Both PCE-1/RX and OTX/CRX Interactions Are Necessary for Photoreceptor-specific Gene Expression*

RX, a homeodomain-containing protein essential for proper eye development (Mathers, P. H. Grinberg, A., Mahon, K. A., and Jamrich, M. (1997) Nature 387, 603–607), binds to the photoreceptor conserved element-1 (PCE-1/Ret 1) in the photoreceptor cell-specific arrestin promoter and stimulates gene expression. RX is found in many retinal cell types including photoreceptor cells. Another homeodomain-containing protein, CRX, which binds to the OTX element to stimulate promoter activity, is found exclusively in photoreceptor cells (Chen, S., Wang, Q. L., Nie, Z., Sun, H., Lennon, G., Copeland, N. G., Gillbert, D. J. Jenkins, N. A., and Zack, D. J. (1997) Neuron 19, 1017–1030; Furukawa, T., Morrow, E. M., and Cepko, C. L. (1997) Cell 91, 531–541). Binding assay and cell culture studies indicate that both PCE-1 and OTX elements and at least two different regulatory factors RX and CRX are necessary for high level, photoreceptor cell-restricted gene expression. Thus, photoreceptor specificity can be achieved by multiple promoter elements interacting with a combination of both photoreceptor-specific regulatory factors and factors present in closely related cell lineages.

Our goal is to understand the basic biochemical mechanisms of tissue-restricted gene expression. The vertebrate retina offers a distinct advantage in studying gene expression since it contains large numbers of well defined cell types, such as photoreceptor cells. Photoreceptor cells are highly specialized to convert light energy into neuronal impulses, a process which involves many proteins, several of which have been well characterized such as opsin, interphotoreceptor retinoid-binding protein (IRBP), 1 and phosphodiesterase (4). Arrestin, also known as S-antigen, accumulates to high levels in the outer segments of retinal photoreceptor cells, where it down-regulates the phototransduction cascade (5).
The transcriptional efficiency and tissue specificity of a given gene are determined by the specific combination of regulatory factors that assemble on the gene's regulatory elements to form the transcription initiation complex. Several homeodomain proteins such as Pax6 (6), RX (1,7), and CRX (2, 3) which profoundly affect vertebrate eye development, bind to promoter elements of several eye-specific genes, and may aid in the activation of these genes. The RX homeobox gene has been found to be essential for normal vertebrate eye development, and its mis-expression has profound effects on eye morphology (1). Xenopus embryos injected with RX mRNA develop ectopic retinal tissue and exhibit hyperproliferation in the neural retina. Mouse embryos homozygous for a null allele of this gene fail to form optic cups and therefore do not develop eyes. It is clear that the RX gene family plays a crucial role in the establishment and/or proliferation of retinal progenitor cells. Other homeobox proteins, especially the transcriptional regulatory protein Pax6, play an important role in vertebrate and invertebrate eye formation. Mutations which affect the function of Pax6 result in severe eye malformations known as Aniridia in humans and small eye syndrome in mice (8).
Another homeodomain protein, CRX, binds to the OTX regulatory sequence found upstream of several photoreceptor cellspecific genes and can stimulate transcriptional activity of the opsin and IRBP promoters in non-retinal cells (2). Moreover, mutation of the CRX gene induces a cone-rod dystrophy in humans (9). Thus, CRX is a novel photoreceptor cell-specific factor and plays a crucial role in the differentiation of photoreceptor cells.
We have previously identified the photoreceptor conserved element 1 (PCE-1) consensus sequence which is found in the functionally important regulatory regions of all known photoreceptor cell-specific genes. The same nuclear regulatory factors which bind to the arrestin PCE-1 site also recognize PCE-1 sites in the promoter regions of other vertebrate photoreceptorspecific genes (10). A factor-binding site with the same core sequence as PCE-1 has also been identified as Ret1 in the 5Ј-flanking region of the rat opsin gene (11). These results suggest that the PCE-1 and Ret1 sites are structurally and perhaps functionally the same.
Recently, a factor designated Erx was found to bind to the Ret 1 site (12). Here we have isolated a human homeodomaincontaining protein RX, which binds to the PCE-1/Ret1 site and activates the TATA-less arrestin promoter (10,13) and IRBP promoter. Our study established that RX is a transcriptional regulatory protein which binds to a regulatory region found in many retina-specific genes and up-regulates expression of these genes. Furthermore, our study suggested that two individual elements (PCE-1 and OTX) and two different factors (RX and CRX) are necessary for stimulating expression of photoreceptor cell-specific genes. * 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) AF115392.

