The Upstream Region of the Rpe65 Gene Confers Retinal Pigment Epithelium-specific Expression in Vivo and in Vitro and Contains Critical Octamer and E-box Binding Sites*

RPE65 is essential for all-trans- to 11-cis-retinoid isomerization, the hallmark reaction of the retinal pigment epithelium (RPE). Here, we identify regulatory elements in the Rpe65 gene and demonstrate their functional relevance to Rpe65 gene expression. We show that the 5′ flanking region of the mouse Rpe65 gene, like the human gene, lacks a canonical TATA box and consensus GC and CAAT boxes. The mouse and human genes do share several cis-acting elements, including an octamer, a nuclear factor one (NFI) site, and two E-box sites, suggesting a conserved mode of regulation. A mouseRpe65 promoter/β-galactosidase transgene containing bases –655 to +52 (TR4) of the mouse 5′ flanking region was sufficient to direct high RPE-specific expression in transgenic mice, whereas shorter fragments (–297 to +52 or –188 to +52) generated only background activity. Furthermore, transient transfection of analogous TR4/luciferase constructs also directed high reporter activity in the human RPE cell line D407 but weak activity in the non-RPE cell lines HeLa, HepG2, and HS27. Functional binding of potential transcription factors to the octamer sequence, AP-4, and NFI sites was demonstrated by directed mutagenesis, electrophoretic mobility shift assay, and cross-linking. Mutations of these sites abolished binding and corresponding transcriptional activity and indicated that octamer and E-box transcription factors synergistically regulate the RPE65 promoter function. Thus, we have identified the regulatory region in theRpe65 gene that accounts for tissue-specific expression in the RPE and found that octamer and E-box transcription factors play a critical role in the transcriptional regulation of theRpe65 gene.

RPE65 is essential for all-trans-to 11-cis-retinoid isomerization, the hallmark reaction of the retinal pigment epithelium (RPE). Here, we identify regulatory elements in the Rpe65 gene and demonstrate their functional relevance to Rpe65 gene expression. We show that the 5 flanking region of the mouse Rpe65 gene, like the human gene, lacks a canonical TATA box and consensus GC and CAAT boxes. The mouse and human genes do share several cis-acting elements, including an octamer, a nuclear factor one (NFI) site, and two E-box sites, suggesting a conserved mode of regulation. A mouse Rpe65 promoter/␤-galactosidase transgene containing bases -655 to ؉52 (TR4) of the mouse 5 flanking region was sufficient to direct high RPE-specific expression in transgenic mice, whereas shorter fragments (-297 to ؉52 or -188 to ؉52) generated only background activity. Furthermore, transient transfection of analogous TR4/luciferase constructs also directed high reporter activity in the human RPE cell line D407 but weak activity in the non-RPE cell lines HeLa, HepG2, and HS27. Functional binding of potential transcription factors to the octamer sequence, AP-4, and NFI sites was demonstrated by directed mutagenesis, electrophoretic mobility shift assay, and cross-linking. Mutations of these sites abolished binding and corresponding transcriptional activity and indicated that octamer and E-box transcription factors synergistically regulate the RPE65 promoter function. Thus, we have identified the regulatory region in the Rpe65 gene that accounts for tissue-specific expression in the RPE and found that octamer and E-box transcription factors play a critical role in the transcriptional regulation of the Rpe65 gene.
All-trans-to 11-cis-isomerization of vitamin A is an obligate and tissue-specific enzymatic step in the renewal of 11-cisretinal, the universal chromophore of rhodopsin and other visual pigment proteins, in the visual cycle (1) of the retinal pigment epithelium (RPE). 1 Several components, including 11-cis-retinol dehydrogenase (2), cellular 11-cis-retinaldehydebinding protein (CRALBP) (3,4) and lecithin:retinol acyltransferase (5,6), all essential to the visual cycle activity, are found highly expressed, but not exclusively expressed, in the RPE. However, the retinol isomerase activity (7)(8)(9), central to 11-cischromophore synthesis, is expected, mechanistically, to be highly tissue-specific. A tissue-specific component of the RPE, RPE65 (10 -12), which copurifies with 11-cis-retinol dehydrogenase (2), appears to play a crucial role in retinoid isomerization. Thus, in the Rpe65-deficient mouse (13), rod photoreceptor function is abolished due to lack of the 11-cis-retinal chromophore. Furthermore, mutations in the human RPE65 gene cause several forms of severe early onset blindness (14 -17). Clearly, RPE65 is essential to the visual cycle in general and to all-trans-to 11-cis-retinoid isomerization in particular.
