Retinal Neuron Activity of ETS Domain-binding Sites in a Nicotinic Acetylcholine Receptor Gene Cluster Enhancer*

Nicotinic acetylcholine receptors (nAchRs) mediate amacrine to ganglion cell synaptic transmission in the developing mammalian retina. The clustered neuronal nAchRs subunit genes, 3 and 4, are expressed in amacrine and ganglion cells where they are used to assemble functional receptor subtypes. The transcriptional mechanisms underlying expression of these subunits in retina are not yet known but may involve enhancers that are selectively active in retinal neurons. We previously identified a neuron-selective enhancer, 43 , whose activity in neural cell lines is dependent on ETS domain-binding sites. To determine whether 43 is active in retinal neurons that express the 3 and 4 genes, we investigated 43 activity in primary dissociated rat retinal cultures. We found that 43 is selectively active in retinal neurons compared with retinal non-neuronal cells. This activity was derived primarily from amacrine and ganglion neurons, which are the retinal neuron cell types that express the clustered genes. Moreover, 43 was selectively active in retinal neurons compared with cerebral cortical neurons suggesting that it is not a panneuronal enhancer. ETS factor-binding sites in the enhancer are required for its retinal neuron activity. These findings suggest that ETS factor interactions with 43 control retinal neuron expression of certain nAchR subtypes.

Nicotinic acetylcholine receptors (nAchRs) 1 are a family of ligand-gated ion channels that are expressed in numerous central and peripheral neuron populations (1). In many instances, different neuronal populations express certain receptor subtypes that have distinct gating and channel properties as well as distinct physiological functions. Diverse gating and channel properties are produced through homomeric and heteromeric assembly of different receptor subunit combinations (2). Localization of nAchRs to either the postsynaptic, presynaptic, or preterminal subcellular compartments is a mechanism that confers distinct physiological roles for particular subtypes (3)(4)(5).
Several neuronal nAchR subunit genes are expressed in the retina including the ␣3 and ␤4 genes, which are clustered in the genome, but not ␣5, the other member of the cluster (6 -12). The ␣3 and ␤4 subunits are likely to be assembled into at least one kind of retinal neuron nAchR subtype (13,14). Moreover, recent evidence (15)(16)(17) has implicated an ␣3-containing receptor in propagation of spontaneous action potential waves that are important for establishing patterns of retinal neuron synaptic connections. Before a neuron can assemble and assign a specific physiological function to a particular nAchR subtype, however, the appropriate combination of subunit mRNAs must be expressed in that cell. In contrast to the growing understanding of the expression and function of nAchRs in retinal neurons, nothing is known about the transcriptional mechanisms that direct expression of ␣3 and ␤4 genes to these cells.
The ␣3 gene promoter does not appear to contain cell typespecific information, which has led us to hypothesize that the clustered nAchR genes are under transcription control of neuron-selective enhancers (18). In support of this hypothesis, we identified an enhancer located in the ␤4 3Ј-untranslated region about 2.5 kb upstream of the ␣3 gene, which is currently the only known enhancer element in the cluster (19). Two distinct cis elements within the ␤43Ј enhancer that bind ETS domain factors are required for its activity in neural cell lines (18). Interest in this enhancer as an essential component of regulatory information in the cluster arises from its differential activity in different cell types. First, it displays neural cell typespecific activity in a manner that correlates well with expression of the endogenous clustered genes in cell lines. Second, transfection experiments in dissociated peripheral primary neuron/non-neuronal cell cultures have shown that expression of reporter genes that are controlled by ␤43Ј is largely limited to neurons (18). These characteristics support the idea that ETS domain factor interactions with ␤43Ј are important for neuron-selective transcriptional control of one or more genes in the cluster.
Because the enhancer is located between the ␣3 and ␤4 promoters and the ␣3 and ␤4 genes are expressed in retina, ␤43Ј may control retinal neuron expression of these genes. However, it is not yet known whether ␤43Ј is active in retinal neurons. Here we have transfected dissociated rat primary neuron/non-neuronal retinal cell cultures with luciferase reporters to determine whether ␤43Ј is active in retinal neurons. The data presented suggest that ␤43Ј is a retinal neuron enhancer that may control expression of the cluster genes in these cells through interactions with ETS domain factors.

Plasmids
The pGL3 vector series (Promega Corp., Madison, WI) was used to prepare luciferase reporter constructs. The rat ␤43Ј-(1-90) sequences and mutations or deletions of it were prepared using three-way ligations of vector and synthetic oligonucleotides as described (18). The AdMLP construct contains sequences Ϫ55/ϩ10 of the adenovirus 2 major late promoter cloned upstream of the luciferase reporter. ␤43Ј-(1-90) has identical activity to ␤43Ј-(1-107) (18). For the green fluorescent protein (GFP) reporter construct, the GFP reporter was cloned from enhanced GFP (Stratagene, La Jolla, CA; gift of T. Large) in place of the luciferase gene in the ␤43Ј/␣3P construct using restriction sites. All other reporters used in this study were described previously (18).