EXPERIMENTAL PROCEDURES
Southwestern Screening of cDNA Library for PCE-1 Site Binding Proteins-The random-primed gt11 human retina library (CLONTECH Laboratories, Palo Alto, CA) was screened by the Southwestern method (14). Oligonucleotides were synthesized by the Life Technologies, Inc. Custom Primers Service (Grand Island, NY). The oligonucleotides containing mouse PCE-1 sites were annealed, 5Ј end-labeled with [␥-32 P]ATP (NEN Life Science Products, Boston, MA) and T4 polynucleotide kinase, and concatenated with T4 DNA ligase. The average final length of probe 1 and probe 2 was approximately 200 bp. Probe 1 was GCGCAAGCTT-TCAATTAGCTATTTTCGAAAGTTAATCGATAACGCG and Probe 2, CCAAAAGCTTTCAATTAGCTATTTTCGAAAGTTAATCGATAAGGTT.
Bacteriophage (2 ϫ 10 5 per dish) were plated with Escherichia coli Y1090 on 120-cm 2 LB plates and incubated at 42°C for 3.5 h. Nitrocellulose filters (BA-S 85 TM ; Schleicher & Schuell Inc., Keene, NH) impregnated with 10 mM isopropylthio-␤-D-galactopyranoside were then overlaid onto the cultures and incubation continued at 37°C overnight. The filters were lifted, air-dried, and submerged for 10 min in 250 ml of binding buffer (12 mM HEPES, pH 7.9, 12 mM KCl, 0.6 mM MgCl 2 , 1.2 mM dithiothreitol) containing 6 M guanidine hydrochloride. The filters were washed for 10 min with gentle shaking in the same volume of the binding buffer without guanidine hydrochloride. To maximize renaturation of the proteins, we repeated this cycle (binding buffer plus 6 M guanidine hydrochloride followed by binding buffer alone) four times. The filters were blocked with 250 ml of binding buffer containing 5% nonfat skim milk for 4 h at 4°C with gentle shaking. They were incubated with gentle shaking at 4°C overnight with labeled DNA probe (10 6 cpm/ml) in 250 ml of binding buffer supplemented with 0.25% skim milk. The filters were washed 3 times with 500 ml of binding buffer containing 0.25% skim milk and 50 mM KCl for 10 min at room temperature. The filters were dried on 3MM paper and exposed to Fuji x-ray Film (Fuji Film Co., Tokyo, Japan) overnight at Ϫ70°C.
Screening of cDNA by the Plaque Hybridization Method-The 5Ј fragment (479 bp) and 3Ј fragment (856 bp) of the hRx cDNA were 32 P-labeled with T7 QuickPrime TM Kit (Amersham Pharmacia Biotech) and used to screen the human retina cDNA library. Bacteriophage (5 ϫ 10 5 ) were plated with E. coli strain Y1090 onto 120-cm 2 plates and grown 37°C for 8 h. Detailed procedures were described elsewhere (15).
Production and Purification of GST-RX and GST-CRX Fusion Proteins-We produced RX and CRX as GST fusion proteins in an E. coli expression system. Because there was a stop codon upstream of the translation initiation codon of hRx, a PCR primer (TCCCGGAATTC-CCCATGCACCTG) was designed with an EcoRI site just upstream of the translation initiation codon of the hRx cDNA, and a second PCR primer (GGACCTCTGGTAGGTTGACCTTG) was designed to introduce EcoNI site into the 3Ј non-coding region of the hRx cDNA. These primers were used to PCR amplify 10 ng of DNA containing the entire coding region of hRx cDNA (16). The PCR amplification product (548 bp) was blunt end ligated into the HincII site of the pBlueScript II SK ϩ (Stratagene, La Jolla, CA). After confirming the fidelity of the PCR amplification by DNA sequence analysis, the fragment between the XhoI site in the pBlueScript and EcoNI site in the PCR product was subcloned such that it replaced the XhoI/EcoNI fragment in the original hRx cDNA/pBlueScript construct. The EcoRI fragment of this new hRx cDNA was subcloned into the EcoRI site of pGEX4T-1 (Amersham Pharmacia Biotech), and the final ligation creates an in-frame fusion of the GST and hRx coding sequences. The cDNA for human CRX (pro-vided by Dr. D.J. Zack, Johns Hopkin University) was subcloned into the BamHI and XhoI sites of pGEX4T-1 to create a similar in-frame fusion between GST and hCRX.