RPE65 is the major protein of the RPE microsomal membrane fraction. The bovine (10), human (18), dog (19), rat (20), and salamander (21) cDNAs have been cloned, as have the human (18) and mouse genes. 2 RPE65 is specific to the vertebrate RPE and is also highly conserved at the level of protein sequence. Previous data suggest a complex transcriptional and translational regulation of RPE65. At the transcriptional level, our knowledge is limited (22), and we lack functional evidence concerning the transcriptional elements involved in the activation of the gene and in its specific expression in the RPE. Transcription of Rpe65 appears to be developmentally regulated, with the protein first appearing at about postnatal day 4 in the rat (11), coincident with the first appearance of the photoreceptor outer segments. Reverse transcription-polymerase chain reaction (reverse transcription-PCR) analysis of RPE65 in embryonic and newborn rat suggests a biphasic induction of RPE65 mRNA expression (20). At the level of translation, we have found that a 170-nucleotide region of the RPE65 3Ј untranslated region acts as a translational inhibition element (23). Also, when RPE cells are explanted into culture, they lose expression of RPE65 protein within 2 weeks, although the expression of RPE65 mRNA can continue (10,11).
Here, we present the sequence of the 5Ј flanking region of the mouse Rpe65 gene and indicate its similarity to the corresponding human gene region. We have generated transgenic mice containing Rpe65 promoter-reporter constructs and show that the Rpe65 5Јflanking region -655 to ϩ52 can drive lacZ reporter gene expression specifically in the RPE. In addition, we * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF271297.
‡ To whom correspondence should be addressed: NEI-LRCMB, National Institutes of Health, Bldg. 6 show that this fragment also displays a high transcriptional activity in D407 RPE cells in vitro. Furthermore, by directed mutagenesis, electrophoretic mobility shift assay (EMSA), and cross-linking, we demonstrate functional binding of transcription factors to an octamer sequence and AP-4 and NFI sites and show their importance to the transcriptional regulation of the mouse Rpe65 gene.

EXPERIMENTAL PROCEDURES
DNA Cloning and Sequence Analysis-A P1 clone containing the complete mouse Rpe65 gene was isolated (Genomic Systems, St. Louis, MO). Restriction fragments containing the 5Ј region of the mouse Rpe65 gene were identified by Southern blot hybridization to a random-primed 32 P-labeled bovine cDNA 5Ј end probe (10). pBluescript subclones containing the 5Ј region of the Rpe65 gene were sequenced. One such clone, E1-12, was found to contain the first three exons of mouse Rpe65, as well as 2.8 kilobase pairs of 5Ј flanking region. This was compared with the sequence of the 5Ј flanking region of the human RPE65 gene, obtained in the same way (15), using the GeneWorks 2.5 and MacVector 6.5 sequence analysis programs (Oxford Molecular, Beaverton, OR).
For transient transfection luciferase assay, constructs TR3 and TR4 were inserted into the plasmid pGL3-Basic (Promega, Madison, WI). The forward primers were the same as those used to amplify TR3 and TR4 fragments during the production of transgenic mice. The common reverse primer also overlapped with the primer used before, but it lacked four bases at the 3Ј end.
Site-directed Mutagenesis of the Mouse Rpe65 Promoter-Mutations were introduced by DNA amplification using QuickChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA). A total of four sitedirected mutants using the pTR4luc plasmid as a template were generated. These were named m1AP4, mNFI, m2AP4, and mOct. 11 additional constructs, as well, containing combinatorial mutations in two, three, or all of the cited elements were created using single, double, or triple mutants as a template, respectively. Mutated oligonucleotides used for DNA amplification are shown in Table II. The introduction of mutations was verified by DNA sequencing.
