A Short, Highly Active Photoreceptor-Specific Enhancer/Promoter Region Upstream of the Human Rhodopsin Kinase Gene

Abstract

purpose. Rhodopsin kinase (Rk or GRK1) is a photoreceptor-specific enzyme that mediates adaptation of photoreceptors to light and protects these cells against light-induced injury. This study examined the transcriptional mechanisms that maintain physiologic levels of this essential enzyme in photoreceptors.

methods. The 2.0-kb region flanking the 5′ end of the human Rk gene was isolated, mapped, and sequenced. The sequence was fused upstream of the luciferase gene and was tested for promoter activity in retinoblastoma cells by transient transfection. Transcriptionally active segments were identified by deletion and site-directed mutagenesis. Transgenic mice were generated that carried the immediate 5′ flanking segment linked upstream of the enhancerless green fluorescent protein (GFP) gene. GFP expression was analyzed by RT-PCR, fluorescence microscopy, and immunohistochemistry.

results. Mapping and sequence analysis uncovered a TATA-less promoter with several recognizable elements concentrated proximally. A conserved putative homeodomain response element H1 and a GC- and a GA-rich motif were noted within a 0.11-kb region. Another putative homeodomain binding site, H2, and a stretch of C-rich repeats were present distally. Mutagenesis in conjunction with transient transfection in retinoblastoma cells identified the 0.11-kb region and H1 sequence as the key active enhancer–promoter sequences. The distal sequences were inhibitory. Transgenic mice that carried the 0.11-kb DNA segment with the GFP gene linked downstream showed GFP transcript, fluorescence, and immunoreactivity that were restricted to photoreceptors.

conclusions. The experiments defined a short, highly active photoreceptor-specific enhancer–promoter region upstream of the Rk gene. The H1 element contributed substantially but not exclusively to the transcriptional activity of the region. The findings support a transcriptional basis for photoreceptor-specific expression of Rk.

Rhodopsin kinase (Rk or GRK1) is a photoreceptor-specific G protein-dependent receptor kinase (GRK) that plays a key role in the recovery of photoreceptors ^{1 2 3 4} and protects them against light-induced damage. ^{5 6 7 8} By catalyzing light-dependent phosphorylation of opsins and directing these light-activated intermediates into an inactive complex with arrestin, Rk clears cells of the bleached intermediates that would otherwise maintain the photoreceptor in a perpetual on state and rapidly restores photoreceptor sensitivity to its original baseline level for subsequent stimulation. ^{8 9 10} Defects in the human rhodopsin kinase gene cause the Oguchi form of stationary night blindness characterized by a metallic sheen of the light-adapted fundus and marked delay in dark adaptation, even after a brief exposure to light. ^{10 11 12} Comparable physiological abnormalities are demonstrable in mice without Rk function, together with a marked increase in propensity toward light-induced photoreceptor apoptosis. ^{5 6 7 8} Rk-deficient photoreceptors are especially vulnerable to bright-light–induced apoptosis, and their vulnerability is markedly accentuated by the presence of additional mutations in the phototransduction deactivation pathway or in other unidentified genes that mediate normal resistance to light. Given at least a partial role for Rk in countering photostress, factors that affect the levels of Rk expression in the eye could have a significant impact on the rate at which photoreceptor cell loss occurs in patients already at risk for retinal degeneration caused by mutations in one or more photoreceptor-specific genes. Despite the availability of extensive information on the biochemistry of this enzyme and others in the GRK superfamily, the molecular mechanisms that govern the basal levels and distribution of these gene products in normal and pathologic states remain obscure. 

Previous studies have suggested a possible transcriptional basis for the differential expression of GRKs or Rk in various tissues. Levels of GRK proteins, including retina and pineal-restricted GRK1, ¹³ cone-specific GRK7, ^{14 15} and testis-specific GRK4, ¹⁶ generally parallel their tissue transcript levels, suggesting that differential promoter responsiveness may be a key factor in determining tissue-specific distribution of the protein. In the eye, Rk protein and transcript are both confined exclusively to the rod and cone photoreceptors, further consistent with this notion. ^{8 17} However, unlike other photoreceptor-specific genes, Rk transcript levels and expression appear largely unperturbed by the functional disruption of the two known photoreceptor transcription factors, the paired homeodomain protein Crx ^{18 19} and the leucine zipper protein Nrl, ²⁰ both of which are essential for driving expression of most of the photoreceptor-specific genes. The apparent relative insensitivity of the Rk gene to these transcriptional regulators raises the possibility of an independent set of unidentified transcriptional or posttranscriptional mechanisms that govern Rk and Rk-like gene expression. 