␤43Ј reporters prepared using synthetic oligonucleotides were confirmed by dideoxy sequencing using Sequenase reagents according to the manufacturer's protocol (Amersham Biosciences); all other constructs were confirmed by restriction digests. Oligonucleotides were synthesized by Invitrogen, and plasmids for transfections were prepared using Qiagen reagents (Qiagen, Santa Clarita, CA). At least two different plasmid preparations were used for each construct, and all constructs were tested in at least three independent experiments.

Cell Culture and Transfection
Retinal Cultures-Retinas were dissected from P1 Sprague-Dawley rat pups (Zivic Miller, Portersville, PA) and dissociated in dispase/ collagenase or 5 mg/ml dispase alone (in the course of these experiments it was found that cell viability was much better when dissociation was carried out with dispase alone) for 12-15 min. Following a rinse in serum-containing medium, retinas were triturated with a fire-polished Pasteur pipette in serum-containing medium with 3.5% bovine serum albumin (Invitrogen), and ϳ5 ϫ 10 5 cells were plated into each well of poly-L-lysine-and laminin-coated 24-well plates or onto 12-mm glass coverslips (Fisher). Cells for 6 -8 wells of a 24-well plate were typically obtained from each animal (two retinas). Cultures were allowed to grow for 3 or 7 days before transfection, with fresh media added every 3 days. Cultured neurons had similar morphologies at 3 and 7 days in vitro, but we found that transfection efficiencies were much higher after 7 days in vitro (compare numbers in Tables I and II). Cultures used for immunocytochemistry (ICC) were grown in Dulbecco's modified Eagle's medium (Celox, St. Paul, MN) supplemented with 10% heat-inactivated fetal bovine serum (HyClone, Logan, UT), penicillin and streptomycin (Invitrogen). Cultures used for luciferase assays were plated in the same media but changed to serum-free medium after about 24 h. In serumfree cultures, non-neuronal cells did not proliferate extensively; transfection efficiency was less variable, and neurons were transfected much more efficiently than non-neuronal cells. ICC results under serum-free conditions indicate that greater than 90% of the reporter-expressing cells are neurons, even when the SV40P/SV40E (which is expressed at high levels in non-neuronal cells, see below) is used. Serum-free conditions were therefore used to analyze ␤43Ј elements using quantitative reporter assays. Serum-free medium consisted of Dulbecco's modified Eagle's medium supplemented with insulin/transferrin/selenium (Sigma), penicillin/streptomycin, 0.1 mg/ml sodium pyruvate (Sigma), bovine serum albumin (1.5%), and 10 ng/ml recombinant human brainderived neurotrophic factor (Peprotech, Rocky Hill, NJ) to increase survival of retinal ganglion cells (20). Growth of cultures in human brain-derived neurotrophic factor did not change relative reporter activities. Calcium phosphate transfections were performed essentially as described (21) including in some cases a 4-min treatment with 2% dimethyl sulfoxide (Me 2 SO) at the end of the transfection period to increase transfection consistency (21). The Me 2 SO treatment was nonessential for cultures grown under serum-free conditions. Serum-free cultures were typically incubated with the calcium phosphate precipitate 35-45 min. Two g of reporter DNA were used per transfection for luciferase assays and 4 g for ICC. In some experiments, an internal control plasmid (RSV-␤-gal) was co-transfected; ␤-galactosidase correction of luciferase data produced similar results to those reported here (data not shown).
Cortical Cultures-Cerebral cortices from P1 or E18 rats (Zivic Miller) were dissected free of meninges, dissociated in dispase/collagenase, rinsed in serum-containing medium, and triturated. Cultures were plated on poly-L-lysine and laminin, grown overnight in serum-containing medium, and changed to serum-free medium. Serum-free medium for cortical cultures consisted of Neurobasal medium with B-27 supplement (both from Invitrogen) and penicillin/streptomycin. Cortical cultures were transfected after 5 days in vitro using calcium phosphate as described for retinal cultures. Similar to retinal cultures, neurons were transfected much more efficiently than non-neuronal cells in these culture conditions so that nearly all of the luciferase activity of reporters is derived from neurons.

Luciferase Assays
Luciferase assays were carried out according to the manufacturer's protocol (Promega Corp., Madison, WI) 24 h after transfection. Cells were lysed in 75 l of lysis buffer per well of 24-well plates, and 10 l of the lysate were assayed in 50 l of luciferin substrate using a Lumat LB9501 luminometer (EG & G Berthold, Nashua, NH).