Production and Analysis of Transgenic Mice-DNA constructs were microinjected into the male pronuclei of single cell FVB/N mouse embryos, which were implanted into pseudopregnant CD1 foster mothers, using standard techniques. Transgenic founder mice and their progeny were identified by PCR of a region common to all of the transgenes. For some founders, copy number was estimated by Southern blot analysis of PstI-digested genomic DNA hybridized with a lacZ gene probe. Transgenic founders were bred to CD1 mice to generate F1 progeny. ␤-Galactosidase (␤-gal) reporter gene activity was assayed using the chemiluminescent Galacto-Light Plus assay (Tropix/PE Applied Biosystems, Bedford, MA). Eyes were dissected into three parts: the anterior segment, comprising the cornea, iris, and ciliary body; the posterior segment, comprising the retina, RPE, choroid, and sclera; and the lens. Noneye tissues assayed were brain, liver, lung, heart, kidney, and spleen. For histochemistry, tissues were fixed for 1 h in 4% paraformaldehyde in phosphate-buffered saline and washed three times in 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (X-gal) rinse buffer (100 mM sodium phosphate, pH 7.3, 2 mM MgCl 2 , 0.01% sodium deoxycholate, 0.02% Nonidet P-40). The eyes were stained overnight in a solution of 2 mg/ml X-gal in X-gal rinse buffer containing 5 mM each potassium ferrocyanide and potassium ferricyanide. After staining, tissues were postfixed in 4% paraformaldehyde and embedded in methacrylate. Sections were cut at a thickness of 5 m, counterstained with neutral red, and evaluated for presence of blue product. In addition, stained eyes were postfixed in 4% paraformaldehyde and dissected to remove anterior segment, lens, and retina. The resultant eyecup was quartered and flat-mounted in 50% glycerol for an en face preparation of the RPE/choroid/sclera complex.
Cell Culture and Transient Transfections-The human retinal pig-ment epithelium cell line D407 was obtained from Richard C. Hunt (25) and grown in high glucose Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 3% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM glutamine. HeLa, HepG2, and HS27 cells were obtained from American Type Culture Collection (Manassas, VA) and grown in the same medium as used for D407 except that the concentration of fetal bovine serum was 10%. Approximately 2.5 ϫ 10 5 cells were plated onto six-well tissue culture dishes and allowed to grow for 48 -72 h (until 80 -90% confluent). To correct for differences in transfection efficiency, 2 g of each luciferase plasmid and 90 ng of pSV40/␤-gal were added to the cells in a solution containing Superfect transfectant (Qiagen, Chatsworth, CA). Luciferase and ␤-gal reporter gene activities were assayed using the Dual-Light reporter gene assay (Tropix/PE Applied Biosystems, Bedford, MA). The ratio of luciferase activity to ␤-gal activity in each sample served as a measure of normalized luciferase activity. Experiments were performed in triplicate at least four times.
EMSA-Nuclear extracts from D407 and freshly dissected RPE bovine cells were prepared by the method of Dignam et al. (26). For EMSA, double-stranded oligonucleotides (Table II) were labeled with polynucleotide kinase and [␥-32 P]dATP (6000 Ci/mmol). Approximately 15 g of nuclear extract were added to binding buffer (33 mM Tris-HCl, pH 7.5, 166 mM NaCl, 1.6 mM dithiothreitol), 4 g of poly(dI-dC), 0.04% Nonidet P-40, 8% glycerol, and 32 P-labeled probe (30,000 -50,000 cpm) and incubated at room temperature 30 min. For the competition assay, a 50 -2000-fold molar excess of unlabeled wild type, mutant, or nonspecific competitor oligonucleotide was used along with the labeled probe. The DNA-protein complexes were resolved on 5% polyacrylamide gels in 0.5ϫ Tris borate-EDTA buffer and visualized by autoradiography. For antibody supershifts, nuclear extracts were incubated with 1 l of CTF/NFI polyclonal antibody (provided by Naoko Tanese) or 4 l of ␣-Oct-1 monoclonal antibody (provided by Winship Herr, isotype IgG1, ) for 1 h at room temperature prior to addition of labeled probe.
Corresponding preimmune serum was used as a control in the NFI supershift assay. Medium containing Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics and an Oct-2 antibody (provided by Winship Herr) with the same isotype as the ␣-Oct-1 antibody were used as controls in the supershift assays.
