The bZIP transcription factor Nrl stimulates rhodopsin promoter activity in primary retinal cell cultures.

In vitro DNA binding assays and transient transfection analysis with monkey kidney cells have implicated Nrl, a member of the Maf-Nrl subfamily of bZIP transcription factors, and the Nrl response element (NRE) in the regulation of rhodopsin expression. We have now further explored the role of the NRE and surrounding promoter elements. Using the yeast one-hybrid screen with integrated NRE and flanking DNA as bait, the predominant clone obtained was bovine Nrl. Recovery of truncated clones in the screen demonstrated that the carboxyl-terminal half of Nrl, which contains the basic and leucine zipper domains, is sufficient for DNA binding. To functionally dissect the rhodopsin promoter, transient expression studies with primary chick retinal cell cultures were performed. Deletion and mutation analyses identified two positive regulatory sequences: one between −40 and −84 base pairs (bp) and another between −84 and −130 bp. Activity of the −40 to −84 region was shown to be largely due to the NRE. On co-transfection with an NRL expression vector, there were 3-5-fold increases in the activity of rhodopsin promoter constructs containing an intact NRE but little or no effect with rhodopsin promoters containing a mutated or deleted NRE. Nrl was more effective than the related bZIP proteins, c-Fos and c-Jun, in stimulating rhodopsin promoter activity. The −84- to −130-bp region acted synergistically with the NRE to enhance both the level of basal expression and the degree of Nrl-mediated trans-activation. These studies support Nrl as a regulator of rhodopsin expression in vivo, identify an additional regulatory region just upstream of the NRE, and demonstrate the utility of primary retinal cell cultures for characterizing both the cis-acting response elements and trans-acting factors that regulate photoreceptor gene expression.

Rhodopsin is the visual pigment of vertebrate rods, and its activation by light initiates the phototransduction process (1). In recent years there has been increasing interest in understanding the mechanisms regulating rhodopsin gene expression, motivated both by the fundamental importance of the protein in visual transduction and by its role as a prototype for the study of photoreceptor-specific gene regulation (2). In addition, since mutations in the rhodopsin gene can lead to retinal degeneration (3,4), understanding of its regulation will have implications for designing effective retinal gene therapy strategies (5).
Nuclear runoff experiments have demonstrated that the expression of rhodopsin is primarily regulated at the transcriptional level (6,7). Transgenic studies have defined two regulatory regions, a rhodopsin proximal promoter region (RPPR) and a more distal rhodopsin enhancer region (RER). 1 The rhodopsin proximal promoter region, located within Ϫ176 to ϩ70 and Ϫ400 to ϩ80 bp in the bovine (34) 2 and murine (8) genes, respectively, provides low level, yet photoreceptor-specific, expression. The RER, a 100-bp sequence located approximately 2 kilobases upstream from the mRNA initiation site, acts as an enhancer (9). Biochemical analysis has identified a number of putative cis-acting DNA regulatory elements; they include the Ret1-PCE1 (10,11), Ret2, Ret3 (12), Ret4 (13), and Mash-1 (14) binding sites and a site homologous to the Drosophila glass response element (15).
Sequence analysis also disclosed the presence of a putative binding site for the Maf/Nrl family of transcription factors in the rhodopsin promoter. Nrl is an evolutionarily conserved basic motif-leucine zipper (bZIP)-containing DNA binding protein, which is homologous to the v-Maf oncogene product (16). The Maf-Nrl subfamily has been implicated in cell type-specific regulation of genes in tissues as diverse as the hematopoietic system (17,18), cerebellum (19), and developing hindbrain (20 -22). Maf and Nrl proteins bind an extended AP-1-like sequence, and form both homodimers and heterodimers with other bZIP family members (23)(24)(25)(26)(27). Nrl itself is specifically expressed in retinal cells, including photoreceptors, in which its expression precedes that of rhodopsin during development (28). Recently, we have shown by electrophoretic mobility shift analysis that Nrl can bind in vitro to an oligomer containing the Nrl response element (NRE) present in the rhodopsin promoter, and that cotransfection of CV-1 monkey kidney cells with an Nrl-expression vector can stimulate expression of rho-dopsin promoter-reporter fusion constructs (29).