All ICC/IHC data were analyzed on a Nikon microscope. Cell counts were performed directly from immunostained coverslips. Photomicrographs were digitally processed from color slides or black and white prints using Adobe Photoshop.
RNase Protection RNA was isolated from retinal cultures grown 4 days in serum-free medium using RNazol (Tel-Test, Friendsville, TX); 10 -20 g of RNA was used for each protection reaction. PC12 cell RNA was used as a positive control, and Rat2 RNA was used as a negative control for both Pet-1 and ␣3 (23). RNase protection was carried out using the RPAII kit (Ambion, Dallas Center, IA) according to the manufacturer's instructions. The ␣3 (24) and Pet-1 (23) probes were prepared as described, including gel purification.

RT-PCR
RNA was isolated from retinal cultures as described above. Reverse transcription was carried out with 1-4 g using Moloney murine leukemia virus-reverse transcriptase (Invitrogen) and 35 cycles of PCR at 60 (ERM, ER81) or 55°C (PEA3) using a PerkinElmer Life Sciences thermal cycler. Primers were constructed from the mouse sequences as

Retinal Nuclear Extracts
Mini-nuclear extracts were prepared from retinal cultures that had been grown in vitro for 4 days essentially as described (27). Briefly, cells were rinsed once in phosphate-buffered saline (PBS), removed from the substrate with trypsin, collected by centrifugation, and washed again in PBS. Cells were resuspended in Buffer A (10 mM HEPES, pH 7.9; 10% glycerol; 10 mM KCl; 0.1 mM EDTA; 0.1 mM EGTA; 1 mM dithiothreitol; 0.5 mM phenylmethylsulfonyl fluoride; 4 g/ml leupeptin; 1 g/ml aprotinin; 1 g/ml pepstatin). Nonidet P-40 (Roche Molecular Biochemicals) (final concentration 0.5%) was added immediately, and the cells were mixed by inverting several times. Nuclei were pelleted by brief centrifugation at 4°C and extracted with Buffer C (20 mM HEPES, pH 7.9; 0.4 M NaCl; 1 mM EDTA; 1 mM EGTA; 1 mM dithiothreitol; protease inhibitors). Extracts were microcentrifuged at 4°C for 10 min, and the supernatant was saved. Protein concentrations were determined using the BCA assay (Bio-Rad) and were typically 2-6 g/l. One to three 60-mm dishes of retinal cells grown in serum-free medium were routinely used for extract preparation, yielding 50 -150 l of nuclear extract.

Retinal Neuron Cultures Express the ␣3
Gene-To confirm that ␣3-expressing neurons were present in primary retinal cultures, RNase protection was carried out with RNA from cultures grown 4 days in vitro. ␣3 mRNA was clearly detected in RNA samples obtained from several different independent cultures, consistent with the presence of retinal ganglion and amacrine cells (Fig. 1).
␤43Ј Directs Neuron-selective Gene Expression in Retinal Cultures-By having confirmed that retinal cultures are a valid system in which to assay transcriptional elements present in the neuronal nAchR gene cluster, we set out to investigate whether ␤43Ј is active in retinal neurons. Dissociated retinal cultures were composed of about 10% neuron-specific tubulin-immunoreactive neurons after 3 days in vitro and only 2-3% after 7 days in vitro because of extensive non-neuronal cell proliferation. The majority of the non-neuronal cells were immunoreactive for glial fibrillary acidic protein, suggesting the presence of Mü ller glial cells, although a small number of Thy-1-positive fibroblasts were also present (data not shown). When retinal cultures were transfected after 3 days in vitro with a luciferase reporter carrying ␤43Ј-(1-90) upstream of the ␣3 promoter (␤43Ј/␣3), nearly 100% of the luciferase-expressing cells were neurons (Fig. 2, A, B, and E, and Table I). The same result was obtained with ␤43Ј in reverse orientation ( Fig. 2E and Table I). To demonstrate that non-neuronal cells can be efficiently transfected in these cultures, we scored luciferasepositive neurons and non-neuronal cells after transfection of a luciferase reporter carrying the SV40 promoter and SV40 enhancer (SV40E/SV40P). In clear contrast to that found for ␤43Ј/␣3, approximately equal numbers of neurons and nonneuronal cells were luciferase-positive over the course of 15 independent experiments with the SV40E/SV40P reporter ( Fig.  2, C, D, and E). The difference in the percentage of neurons expressing the luciferase reporter driven by ␤43Ј versus SV40E was highly significant (Table I). Because the ratio of luciferasepositive neurons and non-neuronal cells is ϳ1:1 in SV40E/ SV40P transfections, rather than the 1:9 that would be expected based on the composition of the cultures, it is possible that SV40E/SV40P also has a cell type-specific bias toward expression in neurons. However, we cannot distinguish this possibility from an alternative explanation, which is a bias toward transfection of neurons in these experiments. Nevertheless, the finding that both neurons and non-neuronal cells express the SV40 reporter demonstrates that both types of cells were transfected. Thus, the highly significant difference in cell type-specific activity between SV40E/SV40P and ␤43Ј/␣3 supports the idea that the ␤43Ј enhancer is a transcriptional ele-ment that together with the ␣3 promoter is able to drive retinal neuron-specific gene expression.