These two GST fusion constructs were expressed in the E. coli strain BL21. The bacteria were cultured in 25 ml of LB medium with 0.1 mg/ml ampicillin at 37°C for 3-4 h (to an A 600 of 0.6), then 0.2 mM isopropylthio-␤-D-galactopyranoside was added and the incubation continued for an additional 4 -6 h. The bacteria were harvested and resuspended in 1 ml of bacteria lysis buffer (50 mM Tris, pH 8.0, 200 mM NaCl, 1.5 mM EDTA, pH 8.0). The bacterial suspensions were treated with 1 mg/ml lysozyme, 5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride and sonicated for 1 min. After centrifugation at 20,000 ϫ g for 20 min, the supernatant was mixed with 30 l of glutathione-Sepharose 4B (Amersham Pharmacia Biotech) and incubated for 1 h at room temperature. The Sepharose was washed 3 times with bacteria lysis buffer, then mixed with glutathione reducing buffer (Amersham Pharmacia Biotech). After centrifugation, the supernatant containing the GST fusion protein was kept at Ϫ70°C.
Generation of Antibodies to GST-RX-An albino rabbit (Harlan Sprague-Dawley) and a mouse (BALB/c) were immunized with the GST-RX fusion protein emulsified with CFA (Life Technologies, Inc.) by subcutaneous injection in the back with 2 mg/0.5 ml and 0.5 mg/0.2 ml, respectively. Two and four weeks later booster injections of GST-RX emulsified with IFA (Life Technologies, Inc.) were given. Five weeks after the first injection the animals were euthanized and their serum collected. The anti-RX antibodies were purified on an immobilized protein-A agarose column following a Pierce instruction manual (Pierce, Rockford, IL).
Immunohistochemistry-The eyes from adult Long-Evan's rats (body weight about 200 g) were immersed in fixative (4% paraformaldehyde in phosphate-buffered saline, pH 7.4) at 4°C for 10 min. Then the eyes were cut and retinae were removed from the anterior segment, lens and vitreous body and kept in the fresh fixative at 4°C for overnight. The tissues were embedded in paraffin and thin sections (5 m) were obtained. These sections were treated with methanol for 30 min at Ϫ20°C to increase membrane permeability, followed by 4% hydrogen peroxide treatment to block endogenous peroxidase activity. The sections were incubated overnight at 4°C with a 1:500 dilution of mouse anti-RX polyclonal Ab alone or with the same antiserum which had been neutralized by preincubation with 200 g of GST-RX protein at 4°C for 24 h. The sections were then incubated with biotinylated goat antirabbit IgG antibody (Sigma). Color development was done with ExtrAvidin Peroxidase (Sigma) and 3-amino-9-ethylcarbazole (AEC substrate kit, Vector Laboratories, Burlingame, CA) according to the manufacturers' instructions.
Electrophoretic Mobility Shift Assays (EMSA)-Oligonucleotide probes listed in Table I were annealed and their 5Ј-ends were labeled with [␥-32 P]ATP and T 4 polynucleotide kinase (Promega, Madison, WI). The standard binding reaction was carried out in 20 l of a mixture containing 12 mM HEPES (pH 7.9), 12 mM KCl, 0.6 mM MgCl 2 , 1.2 mM dithiothreitol, 10% glycerol, 2 g of poly(dI-dC), 5 fmol of the 32 P-labeled probe, and 8 ng to 1 g of purified GST-RX fusion protein. In competition assays, 8 fmol to 1 pmol (1.6 -200 fold molar excess) of the competitor DNA was added to the standard mixture. After incubation on ice for 30 min, the samples were loaded onto 5% polyacrylamide gels in 0.5 ϫ Tris borate-EDTA (TBE) buffer, and electrophoresed at 10 V/cm for 2 to 3 h at room temperature. The gel was then dried and autoradiographed at Ϫ70°C.
For Ab supershift analysis, 1 l of rabbit Ab to RX was added to the standard mixture halfway through the reaction (after the first 15 min of incubation). Nuclear extract from bovine retinas was prepared by the method of Gorski et al. (18).