In this report we provide more direct evidence supporting a role for Nrl in the regulation of rhodopsin gene expression. In the first part of the study, using the yeast one-hybrid system, we demonstrate that Nrl can bind to the NRE in vivo and regulate expression of a reporter gene. We then use transient expression analysis of primary chick retinal cultures (30) to further analyze Nrl, the NRE, and related cis-elements in the rhodopsin promoter. The cells in these primary cultures, unlike those in transformed lines, express a highly differentiated phenotype resembling closely their in vivo counterparts (31,32). Using transfection and cotransfection studies, we show that Nrl can activate rhodopsin expression in retinal cells in an NRE-dependent manner and identify a positive regulatory region that is located just upstream of the NRE and acts synergistically with it.

EXPERIMENTAL PROCEDURES
Yeast One-hybrid Screening-A retinal cDNA fusion library for onehybrid screening was generated from poly(A) ϩ bovine retinal RNA. Oligo(dT)-primed cDNA was directionally inserted into the EcoRI and XhoI sites of plasmid pACTII (Steve Elledge, Baylor College of Medicine) just downstream of the GAL4 activation domain. The ligation products were electroporated into Escherichia coli strain DH10B and plated onto LB/carbenicillin plates. The average transformation efficiency was 10 10 transformants/g of DNA. Approximately 2 ϫ 10 6 independent clones were obtained. Testing of individual colonies revealed that the average insert size was 2 kilobases. Bacteria from the plates was harvested by scraping, and DNA was prepared using Qiagen Maxi columns according to the manufacturer's directions. The resulting plasmid library was amplified twice before use. Bait constructs containing tetramers of the Ϫ73to Ϫ30-bp sequence from the bovine rhodopsin promoter in the vectors pHISi-1 and pLacZ were prepared (Clontech). The pHISi-1 and pLacZ bait constructs were then linearized with XhoI and NcoI, respectively, and integrated into the genome of YM4271. The resulting strain with the integrated pHISi-1-bait construct was used for library screening according to Clontech's protocol. Positive library clones were defined as those that, on retransformation, led to growth of pHISi-1-bait cells on His Ϫ medium containing 60 mM 3-amino-1,2,4-triazole and generated a blue color on 5-bromo-4-chloro-3-indoyl-␤-D-galactopyranoside-containing plates on transformation of cells harboring the pLacZ-bait construct. Library plasmid DNA minipreparations were prepared using standard methods (33).
Preparation of Constructs-Reporter constructs were generated by cloning appropriate DNA fragments into the multiple cloning site of the pGL2-basic luciferase-containing plasmid (Promega, Madison, WI). For the generation of bRho-2174, gBR200pLacF (34) was digested with Asp-718 and BamHI, and the appropriate gel-purified fragment was ligated into pGL2 predigested with Asp-718 and BglII. To generate mRho-4907, the plasmid gMR4-4, containing the mouse rhodopsin upstream sequence (34), was digested with SalI, filled in with Klenow, and digested with XhoI, and the appropriate gel-purified fragment was ligated into pGL2 that had been digested with NheI, filled in, and then digested with XhoI. For mRho-1609, -1488, and -270, mRho-4907 was digested with Asp-718 and NheI, MscI, and ApaI, respectively, and then filled in and recircularized with ligase. The bovine and murine proximal element deletion constructs were generated by polymerase chain reaction amplification of appropriate regions of bRho-2174 and mRho-4907, respectively; for each construct, a unique upstream primer containing an XhoI site and a common downstream primer in the luciferase gene (GLprimer 2; Promega) were used. The polymerase chain reaction products were then digested with XhoI and HindIII and ligated into pGL2 predigested with XhoI and HindIII. hRho-85 was generated similarly, using gJHN7 (35) as a template for the polymerase chain reaction. The construction of pMT-NRL was described previously (29). The human collagenase promoter construct and c-Fos and c-Jun expression vectors (36) were gifts from Dr. Daniel Nathans (The Johns Hopkins University); CMV-lacZ was a gift from Dr. Kuan Teh-Jeang (NIAID); and RSV-Luc was a gift from Dr. Mark Showers (Harvard University).
DNA for transfection was obtained using the alkaline lysis protocol and double banding through CsCl 2 (37). The supercoiled DNA was precipitated twice with ethanol, resuspended in 2-4 ml of TE (10 mM Tris-HCl, pH 7.4, 1 mM EDTA) and filtered through a Microcon-100 column (Amicon, Beverly, MA).