To determine whether the ␣3 promoter is required for neuron-selective expression in retinal cultures, assays were performed with reporters, ␤43Ј/SV40P and ␤43Ј/AdMLP, in which the ␣3 promoter was replaced with either the adenoviral major late (AdMLP) or the simian virus 40 (SV40) promoters. Both of these promoters are essentially TATA boxes and therefore are not expected to direct cell type-specific transcription on their own. In cultures transfected after 3 days in vitro with either reporter, the majority of the luciferase-expressing cells were neurons regardless of ␤43Ј orientation ( Fig. 2E and Table I). When retinal cultures were transfected with a reporter carrying the ␣3 promoter and SV40 enhancer (SV40E/␣3P), more non-neuronal cells than neurons expressed detectable luciferase, similar to what was seen with SV40E/SV40P (Fig. 2E, Table I). This further confirms that both neurons and nonneuronal cells were being transfected but that ␤43Ј-driven reporters are preferentially expressed in neurons. Thus, the neuron-selective activity of ␤43Ј does not require the ␣3 pro-  (Table I) with the indicated reporters. The percentages were averaged to produce the bar graphs. Cultures were maintained for 3 days in vitro before transfection. Arrows indicate orientation of the enhancer. Error bars are S.E. moter, and this promoter is not sufficient to direct a neuronal pattern of gene expression in retinal cultures. We conclude that the ␤43Ј enhancer is both necessary and sufficient to drive neuron-selective expression in these cultures.
We next investigated whether ␤43Ј neuron selectivity is maintained under conditions in which the ratio of non-neuronal cells to neurons is increased in cultures. To test this, we transfected cultures with the various reporters after 7 days in vitro; this incidentally resulted in higher transfection efficiencies (compare luciferase-positive cells in Tables I and II). As before, neither SV40E/SV40P nor SV40E /␣3P supported neuron-selective expression of luciferase. In contrast, when ␤43Ј/ ␣3P was transfected ϳ90% of the luciferase-expressing cells were neurons regardless of ␤43Ј orientation ( Fig. 3 and Table  II). Thus neuron selectivity was maintained even when neurons constituted only a small percentage of the cells in culture. Significantly more neurons were luciferase-positive in cultures transfected with either ␤43Ј/SV40P or ␤43Ј/AdMLP than with SV40P/SV40E. The percentage of luciferase-positive cells that were neurons, however, was lower in transfections with ␤43Ј/ SV40P or ␤43Ј/AdMLP than in cultures transfected with ␤43Ј/ ␣3P ( Fig. 3 and Table II); this difference was statistically significant in some cases. These results suggest that the discrimination of ␤43Ј between neurons and non-neuronal cells may be supported by its natural promoter.
␤43Ј could achieve neuron selectivity by silencing promoter activity in non-neuronal cells. To test this idea, we transfected retinal cultures with reporters driven by both the ␤43Ј and SV40 enhancers and the SV40 promoter. If ␤43Ј silences promoter activity in non-neuronal cells, this construct should have similar neuron specificity to reporters driven by ␤43Ј alone. We found, however, that this reporter is highly expressed in both neurons and non-neuronal cells (Table II), suggesting ␤43Ј does not silence promoter activity in non-neuronal cells.
␤43Ј Activity in Retinal Ganglion and Amacrine Neuron Cell Types-Both retinal ganglion cells and amacrine cells in the ganglion cell and inner nuclear layers of the retina express ␣3 and ␤4 mRNA from about E13 into maturity (7)(8)(9)12). To determine whether ␤43Ј-driven reporters are also expressed in amacrine and ganglion cells, we quantitated colocalization of retinal cell type markers with ␤43Ј-driven GFP. Two retinal ganglion cell markers were used, L1 and Thy-1. The localization of ␤43Ј-driven GFP to ganglion cells was first tested by immunostaining against the cell adhesion molecule L1, which is specific for ganglion cells in the retina (28). L1 immunoreactivity was largely restricted to neurites and growth cones, but processes that were positive for both L1 and ␤43Ј-driven GFP could be clearly identified (Fig. 4, A and B). Because of the density of L1-expressing neuritic processes, and the length of the processes, we were unable to quantify L1-GFP colocalization, which would require identification of the cell body producing the neurite. Thy-1 has been used extensively for the identification of retinal ganglion cells (29). We found that 45-55% of the neurons expressing ␤43Ј-driven GFP could be identified as retinal ganglion cells by Thy-1 expression (Fig. 4, C and D). Together, these data are consistent with the L1 colocalization and support the conclusion that a large proportion of the neurons expressing ␤43Ј-driven GFP are retinal ganglion cells.