CAT constructs containing PCE-1 sites or OTX sites upstream of the basal herpes simplex virus thymidine kinase (tk) promoter were generated by inserting double-stranded oligonucleotides into the BstXI site of pBLCAT5 (ATCC, Manassas, VA). Oligonucleotides containing either the PCE-1 or OTX element were chemically synthesized and annealed and ligated as described above. The elements were: PCE-1 CTAGAAA-AGCTTTCAATTAGCTATTTTTTCGAAAGTTAATCGATAAGATC and OTX CTAGACTGCTGAGCTTAATCAGGCCTTGACGACTCGAATTA-GTCCGGAGATC. Selected clones were sequenced and only clones having exactly two tandemly repeated PCE-1 or OTX elements in the correct orientation were chosen for further experiments.
To confirm that RX binds to the PCE-1 site and activates the natural arrestin promoter, the PCE-1 site in the mouse arrestin promoter/CAT construct m-ARR-209-CAT was mutated. The m-ARR-209-CAT was cleaved with HindIII and EcoNI and the fragment was replaced with a double-stranded oligomer containing the mutated sequence as follows: AGCTTTCAACCAGCTATTCCTCTCTTTGCACCTTGAAGTTGGTC-GATAAGGAGAGAAACGTGGAACG. The resulting construct was called m-ARR-209-PCE-1-mut-CAT.
The mammalian expression plasmid pcDNA3 (Stratagene), bearing the human cytomegalovirus promoter-enhancer, was used to generate RX and CRX expression vectors for co-transfection studies. cDNA fragments containing the complete coding regions of human RX (EcoRI/ EcoRI fragment) and human CRX (BamHI/XhoI fragment) were subcloned into EcoRI-digested and BamHI-and XhoI-digested pcDNA3, respectively.
Cell Culture-Neuroretinal cells from day 14 embryonic chicken retinae were incubated in 0.25% trypsin at 37°C for 20 min. Then 5 ϫ 10 6 cells were cultured in Y-199 medium with 10% fetal calf serum, penicillin (100 units/ml), and streptomycin (100 g/ml) at 37°C in polyornithine-coated 60-mm tissue culture plates. Detailed procedures were described elsewhere (19). Human embryonic kidney cells (293 cells) were cultured for 1 to 2 days in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, then they were co-transfected by the calcium phosphate precipitation method with three constructs: (i) 0.04 -5.0 g of an RX or CRX expression vector, (ii) 10 g of promoter/CAT reporter construct, and (iii) 5 g of pSV-␤-galactosidase vector. After a 48-h incubation, cells were harvested and extracts were assayed for CAT activity (20). All CAT values were normalized to ␤-galactosidase activity. The results presented represent the mean of at least three independent experiments.

RESULTS
Cloning of the Human RX Homeodomain Protein-We isolated cDNAs for seven individual factors from a human retinal library by Southwestern screening with the PCE-1 probes (probe 1 and probe 2; see "Experimental Procedures"), and a clone which bound strongly to both probes was selected for further study. Sequencing revealed that the clone contained a region homologous to the highly conserved DNA binding con-sensus (homeobox) in homeodomain proteins. The full size of this clone was 1790 bp long and encoded a 346-amino acid protein, which we have named human RX (GenBank accession number AF115392). This protein shares 85% identity (Fig. 1A) with mouse RX (1, 7) and 100% identity between 97 and 162 with the partial human RX sequence isolated by Mathers et al.
(1) (GenBank™ accession number AF001911). The nucleotide sequence between 295 and 493 bp of the partial human hRx cDNA (1) shares 96% homology with the 478 to 676-bp region of our hRx. The sequence between 1 and 294 bp of the partial human Rx cDNA (1) shares 95% homology with a Int-1 sequence of our hRx-1 cDNA (see below). Thus, the human partial sequence of Rx published by the Mathers et al. (1) is part of hRx-1 cDNA. Two additional clones similar to hRx (hRx-1 and hRx-2) were isolated from the same library by a DNA hybridization method using an hRx (cDNA) probe. Nucleotide sequence of hRx-1 (2,230 bp) and hRx-2 (1,846 bp) was identical in sequence to hRx except for a 303-bp insertion between bp 484/485 in hRx-1 and two insertions of 303 and 159 bp between bp 484/485 and bp 742/743, respectively, in hRx-2 (Fig. 1B). We believe that these cDNAs might be products of alternative RNA splicing. A functional role of these two clones is presently unknown. It does, however, seem unlikely that these cDNA clones could encode a homeodomain-containing protein, since the first inserted sequence (Int-1) contains two in-frame stop codons upstream of the homeodomain, and the second insertion (Int-2) also contains two stop codons.