Cross-linking-For cross-linking experiments, thymidines were substituted by 5-bromo-deoxyuridine in one oligonucleotide strand. After incubation with the probe in the same conditions described above, samples were cross-linked for 10 min under UV radiation in a Stratalinker. For molecular weight determinations, samples were electrophoresed on a 12% polyacrylamide-Tris-HCl gel (Bio-Rad) in 1ϫ Trisglycine-SDS buffer.

Sequence Conservation of the Mouse and Human 5Ј Flanking
Regions-We have sequenced approximately 2.8 kilobase pairs of 5Ј flanking region upstream of the putative transcription start site of the mouse Rpe65 gene ( Fig. 1A; numbered ϩ1 based on homology with the human gene (22)) and searched for sequences evolutionarily conserved between the mouse and human Rpe65/RPE65 genes 5Ј flanking regions. When 2.8 kilobase pairs of the 5Ј flanking region of the mouse and human Rpe65/RPE65 genes were compared by dot matrix analysis, a diagonal of similarity was seen distally to approximately -1200 with a short region of similarity closer to the 5Ј end of each segment (data not shown). The proximal 628 and 581 nucleotides of the human and mouse 5Ј flanking regions, respectively, and 5Ј untranslated regions were compared by ClustalW alignment (Fig. 1B). Many conserved blocks of sequence were noted between the two, including NFI (at -178 to -165), octamer (at -498 to -491), and two E-box consensus (at -84 to -79 and -257 to -252) binding sites. The overall homology is over 70%. A possible site homologous to the human CRALBP gene (27) element, noted by Nicoletti et al. (22) in the human RPE65 5Ј flanking region is present at -72 to -63 in the mouse Rpe65 5Ј flanking region. This element, however, is not represented in the mouse CRALBP gene promoter. The NFI site in both genes has a similar intervening nonconsensus sequence (AAA(T/C)A) only seen previously in the human RPE65 gene (22) (22), the mouse Rpe65 transcriptional start site is numbered ϩ1. A nonconsensus possible TATA box is underlined and labeled (?TATA). Various consensus binding sites (AP-4, NFI, and octamer) are underlined and labeled. The 5Ј ends of fragments used to generate transgene constructs (TR2, TR3, and TR4) are indicated by asterisks. Each fragment is indicated by an underlined boldface designation (TR2-TR4). This sequence has been submitted to GenBank TM (accession number AF271297). This sequence has been scanned against the GenBank data base (July 1999), and the only sequence with significant relatedness identified was the 5Ј untranslated region of the rat RPE65 cDNA (AF035673). B, optimal alignment of mouse and human Rpe65/RPE65 proximal promoter sequences. Conserved nucleotides are boxed and shaded, and consensus-binding elements for transcription factors are indicated. Putative Ret1/PCE1, TATA box, and human CRALBP, as well as AP-4, NFI, and octamer gene elements are also indicated. consensus binding half-sites (TGGA-N5-GCCA) match perfectly with the CTF/NFI family consensus. The two E-box sites are consensus basic helix-loop-helix (HLH) protein binding sites, although Nicoletti et al. (22) specifically ascribed them to AP-4. An octamer sequence was also identified in both promoters; they differed by only one nucleotide from the consensus octamer sequence ATGCAAAT. Both human and mouse genes lack consensus GC and CAAT boxes. A possible TATA box was identified at -27 to -20 in the human gene (22), although, as in the mouse gene, it is somewhat deviated from consensus (28,29). In addition, sequences similar to motifs found in other retina-specific genes, including interphotoreceptor retinoidbinding protein (IRBP) (30) and CRALBP (27), were identified. Although similar sequences were found in the human RPE65 gene proximal promoter (22), it is not yet clear what role, if any, these play in the overall activity of the Rpe65 gene.