Transfections-Embryonic day 8 (E8) and E17 chick retinal cultures were prepared as described previously (32,38). Photoreceptors in these cultures develop and maintain a polarized phenotype, form outer segment processes, express a number of photoreceptor-specific genes, including rhodopsin, undergo photomechanical movements in response to light, and respond to neurotransmitters and retinoids (31,32). Conditions for transient expression were optimized using E8 retinal cultures. Pilot studies revealed that the calcium phosphate method yielded higher transfection efficiencies than transfection with Transfectam (Promega), TransfectACE (Life Technologies, Inc.), Lipofectin (Life Technologies), LipofectAMINE (Life Technologies), or calcium phosphate with glycerol shock. The calcium phosphate procedure was then optimized with respect to the amount of reporter and carrier DNA, the volume of calcium phosphate precipitate, cell density, the number of days in vitro prior to transfection, and the interval between transfection and harvesting (data not shown). These studies resulted in the protocol described below. Using this protocol and a Rous sarcoma virus long terminal repeat-luciferase reporter construct, transfection efficiencies as high as 15% could be obtained (data not shown). E8 cultures were transfected after 8 days in vitro. E17 cultures were transfected after 3-5 days in vitro. DNA in 80 l of solution I (0.25 M CaCl 2 ) was added dropwise to 80 l of solution II (25 mM HEPES, pH 7.1, 140 mM NaCl, and 14 mM Na 2 HPO 4 ) while vortexing at a slow speed. The mixture was incubated at room temperature for 15 min and then added dropwise to the appropriate cell culture dish. For cotransfection assays, all plasmids were added to solution I before mixing with solution II. mRho-4907 was transfected at 9 g of DNA/plate; for all other reporter constructs, equivalent molar amounts of DNA were used. Each transfection included 1 g of CMV-lacZ as an internal control, to measure and correct for differences in transfection efficiency. Transfections were performed in duplicate during each experiment, and all experiments were done three independent times.
Luciferase Assays-Cell lysates were prepared 40 -48 h after transfection by scraping the cells from individual plates into 200 l of lysis buffer following the manufacturer's instructions (Promega). The lysates were either used immediately for luciferase and ␤-galactosidase assays or stored on ice for a few hours. Twenty l of cell lysate was mixed with 50 l of reconstituted luciferin (Promega). Luciferase activity was measured with a TD-20e luminometer (Turner Designs, Inc., Mountain View, CA) with an integration time of 10 s.
␤-Galactosidase Assays-Fifty l of cell lysate was mixed with an equal volume of 8 mg/ml chlorophenyl red-␤-D-galactopyranoside (Boehringer Mannheim) in duplicate in a 96-well plate format, and reactions were measured using an EL312e enzyme-linked immunosorbent assay reader (Bio-Tek Instruments, Inc., Winooski, VT). Serial dilution of a ␤-galactosidase enzyme standard on the same plate was used to determine the linearity of the reaction and to calculate concentrations based on the rate of reaction (Kineticalc software; Bio-Tek Instruments). The resulting ␤-galactosidase values were used to correct the luciferase values for transfection efficiency.

RESULTS
Nrl Binds to Chromosomal DNA Containing the NRE-Sequence comparison of the mouse (34), rat (10), Chinese hamster (39), cow (40), and human (35) rhodopsin proximal promoter regions demonstrate that the NRE and flanking DNA are highly conserved (Fig. 1). The NRE itself, spanning the bovine sequence between Ϫ68 and Ϫ56 bp, is conserved in 12 of its 13 nucleotide positions. We had previously shown that an oligomer pair containing the rhodopsin NRE can bind in vitro to proteins in a bovine retinal nuclear extract and that Nrl is one of the proteins in the bound complex (29). To identify transcription factors that can bind to the rhodopsin NRE and its flanking DNA in vivo, in the context of chromatin structure, we used the yeast one-hybrid system (19,41,42). A bovine retinal cDNA-GAL4 activation domain fusion library was generated and transformed into yeast cells containing as bait a stably integrated tetramer of the bovine rhodopsin promoter sequence from Ϫ73 to Ϫ30 bp cloned upstream of a HIS3 reporter construct. A total of 1.7 ϫ 10 5 yeast transformants were screened. Thirty-one positive colonies that grew on His Ϫ plates were identified in the first round. On retransformation, plasmids from 11 of these 31 clones were still positive with both the HIS3 and lacZ reporter assays. Sequence analysis revealed that seven of the clones contained full-length Nrl open reading frames, whereas three others contained 5Ј-Nrl truncations. The 11th clone was highly homologous to aspartate aminotransferase and probably represented a false positive. In a second round of library screening, involving 3.2 ϫ 10 6 transformants, of 44 confirmed positives that have been characterized, 42 represent Nrl clones. The smallest 5Ј-truncation, present in four identical clones, encodes a fusion protein in which the Nrl sequence begins at residue 129 (see Fig. 2), demonstrating that residues 129 -236 are sufficient for DNA binding. Fig. 2 shows the predicted amino acid sequence of the bovine Nrl protein, as determined by the sequence of one of the fulllength library clones and its comparison with the previously determined human and murine Nrl sequences (16,43). The bovine sequence is 92 and 85% identical to the human and murine sequences, respectively. Interestingly, the majority of the sequence differences are present in the immediate carboxylterminal region, just beyond the leucine zipper motif.