Expression Pattern of RX in Various Tissues-We studied the expression pattern of RX in various tissues by immunoblot analysis. Specificity of the this procedure was demonstrated on duplicate immunoblots probed with the same RX-specific Abs ( Fig. 2A) which had been preincubated with purified GST-RX (Fig. 2B) or with preimmune serum (data not shown). Two predominant bands in the 37-kDa molecular mass region were detected in retina ( Fig. 2A), but only one of these was RX ( Fig.  2A, shown by an arrowhead) since this band could be eliminated by addition of GST-RX to anti-RX Ab (Fig. 2B). This RX-specific band was also detected in iris ( Fig. 2A). The specificity of the RX Abs for GST-RX but not the related GST-CRX is shown in Fig. 2A. These results suggest that RX is present in neural retina and iris, but not detectable lens, brain, and liver. The expression pattern of RX in the retina was further investigated by immunohistochemistry using the affinity purified rabbit anti-RX Ab (1/500 dilution). This analysis revealed that RX was present in the outer nuclear layer (nucleus of photoreceptor cells), in addition, it was present in neuroretinal cells including ganglion cells and the inner nuclear layer (Fig. 3A).
In control experiments, RX Abs which had been preincubated with purified GST-RX failed to stain any of these cell types (Fig. 3B).
RX Binds to the PCE-1 Element-We investigated the binding properties of RX to the PCE-1 element using an EMSA. Double-stranded oligonucleotides containing the PCE-1 element of the mouse arrestin gene (10) and the mouse OTX element (m-OTX) (see Refs. 2 and 3 and Table I) were 5Ј-end labeled with [ 32 P]. Fusion proteins of GST with RX (GST-RX) or with the human CRX (GST-CRX) were produced in an E. coli expression system and purified by glutathione column chromatography.
Strong binding was observed when the m-PCE-1 element was incubated with the GST-RX protein at higher concentrations (Fig. 4A). In contrast, less strong binding of GST-CRX to m-PCE-1 was observed. Conversely, binding was observed between m-OTX and GST-RX only at higher concentrations, but stronger binding was observed between m-OTX and GST-CRX at those concentrations. In the control experiments, GST bound to neither the m-PCE-1 site nor the m-OTX site.
Binding of RX to the m-PCE-1 element was decreased sig-nificantly by a 40-fold molar excesses of unlabeled m-PCE-1 competitor, but was only slightly diminished by a 200-fold molar excesses of unlabeled m-OTX competitor (Fig. 4B). Complex formation between RX and PCE-1 sites was unaffected by a large excess of an unlabeled, unrelated oligonucleotide containing the chicken ovalbumin upstream promoter transcription factor (COUP-TF) site. Conversely, binding of CRX to m-OTX was inhibited slightly by a large molar excess of unlabeled m-PCE-1, and strongly by a large molar excess of unlabeled of m-OTX, but was unaffected by the COUP competitor. These results indicated that the stronger binding was between m-PCE-1 and RX and between m-OTX and CRX, and that the affinity of CRX for m-PCE-1 and the affinity of RX for m-OTX are much weaker. We have therefore concluded that PCE-1 and OTX are the binding sites for RX and CRX, respectively. We used an antibody (Ab) supershifting technique in combination with EMSA to confirm that the PCE-1 elements bind to RX contained in retina extracts. In a previous publication (10), when m-PCE-1 was incubated with bovine retinal nuclear extract, EMSA revealed a complex designated Bp1 (see also Fig.  5, lane 1, indicated by an arrow). Initially we generated complexes between m-PCE-1 and bovine nuclear extract followed by addition of Abs (Fig. 5, lanes 1 and 3, an arrow and an open  arrowhead, respectively). Similarly, we generated complexes between m-PCE-1 and purified RX, then added RX-specific Abs (Fig. 5, lanes 4 and 5, an arrowhead and an open arrowhead,  respectively). The RX-specific Abs recognized both purified GST-RX and RX from bovine nuclear extracts bound in PCE-1 complexes. The complexes formed between m-PCE-1 and purified GST-RX were smaller than those formed with the nuclear extract, but both of these complexes were supershifted by RXspecific Abs (Fig. 5, lanes 3 and 5, open arrowheads, respectively). We therefore concluded that the Bp1 complexes formed between