In this study, we examined the role of transcriptional mechanism in governing the photoreceptor-restricted expression of Rk. Structural analysis, mutagenesis, and in vitro and in vivo functional studies were used to define a short, robustly active 0.11-kb DNA segment upstream of the start site that is sufficient for mediating photoreceptor-specific expression. A putative homeodomain binding site was located in the core of this segment that was responsible for part but not all of the transcriptional activity. The findings in this study highlight the importance of transcriptional mechanisms in regulating the cellular distribution of this photoreceptor-specific GRK in the eye. 

Materials and Methods

Human retinal RNA was isolated from retinas of donor eyes provided by the Upstate New York Transplant Service. The protocol adhered to the provisions of the Declaration of Helsinki for research involving human tissue. Human cell lines and luciferase expression plasmids pXP1 and pXP2 ²¹ were obtained from ATCC (Manassas, VA). All procedures involving animals were approved by the institutional animal care and use committee and were performed according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Structure of the Rk Promoter Region

Isolation of the human Rk genomic clone G1, which contains the 5′ flanking sequence, the first two coding exons of the human Rk gene, and the 5′ portion of the third exon gene has been described. ²² After mapping the 16-kb DNA insert by partial-restriction digestion, the 2.0-kb EcoRI-SmaI fragment containing the flanking sequence upstream of the initiation codon was subcloned into a vector (BlueScribe, p-2.0; Stratagene, La Jolla, CA) and characterized by sequencing in both directions by an automated or manual ³³P-dideoxy nucleotide–based cycle sequencing approach with a commercial DNA polymerase (ThermoSequenase; Amersham Pharmacia Biotech, Piscataway, NJ). Sequence comparisons and analysis were performed by using BLAST (www.ncbi.nlm.nih.gov/blast/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD) or GCG BestFit algorithms (Accelrys, Princeton, NJ). 

The position of the transcription start site was determined by sequence analysis of Rk cDNA clones extending farthest into the 5′ untranslated region. In addition to sequencing previously retrieved cDNA clones from a random primer–generated human retinal cDNA library (Stratagene) additional Rk cDNA clones were generated by extension of Rk primers on a human retinal RNA template using an RNA ligase-mediated rapid amplification of cDNA ends kit (RLM-RACE; Ambion, Austin, TX). The inner and outer antisense primers used for the race amplification were complementary to codons 1 to 6 and 22 to 28, respectively. ²² After cloning into a vector (pGEM-T; Promega, Madison, WI), the cDNAs were sequenced. 

The position of start sites in relation to the 2.0-kb EcoRI-SmaI genomic sequence was further clarified by S1 nuclease and RNase mapping. ^{23 24} To generate the antisense single-stranded S1 probe from this region, a ³²P end-labeled primer complementary to positions +87 to +67 was cycle extended on a single-stranded m13mp18 template carrying the EcoRI-SmaI genomic fragment using the DNA polymerase (ThermoSequenase; Amersham Pharmacia Biotech). After the digestion of the product with ApaI, the 0.2-kb end-labeled single-stranded probe (−111 to +87) was retrieved from a denaturing polyacrylamide gel and hybridized to human retinal or control RNA (35 μg) in formamide-containing buffer. ²⁴ The hybridization mixtures were then incubated with 300 units of S1 nuclease in 1× S1 buffer for 1 hour at 37°C, and the protected hybrids were fractionated by electrophoresis on 6% denaturing polyacrylamide gel before autoradiography. For RNase mapping of the start site, antisense RNA probes corresponding to the exact same 0.2-kb genomic region were transcribed in vitro from an ApaI-linearized p-2.0 template in the presence of [α-³²P] UTP (MaxiScript kit; Ambion) and hybridized to 20 μg of total human retinal or control RNA before RNase A/T1 digestion (RPAIII kit, Ambion), electrophoresis, and autoradiography. 

