Analysis of the VMD2 promoter and implication of E-box binding factors in its regulation.

The retinal pigment epithelium (RPE) is crucial for the normal development and function of retinal photo-receptors, and mutations in several genes that are preferentially expressed in the RPE have been shown to cause retinal degeneration. We analyzed the 5'-up-stream region of human VMD2, a gene that is preferentially expressed in the RPE and, when mutated, causes Best macular dystrophy. Transgenic mouse studies with VMD2 promoter/lacZ constructs demonstrated that a-253 to +38 bp fragment is sufficient to direct RPE-specific expression in the eye. Transient transfection assays using the D407 human RPE cell line with VMD2 promoter/luciferase reporter constructs identified two positive regulatory regions, -585 to -541 bp for high level expression and -56 to -42 bp for low level expression. Mutation of a canonical E-box located in the -56 to -42 bp region greatly diminished luciferase expression in D407 cells and abolished the bands shifted with bovine RPE nuclear extract in electrophoretic mobility shift assays. Independently a candidate approach was used to select microphthalmia-associated transcription factor (MITF) for testing because it is expressed in the RPE and associated with RPE abnormalities when mutated. MITF-M significantly increased luciferase expression in D407 cells in an E-box-dependent manner. These studies define the VMD2 promoter region sufficient to drive RPE-specific expression in the eye, identify positive regulatory regions in vitro, and suggest that MITF as well as other E-box binding factors may act as positive regulators of VMD2 expression.

The retinal pigment epithelium (RPE) 1 is a monolayer of cuboidal cells located between the photoreceptors and choroid of the eye. It has many specialized functions that support and nourish photoreceptors, including important roles in retinoid metabolism (visual cycle), phagocytosis of shed photoreceptor outer segments, maintenance of the blood-retina barrier, movement of ions and water, and synthesis and transport of substances that constitute the interphotoreceptor matrix (1). The importance of the RPE in maintaining retinal photoreceptors is highlighted by the Royal College of Surgeons rat that exhibits a markedly reduced capacity for phagocytosis of outer segments by the RPE that results in the degeneration and loss of photoreceptor cells (2,3). In addition, in humans, RPE dysfunction has been implicated in the pathogenesis of age-related macular degeneration, which is the leading cause of irreversible blindness in elderly people in western countries (4,5).
Mutations in several genes that are specifically or preferentially expressed in the RPE, such as RPE65, RLBP1, RGR, TIMP3, and VMD2, are associated with inherited human retinal dystrophies (6 -15). Mutations in VMD2 result in Best disease (vitelliform macular dystrophy (VMD)), an autosomal dominant, juvenile onset macular dystrophy characterized by a striking accumulation of lipofuscin-like material within and beneath the RPE (16 -18). VMD2 encodes a multispan transmembrane protein, bestrophin, that is expressed preferentially in the RPE and functions as an oligomeric chloride channel that is thought to be responsible for the characteristic abnormality observed in the electrooculogram in patients with Best disease (19,20).
Despite the key role of the RPE in vision and the importance of the genes that are specifically or preferentially expressed in the RPE, relatively few studies have focused on the regulation of RPE gene expression. Perhaps the best studied model of RPE gene regulation to date has been the RPE65 gene (21,22), which is specifically expressed in the RPE and cone photoreceptors (23). A combination of transgenic and cell culture experiments have identified a murine promoter region sufficient for RPE-specific expression and suggested that octamer and E-box binding transcription factors may play important regulatory roles (21). The human RPE65 5Ј-upstream region has also been studied using sequence analyses, cell transfection assays, and DNase I footprinting (24). Analysis of the 5Ј-up-stream region of the human cellular 11-cis-retinaldehyde-binding protein gene (RLBP1), which is preferentially expressed in RPE and Muller cells, suggested that a photoreceptor consensus element-1 (PCE-1, CAATTAG) located in the proximal promoter region might act as a positive regulator of RPE gene expression (25). This is interesting because PCE-1 was generally considered a photoreceptor element, and a sequence similar to PCE-1 is also present within the proximal promoter region of both human and murine Rpe65; however, its biological relevance has not yet been experimentally tested. Promoter regions have also been analyzed for genes that are expressed in the RPE but are not RPE-specific. For example, transgenic studies demonstrated that 270 bp of the murine tyrosinase gene (Tyr) upstream region is sufficient to direct cell typespecific expression and developmental regulation in melanocytes and the RPE (26). At the biochemical level, the human TYR promoter can be bound in vitro by microphthalmia-associated transcription factor (MITF) with binding mediated by an M-box, which contains a core CATGTG E-box motif (27,28). Consistent with this binding, MITF is sufficient to direct pigment cell-specific transcription of TYR (28,29). MITF is a member of the basic helix-loop-helix leucine zipper family of transcription factors and expressed in several cell lineages including melanocytes, RPE, osteoclasts, and mast cells (27, 30 -34). Mice with mutations in Mitf (Mitf mi /Mitf mi ) show loss of pigmentation, microphthalmia, and early onset of deafness. The RPE in these mice forms a multilayered structure resembling the neural retina, suggesting a critical role of MITF in RPE development (35)(36)(37)(38).
