣1 nAChR gene is not expressed in this tissue, the transfected ␣1KK fragment displayed detectable promoter activity, which was strongly enhanced by ectopic expression of MyoD (Fig. 5A). Low levels of ␤3RS promoter activity were detected in glia selected from optic tectum or neuroretina or in embryonic fibroblasts (Fig. 5B). We believe that this transactivation resulted from the presence of residual populations of FIG. 4. Trans-activation of the ␤3 and ␣1 promoters by different neuronal bHLH proteins. A, the ␤3RS promoter and the core promoter of the muscle nAChR ␣1 subunit (␣1KK fragment, Fig. 7A) were fused to the CAT gene, and the constructs (5 g) were co-transfected with the expression vectors for CASH-1, NeuroD, NeuroM, and CTF-4 (3.5 g) into cells isolated from E8 optic tectum and E9 telencephalon. The insertionless EMSV plasmid (3.5 g) serves as negative control. Cells were assayed for CAT activity 48 h after transfection. Activity obtained upon co-transfection with SV-CAT (5 g) plus EMSV (3.5 g) is arbitrarily set at 100 for each cell type. Activities of the ␤3 and ␣1 promoters are given relative to this value. B, the DNA binding affinity of purified CTF-4 (0.3 or 0.5 g/assay) was tested on the uncleaved (c) or cleaved (PvuII, EcoNI) ␤3RS fragment. Note that cleavage of the E-box by PvuII abolished binding, whereas cleavage by EcoNI elsewhere in the fragment had no effect. C, the base pairs flanking the E-box do not influence trans-activation by neuronal bHLH proteins. Tectal cells (E8) were co-transfected with either ␤3RS-CAT (TGA-CAGCTGATG) or SF-E-CAT (TCCCAGCTGGCC) together with the expression vectors for CASH-1, NeuroM, NeuroD, or CTF-4. The two promoters work equally well in neuroretina and cannot be activated in optic tectum by any of the tested bHLH proteins.
neurons, since no X-gal-positive glial or fibroblast-like cells were detected when these cells were co-transfected with ␤3RS-lacZ and MyoD. Taken together, these results indicate that the ␤3 promoter is regulated by a bHLH protein that can be substituted by MyoD and that trans-activation of ␤3RS by MyoD is restricted to neurons, suggesting that additional neuron-specific co-activators are required. This view is further supported by the fact that there is no expression of the ␤3 gene in muscle (15) and no activity of the ␤3RS promoter in transfected myotubes (Fig. 7C), despite the presence there of MyoD and other myogenic bHLH factors.
Influence of MyoD on ␤3 Promoter Activity in the Developing Retina-We investigated the effect of MyoD on ␤3RS promoter activity within the domain of ␤3 expression. Transcription of ␤3 in neuroretina is first detected on E4, whereupon activity of the promoter rapidly increases and peaks on E5, decreasing later to relatively low levels (11). We wanted to determine whether ectopic MyoD could activate the ␤3 promoter earlier in development, increase the peak value at E5, or maintain a high level of activity at later stages. The MyoD expression vector was co-transfected with ␤3RS-CAT or with ␤3RS-lacZ into retinal cells isolated at different stages between E4 and E13. MyoD had no influence on promoter activity at early stages of development, but it enhanced CAT synthesis in the developed retina (E8 -E13) without modifying the proportion of ␤-galactosidasepositive cells ( Fig. 6; Table I). The levels of endogenous ␤3 mRNA were affected in the same way, with a MyoD-induced increase on E13 but not on E5 (data not shown). In contrast, MyoD strongly stimulated promoter activity of the ␣1KK fragment both in E5 and E13 retinal cells, indicating that functional MyoD was indeed synthesized in these cells (Table I). We interpret these observations as suggesting that endogenous neuronal bHLH protein(s) for which MyoD can substitute are not limiting in early retina, whereas later in development, ectopic expression of MyoD compensates for decreased amounts of the endogenous bHLH protein(s).