Amacrine-specific markers were not available, but the monoclonal antibody VC1.1 recognizes an epitope present on horizontal and amacrine cells, but not retinal ganglion cells, in the retina (30). Many VC1.1 immunoreactive neurons were observed in our retinal cultures. Approximately 30% of the ␤43Јdriven GFP-expressing neurons were immunoreactive for VC1.1 (Fig. 4, E and F). The number of horizontal cells in the mammalian retina is far smaller than the number of amacrine cells (31), which suggests that the majority of VC1.1-positive GFP-expressing cells are amacrines. Consistent with this, calbindin immunoreactivity, a marker for horizontal cells (30,32), was detected in about 1% of GFP-expressing cells (Fig. 4, G and  H). Together, these results suggest that at least 75-85% of the neurons expressing the ␤43Ј/␣3P-GFP reporter were ganglion (Thy-1ϩ) and amacrine (VC1.1ϩ, Calbinden-) cells. The remaining neurons are likely to be bipolar neurons or ganglion cells that were undetected due to the weak immunoreactivity observed with Thy-1. Although occasional transfected photoreceptors were observed in retinal cultures, these cells have distinct morphologies and were not included in the analysis of retinal neuron subtype markers.
Magnitude of ␤43Ј Transcriptional Activation in Different Neuronal Cell Types-As an additional measure of the neuronselective activity of ␤43Ј, we quantitated the magnitude of ␤43Ј-enhanced luciferase expression in either dissociated retinal or cerebral cortical cultures. We compared the ratio of ␤43Ј/␣3P-driven luciferase activity to ␣3-driven luciferase activity across cultures. In the conditions used here for retinal and cortical cultures, most of the transfected cells were neurons (see "Experimental Procedures"). Therefore, luciferase activity is derived primarily from neurons, and as presented in  (Table II), and the percentages were averaged. Error bars are S.E. Figs. 2-5, this activity is primarily from ganglion and amacrine cells in retinal culture transfections. Quantitation of luciferase activity indicated that the enhancer increased ␣3P reporter gene expression in cerebral cortical cultures, which contain many different types of neurons, by only a few fold. A small number of cortical neurons express ␣3, and therefore ␤43Ј activity observed in cortical cultures could arise from a small number of neurons in which the enhancer is highly active, or a low level of activity in many different types of neurons. In contrast, a 25-fold mean stimulation was detected in dissociated retinal cultures isolated from P1 rats (Fig. 5). In some experiments the stimulation was as high as 60-fold. This comparison indicates that ␤43Ј is not equally active in different types of neurons, and therefore ␤43Ј is not likely to be a panneuronal enhancer.

ETS Domain-binding Sites in ␤43Ј Are Required for Its Retinal Neuron Activity-The data presented in Figs. 1-5 and
Tables I and II support the hypothesis that ␤43Ј is a retinal neuron enhancer involved in controlling retinal neuron-specific nAchR gene expression. This raises the question of what elements within the enhancer are important for its activity and what transcription factors interact with these elements. Analysis of various fragments and point mutations of ␤43Ј in PC12 cells has suggested that it is composed of at least two interact-ing cis elements that can bind ETS domain factors (18). The enhancer is much stronger in retinal neurons than in either PC12 cells (18) or cortical neurons (Fig. 5). Therefore, we investigated whether the same cis elements are responsible for enhancer activity in retinal neurons or whether additional ␤43Ј elements that are silent in PC12 cells and cortical neurons augment enhancer activity in retinal neurons. We first prepared various deletions of the enhancer through either of its two repeats while leaving the 6-bp spacer region intact (Fig. 6). These were then tested for stimulation of reporter activity in the context of the AdMLP after transfection into retinal cultures. Neither of the isolated repeats increased reporter activity when placed upstream of the minimal promoter (Fig. 6,  1-37, 1-43, and 38 -80). In contrast, reporters carrying different deletions of repeat 1 without altering repeat 2 had no effect on enhancer activity (Fig. 6, 11-90, 21-90, and 31-80). However, when deletions were made at the end of both repeats, enhancer activity was reduced to about 50% (Fig. 6, 24 -52). These results are similar to those obtained in the PC12 cell line, and therefore, it is not likely that the enhancer uses additional cis elements to stimulate high levels of transcription in retinal neurons.