To gain further insights into the molecular mechanisms mediating RPE-specific gene regulation, we analyzed VMD2 as a model system. Using a combination of transgenic and transient transfection approaches, we found that a fragment as small as Ϫ253 to ϩ38 bp is sufficient to direct RPE-specific expression in the eye, and two regions, Ϫ585 to Ϫ541 bp and Ϫ56 to Ϫ42 bp, are important for high level and low level expression in vitro, respectively. We also present evidence that MITF as well as other E-box binding factors may act as positive regulators of VMD2 expression.

EXPERIMENTAL PROCEDURES
Cloning, Sequencing, and Comparison of the 5Ј-Upstream Region of Human and Murine Vmd2-A P1 human genomic library (Genome Systems, St. Louis, MO) was screened with a 32 P-labeled bovine Vmd2 cDNA clone according to the company's instructions. P1 plasmid DNAs were purified from purchased Escherichia coli clones and analyzed by Southern hybridization to confirm that the clones contained the 5Јflanking region of VMD2. A 129/SvJ mouse genomic library in Lambda FIX II vector (Stratagene, La Jolla, CA), which was a kind gift from Dr. Se-Jin Lee (The Johns Hopkins University), was screened with the same bovine Vmd2 cDNA probe according to standard methods (39). Both P1 and phage DNAs were directly sequenced using the Thermo Sequenase cycle sequencing kit (USB Corp., Cleveland, OH) with 2.5% dimethyl sulfoxide added in the reactions. Later the 5Ј-flanking sequences of both human and murine Vmd2 were also obtained from the GenBank TM at National Center for Biotechnology Information (accession number NT_033903 for human and NT_039687 for mouse).
Computer programs, GeneWorks 2.5 (Oxford Molecular Group, Beaverton, OR) and Vector NTI version 7 (Invitrogen) were used for sequence alignment. To look for the presence of known transcription factor binding sites, MatInspector professional version 6.0 was used with the database of Matrix Family Library version 3.0 (Genomatix, Munich, Germany) (40).
Primer Extension-Human donor eyes were obtained from eye banks through the National Disease Research Interchange. Eyes were dissected equatorially, the retina was removed, RPE cells were collected by gentle scraping, and total RNA was extracted using TRIzol reagent (Invitrogen) (41).
Primer extension was performed according to standard procedures (42) using 10 g of human RPE total RNA as template with two primers complementary to VMD2 mRNA sequence. The locations of Primer A (5Ј-TAAGTGATGGTCATGGCCAGGCAGTGG-3Ј) and Primer B (5Ј-AGGTGGGGTTCCAGGTGGGTCCGATGATCC-3Ј) are shown in Fig.  1B. The end-labeled primers were annealed to RNA at 65°C for 90 min and extended using Thermoscript reverse transcriptase (Invitrogen) at 65°C for 1 h.
To construct an expression vector, a human MITF-M cDNA was generated by reverse transcription (RT)-PCR using human RPE total RNA as template with a forward primer containing an EcoRI site (5Ј-ACTGAATTCATTGTTATGCTGGAAATGCTAGA-3Ј) and a reverse primer containing a HindIII site (5Ј-AGAAAGCTTGAACAAGTGT-GCTCCGTCTCTTC-3Ј). The cDNA fragment was then inserted into EcoRI/HindIII sites downstream of CMV promoter in pcDNA3.1(Ϫ)/ Myc-His B vector (Invitrogen).
Generation of Transgenic Mice-The transgenic constructs were microinjected into mouse one-cell embryos (B6/SJL F2 hybrid) at the Transgenic Mouse Core Facility of The Johns Hopkins University School of Medicine as described previously (43). Mouse pups were screened to determine positive transgenic founders by both Southern blot analysis and PCR of tail DNAs. Ten micrograms of mouse tail DNAs were digested with BamHI and hybridized with a 32 P-labeled 3.1-kb lacZ fragment according to standard procedures for Southern blotting (39). Primers for PCR were forward 5Ј-ACATCAGCCGCTA-CAGTCAA-3Ј and reverse 5Ј-GCGAGATGCTCTTGAAGTCT-3Ј. Transgenic founders were mated with wild-type BALB/cJ mice (The Jackson Laboratory, Bar Harbor, ME) to generate progeny with albino background.