Transient Transfection Assays of the 5′ Flanking Sequence Upstream of the Human Rk Gene

A series of Rk-Luc plasmids was constructed carrying the human Rk gene 5′ flanking sequences inserted upstream of the firefly luciferase gene in the pLuc vectors. ²¹ To make the ends of the human genomic sequence compatible with the polylinker region of the pLuc vectors, the 5′ ends of the 2.0-kb human EcoRI-SmaI and 0.2-kb ApaI-SmaI fragments were modified by insertion of a unique XhoI recognition sequence into either the EcoRI or ApaI site in p-2.0. The 2.0- or 0.2-kb XhoI-SmaI cassettes were then retrieved and inserted in forward or reverse orientation upstream of the luciferase gene in pXP1 or pXP2 to generate p-2.0Luc, p-2.0Luc(−), p-0.11Luc, and p-0.11Luc(−). Plasmids p-2.0h1Luc and p-0.11h1Luc were constructed from p-2.0Luc and p-0.11Luc, respectively, by replacing the wild-type H1 motif (TCTAATC; −29 to −23) with an inactive h1 sequence (AGATCTC). ^{25 26} Mutagenic oligonucleotide pairs corresponding to positions −36 through −4 were used in conjunction with a pfu-mediated mutagenesis kit (QuickChange; Stratagene) to introduce the substitutions into this region. After sequencing the segment to ensure accurate mutagenesis, the mutagenized region was then used to replace the corresponding wild-type segment to generate the final mutant Rk-Luc plasmids. An additional series of truncated Rk-luciferase constructs without segments of the 5′ end of the human flanking sequence was generated by unidirectional partial exonuclease III/S1 digestion starting from p-2.0Luc. After linearization with two adjacent restriction enzymes, XhoI and BglII, and fill-in protection of the BglII site with α-phosphorothioate dNTPs, the DNA was nuclease treated (Erase-a-Base kit; Promega) for various periods and recircularized for cloning. Nested deletions of the 2.0-kb region with segments missing from the 3′ end were generated by nuclease degradation, starting with SalI-HindIII–digested p-2.0Luc protected at the HindIII end. 

For transient transfection studies, low-passage WERI-RB1 retinoblastoma and Jurkat cell lines were grown in suspension in RPMI 1640 medium containing 10% fetal calf serum, and exposed to 2.5 μg of the above Rk-Luc constructs, together with 0.5 μg of pCMV-LacF in the presence of lipofection reagent (GenePorter I; Gene Therapy Systems, Inc., San Diego, CA) for 45 minutes in serum-free medium at 37°C. After subsequent recovery in serum-supplemented medium for 48 to 72 hours, the cells were harvested, lysed, and fluorometrically assayed for luciferase and β-galactosidase activities, with kits supplied by Tropix (Bedford, MA). Relative luciferase activities (RLA) for each construct were calculated by dividing the raw light units by the nuclear–cytoplasmic β-galactosidase activity and subsequently by the relative activity of promoterless pLuc. At least three separate transfections were performed with each of the Rk-Luc constructs, and the average ± SEM RLA was recorded for each. 

Generation of Rk-GFP Transgenic Mouse Lines

A composite plasmid was developed containing the 5′ flanking sequence of the human Rk gene arranged in series with the SV40 splice donor–acceptor site, the GFP gene, and the SV40 poly(A) sequentially downstream. The plasmid was constructed by incorporating the XbaI-SalI sequence from pTR-UF5 ²⁷ into the polylinker region of p-2.0. No viral enhancer sequences were present in the construct. The 0.11GFP (1.3-kb) and 2.0GFP Rk-GFP (3.2-kb) fragments were then retrieved from the composite plasmid by digestion with ApaI-SalI and XhoI-SalI, respectively, and used as transgenes after a purification step (Qiaquick; Qiagen, Valencia, CA). Transgenic mouse lines were generated by pronuclear injection of the transgene into BCF2 embryos (C57BL/10Ros-p^d x C3H/HeRos F2) followed by implantation into pseudopregnant mice. ²⁸ The progenies were screened for the presence of the GFP gene by PCR amplification of tail DNA for 30 cycles. ²² Sense and antisense PCR primers were complementary to GFP sequence at positions 267-286 and 417-398, respectively (GenBank accession no. U50963; http://www.ncbi.nlm.nih.gov/GenBank; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD). Transgene copy numbers were estimated by the comparing the intensity of GFP band amplified from Rk-GFP mice with those carrying Ren-GFP. ²⁸ Transgenic founders were crossed with C57BL/6 to propagate and maintain the lines. 