Analysis of Rpe65 Promoter-driven LacZ Activity in Transgenic Mouse Tissues-To identify sequence elements and transcriptional factors responsible of Rpe65 expression, we first determined the minimal Rpe65 promoter sequence necessary for the in vivo specific expression of the Rpe65 gene in the RPE. We made three constructs containing the upstream sequence of the Rpe65 gene coupled to the lacZ gene/SV40 poly(A) signal sequence, and we analyzed their corresponding ␤-gal activity in transgenic mice. These contained the following sequence positions (construct name in parentheses): -188 to ϩ52 (TR2), -297 to ϩ52 (TR3), and -655 to ϩ52 (TR4). Microinjection of these constructs into fertilized oocytes resulted in the generation of several founder lines for each construct. Copy number was estimated by comparing the hybridization signal of probe with genomic DNA from transgenic mice to serial dilutions of a known quantity of the relevant linearized construct. The number of founders analyzed and the number of copies and ␤-gal expression levels of each construct are presented in Table I. Although RPE65 has been shown to be expressed specifically in the RPE of the eye and is not found in nonocular tissues (11), we surveyed expression of RPE65 promoter-lacZ gene constructs in a variety of tissues. We assayed eye tissues (the anterior segment, comprising the cornea, iris and ciliary body; the posterior segment, comprising the retina, RPE, choroid, and sclera; and the lens) and noneye tissues (brain, heart, lung, liver, kidney, and spleen) from transgenic and nontransgenic F1 or F2 littermates for ␤-gal activity. We found that the constructs pTR2lacZ (-188/ϩ52, not shown) and pTR3lacZ (-297/ϩ52) produced only background ␤-gal activity in control or transgenic animals in all tissues assayed ( Fig. 2A). However, the longest construct pTR4lacZ (-655/ϩ52) exhibited an average of about 15-fold higher ␤-gal activity than the control in the posterior segment, but no significant difference in other ocular and nonocular tissues (Fig. 2B). The values obtained for TR3 (progeny of founder 12) are shown in Fig. 2A, and those obtained for TR4 (progeny of founder 3) are shown in Fig. 2B. Concerning TR4, very similar values were obtained for F1 progeny of founder 6 and 184, but founder 190 had values 75% lower than these.
Discernible staining was seen only in the eyes of pTR4lacZ (-655/ϩ52) transgenic animals (founders 3, 6, and 184). At the gross level, staining of the whole eyes revealed a punctate pattern (Fig. 3A). In sections of these eyes examined by light microscopy, the blue X-gal product was seen to be restricted in its distribution to the RPE cells of transgenic mice (Fig. 3C), whereas none was present in nontransgenic littermates (Fig. 3,  B and D). No staining of lens, anterior segment, or neural retina was observed in transgenic and nontransgenic animals (data not shown). Staining of the RPE was patchy, however, and this was best appreciated by en face light microscopy of transgenic RPE/choroid/sclera flat mount (Fig. 3E). Again, no staining was present in nontransgenic littermate controls (Fig.  3F). Although most, if not all, cells of the RPE demonstrated some level of X-gal staining in the cytoplasm, about 15% of cells were much more highly stained and filled with blue product.
Analyses of fixed TR4 noneye tissues stained with X-gal did not reveal any detectable staining except for the founder 6 progeny, which showed an ectopic ␤-gal expression in the cerebellum. RPE65 is not expressed in brain (11) or cerebellum. 3 Analysis of the Rpe65 Promoter-driven Luciferase Activity in Vitro-To better understand the transcriptional regulation of the Rpe65 gene, we searched for a cellular model capable of activation of the mouse Rpe65 promoter. Although only traces of RPE65 mRNA are detected by PCR in nonconfluent cultures of the human RPE cell D407 (data not shown), it has been demonstrated that these cells are able to activate a human RPE65 promoter (22). Thus, to test its activity and specificity in vitro, the promoter fragment TR4 (-655 to ϩ 52) was cloned 3 T. M. Redmond, unpublished data.  into the luciferase reporter vector pGL3-Basic and transfected into D407 cells. To show the cellular specificity of the vector, it was also transfected into the non-RPE cell lines HeLa, HepG2, and HS27. None of these latter cell lines or the tissues from which they are derived (cervix, liver, and skin, respectively) express the RPE65 mRNA.
Our results show that luciferase activity generated by pTR3luc was very low and was similar in all of the cell lines tested. In HeLa, HepG2, and HS27, pTR4luc activity was also similar to that produced by pTR3luc (2.6, 2.2, and 2.6 times higher, respectively, than luciferase activity generated by pGL3-Basic alone). In contrast, in D407 cells, pTR4luc generated activity 10 -15 times higher than that generated by control plasmid alone (Fig. 4).