Histology, X-Gal Staining, Flat Mount, and RT-PCR Analysis for Transgenic Mice-Mice were euthanized, eyes were enucleated, and the whole eyes were fixed at 4°C for 1 h in 2% paraformaldehyde and 0.25% glutaraldehyde in phosphate-buffered saline (PBS) for eye sections or in 0.5% glutaraldehyde in PBS for RPE/choroid flat mounts. For histochemical staining with 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal), eyes were then cryoprotected in 25% sucrose in PBS at 4°C for 12-18 h, mounted in OCT medium (Sakura Finetek, Torrance, CA), and cut at 10 m on a cryostat. Sections were stained in 1 mg/ml X-gal, 5 mM K 3 Fe(CN) 6 , 5 mM K 4 Fe(CN) 6 ⅐3H 2 O, 1 mM MgCl 2 in PBS at 37°C for 12-48 h (43). Sections of non-transgenic mice were also stained by hematoxylin and eosin. For RPE/choroid flat mounts, fixed whole eyes were washed in PBS and stained in the X-gal solution at 37°C for 24 -48 h. The stained eyes were cut at the equatorial zone, the anterior portion was removed, the retina was removed, and the RPE/choroid eye cup was cut from the periphery by a pair of microsurgical scissors.
To check the expression of lacZ reporter in other tissues, RT-PCR was performed using 1 g of total RNA extracted by TRIzol from liver, brain, kidney, spleen, testis, and eye of each founder. First strand cDNA was synthesized with oligo(dT) primer and Superscript II reverse transcriptase (Invitrogen), and PCR was performed using 35 cycles with the FIG. 1. The 5-upstream region of human VMD2. A, transcription start site of the VMD2 determined by primer extension. The extension products obtained using 10 g of total RNA extracted from human RPE cells with two primers complementary to VMD2 mRNA sequence were analyzed next to a DNA sequence ladder. The locations of Primer A and Primer B are shown in B. B, nucleotide sequence of the 5Ј-upstream region of human VMD2. The transcription start site determined by primer extension is numbered ϩ1, and the numbers on the left are the nucleotide positions relative to it. Intron 1 is shown by dotted line, and its sequence (1343 bp) is not presented. Ebox, NF1/CTF1, octamer, CEBP, PCE-1like, and Crx/Otx consensus binding sites are underlined and labeled. Three EMSA probes that cover the Ϫ585 to Ϫ541 bp region (Probes 23, 24, and 25) and the probe that contains E-box 1 (Probe Ϫ55/ Ϫ34) are indicated by dotted underlines. The locations and directions of Primer A and Primer B used for primer extension are shown by arrows (5Ј 3 3Ј). The initiation ATG is also underlined. same primer set as used for the tail DNA screening. As a control for RNA and RT reaction, a cDNA fragment of ribosomal protein S16 was also amplified using forward primer 5Ј-CACTGCAAACGGGGAAATG-G-3Ј and reverse primer 5Ј-TGAGATGGACTGTCGGATGG-3Ј.
Cell Cultures and Transient Transfections-Nine human cell cultures were screened for the expression of VMD2 by RT-PCR using 1 g of total RNA, and a 269-bp VMD2 fragment was amplified using forward primer 5Ј-CATAGACACCAAAGACAAAAGC-3Ј and reverse primer 5Ј-GTGCTTCATCCCTGTTTTCC-3Ј. As a control, S16 was used as described above. Based on these results, three human RPE cell lines, D407 (44), ARPE19 (45), and telomerase-immortalized hTERT-RPE1 (46) (Clontech, Palo Alto, CA), and a human neuroepithelioma cell line, SK-N-MC (47), were selected. Each cell line was cultured in the medium suggested in the references. Cells were transfected at 40 -48 h after plating into 60-mm dishes at 70 -80% confluency using LipofectAMINE PLUS reagent (Invitrogen). Plasmid DNA for each dish included 9 g of a luciferase construct and 1 g of pCMV-lacZ as an internal control for transfection efficiency. For co-transfection studies, 0, 0.5, or 2.5 g of a human MITF-M expression vector were added to each DNA mixture, and the total amount of expression vector was adjusted to 2.5 g by adding pcDNA3.1 vector when necessary. Transfections were performed six independent times in duplicate each time. Cell lysates were prepared at 48 -60 h after transfection using 300 l of Reporter Lysis Buffer (Promega). Luciferase and ␤-galactosidase activities were measured as described previously (48) except that Softmax Pro 2.6.1 program with SpectraMax Plus 96-well plate reader (Molecular Devices, Sunnyvale, CA) was used for ␤-galactosidase assay.
Two oligonucleotide probes were made corresponding to Ϫ55 to Ϫ34 bp (Fig. 4A, underlined), one with wild-type sequence and the other with mutation of E-box 1 (CACGTG to ACCGTA). For cold oligomer competition experiments, 1-, 10-, and 100-fold molar excess of unlabeled annealed oligonucleotides, wild-type or E-box 1 mutation, were added to binding reactions containing 0.5 g of poly(dI-dC), and the mixtures were kept on ice for 10 min before adding 32 P-labeled wild-type probe.

Cloning and Sequence Comparison of the 5Ј-Upstream Region of Human and Murine
Vmd2-Genomic clones corresponding to human and murine Vmd2 were obtained as described under "Experimental Procedures." The transcription start site for VMD2 was determined as 111 bp upstream from the initiation ATG by 5Ј-rapid amplification of cDNA ends using human RPE RNA (data not shown). This result was confirmed by primer extension using two independent primers (Fig. 1A).