NRE Acts as a Positive Regulatory Element in Retinal Cells-Procedures for transient expression analysis using primary chick retinal cultures were developed (see "Experimental Procedures") to examine whether Nrl can trans-activate rhodopsin promoter activity in retinal cells and to explore the ability of DNA elements flanking the NRE to modulate Nrl activity. To determine possible differences in promoter activity associated with the differentiation stage of the retinal cells, transfections were performed with chick retinal cultures derived from both E8 and E17. E8-derived cultures are free of glial and retinal pigment epithelial cells. They contain a mixture of photoreceptors and nonphotoreceptor neurons; the photoreceptors, which constitute approximately 10 -20% of the cell population, differentiate largely in vitro (31,44). In E17 cultures, photoreceptors represent 30 -60% of the population, and many aspects of their differentiation occur in vivo, prior to their isolation for culture.
Chick retinal cells were transfected with a series of luciferase fusion constructs containing DNA from the region upstream of the bovine (bRho) and murine (mRho) rhodopsin. (The nomenclature used for reporter constructs includes abbreviations for the species and gene name, followed by a number that refers to the position of the 5Ј-end of the construct relative to the mRNA start site.) The bRho-2174, bRho-225, bRho-176, mRho-4907, mRho-1609, mRho-1488, and mRho-270 constructs demonstrated similar levels of activity, ranging between 66 and 107 relative light units (Fig. 3A). These results demonstrate that the bovine and murine rhodopsin proximal promoter regions, consisting of less than 300 bp of upstream DNA, are as active in chick retinal cultures as longer constructs containing the RER. (The RER spans the region from Ϫ2044 to Ϫ1943 and from Ϫ1575 to Ϫ1477 bp in the cow and mouse genes, respectively (9).) To look specifically at the activity of the NRE and flanking DNA, the additional constructs bRho-38, bRho-84, bRho-130, mRho-40, mRho-84, mRho-130, mRho-176, and mRho-222 were generated. As shown in Fig. 3, B and C, the bovine (murine) constructs showed evidence of positive regulatory elements between Ϫ38 (Ϫ40) and Ϫ84 (Ϫ85) bp, which contain the NRE, and between Ϫ84 (Ϫ85) and Ϫ130 bp. The activity of the Ϫ84to Ϫ130-bp region, particularly for the bovine constructs, is significantly greater than the NRE-containing region. The results also suggest the possibility of a negative element located between Ϫ130 and Ϫ176 bp. The importance of the NRE within the Ϫ38to Ϫ84-bp sequence was confirmed by the finding that bRho-225(m62-66), which contains a mutated NRE, demonstrated significantly reduced reporter activity (Fig. 3B).
Nrl Activates Expression of Transfected Rhodopsin Promoters via the NRE-To further investigate the functional role of Nrl in rhodopsin regulation, E8 and E17 retinal cell cultures  were cotransfected with bRho constructs together with an expression vector containing the human NRL cDNA (pMT-NRL). Cotransfection of pMT-NRL resulted in induction of reporter gene activity with constructs that contained an intact NRE but had little effect with constructs that lacked the NRE or contained a mutated NRE, in both E8 (Fig. 4A) and E17 (Fig. 4B) cells. In E8 cultures, but not in E17 cultures, the Rho-130 constructs showed substantially greater stimulation with pMT-NRL cotransfection than did the Rho-84 constructs, even though there are no NRE-like sequences between Ϫ84 and Ϫ130 bp. This suggested that in the E8 cultures there are factors that can interact with sequences in the Ϫ84to Ϫ130-bp region and increase NRL-mediated trans-activation. The human reporter construct hRho-85, which contains a homologous NRE, also showed approximately 3-fold stimulation on cotransfection with pMT-NRL (Fig. 4B), as did analogous murine rhodopsin constructs (data not shown).