bovine retina nuclear extract and the m-PCE-1 element contained RX factor. As reported previously (10), m-PCE-1 and a bovine nuclear factor generated a second complex, Bp2 (Fig. 5, arrow). Interestingly, the Bp2 complex failed to supershift upon addition of RX-specific Abs suggesting that it does not contain RX. We speculated that the Bp2 may be composed of m-PCE-1 element and other homeodomain proteins such as CRX (2, 3), Pax6 (6, 8), or Erx (12), since bovine retinal extract is derived from various cell types. Two Nucleotide Differences in the PCE-1 Sequence Determine RX versus CRX Binding Specificity-We next investigated weather the RX factor binds to the PCE-1 element in the murine and bovine arrestin promoters. One PCE-1 element is present in the mouse arrestin gene at position Ϫ15 to Ϫ21 (CAATTAG). The bovine arrestin gene, however, has three PCE-1 sites associated with it; one in the 5Ј-flanking region (Ϫ1830 CAATTAG Ϫ1824) and two in the first intron (ϩ561 GTAATTG ϩ567 and ϩ656 TTAATTGϩ663) (see below). Complete sequences of all oligonucleotides used in the following EMSA experiments are shown in Table I. We conducted EMSA using purified GST-RX and oligonucleotide probes containing the identical mouse (Ϫ21 CAATTAG Ϫ15) and bovine upstream (Ϫ1830 CAATTAG Ϫ1824) PCE-1 site, and one bovine first intron PCE-1 site (ϩ663 CAATTAA (rev) ϩ656). These oligonucleotides bound strongly to GST-RX to form complexes (Fig. 6, lanes 1-10) suggesting that the PCE-1 core consensus (CAATTA(G/A)) was an essential binding element and the flanking sequence contributed minimally to the observed binding activity.
Although the consensus PCE-1 and OTX sequences are similar, the difference is in two nucleotides at their 5Ј-ends: the PCE-1 consensus sequence is CAATTAG and OTX is TGAT-TAG. Oligonucleotide pair m-PCE-1-mut was synthesized in which the two 5Ј nucleotides of the mouse PCE-1 core sequence were altered (CAATTAG 3 TGATTAG) but the sequences within the core and adjacent remained unchanged. The RX bound much more weakly to m-PCE-1-mut than to wild type  2-5, 7-10, and 12-15) or GST-CRX (lanes 17-20). In control experiments, 1000 ng of GST protein was used (lanes 1, 6, 11, and 16). m-PCE-1 (Fig. 6, lanes 12-15). Interestingly, the m-PCE-1-mut bound strongly to CRX (Fig. 6, lanes 17-20) with an affinity similar to that of m-OTX. Taken together, our results show that PCE-1 and OTX are two distinct elements, with RX binding specifically to PCE-1 and CRX binding specifically to OTX and that a small sequence difference adjacent to the core consensus sequence could contribute significantly to cell type-specific gene expression (see "Discussion").
Activation of Arrestin Promoter by RX-Next, we investigated whether RX can stimulate expression of the mouse arrestin promoter. Initially we co-transfected a construct containing a hRx cDNA with a construct containing the mouse arrestin promoter and a CAT reporter gene (m-ARR-209-CAT) into embryonic day 14 chicken retina cells. To normalize for efficiency of transfection, a plasmid containing the ␤-galactosidase gene under the control of the SV40 promoter and enhancer (pSV␤galactosidase) was transfected simultaneously. Up to 2.2-fold transactivation of the arrestin promoter was produced by cotransfection of RX into the retinal cells (data not shown). However, since embryonic retina express large amounts of endogenous RX (1, 7) it is likely that these cells were already saturated with endogenous RX factor and therefore addition of the exogenous RX would not significantly increase promoter activity.
To eliminate this complication, we used human embryonic kidney (293) cells which exhibit no endogenous expression of RX factor. The 293 cells were co-transfected with the hRx cDNA and the mouse arrestin promoter/CAT reporter construct (see Fig. 7A, m-ARR-209-CAT). As expected co-transfection of hRx stimulated promoter activity in a dose-dependent manner reaching levels of 6.4-fold induction. Interestingly, CRX also enhanced reporter activity in a dose-dependent manner reaching 10.1-fold induction. Since the reporter construct, m-ARR-209-CAT, contained a PCE-1 site and two OTX sites in the promoter region and one OTX site upstream of the CAT gene (see Fig. 7A), it is reasonable that RX and CRX both transactivated the construct.