Mutational Analysis of the Rpe65 Promoter-To identify functionally important elements in the Rpe65 promoter, mutated derivatives of the TR4 promoter fragment (-655 to ϩ52) cloned into pGL3-Basic (pTR4luc) were constructed and transfected into D407 cells (oligonucleotides used for directed mutagenesis are shown in Table II). The mutation (TGGAAAAT-AGCCAA f TGGAAAATATAAAA) introduced into the NFI site (31) (located at position -178 to -165) had only a minor positive effect on promoter activity (Fig. 5A). In contrast, mutations of the core sequence of the two potential AP-4 binding sites, 1AP-4 (TCAGCTGAGG f TCCATCGAA) and 2AP-4 (TCAGCTCAGG f TCATTAATT), located at positions -84 to -79 and -257 to -252, reduced promoter activity by 67 and 60%, respectively (Fig. 5A). In addition, mutations of the octamer sequence (ATGCAAAG f CCACACCA), located at position -498 to -491, reduced promoter activity by 67%.
Because none of these individual mutations completely abolished promoter activity, we next constructed combination mutants. The double combination of mutations in the 1AP-4 and 2AP-4 or octamer site had a greater effect than either mutation alone (Fig. 5B), dramatically reducing promoter activity by about 78%, although the double mutant retained measurable promoter activity (3.5-fold the level of vector alone). Interestingly, a vector containing mutated 1AP-4, 2AP-4, and octamer sites further lowered luciferase activity (85% reduction compared with the wild type vector; Fig. 5C). This suggested that the E-box and octamer sequences are critical Rpe65 promoter elements. Finally, combinations that included the mutated NFI site together with an octamer mutation increased the luciferase activity obtained, compared with the values when this site was normal (Fig. 5, B and C).
Specific Interactions of D407 and Bovine-RPE Nuclear Proteins with Elements in the Rpe65 Promoter-To determine whether transcription factors binding to the potential NFI, octamer, and AP-4 sites in the Rpe65 promoter were present in vivo and in vitro, we performed EMSA using bovine-RPE nuclear proteins from freshly dissected bovine RPE tissue and D407 cells. A pattern consisting of three specific complexes was observed when either bovine RPE or D407 nuclear extracts were incubated with a labeled probe containing the NFI site. Each of these three complexes showed a similar mobility between the two nuclear extracts, and these were inhibited by addition of a 370-fold molar excess of cold competitor (Figs. 6B and 7B). The constructs pTR3luc and pTR4luc were transiently transfected into D407, HeLa, HepG2, and HS27 cell lines, and luciferase reporter gene activity was assayed. The activities of the luciferase reporter gene were expressed as fold relative to the activity of pGL3-Basic (which was assigned an activity value of 1.0). The means and S.E. are shown (n Ն 4).
of cold competitor but not in the presence of a similar excess of mutated or nonspecific competitor.
Incubation of D407 nuclear extract with the Oct-1 probe resulted in the formation of two complexes. The lower molecular weight complex was competed away by a 250-fold molar excess of cold wild type competitor but not by cold mutant or irrelevant competitor, whereas the higher molecular weight complex was not inhibited by cold mutant competitor and was judged to be nonspecific (Fig. 7D). Only one complex was observed when the bovine-RPE nuclear extract was incubated with the Oct-1 probe (Fig. 6D); this complex migrated with the same mobility as the smaller one observed with the D407 nuclear extract. No complex was seen when any of the labeled mutant oligonucleotides were incubated with either of the nuclear extracts. These results suggest that these sites bind proteins involved in the regulation of the Rpe65 promoter.
Transcription Factors CTF/NFI and ␣-Oct-1 Bind to the RPE65 Promoter-Antibodies against CTF/NFI and ␣-Oct-1 were used to characterize the transcription factors involved in the binding of the D407 nuclear extracts with the NFI and Oct-1 probe, respectively. The polyclonal CTF/NFI antibody completely supershifted the three complexes seen with NFI in the absence of antibody (Fig. 8A, panel 1). However, although a supershifted band was also detected in the presence of the monoclonal ␣-Oct-1 antibody and the corresponding probe (Fig.  8A, panel 2), most of the complex was not supershifted, indicating that some other factor(s) may also be implicated in the binding to the octamer site. Alternatively, this result may be due to the restrictive epitope recognition of monoclonal anti-bodies. Supershift was absent when the samples were incubated with the corresponding preimmune serum (Fig. 8A, panel  1) or media containing 10% fetal bovine serum (Fig. 8A, panel  2), respectively, for each probe. A similar negative result was observed when an Oct-2 antibody with the same isotype as the ␣-Oct-1 antibody was used as a control (data not shown).