A Hybrid ␣1/␤3 Promoter Behaves Like the ␤3 Promoter-Our experiments highlight the remarkable capacity of the ␤3RS promoter to discriminate between different members of the bHLH family, in striking contrast to the ␣1KK promoter, which was activated by all five of the bHLH proteins we tested. Because sequences in the vicinity of the E-box may play an important role in the recognition of a specific bHLH factor, we constructed a hybrid promoter from portions of the ␣1KK and ␤3RS fragments. The ␣1KK promoter contains two E-boxes. The distal ␣1 E-box and the ␤3 E-box have the same sequence (CAGCTG), and we took advantage of the fact that it is a PvuII restriction enzyme recognition site to fuse the two promoters at this level (Fig. 7A). In the hybrid promoter, termed ␣1/␤3, the 56 bp upstream of the reconstituted CAGCTG E-box come from the ␣1 promoter, while the 69 bp downstream of it come from the ␤3 promoter. Promoter activity of the hybrid was compared with that of ␣1 and ␤3 in myotubes and in CEFs (Fig. 7C). As FIG. 5. Trans-activation of the ␤3 promoter by MyoD. A, the wild-type ␤3RS and E-box mutant (E-box*) promoters fused to the CAT gene were co-transfected with a MyoD expression vector (EMSV-MyoD) into cells isolated from E8 optic tectum, E9 telencephalon, and E5 retinal pigment epithelium. MyoD trans-activated the ␤3 promoter in tectal and telencephalic cells but not in the pigment epithelium. We used the ␣1KK promoter to ascertain that functional MyoD was synthesized in the transfected pigmented epithelial cells. Although the ␣1 gene is not expressed in the eye, the transfected ␣1KK fragment displays ubiquitous activity in retina, and MyoD strongly stimulates this activity. B, the wild-type ␤3RS-CAT promoter construct was co-transfected with either the MyoD (ϩ) or the insertionless (EMSV) expression vector (Ϫ) into different neuronal and non-neuronal cell types. The CAT activity obtained upon co-transfection with SV-CAT plus EMSV is arbitrarily set at 100 for each cell-type, and ␤3 promoter activities are given relative to this value. C, the MyoD (ϩ) or control expression vector (Ϫ) was transfected into cells isolated from E8 optic tectum (OT) and E5 retinal pigment epithelium (PE). Cells were collected, and RNA was extracted 48 h after transfection. Total RNA (3 g/lane) was fractionated by gel electrophoresis, blotted to nylon membranes, hybridized with a 32 P-labeled ␤3 probe, and autoradiographed for 10 days. Before hybridization, the membranes were stained with methylene blue to check that RNA loads were similar in all lanes. Note that endogenous ␤3 nAChR mRNA was produced in tectal cells transfected by MyoD and that it had the same size as the ␤3 mRNA detected in untreated E12 neuroretina (NR). Arrows, ribosomal RNA markers.

Cell types
Control (5) expected, the ␣1 promoter had a strong activity in myotubes and was weak in CEFs. In contrast, the ␤3 and ␣1/␤3 promoters were completely silent in both cell types. In co-transfections, MyoD consistently and strongly enhanced activity of the ␣1 promoter but did not influence the ␣1/␤3 promoter. In tectal and telencephalic neurons, the ␤3 and ␣1/␤3 promoters were completely silent and could not be trans-activated by CASH-1, NeuroM, NeuroD, or CTF-4 ( Fig. 4 and data not shown). The hybrid promoter was only found to be active in retinal cells, where it reached an activity level somewhat lower than ␤3 (Fig.  7B). It is known from a previous study that the distal E-box of the ␣1 promoter is sufficient to drive reporter gene transcription in myotubes (30), yet when it is flanked in 3Ј by sequences from the ␤3 promoter, its activity becomes restricted to retinal cells. Indeed, the hybrid promoter behaves much like the ␤3RS or SF-E fragments (Fig. 2), suggesting that the 69-bp sequence located downstream of the ␤3 E-box contains essential elements that act in concert with the E-box to confer stringent specificity upon the ␤3 promoter.

DISCUSSION
A short fragment, 75 bp in length and located immediately upstream of the transcription start site, is sufficient to generate the neuron-specific expression pattern of the ␤3 nAChR gene. Inquiring into the underlying mechanisms, we present evidence that the ␤3 promoter is positively regulated by an E-box acting in concert with a neighboring CAAT box. The ␤3 promoter appears to have a simple structure and be devoid of the regulatory complexities resulting either from inhibitory DNA elements that prevent expression in non-neuronal cell types (reviewed by Schoenherr and Anderson (2)) or from multipartite elements whose active combinations vary in the course of the development (31). Very few neuron-type specific promoters have been characterized in detail, and we do not know whether the simple organization of the ␤3 cis-regulatory domain is a common feature of genes whose expression is confined to restricted subsets of neurons.
The finding that an E-box is a key regulator of the ␤3 gene emphasizes the role of bHLH factors in neural transcriptional control. Several members of the bHLH family are transiently expressed in the developing CNS. ASH-1 is widely expressed in proliferating precursor cells (26). ASH-1 and neurogenin-1 (ATH-4C) exhibit complementary domains of expression in the neuroepithelium, suggesting that these early bHLH genes might be associated with specification of cell identity (32). The widespread expression of NeuroM and NeuroD in postmitotic cells at distinct times in neural development suggests that they do not define functionally distinct neuronal phenotypes but rather successive stages of a cell's life course (21,33). Transcription of the ␤3 gene is activated in a small subset of proliferating retinal cells, and then it is continuously expressed during cell differentiation and in the mature ganglion and amacrine cells (11). Although ASH-1, NeuroM, and NeuroD are expressed in subsets of retinal cells and instances of overlapping expression between these factors and ␤3 have been detected, we found that the ␤3 promoter was also active in cells that do not express these bHLH genes. 4 Co-transfection experiments have confirmed that the ASH-1, NeuroM, and NeuroD proteins do not control transcription of the ␤3 gene. None of the known neuronal bHLH proteins is present throughout the period of ␤3 expression, and, although we cannot rule out the possibility that ␤3 is sequentially regulated by distinct bHLH proteins during development, we favor the idea that it is regulated by a particular, unidentified bHLH protein whose expression is associated with specific neuronal phenotypes.