Next we investigated whether mixture of both RX and CRX factors might act synergistically to transactivate the m-ARR-209-CAT construct. The promoter activity of the construct was transactivated by CRX (3.1-fold) or RX (1.8-fold) individually, but together they transactivate more than 6-fold with 0.2 g of RX and 0.2 g of CRX (Fig. 7A). This moderate synergistic transactivation was not stimulated by higher concentrations of RX (1.0 or 5.0 g) in the presence of 0.2 g of CRX (Fig. 7A). These results demonstrate that the appropriate mixture of RX and CRX can stimulate arrestin promoter activity.
To further investigate whether a mutation in the PCE-1 site of the natural mouse arrestin promoter (m-ARR-209-CAT) affects transactivation of this promoter, we altered the mouse PCE-1 core CAATTAG to CAACCAG (m-ARR-209-PCE-1-mut-CAT). Neither RX nor CRX bound detectably to this mutant oligonucleotide (data not shown). Promoter activity of the reporter construct containing this mutation was not transactivated (approximately 2-fold) by RX alone (Fig. 7B), but was activated 4.3-fold by CRX alone. However, co-transfection with RX and CRX together did not further activate the promoter (Fig. 7B). These results demonstrate that RX and CRX binding to the PCE-1 and OTX elements, respectively, transactivate the natural mouse arrestin promoter.
RX Also Activates the Mouse IRBP Promoter-To determine whether this phenomenon was unique to the arrestin promoter, we investigated RX transactivation of a promoter for photoreceptor-specific gene, mouse IRBP. The mouse IRBP promoter region (Ϫ1783 to ϩ101) in the m-IRBP-1783-CAT construct (see Fig. 8 and Ref. 19) contains a PCE-1 site (Ϫ1410 CTA-ATTG Ϫ1404) and an OTX site (Ϫ57 GGATTAA Ϫ51). The construct was co-transfected into 293 cells with either an hRx cDNA construct or a CRX cDNA construct. RX transactivated m-IRBP-1783-CAT promoter up to 14.8-fold in a dose-dependent manner (Fig. 8). On the other hand CRX transactivated the promoter up to 37.0-fold, more than twice that observed with RX. These results suggest that CRX and RX can stimulate a variety of photoreceptor-specific promoters.
RX Activated the PCE-1 Site-To further confirm that RX binding to the PCE-1 site activates an arrestin promoter, we generated two heterogeneous promoter constructs (Fig. 9). One construct had two tandem repeats of the PCE-1 site and the other had two tandem repeats of the OTX site upstream of the weak HSV/tk promoter which drives CAT reporter gene expression. Each of these reporter constructs was co-transfected into 293 cells with constructs expressing either RX or CRX, and promoter activation was quantified by CAT assay. Promoter activity of the tk promoter alone was not affected by RX or CRX (1.4-and 1.2-fold induction, respectively, Fig. 9). Promoter activity of the construct containing two tandem repeats of PCE-1 was stimulated by up to 3.9-fold by RX, but unaffected by CRX. Conversely, the promoter activity of the construct containing two tandem repeats of OTX was stimulated up to 2.5-fold by CRX, but it was unaffected by RX. This result clearly demonstrates that RX acts through PCE-1 site to activate transcription, and CRX acts through the OTX site to activate transcription. . It is an essential factor for normal eye development and its misexpression has profound effects on eye morphogenesis (1). The present study established that in addition to the expression seen during development, a low level of RX, comparable to that of other known homeodomain proteins, was present in cells of the adult retina. Double shift assays indicated that there is sufficient RX in adult retina to bind to PCE-1 element (see Fig. 5). Interestingly, expression of most photoreceptorspecific genes is initiated after RX was dropped at lower levels (1). Our results showed that the RX factor, known to be essential for eye morphogenesis, binds to, and can act through, the PCE-1 site.
RX Binds to PCE-1 Sites and May Stimulate Retina-specific Promoters-The finding that RX is expressed in most cell types in the retina suggested that the promoters of a variety of retina-specific genes might contain PCE-1 elements. Computer searches of genes expressed in the retina revealed the presence of at least one PCE-1 site in all known retina-specific genes, and multiple PCE-1 sites in the 5Ј-flanking region of several of these genes (10). Since RX binds tightly to the PCE-1 element and regulates promoter activity in our assay systems, it is reasonable to speculate that RX binds to the PCE-1 sites of retinal-restricted genes and enhances retinal specific promoter activity in vivo.