The 5Ј-upstream regions of human and murine Vmd2 were sequenced 2.3 and 5.7 kb from the transcription start site, respectively. Within 2.3 kb of the 5Ј-flanking region of human VMD2, two repetitive sequences are present: Alu at Ϫ2391 to Ϫ2113 bp and L1 at Ϫ1471 to Ϫ1280 bp. In murine Vmd2, the B1 repeat is located at Ϫ396 to Ϫ267 bp (assuming the transcription start site based on homology to the human gene). Sequence alignment of these 5Ј-flanking regions of the two species demonstrated many blocks of conserved sequences par-ticularly in three regions: Ϫ1579 to Ϫ1523, Ϫ664 to Ϫ530, and Ϫ244 bp to the 5Ј-untranslated region (data not shown).
The VMD2 5Ј-upstream region contains several consensus transcription factor binding sites, including three E-box, E-box 1 (Ϫ47 to Ϫ42 bp), E-box 2 (Ϫ69 to Ϫ64 bp), and E-box 3 (Ϫ847 to Ϫ842 bp); NF1/CTF1 (Ϫ782 to Ϫ769 bp); two octamer (Ϫ420 to Ϫ413 and Ϫ591 to Ϫ584 bp); CEBP (Ϫ565 to Ϫ556 bp); and PCE-1-like (Ϫ174 to Ϫ168 bp) binding sites (Fig. 1B). Four Crx/Otx-like homeodomain protein binding sites are also present at Ϫ81 to Ϫ76, Ϫ127 to Ϫ122, Ϫ1088 to Ϫ1083, and Ϫ1553 to Ϫ1548 bp. Although many blocks of conserved sequences are seen in the 5Ј-upstream region between the two species, some of the above mentioned consensus sequences are not found in murine Vmd2. For instance, among three E-box sites found in VMD2, only E-box 2 is conserved in the murine gene, and neither the NF1/CTF1, CEBP, nor one octamer (Ϫ420 to Ϫ413 bp) binding site is well conserved. In contrast, a PCE-1-like sequence is identified in both species, and it differs by only one nucleotide from the consensus sequence CAATTAG. Interestingly all three E-box sites, of which a consensus is CANNTG, are flanked by 5Ј ANT residues, creating ANTCANNTG, similar to the M-box in which a core CATGTG motif is flanked by specific 5Ј residues. M-box is found in the promoter region of pigment-related genes, such as Tyr, tyrosinase-related protein-1 (Tyrp-1), and tyrosinase-related protein-2 (Tyrp-2) (27). Consensus TATA and CAAT boxes are not present at typical positions, Ϫ20 to Ϫ30 and Ϫ50 to Ϫ130 bp, respectively, in either human or murine gene.
Definition of a VMD2 Promoter Region Sufficient to Drive RPE-specific Expression in the Eye Using Transgenic Mice-To identify a VMD2 promoter region sufficient to direct RPEspecific expression in vivo, transgenic mice were generated with constructs containing various 5Ј-upstream regions of the gene fused to a lacZ reporter (VMD2 promoter/lacZ) ( Fig. 2A). ␤-Galactosidase activity was analyzed by X-gal staining of eye sections and RPE/choroid flat mounts, and ocular and extraocular RNA expression was assayed by RT-PCR. The transgene was expressed in the RPE in two of three, two of five, four of seven, and five of six lines with Constructs 1, 2, 3, and 4, respectively. Expression within the eye was RPE-specific for all expressing lines except one of the Construct 4 lines that showed weak expression in the inner nuclear and ganglion cell layers in addition to the RPE. None of the RPE-negative lines showed ectopic expression elsewhere in the eye. Although there were variations in the staining patterns, all positive RPE/choroid flat mounts demonstrated to some degree a spotty pattern of transgene expression (Fig. 2B) in some ways similar to the patchy pattern we had noted previously with rhodopsin promoter/lacZ transgenic mice (43). Interestingly X-gal staining was observed more intensely and densely in the central RPE but excluded from the peripheral RPE close to the ciliary body. Histochemical X-gal staining of eye sections from these positive lines also showed a patchy pattern of staining (Fig. 2C). RT-PCR analysis for lacZ expression in different tissues revealed that transgenic lines that showed strong X-gal staining in the RPE had a tendency to have a more tissue-restricted pattern in the eye, brain, and testis (data not shown). Conversely transgenic lines that had weak or no X-gal staining in the RPE yielded RT-PCR products in a less tissue-restricted manner. There were substantial variations in the levels and patterns of lacZ expression among different lines of the same constructs, indicating a strong effect of insertion sites of the transgene. Nonetheless, since endogenous VMD2 is also expressed weakly in brain and testis in addition to the RPE (14), several transgenic lines seemed to mimic the expression pattern of endogenous VMD2. These transgenic results indicate that a fragment luciferase constructs containing the same 5Ј-upstream fragments as used for transgenic mouse studies and pGL2-Basic vector as background control were transiently transfected into the indicated cell lines. Luciferase activities were normalized by ␤-galactosidase activities, and relative luciferase activities were calculated as a ratio of the normalized luciferase activity with constructs containing VMD2 promoter fragments to that with pGL2-Basic vector. The values represent the means and S.D. (bar), but in some cases, the S.D. was too small to be shown by bars. C, Transfection II. Seven luciferase constructs containing different lengths of the 5Ј-upstream fragments as well as the three constructs used for Transfection I and pGL2-Basic vector were transiently transfected into D407 and SK-N-MC cells. Relative luciferase activities were calculated and presented as described in B. The mean values for the construct containing the Ϫ41 to ϩ38 bp fragment were too small to be shown in both cell lines. D, Transfection III. To further analyze the Ϫ104 to Ϫ42 bp region, three more luciferase constructs as well as the two constructs used for Transfection II and pGL2-Basic vector were transiently transfected into D407 and SK-N-MC cells. Relative luciferase activities were calculated and presented as described in B.