To determine whether the ETS domain-binding sites present in the enhancer are required for its activity in retinal neurons, we tested reporters carrying base substitutions in these sites  ( Fig. 7). Each of the three sites was eliminated by changing its GGA core, which is required for ETS domain binding (33). Elimination of either ETS site in the repeats had little or no effect on the activity of the enhancer (Fig. 7, reporters 2 and 3) which is consistent with the deletion analysis that showed that most of one repeat can be eliminated without affecting enhancer activity (Fig. 6). In contrast, activity of a reporter (Fig.  7, reporter 4) in which both ETS sites in the repeats were eliminated was reduced by about 55%, which is similar to the level of reporter 24 -52 (Fig. 6). Other dual repeat point mutations had no significant effect on enhancer activity (Fig. 7,  reporters 5 and 6). We then tested a reporter in which the spacer ETS site was destroyed (Fig. 7, reporter 7) and found that this change nearly eliminated enhancer activity.
Expression of ETS Transcripts and ETS-like ␤43Ј-binding Proteins in Retinal Cells-To begin to determine which ETS factors might interact with ␤43Ј in retinal cultures, RNase protection and RT-PCR was carried out for various ETS transcripts. We showed previously that the ETS domain factor Pet-1 can bind the ␤43Ј enhancer (18). RNase protection with RNA obtained from 4-day cultures detected Pet-1 RNA indicating that Pet-1 is expressed in retinal cultures (Fig. 8A). The PEA3 family of ETS factors (ERM, ER81, and PEA3) binds similar sites to Pet-1 (26), and RNA for each of these factors is expressed in the developing mouse retina (25). RNA for all three factors could be amplified from retinal culture (Fig. 8B) indicating that, in addition to Pet-1, ERM, ER81, and PEA3 are potential regulators of ␤43Ј.
To characterize retinal ␤43Ј-binding proteins, nuclear extracts were prepared from retinal cultures and used for EMSA analysis with a probe corresponding to most of the R1 sequence (Table III). Several DNA-protein complexes were reproducibly formed which were competed by an excess of cold probe (Fig. 8, C and D). To identify complexes that require the ETS site for binding, a competitor corresponding to R1 with the ETS site eliminated (R1 ETSm) was tested (Table III). The lowest mobility complex (complex A) (Fig. 8C, lane 4), was no longer effectively competed by R1 ETSm consistent with the presence of an ETS-like factor in this complex. The sequences of R1 and R2 are nearly identical; one of the differences occurs at position ϩ5 of the ETS core site (R1 ϭ G; R2 ϭ A). Therefore, to determine whether similar ETS-like factors bind R2, wild type and R2 ETSm oligonucleotides were also used as competitors (Table III). R2 competed as effectively as R1 (Fig. 8C, lane 5) but R2 ETSm did not (Fig. 8C, lane 6). These results suggest the two repeats contain similar binding sites and probably bind similar proteins.
To pursue further the possibility that complex A contains an ETS factor, a series of EMSAs were carried out using competitors that correspond to ETS sites characterized previously. Four ETS sites, which have been shown to interact with ETS2, GAPB␣/PU1.1, PEA3, and ER81 (Table III), competed for complex A (Fig. 8D, lanes 4, 6, 8, and 10). However, oligonucleotides in which the GGA cores of three of the tested ETS sites were disrupted did not compete (Fig. 8D, lanes 5, 7, and 9). A different ETS site derived from the T-cell receptor ␣ enhancer binds ETS-1 (34,35) and competes for spacer ETS-binding proteins from PC12 cells (18). This site, however, did not compete for complex A (Fig. 8D, lanes 11 and 12). The ability of multiple ETS site sequences to compete for complex A and the dependence on the GGA core for competition strongly suggest that complex A contains an ETS factor. The finding that the T-cell receptor ␣ site, which can bind ETS-1 (34), did not compete for complex A suggests that the ETS factor involved is more related to PEA3, GABP␣, or PU1 than to ETS-1.

DISCUSSION
Elucidating the transcriptional mechanisms that underlie selective expression of certain genes in different neuronal cell types has the potential to help reveal how neuronal cell type diversity is generated in the nervous system. For example, the proper development and maintenance of cholinergic neurotransmitter systems depends on appropriate cell type-specific transcription of particular neuronal nAchR subunit genes in neurons. The clustered neuronal nAchR genes are expressed in numerous but far from all central and peripheral neuronal populations (11, 36 -38). Moreover, the pattern of expression of each gene undergoes significant changes among these populations during embryonic development (12). Unfortunately, for most types of differentiated neurons, including cholinergic types, representative cell lines are not available for gene transcription studies. Primary neural cell culture offers an alternative approach that permits detailed molecular analyses of transcriptional cis elements in normal neurons (18,39). By using primary neural cell cultures that express the clustered nAchR subunit genes, we have investigated the cell type-specific activity and cis elements of the nAchR ␤43Ј enhancer. The main findings presented are as follows. 1) The ␤43Ј enhancer is selectively active in retinal neurons compared with non-neuronal cells and other kinds of central neurons.