RX has been shown by protein immunoblot analysis and immunocytochemistry to be expressed in most retinal cell types (predominantly in the photoreceptor cells). Our results are also in agreement with those of Yu et al. (21), who showed that in transgenic mice carrying four tandem copies of the PCE-1/Ret1 element fused to the lacZ reporter gene, ␤-gal was expressed not only in photoreceptor cells, but also in the ganglion cell layer and ciliary epithelium. The PCE-1 element in the 5Јflanking region of cellular retinaldehyde-binding protein may play a important role in the expression of cellular retinaldehyde-binding protein in RPE cells (22). In addition, present supershift analysis suggested that the PCE-1 site bound to RX as well as some other factor(s). One such factor, Erx, was recently isolated and found to bind to the PCE-1/Ret1 site and activate transcription (12).
RX and CRX Bind to the PCE-1 and OTX Sites, respectively-In our present study, EMSA and competition EMSA assays confirmed that the PCE-1 and OTX sites bound tightly to the FIG. 7. A, transactivation of promoter constructs by RX or CRX, or RX ϩ CRX. 293 cells were co-transfected with 10 g of mouse arrestin promoter (Ϫ209 to ϩ304 bp)/CAT (m-ARR-209-CAT) fusion gene, together with expression vector(s) encoding RX or CRX, or RX and CRX together. RX and CRX (0.2 g) appear to act synergistically in activating the arrestin promoter. B, similarly, 293 cells were co-transfected with 10 g of mouse arrestin mutant promoter (m-ARR-209-PCE-1-mut-CAT) fusion gene, together with expression vector(s) encoding RX or CRX, or RX and CRX. CRX, but not RX, activated the mutant mouse arrestin promoter. All CAT activity values were normalized to ␤-galactosidase activity to control for transfection efficiency.
RX factor and CRX factor, respectively. We also demonstrated the functional transformation of a PCE-1 element (CAATTAG) into an OTX element (TGATTAG) by a simple two-nucleotide substitution (CA 3 TG) adjacent to the core sequence of the m-PCE-1 element. The ATTA core was known to be essential for DNA binding activity of many homeodomain proteins, while nucleotides immediately flanking this core determine the specificity of binding (23). In addition, the S8 factor (24), a mesoderm-specific homeodomain protein, was shown to bind to a sequence element identical to that of PCE-1, and it bound strongly to the mouse and bovine PCE-1 element. 2 The predicted amino acid sequences of human and mouse RX show high degree of homology to S8 (25), aristless (26), Chx10 (27), pax6 (6), and paired (28). The DNA-binding domain of the homeobox is comprised of three regions which presumably me-  9. Transactivation of synthetic promoter constructs by RX and CRX. 293 cells were co-transfected with 10 g of pBLCAT5 bearing the herpes simplex virus thymidine kinase gene (tk) promoter alone (top), pBLCAT5 into which two tandem copies of mouse arrestin PCE-1 element had been inserted (middle) and pBLCAT5 into which two tandem copies of mouse arrestin OTX element had been inserted (bottom), together with the indicated amount of RX or CRX expression vector. diate interaction of the homeobox protein with specific DNA elements. The second and third helical region are arranged in a helix turn helix motif which is highly conserved among species, and the ninth position of the third helix is a major determinant of the DNA binding specificity of homeodomain proteins (29). These third helical regions of RX and S8 have identical sequences including a glutamine at the ninth position. In contrast, CRX, which belongs to the OTX family of homeodomain proteins, has a lysine at the ninth position of the third helix (2, 3). Thus, protein sequence of DNA-binding domains further supports the notion that RX and CRX bind to, and exert their effects through, different sites.
Our transgenic studies indicated that both PCE-1 and OTX elements are necessary for high levels of photoreceptor-specific gene expression. 2 However, neither element alone could stimulate high level photoreceptor-restricted gene expression, but the two elements together could induce high level photoreceptor-restricted gene expression. The transgenic and present in vitro studies with Rx and CRX factors suggest that tissue specificity can be achieved by multiple promoter elements interacting with a combination of both highly cell type-specific regulatory factors and factors present in closely related cell lineages.