as small as Ϫ253 to ϩ38 bp is sufficient to direct RPE-specific expression in the eye. This promoter fragment should be useful for RPE-specific delivery systems utilizing transgenic mice and viral vectors. The positive VMD2 promoter/lacZ transgenic lines may be useful for RPE transplantation studies in that lacZ can be used as a marker of donor RPE cells.
Definition of Regulatory Regions within the VMD2 Promoter-To define the regulatory elements responsible for VMD2 expression in the RPE, transient transfection analysis with cultured human cells was undertaken. We first assessed a variety of cell lines for expression of endogenous VMD2 since it seemed likely that cells expressing the endogenous gene would more closely approximate the in vivo VMD2 transcriptional machinery and therefore provide biologically more meaningful results. Three RPE (D407, ARPE19, and hTERT-RPE1), three neuroblastoma (IMR-32, SK-N-AS, and SK-N-DZ), a neuroepithelioma (SK-N-MC), and a transformed primary embryonic kidney (293) cell line as well as primary human RPE cell cultures were screened by RT-PCR (data not shown). Based on these results, D407, ARPE19, hTERT-RPE1, and SK-N-MC, which express VMD2 at medium, low, no, and high levels, respectively, were chosen as host cells (Fig. 3A).
Fourteen constructs containing fragments of VMD2 5Ј-upstream region fused to a luciferase reporter gene (VMD2 promoter/luciferase) were tested in three sequential sets of experiments. The first set of experiments was carried out with constructs containing the four fragments as used for transgenic mouse studies (Fig. 3B). The promoter activities in four cell lines were well correlated overall with the levels of endogenous VMD2 expression analyzed by RT-PCR. Although in SK-N-MC the fragments of Ϫ2948 to ϩ38 and Ϫ585 to ϩ38 bp showed similar promoter activities, in D407 the activity of Ϫ2948 to ϩ38 bp fragment was approximately one-third of that of Ϫ585 to ϩ38 bp. In both D407 and SK-N-MC, deletion of the region between Ϫ585 and Ϫ425 bp resulted in an 80 -85% decrease in the promoter activities compared with that with Ϫ585 to ϩ38 bp. The construct containing Ϫ253 to ϩ38 bp still directed luciferase expression 12-fold higher than the background by pGL2-Basic vector in D407 cells. After these experiments, we decided to use only D407 and SK-N-MC to narrow down the regions responsible for the promoter activities (Fig. 3C). Deletion of the region between Ϫ585 and Ϫ541 bp decreased the promoter activity down to ϳ39 and ϳ17% of that with Ϫ585 to ϩ38 bp in D407 and SK-N-MC, respectively. The fragment Ϫ104 to ϩ38 bp still retained the activity to drive luciferase expression about 30 times higher than the background by pGL2-Basic vector in both cell lines. Interestingly the construct containing the fragment Ϫ41 to ϩ38 bp consistently demonstrated a very low luciferase activity of only 10 -20% of the background in both cell lines, indicating that the region between Ϫ104 and Ϫ42 bp contains elements contributing to a 150 -300-fold difference in the promoter activities (Fig. 3C). Therefore, the third set of experiments was performed to further narrow down the regulatory sequences in this Ϫ104 to Ϫ42 bp region (Fig. 3D). The fragment Ϫ56 to ϩ38 bp still induced luciferase activity about 10 times higher than that by pGL2-Basic vector, indicating that the region between Ϫ56 and Ϫ42 bp is responsible for the promoter activity that is 50 -100fold higher than that of the Ϫ41 to ϩ38 bp fragment in both cell lines. These results suggest that basal promoter elements exist in the region Ϫ56 to ϩ38 bp, and positive regulatory elements that are important for VMD2 expression are located in two regions at Ϫ585 to Ϫ541 bp for high level expression and at Ϫ56 to Ϫ42 bp for low level expression.