2) The majority of retinal neurons expressing the ␤43Ј-driven reporter genes are amacrine and ganglion cells, which are the types of neurons that express the clustered nAChR subunit genes. 3) Retinal neuron activity of the enhancer depends on at least two different ETS domain-binding sites. 4) In contrast to many other cis elements identified in neuronal genes (40 -44), ␤43Ј does not appear to use negative elements to direct retinal neuron-selective expression. These results are discussed in terms of a model in which ETS domain binding to ␤43Ј regulates transcription of one or both of the ␣3 or ␤4 genes as a first step in controlling expression of certain neuronal nAchR subtypes at cholinergic synapses of the retina.
Selective Activity of ␤43Ј in Retinal Neurons-To begin to understand transcriptional control within the cluster, much attention has focused on the promoter regions of these genes. A common feature of these promoters across species is the presence of several Sp1 sites embedded in a G ϩ G-rich region that initiates transcription at numerous nucleotides (45)(46)(47)(48)(49)(50). Thus, it is not surprising that the differential activity of these promoters in different cell lines does not display a strict concordance with expression of the endogenous clustered genes in these lines. These characteristics make it difficult to account for the complex and dynamic neuronal expression patterns of the clustered nAchR genes solely through promoter-directed transcriptional events and have led us to hypothesize that enhancers are operating in the cluster to influence promoter activity in certain neuronal populations.
The identification of the ␤43Ј enhancer provides support for this idea because its activity among cell lines correlates well with expression of the endogenous clustered genes in these lines (19). Moreover, in physiologically more relevant primary neuron transfections, we have found that ␤43Ј is selectively active in neurons versus non-neuronal cells (18), and as presented here it is strikingly more active in retinal neurons than in cerebral cortical neurons. The weak activity of ␤43Ј in cortical cultures is significant because it strongly suggests that ␤43Ј is not a pan-neuronal enhancer that is equally active in all types of differentiated neurons. The data presented suggest instead that ␤43Ј can direct two levels of cell type-selective transcription. First, it is largely inactive in non-neuronal cells. Second, it confers higher activity in only certain types of neurons.
Many vertebrate neuron-specific genes are regulated by cell type-specific silencers that repress their expression in nonneuronal cells (43,51). In contrast, several lines of evidence suggest that ␤43Ј does not use silencers to direct neuron-selective expression. First, sequences similar to the neuron-restrictive silencer element (43) and the GAP-43 repressor element (44) are not present in the enhancer. Second, a reporter containing both ␤43Ј and SV40 enhancers and the SV40 promoter is strongly expressed in both neurons and non-neuronal cells of retinal cultures (Table II), suggesting that ␤43Ј does not silence promoter activity in non-neuronal cells. Third, if activity is controlled by a strong cell type-specific repressor, certain truncations or mutations of the enhancer sequence would be expected to increase reporter activity in mixed cultures by disrupting this element. However, none of the enhancer truncations or mutations result in increased activity (Figs. 6 and 7), even in cultures grown in serum, which contain larger numbers of non-neuronal cells (data not shown). In keeping with these data, the 24 -52 truncation is neural cell typespecific (18), and preliminary analysis of truncated enhancers with reduced activity in retinal cultures suggests they are still neuron-selective. 2 Thus, our data strongly suggest that ␤43Ј is composed of a combination of cis elements that are preferentially active in certain neurons rather than suppressed in non-neuronal cells. The neuron selectivity and differential activity of the enhancer among different neuron populations may arise, at least in part, through an optimal expression level of a combination of appropriate cell type-restricted ETS factors and other factors that together cooperatively bind the enhancer. Non-neuronal cell types and certain kinds of neurons such as those in cortical cultures may not express all of the ␤43Јbinding proteins required for cooperativity, or their expression levels may not be optimal for cooperative interactions.
ETS Factor Function in Retinal Neurons-Activity of ␤43Ј in the PC12 neural cell line depends on two different ETS domainbinding sites. In addition, mobility shift assays with different ETS-binding site competitors and protein-DNA complex-blocking antibodies suggest that each of these sites binds different ETS domain factors expressed in this line (18). The results presented here show that these same ETS-binding sites are required for retinal neuron activity of ␤43Ј. Elimination of either repeat ETS site had little or no effect on ␤43Ј activity, whereas elimination of both sites reduced enhancer activity to less than half of wild type. Significantly, elimination of the spacer ETS site nearly destroyed enhancer activity in retinal neurons. Together with activities determined for various ␤43Ј fragments in retinal neurons, these results suggest that ␤43Ј neuronal activity is the result of interactions among an ETS site in the spacer and redundant ETS sites in the repeats. As alluded to above, it is possible that additional undefined cis elements are present within the enhancer and are necessary along with the ETS sites for retinal neuron activity.