Nrl, c-Fos, and c-Jun Differentially Trans-activate the Rhodopsin Promoter-Since the NRE is an extended AP-1 element, which binds to c-Fos-c-Jun heterodimers, we investigated whether c-Fos and c-Jun could also trans-activate the rhodopsin promoter. As shown in Fig. 5A, cotransfected Nrl, c-Fos, and c-Jun expression plasmids stimulated reporter gene activity from the bRho-130 construct by 3.2-, 1.9-, and 0.9-fold, respectively. When the expression constructs were cotransfected as pairs, Nrl-c-Fos and Nrl-c-Jun were both more potent than c-Fos-c-Jun. In contrast, in control experiments with a collagenase reporter construct (hCol-71), which contains a canonical AP-1 site but lacks an NRE, the maximal stimulation was seen with c-Fos, and Nrl was only minimally effective (Fig. 5B). Taken together, these results indicate that in primary chick retinal cells Nrl preferentially trans-activates the rhodopsin NRE compared with c-Fos and c-Jun. The results are also consistent with in vitro studies that have shown that the central cytidine in the consensus Maf binding site (TGCTGACT-CAGCA) is important in determining the specificity of DNA interaction with bZIP proteins (25). Maf, Jun homodimers, and Fos-Jun heterodimers were all found to bind equally well to the consensus sequence; however, when the central cytidine nucleotide was altered (TGCTGATTCAGCA), as occurs in the rhodopsin NRE, Maf still bound strongly to the oligomer, whereas the binding of both Jun homodimers and Fos-Jun heterodimers was abolished.

DISCUSSION
Using the yeast one-hybrid system together with transient expression analysis of primary chick retinal cultures, we have obtained evidence supporting an important in vivo role for the bZIP transcription factor Nrl in the regulation of rhodopsin expression. Independent one-hybrid assays consistently demonstrated that Nrl is the predominant gene obtained using the rhodopsin Ϫ73to Ϫ30-bp sequence as bait, thus suggesting that Nrl may indeed be the primary transcription factor bind- FIG. 3   FIG. 3. Expression of murine and bovine rhodopsin constructs in E17 cultures. E17 chick retinal cultures were transfected with equal molar amounts of the indicated constructs by the calcium phosphate method as described under "Experimental Procedures." The constructs in A were designed to assess the activity of the RER and other upstream regions, whereas the bovine and murine constructs shown in B and C, respectively, were designed to dissect the rhodopsin proximal promoter region (RPPR). In B, the construct bRho225(m62-66), which contains a mutated NRE in which the central 5 bp (CTGAT) have been mutated to GTCGAG, is also included. Relative light units (R.L.U.) were corrected for transfection efficiency with a ␤-galactosidase internal control and represent the ratio of the activity of each construct relative to the activity of the promotorless construct pGL2-basic. The values are means of three independent experiments performed in duplicate. Bars, 1 S.E.
ing to this region in vivo. The observation that Nrl can bind to the NRE and flanking DNA in the one-hybrid system adds to our previous demonstration that Nrl can bind to the rhodopsin NRE in an in vitro eletrophoretic mobility shift assay (29) by demonstrating that the interaction can take place in an in vivo situation in the context of chromatin structure, which is often more stringent than naked DNA. In addition, since in some of the full-length constructs the Nrl coding region is out of frame with the GAL4 activation domain, the activation of reporter gene expression is presumably due to the activity of the Nrl trans-activation domain itself. In an analogous analysis of c-Maf, it was similarly found that the exogenous activation domain was not needed when the full-length coding region was used (19).
The one-hybrid analysis also provided a deletion analysis of the regions of Nrl required for DNA binding. The finding that the smallest 5Ј-truncations obtained in the screen corre-sponded to amino acid residues 129 -236 is consistent with the structure of Nrl and previous biochemical studies with the Maf-Nrl family, which indicate that in addition to the basic region and leucine zipper motifs, a highly conserved sequence just amino-terminal to the basic region (extended homology domain; see Fig. 2) is also required for specific DNA binding (19,23,24,45). Consistent with the model that the transactivating activity is associated with the amino-terminal part of the molecule, the Nrl coding region in the deletions was in frame with the GAL4 activation domain.