A Proximal E-box Is Important for VMD2 Promoter Activity-Since the Ϫ56 to Ϫ42 bp regulatory region contains a

FIG. 4. Effect of mutation in E-box elements on VMD2 promoter activity.
A, sequence of the VMD2 proximal promoter containing two E-box elements. The transcription start site (TS, numbered ϩ1) is indicated by an angled arrow. Two E-box sites, E-box 1 and E-box 2, are highlighted in boldface. Mutations in E-box 1 and E-box 2 are indicated under the boldfaced consensus sequence and designated as m1 and m2, respectively. A probe for EMSA that contains E-box 1 (Probe Ϫ55/Ϫ34) is underlined. B, effect of mutation in E-box elements on VMD2 promoter activity. Luciferase constructs containing mutations in E-box elements in the context of both the Ϫ104 to ϩ38 and Ϫ71 to ϩ38 bp fragments as well as the constructs containing wild-type fragments and pGL2-Basic vector were transiently transfected into D407 cells. Relative luciferase activities were calculated and presented as described in Fig. 3B. The constructs containing mutations of both E-box 1 and E-box 2 are designated as m1m2.
canonical E-box (E-box 1, Ϫ47 to Ϫ42 bp) and another E-box is located nearby (E-box 2, Ϫ69 to Ϫ64 bp) (Fig. 1B), we directly tested whether these E-box sites are important for VMD2 promoter activity in D407 cells. Constructs containing mutations (CANNTG to ACNNTA) in E-box 1 (designated m1), in E-box 2 (m2), or in both E-boxes (m1m2) were generated (Fig. 4A). In the context of both the Ϫ104 to ϩ38 and Ϫ71 to ϩ38 bp fragments, m1 greatly diminished luciferase expression (Fig.  4B). In contrast, m2 decreased luciferase expression down to 60% with the Ϫ104 to ϩ38 bp fragment but did not affect activity with the Ϫ71 to ϩ38 bp fragment. The effect of m1m2 was the same as that of m1. These results suggest that E-box 1 is critical for the activity of the VMD2 proximal promoter in D407 cells.
Bovine RPE Nuclear Extract Contains VMD2 Promoter Binding Activity-To complement these functional studies with biochemical analysis, we scanned the VMD2 upstream sequences from Ϫ610 to Ϫ400 bp and from Ϫ280 to ϩ50 bp (which include the identified Ϫ585 to Ϫ541 and Ϫ56 to Ϫ42 bp regulatory regions) for evidence of protein binding activities. EMSAs were performed with bovine RPE nuclear extract and a total of 26 annealed oligonucleotide probes. Several of the probes yielded prominent shifted bands (full data not shown). The region Ϫ585 to Ϫ541 bp was covered by three separate oligonucleotides, Probes 23 (Ϫ550 to Ϫ521 bp), 24 (Ϫ570 to Ϫ541 bp), and 25 (Ϫ590 to Ϫ561 bp) (Fig. 1B), and two and three prominent bands were obtained with Probes 23 and 24, respectively (Fig.  5A, left panel). Since the sequence from Ϫ565 to Ϫ556 bp was predicted to be a putative CEBP binding site (Fig. 1B), we performed a supershift assay to evaluate whether the binding activity observed with Probe 24 was due, at least in part, to CEBP. Four antibodies against different CEBP family members were tested, and a supershift was obtained with anti-CEBP␤ antibody (sc-150) (Fig. 5A, right panel).
Based on the data supporting a role for the Ϫ56 to Ϫ42 bp region, especially the included E-box 1, we analyzed this region in further detail using mutated probes (Fig. 4A). Three shifted bands were observed with wild-type probe, but all three were lost when E-box 1 was mutated (Fig. 5B, left panel). Consistent with this result, in cold competition experiments with wild-type probe, the three shifted bands were efficiently competed away by wild-type competitor but were unaffected by mutated competitor, indicating that E-box 1 binding activity is present in native bovine RPE (Fig. 5B, right panel). We also performed EMSAs with probes containing E-box 2, but no specific bands were obtained (data not shown).