Our results demonstrate that MyoD is able, upon ectopic expression in central neurons, to induce transcription of both transfected and endogenous ␤3 promoters. MyoD itself is absent from the nervous system, but expression in the developing brain of several other myogenic regulatory genes such as mef-2 and myf-5 suggests that interesting parallels may exist between muscle and neuron differentiation (34,35). Although MyoD acts as an activator of ␤3 in subsets of central neurons, it is incapable of inducing transcription of this gene in non- The distal E-box in ␣1 and the E-box in ␤3 have the same sequence (CAGCTG, a site for the PvuII restriction enzyme) and fragments were fused at this level. In ␣1/␤3, the reconstructed E-box is flanked in 5Ј by 56 bp from ␣1 and in 3Ј by 69 bp from ␤3. B, the ␣1/␤3-CAT and ␤3RS-CAT constructs were transfected in cells isolated from E5 neuroretina, and 48 h later cells were assayed for CAT activity. Note that both promoters have strong activities in retinal cells. C, the ␤3RS-CAT, ␣1KK-CAT, and ␣1/␤3-CAT constructs were co-transfected with the MyoD (ϩ) or the control expression vector (Ϫ) in primary (CEF ϩ Myotubes) or secondary (CEF) cultures of chick embryonic fibroblasts. Primary cultures contained 5-10% myotubes, whereas myotubes were absent in secondary cultures. Note the 15-fold higher promoter activity of the ␣1KK fragment in primary cultures, reflecting its much stronger activity in myotubes than in CEFs. In contrast, the ␤3 and ␣1/␤3 promoters were inactive in myotubes and CEFs. In B and C, SV40 promoter activity is set at 100, and other promoter activities are given relative to this value. neuronal cells, suggesting that induction requires additional co-activators that are exclusively expressed in neurons. In early retina, MyoD has no influence on ␤3 promoter activity, presumably because the bHLH protein, which controls ␤3 transcription (and for which MyoD can substitute), is available in nonlimiting amounts. Since ectopic expression of MyoD does not increase the proportion (about 15%) of retinal cells expressing ␤3, we propose that the endogenous bHLH protein must, like MyoD, synergize with neuronal co-activators that, in the retina, are confined to the subset of neurons forming the domain of ␤3 expression. However, these or similar co-activators are also present in other subsets of neurons elsewhere in the developing CNS, as demonstrated by MyoD's ability to transactivate ␤3 in a fraction of neurons from different regions of the CNS (Table I). Thus, ␤3 expression depends on the presence in the same neuron of both the appropriate bHLH protein and the appropriate co-activators. In the retina, such a combination is predicted to occur in ganglion and amacrine cells.
The ability of the ␤3 promoter to distinguish between different members of the bHLH family determines its stringent neuron-type specificity. The mechanism by which E-boxes discriminate in vivo between related bHLH proteins is poorly understood. We have shown that although CTF-4 binds to the ␤3 E-box with high affinity in vitro, it is not competent to activate ␤3 transcription in vivo. Our experiments did not provide evidence that the base pairs flanking the ␤3 E-box constitute part of a negative cis-acting element that may prevent ASH-1, NeuroD, NeuroM, and CTF-4 from functioning as activators. Thus, the selection mechanism enabling ␤3 to discriminate between different bHLH proteins is probably different from the mechanism proposed by Weintraub et al. (28) for the regulation of the IgH gene. To locate the sequences that confer specificity to the E-box in ␤3, we constructed a hybrid ␣1/␤3 promoter where the E-box was flanked in 5Ј with sequence from ␣1. Whereas the activity of the ␣1 promoter is enhanced in neurons by the four different neuronal bHLH proteins we have tested, the hybrid does not respond to them, behaving like the ␤3RS promoter and thus demonstrating that the 3Ј-flanking sequence is sufficient to confer upon the E-box an ability to discriminate between bHLH factors. Moreover, we have shown that the CAAT box in ␤3 is an essential positive regulatory element and that shortening the distance between the E-box and CAAT box by 1, 2, or 3 bp strongly decreases promoter activity in retinal cells. The E-box presumably cooperates with the CAAT box and, in order to allow appropriate protein-DNA interactions, the distance between the two motifs should be no shorter than 9 bp. Mobility shift experiments indeed suggest that nuclear proteins specifically interact with a very short fragment of the ␤3 promoter that encompasses the E-box and the CAAT box, supporting the view that the ␤3 promoter utilizes a cooperative mechanism consisting of at least two juxtaposed protein-DNA complexes as components of the transcriptional activation process. The different neuronal bHLH proteins we have tested perhaps fail to activate ␤3 because of their inability to interact with a neighboring com-plex. Thus, the finely tuned regulation of the ␤3 gene depends on the availability of the appropriate bHLH factor and coactivators. The identification of other genes regulated by members of the extending family of neuronal bHLH proteins should help determine whether the regulatory mechanism used by the ␤3 gene is shared by other neuron-specific genes.