Our findings raise the question of which ETS domain factors are expressed in retinal amacrine and ganglion cells and which of these interact with the enhancer. There is currently little known about ETS domain gene expression in the vertebrate retina, and to our knowledge nothing is known about ETS factor function in this tissue. One study showed that the erm, er81, and pea3 genes are expressed in the developing mouse neuroretina; however, whether these genes are expressed in amacrine or ganglion cells was not reported (25). mRNA for Pet-1, ERM, ER81, and PEA3 were detected in our cultures, suggesting they are candidates for regulating ␤43Ј. Our retinal cultures contain ␤43Ј-binding proteins that require an intact ETS site for binding and are specifically competed by a variety of ETS sites. Together, these data are an indication, albeit indirect, that ETS domain factors may regulate gene expression in retinal neurons. An important future goal will be to 2 N. Francis and E. S. Deneris, unpublished observations. FIG. 7. Two different ETS domainbindings sites are required for ␤43 activity in retinal neurons. Transient transfections were performed in dissociated primary retinal cultures with the indicated reporters. The various reporters include either the intact ␤43Ј enhancer (filled tandem rectangles linked by thin line) or modified enhancers (indicated by crosses) in which ETS-binding sites and other sequences were eliminated in either the first repeat alone, the second repeat alone, both repeats, or the spacer. Activity of the indicated reporters is presented relative to the activity of a reporter containing an intact enhancer, which is set at 100%. Also shown is the relative activity of the AdMLP in the absence of enhancer sequences. Bars represent the average % of ␤43Ј-(1-80) luciferase activity from at least three independent experiments Ϯ S.E. Asterisks indicate activities significantly different from ␤43Ј-(1-80) by analysis of variance.  PC12 RNA (1st lane) was used as a positive control, and Rat2 fibroblast and yeast total RNAs (2nd and 3rd lanes) were used as negative controls. 20 g of RNA was used for each RNase protection reaction, and the Pet-1-protected fragment is ϳ170 bases long. For PC12 RNA, 2 g of RNA was mixed with 18 g of yeast total RNA. B, mRNA for PEA3 family ETS factors ERM, ER81, and PEA3 are expressed in 4-day retinal cultures. Each set of three lanes consists of a reaction run without template (lanes 1, 4, and 7), without reverse transcriptase (lanes 2, 5, and 8), or with template and reverse transcriptase (lanes 3, 6, and 9). The expected sizes of the PCR products are 1.5 kb (ERM), 1.4 kb (ER81), and 0.6 kb (PEA3). Lane 10 is HindIII digested size markers. C, EMSAs (C and D) were carried out with a radiolabeled probe consisting of most of the R1 sequence (Table III), retinal nuclear extracts, and various oligonucleotide competitors. Several protein complexes were formed by retinal nuclear extracts on the R1 probe (lane 2) which were competed by an excess of cold R1 (lane 3) or R2 (lane 5). R1 or R2 with the ETS site mutated (Table III) did not compete effectively for complex A (lanes 4 and 6). Lane 1 shows probe incubated in the absence of extracts. Competitors were used at a 200-fold molar excess. D, EMSAs were carried out as in C with a variety of oligonucleotide competitors (Table III). Complex A was competed by wild type but not mutated ETS sites. Asterisks mark nonspecific complexes that were formed from certain batches of nuclear extracts. The molar excess of competitors used is as follows: lanes 2-5, 11, and 12 are ϫ200 and lanes 6 -10 are ϫ400. The lanes shown are taken from several gels that have been aligned according to the position of complex A. identify the ETS protein(s) present in complex A and determine their role in regulating nAchR gene expression.
What Is the Biological Role of ␤43Ј?-nAchRs composed of ␤4 and ␣3 subunits are likely to function in the retina to mediate cholinergic synaptic transmission (13,15). Assembly of these heteromeric nAchR subtypes, therefore, depends on coordinate regulation of the ␤4 and ␣3 genes, which suggest the existence of cis elements that are active in retinal neurons. ␤43Ј is currently the only known transcriptional element in or near the clustered nAchR genes that is capable of supporting retinal neuron-selective gene transcription. This characteristic, together with the absence of ␣5 expression in retina, raises the possibility that ETS domain interactions with ␤43Ј are required specifically for ␤4 or ␣3 expression in retinal ganglion and amacrine cells. The location of ␤43Ј between the ␤4 and ␣3 promoters may enable it to coordinately regulate both genes in these neuronal cell types without influencing ␣5 transcription. In vivo molecular genetic approaches are ongoing to determine the biological function of ␤43Ј and ETS factors in the vertebrate retina.