MITF-M Transactivates the VMD2 Promoter-The lower expression level of endogenous VMD2 in D407 compared with native RPE suggests that D407 may lack some of the regulatory factors present in the RPE in vivo. To complement the promoter analyses described above, we also took a candidate approach based on differential expression of transcription factors between D407 and native RPE. As the first factor to test, we chose MITF, which binds to the E-box sequence motif, is expressed in the RPE, plays a critical role in RPE development, and can cause RPE abnormalities when mutated (35,37,38). During the process of comparing expression of the various MITF isoforms, of which there are at least eight (30,31,(51)(52)(53)(54)(55)(56)(57), we found by RT-PCR that although MITF-A and MITF-H were present at similar levels in RPE cell lines and native RPE, MITF-M was readily detectable in native RPE but undetectable  Ϫ561 bp, lanes 7-9). Radiolabeled oligonucleotide probes were incubated with 3 g of the nuclear extract in the presence of 0.1 g (lanes 1, 4, and 7), 0.5 g (lanes 2, 5, and 8), or 2.5 g (lanes 3, 6, and 9) of poly(dI-dC). A nonspecific band indicated by an asterisk was seen in all lanes of scanning EMSA using 26 different probes. Right panel, for supershift assays, the nuclear extract was incubated with 2 l of antibody in the presence of 1 g of poly(dI-dC) at room temperature for 1 h before addition of Probe 24.  (Fig. 4A). Radiolabeled oligonucleotide probes were incubated with 5 g of the bovine RPE nuclear extracts in the presence of 0.1 g (lanes 1 and 4), 0.5 g (lanes 2 and 5), or 2.5 g (lanes 3 and 6) of poly(dI-dC). Three shifted bands indicated by arrows were observed with wild-type probe, but all were lost with mutated probe. in RPE cell lines. 2 Based on this finding, we co-transfected D407 cells with an MITF-M expression vector and the already described VMD2 reporter constructs. MITF-M increased luciferase expression in a dose-dependent manner with both the Ϫ104 to ϩ38 and Ϫ71 to ϩ38 bp constructs (Fig. 6). Similar experiments were performed with ARPE19, hTERT-RPE1, and SK-N-MC cells, but minimal, if any, MITF-M-mediated transactivation was detected (data not shown). We also tested the promoter sequence requirements for transactivation in D407 and found that both E-boxes were necessary (Fig. 6). DISCUSSION We chose to focus on the regulation of VMD2 both because of the functional importance of bestrophin itself (14,15,19) and because its high and selective expression patterns in the RPE make it a good model for the study of RPE gene regulation. The transgenic mouse studies demonstrate that fragment Ϫ253 to ϩ38 bp of the 5Ј-upstream region of VMD2 drives RPE-specific expression in the eye. The transgene expression patterns detected by X-gal staining were spotty, similar to those observed previously with other transgenic promoter constructs, including rhodopsin in the retina (43), Rpe65 in the RPE (21), Ldhc in the testis (58), and Mk6a in the skin (59). Although the meaning of these expression patterns, which do not fully reproduce that of the endogenous gene, is unclear and may simply reflect artifacts related to the ␤-galactosidase reporter (60), there has been speculation that they may reflect transcription factor differences between cells (61). For example, transgenic lines carrying myosin light chain promoter constructs demonstrate an apparently biologically significant expression gradient that is distinct from that of the endogenous gene (62, 63) possibly due to transgene methylation differences between the expressing cells (64).
Interestingly the RPE/choroid flat mounts of the VMD2 mice consistently showed X-gal staining predominantly in the central and midperipheral regions with absence of staining in the peripheral RPE areas close to the ciliary margin. This spatial pattern is intriguingly reminiscent of that recently reported for Hedgehog (Hh) genes (65). In Xenopus, where Hh signaling appears to be essential for the proper RPE differentiation, expression of two Hh family members, banded hedgehog (Xbhh) and cephalic hedgehog (X-chh), is restricted to the central RPE cells, while downstream components of the Hh signaling pathway, such as Gli2, Gli3, and X-Smoothened (X-Smo), are expressed in the peripheral RPE surrounding the ciliary marginal zone. Expression of other RPE markers such as Xotx5 is also excluded from the peripheral RPE. Although molecular differences between mammalian RPE cells have not been demonstrated, our observations together with the Xenopus data suggest that such differences may exist.
To complement and extend the transgenic experiments, transient transfection studies were carried out. A challenge in these studies was that the available RPE cell lines generally do not maintain their original differentiated state as indicated by cell morphology, gene expression patterns, and physiological properties (66,67). To find host cells suitable for transfection analysis, we screened nine human cell cultures for expression of endogenous VMD2 and chose three RPE and a neuroepithelioma cell lines. Consistent with their higher level of endogenous VMD2 expression, the highest reporter activities were obtained with D407 and SK-N-MC cells. Although the transfection results with the D407 and SK-N-MC were generally consistent, one significant difference was observed in the activity of the Ϫ2948 to ϩ38 bp fragment between the two cell lines (Fig. 3B), suggesting the possibility that repressor elements may be present between Ϫ2948 and Ϫ586 bp and that D407 cells may express factors that bind to these elements, while SK-N-MC cells do not.