The lack of available AP-4 antiserum precluded confirmation of AP-4 as a component of the complexes identified by EMSA with the two AP-4 probes. However, we performed UV crosslinking assays using the 1AP-4 probe and a D407 nuclear extract in order to determine the molecular weight of the complexes observed by EMSA. A higher band of 120 kDa and a lower band of 50 kDa were observed with both AP-4 probes (Fig. 8B). Both of the bands were competed away with a 2000fold molar excess of cold probe but not with the same excess of an irrelevant competitor. These sizes correspond to those observed before for AP-4 by Mermod et al. (32). DISCUSSION RPE65 is, to date, the only known RPE-specific component of the RPE-specific all-trans-to 11-cis-retinoid isomerization mechanism known as the visual cycle. To understand the mechanism of this tissue specificity, we have characterized the Rpe65 promoter in vivo and in vitro. In this paper, we show that the -655 to ϩ52 region of the mouse Rpe65 promoter confers tissue-specific expression in vivo and in vitro, and we define elements within this sequence that are crucial to this expression.
Our first goal was to identify a region of the 5Ј flanking  5. Effect of mutations on mouse Rpe65 promoter activity. Putative cis-acting elements identified within the Rpe65 promoter are indicated, as are the locations of the different mutations (ϫ). D407 cells were transfected with pTR4luc (WT) or mutated pTR4luc (m) constructs. In all panels, the promoter activity of the WT and the constructs containing single mutations (A), double mutations (B), and triple or quadruple mutations (m4) (C) are shown. The activities of the luciferase reporter gene were expressed as fold relative to the activity of pGL3-Basic (which was assigned an activity value of 1.0). The means and S.E. are shown (n Ն 4). region of the Rpe65 gene that reliably conferred in vivo RPEspecific expression. Our results from transgenic mice showed that the upstream region of the Rpe65 gene (-655 to ϩ52) confers RPE-specific expression in adult animals, with no expression in nonocular tissues. ␤-Gal expression was also observed in mice at postnatal day 4 (data not shown), correlating with RPE65 expression (11). Histology of adult mouse eyes showed a patchy nature of the RPE-specific expression. This kind of nonhomogeneous spotty staining pattern has been observed before in other transgenic mouse tissues (33). Although ectopic expression was observed in cerebellum of progeny of founder 6, this was likely a result of integration of the transgene into or next to a cerebellum-expressed gene.
Shorter fragments (-297 to ϩ52 and -188 to ϩ52) did not confer activity in transgenic animals, indicating that a crucial element(s) is present in the -655 to -297 interval. It is interesting that although the mRNA for RPE65 has been detected by reverse transcription-PCR in salamander cone photoreceptors (21), we detected no staining of retinal cells, including cone photoreceptors.
Because transgenic animal production is a long-term procedure and unsuited to testing the effects of multiple mutations, we used an RPE cell line to perform in vitro experiments. However, expression levels of the Rpe65 gene are low or nonexistent in RPE cell lines (34) 4  elements necessary for this activation are present. A very low level of RPE65 mRNA was detectable by reverse transcription-PCR in 80% confluent cultures of the human RPE cell line D407 (not shown). Nicoletti et al. (22) reported that D407 is the RPE cell line that showed the highest transcriptional activity when transfected with a reporter human RPE65 promoterluciferase vector. Accordingly, we used this cell line for transfection experiments to determine the minimal promoter sequence and to study the transcriptional elements necessary to induce luciferase expression. Our results showed that, as in vivo, the TR4 fragment induces the highest levels of luciferase expression, indicating that most, if not all, of the positive elements regulating the Rpe65 promoter activity in the RPE are indeed located in the -655 to ϩ52 bp promoter sequence. Consequently, transcription factors binding to the AP-4, NFI, and octamer elements appeared to be obvious candidates for the Rpe65 specific gene expression in the RPE.