In a complementary set of studies, transient expression analyses were carried out to explore the rhodopsin promoter regulatory interactions in the context of retinal cells. Primary chick retinal cell cultures were chosen for these studies, because they allow transfection efficiencies as high as 15% and contain photoreceptor cells that exhibit many of the molecular, structural, and functional properties of their in vivo counterparts (31). These characteristics are particularly appealing, since they allow analysis of the results of transient expression studies in the context of other differentiated cell behaviors, which is seldom possible using transformed cell lines. Mouse retinal cells would theoretically provide another attractive system for The data are represented as -fold stimulation as in Fig. 6. Bars, S.E. transfection analysis, since they also undergo extensive differentiation in culture (46); however, pilot studies with mouse cells demonstrated transfection efficiencies that were so low as to preclude meaningful promoter analysis.
The deletion analysis did not demonstrate any activity associated with the rhodopsin RER. Although the reason for this is unclear, it is perhaps not surprising, given that presently there is no evidence to suggest the existence of a chick version of the RER. Additionally, enhancer regions often require chromatin structure for their activity. Related to this, the RER does not show activity with a bovine in vitro transcription system that also relies on naked DNA templates (13).
The deletion analysis did provide evidence of the presence of cis-acting elements located between Ϫ130 and Ϫ40 bp, a region that is highly conserved among the mouse, rat, Chinese hamster, cow, and human rhodopsin genes. Within this region, two separate domains could be defined, one between Ϫ40 and Ϫ84 bp and another between Ϫ85 and Ϫ130 bp, that contain positive regulatory elements. A third domain, located between Ϫ130 and Ϫ176 bp, may possess a negative regulatory element. At least part of the activity of the Ϫ40 to Ϫ84 bp region derives from the NRE present between Ϫ56 and Ϫ68 bp. The Ϫ85 to Ϫ130 bp region, which contains part of the rhodopsin ret1-PCE1 element (10), is particularly active, stronger than the NRE-containing region. It appears to act synergistically with the NRE because it leads to an increase in the trans-activating activity of co-transfected Nrl, even though it lacks a binding site for Nrl. The synergism could be mediated by a regulatory protein that binds to a sequence between Ϫ85 and Ϫ130 bp and stabilizes the interaction between Nrl and the NRE or interacts with bound Nrl and increases its trans-activating activity. Alternatively, Nrl could independently stimulate the synthesis or activity of another protein that binds to and acts directly through the Ϫ85 to Ϫ130 bp sequence. Regardless of its nature, the synergistic activity appears to be developmentally regulated, since it is seen with E8 but not with E17 cultures (Fig. 4,  A and B). Future studies will better define the mode of interaction and regulatory element(s) within the Ϫ85 to Ϫ130 bp region.
The experiments presented here indicate that Nrl can stimulate rhodopsin expression in an NRE-dependent manner in yeast cells and chick retinal cell cultures, respectively, but they do not by themselves demonstrate that Nrl plays a similar role in vivo. Several lines of evidence, however, are consistent with this possibility and favor the involvement of Nrl compared with other bZIP proteins: 1) in the adult, Nrl is specifically expressed in the retina, including photoreceptors (16); 2) Nrl expression precedes rhodopsin expression during development (28); 3) the position and sequence of the rhodopsin NRE is evolutionarily conserved; 4) supershift experiments with retinal nuclear extracts indicate that the activity that binds to a rhodopsin NRE oligomer contains Nrl, or an immunologically similar molecule, but not Fos or Jun (29); 5) although Fos and Jun have been shown to be expressed in the retina, they have not been identified in photoreceptors (47,48); 6) Nrl is more specific than either Fos or Jun for the rhodopsin promoter; and 7) neither Fos nor Jun was detected in the one-hybrid screen.
It is also clear that Nrl by itself is not sufficient to turn on rhodopsin expression in vivo. Rhodopsin expression is limited to rod photoreceptors, but Nrl is expressed in a number of different retinal and neuronal cell types (16). Additionally, retinoblastoma cell lines express Nrl but not rhodopsin (16,49). Like other genes that are regulated in a cell type-and developmentally restricted manner, control of rhodopsin expression is likely to be mediated by a combinatorial array of transcription factors, some of which are retina-and/or cell type-specific, whereas others are more ubiquitously expressed (2). On the other hand, Nrl may also be involved in the regulation of other photoreceptor-specific genes that contain NRE-like sites, such as the human red and green opsin (50) and bovine interphotoreceptor retinoid-binding protein genes, 3 and nonphotoreceptor genes, such as the quail gene QR1, a developmentally restricted gene expressed in retinal Mü ller cells (51). Future experiments will be necessary to define more precisely this apparently complex role of Nrl in vivo.