Comparison of the transgenic and cell culture results illustrates the strengths and limitations of each approach. From the transfection studies, the Ϫ585 to Ϫ541 bp region appears to contain enhancer elements, while the Ϫ56 to Ϫ42 bp region seems to contain elements that are important for lower level VMD2 expression. From the transgenic data, the region from Ϫ253 to ϩ38 bp is sufficient to drive RPE-specific expression in the eye, and thus it is clear that the Ϫ585 to Ϫ541 bp region is not necessary for RPE expression. The transgenic data is, however, consistent with the Ϫ585 to Ϫ541 bp region having positive regulatory activity as the Ϫ585 to ϩ38 bp mice tended to have more intense X-gal staining than the Ϫ253 to ϩ38 bp mice, although the strength of this evidence is limited due to the small numbers of animals analyzed and the potential for strong effects of transgene insertion sites. Also supporting a regulatory role for the Ϫ585 to Ϫ541 bp region in vivo might be the supershift result showing that CEBP␤ present in RPE nuclear extract can bind to this region.
The existence of an E-box (E-box 1) in the Ϫ56 to Ϫ42 bp region prompted us to explore further its possible biological function and also the function of the nearby E-box 2. As might 2 N. Esumi, unpublished results.
FIG. 6. MITF transactivates the VMD2 promoter via two E-box elements. Luciferase constructs containing either wild-type sequence or mutation of E-box elements, m1, m2, or m1m2, in the context of both the Ϫ104 to ϩ38 and Ϫ71 to ϩ38 bp fragments were co-transfected into D407 cells with 0, 0.5, or 2.5 g of a human MITF-M expression vector. The amount of expression plasmid was adjusted to a total of 2.5 g for every transfection by adding empty pcDNA3.1 vector. Relative luciferase activities were calculated as a ratio of the normalized luciferase activity with MITF to that without MITF (labeled as pcDNA vector). The values represent the means and S.D. (bar). be expected, mutation of E-box 1 greatly diminished luciferase expression in D407 cells, although mutation of E-box 2 did not. These results using D407 cells are consistent with the EMSA data, which showed that native bovine RPE nuclear extracts bind detectably to wild-type E-box 1 but not to mutated E-box 1 or to wild-type E-box 2.
To complement the functional and DNA binding studies, we also took a candidate factor approach. Among the few transcription factors that have been well characterized in the RPE, MITF, a member of the basic helix-loop-helix leucine zipper family, is critical in the development of several cell lineages including RPE, melanocytes, osteoclasts, and mast cells (27, 30 -34). Mutations in Mitf are associated with RPE abnormalities in mice (35)(36)(37)(38)51) and pigmentation defects and hearing loss in human (68 -70). In addition, like other members of the basic helix-loop-helix and basic helix-loop-helix leucine zipper protein families, MITF is an E-box-binding protein. We therefore explored the possible involvement of MITF in the regulation of VMD2. We chose to test the MITF-M isoform because it is expressed at reasonable levels in native mature RPE but is undetectable in D407 cells. 2 MITF-M protein, generated by in vitro transcription and translation, binds to the VMD2 E-box 1 sequence. 2 MITF-M increased reporter gene expression in a dose-dependent manner, and its transactivating effect required both E-box 1 and E-box 2. This result may be due to cross-talk between the two E-box sites as the requirement of occupancy of two E-box sites has been reported for activation of other genes, such as muscle-specific creatine kinase (71), chicken acetylcholine receptor ␣-subunit (72), rat acetylcholine receptor ␤-subunit (73), mouse MyoD1 (74), and human GLI1 (75).
Based on the results presented, it is tempting to speculate that the MITF may be involved in regulating VMD2 expression. However, the situation is rather complex, and a number of questions remain. At one level, the finding that mutation of E-box 2 had different effects in the direct reporter assays and the MITF transactivation assays could be interpreted to mean that the E-box-binding protein(s) in D407 cells are molecule(s) other than MITF. Alternatively the results could reflect differences in activity among MITF isoforms, perhaps similar to the differences in target gene specificity that have been reported between the mast cell isoform MITF-MC and MITF-M (57). If so, the difference in the expression level of MITF-M between RPE cell lines and native RPE could account, at least in part, for the significantly higher expression level of VMD2 in native RPE. Although mice deficient for MITF-M have morphologically normal retina and RPE, the possibility that such mice have alterations in the expression of Vmd2 and other RPE genes has not been explored (76). Since multiple MITF isoforms are expressed in the RPE, they may have different target genes and play a role at different stages and/or conditions. In this regard, MITF-D is another interesting candidate for the regulation of VMD2 as well as RPE development as MITF-D has been reported to be preferentially expressed in the RPE (55). Finally, as one further indication of cellular specificity, we noted that while MITF-M demonstrated significant transactivating activity in D407 cells, it demonstrated only minimal, if any, activity in other cell lines tested, a finding presumably reflecting differential expression of needed co-activators or other interacting proteins between cell lines. Of potential relevance to this finding, there is a Crx/Otx-like homeodomain protein binding site 6 bp distant from E-box 2, and it has recently been shown that OTX2 and MITF co-localize in the nuclei of RPE cells, they physically interact, and their coexpression results in a cooperative activation of QNR71 and Tyr promoters (77). Together these complexities suggest some of the exciting challenges awaiting future efforts directed at clarifying the role of MITF and other E-box-binding proteins in the regulation of VMD2 expression.