Specific and similar binding was detected by EMSA with nuclear extracts obtained either from freshly dissected bovine RPE cells or from D407, indicating that the same nuclear proteins may be involved in the Rpe65 gene expression both in vivo and in vitro. We found that mutations performed both in the AP-4 sites and in the octamer sequence completely disrupted the binding of nuclear proteins to these sites. In addition, these mutations introduced into the pTR4luc vector reduced the transcriptional activity of these potential elements. Although mutations of the upstream and downstream AP-4 sites separately reduced Rpe65 promoter activity by 67 and 60%, respectively, it was diminished by 78% by mutation of both together. Concurrent mutation of the octamer sequence further decreased transcriptional activity. This suggests a synergistic positive action of the transcription factors binding to these three sites in the transcriptional regulation of the Rpe65 gene.
Synergism of HLH and octamer binding proteins with themselves, between HLH and octamer binding proteins, and even of each with other proteins is consistent with previous studies. HLH proteins can act in concert with other HLH proteins to activate transcription. For instance, late transcription of SV40 activated by AP-4 is augmented by the addition of transcription factor AP-1 (32). Also, although MyoD can bind single sites, it must bind to multiple sites or adjacent to other transcription factors to activate muscle-specific genes (35,36). Recently, tissue specific expression of the tyrosine hydroxylase gene has been shown to involve synergy between an HLH motif and an adjacent AP-1 site (37). In addition, the rat insulin gene is apparently transactivated by synergism between the Pan HLH factor and lmx-1, a homeodomain-containing protein (38). In the case of octamer factors, Oct-2 has been shown to bind cooperatively to adjacent heptamer and octamer sites to synergistically activate immunoglobin-chain gene promoters (39).
Numerous other examples of synergy of octamer factors with other transcription factors exist (40 -42).
Given that Oct-1 can participate in tissue-specific expression (43)(44)(45)(46)(47), Oct-1 and AP-4 are potentially the major determinants of Rpe65 promoter activity in RPE cells, both in vivo and in vitro. Their presence alone, however, might not account for the specificity of the promoter, because Oct-1 is ubiquitously expressed and AP-4 is expressed relatively widely. Thus, one possibility is that a threshold level of one of these is required for activation of the promoter. Another possibility is that the octamer sequence is recognized by other octamer factors, because it has been shown that Oct-1 and Oct-2 factors can bind to the same octamer sequence (39). In addition, cell-specific enhancers that contain octamer binding sites tend to bind Oct-1 weakly (48). Furthermore, it has been shown that a cervical cell-specific octamer factor binds a nonconsensus octamer site more strongly than the octamer consensus (49). Thus, it is tempting to speculate that octamer motifs in the Rpe65 promoter may represent the binding site of an RPE-restricted factor, active alone or as a trans-activator by binding to a ubiquitous octamer factor as Oct-1. This might account for the weak supershift observed in the presence of ␣-Oct-1 monoclonal antibody. Finally, it is conceivable that RPE65 expression in non-RPE cells is repressed by a negative-acting factor. One example of such a factor is neuronal restrictive silencer factor/RE1 (represor element 1)-silencing transcription factor (NRSF/REST), required for repression of multiple neuronal target genes in nonneuronal cells (50,51).
The role of NFI in the Rpe65 transcriptional regulation is not clear. Our results indicated that NFI specifically binds to the Rpe65 promoter and also that luciferase reporter activity increases when the NFI site is mutated in combination with the octamer mutation. This might suggest a weakly negative effect of NFI in Rpe65 gene expression.
In summary, we have established that the proximal promoter region of the mouse Rpe65 gene can direct RPE-specific expression of a ␤-gal reporter construct in transgenic mice. We have also found that HLH and octamer-binding factors can synergistically regulate the Rpe65 gene and that mutations in these elements abolish transcriptional activity and prevent binding of the corresponding proteins. An NFI element is also involved but might act in a negative context. Although the regulation of Rpe65 gene expression provides a paradigm of tissue specific gene regulation, further investigation will be necessary to understand these mechanisms fully. In particular, identification and characterization of transcription factors binding to the transcriptional elements described is required. Such experiments are under way.