Analysis of Transgene Expression by RT-PCR

Total RNA was isolated from pooled transgenic tissues with standard CsCl-guanidium isothiocyanate or extraction reagent ²⁴ (TriZol; Invitrogen-Gibco, Grand Island, NY) and analyzed for GFP transcript. Reverse transcription reactions included 3 to 5 μg RNA annealed to random hexamers (50 μg/mL) in 1× reaction buffer (Promega) containing 1 mM dNTPs and 250 U Moloney murine leukemia virus (MMLV) reverse transcriptase in a final volume of 25 μL. The reverse transcriptase was excluded in control experiments to ensure that the bands detected were not the result of contamination by genomic DNA. PCR amplification of the reverse transcription products was performed as described earlier with either the GFP primer pair alone or with mouse actin amplimers in the same reaction. Mouse actin primers, both sense (positions 817-836) and antisense (positions 934-915), were chosen from adjacent exons to differentiate a band resulting from transcript versus genomic DNA amplification (GenBank accession no. NM007393). 

Analysis of Transgene Expression by Fluorescence Microscopy

For gross GFP detection, freshly harvested tissues from transgenic and nontransgenic mice were compared side by side under a dissecting epifluorescence microscope. Digitally recorded images were examined for fluorescence density and distribution. 

Confocal microscopy was used to further assess the distribution and cellular localization of GFP fluorescence. After a brief 10-minute fixation in 0.1 M sodium phosphate (pH 7.4) containing 4% paraformaldehyde at room temperature, dissected retinas and eyecups were washed with 10 mM sodium phosphate-buffered saline (PBS). For viewing as wholemounts, retinas were placed under 50% glycerol, with the outer segments contacting the overlaid coverslip. For identifying the layer of fluorescence, 30-to 50-μm-thick frozen eyecup sections were prepared and examined immediately after fixation under a confocal microscope. 

Immunofluorescent Staining of GFP in Transgenic Eyes

Immunostaining of the eye sections was performed by a modification of methods described elsewhere. ^{14 29} Transgenic mouse eyes were enucleated after death and placed in buffered 4% paraformaldehyde at room temperature for 1 hour. After fixation, the globes were washed in several changes of PBS and cryoprotected overnight at 4°C in PBS containing 30% sucrose. After freezing in optimal cutting temperature (OCT) compound, 10-μm frozen sections were taken onto slides (Plus; Fisher Scientific, Pittsburgh, PA) and stored at −80°C until further use. For immunolabeling, the slides were thawed, refixed, and washed in 10 mM Tris-buffered saline pH 7.4 (TBS) containing 0.1% Triton X-100 at room temperature. After the sections were blocked with 5% bovine serum albumin and 5% goat serum in TBS-Triton X100 at 37°C, they were incubated with a 1:500 dilution of the primary rabbit anti-GFP polyclonal antibody ²⁹ (kindly provided by W. Clay Smith, University of Florida, Gainesville, FL) in TBS-Triton X100 and 5% BSA. For double-staining experiments, the primary antibody cocktail was supplemented with a 1:1000 dilution of monoclonal Rk antibody D11 ³⁰ (a gift from Krzysztof Palczewski, University of Washington, Seattle, WA, also available through Affinity BioReagents, Inc., Golden, CO). After overnight incubation at 4°C, the slides were washed with TBS-Triton X-100 at room temperature and blocked again for 1 hour at 37°C. The sections were then reacted with secondary (red; Alexa-568; Molecular Probes Inc.) fluorescence-conjugated goat anti-rabbit antibody in TBS containing 5% bovine serum albumin, to visualize primary antibody-antigen complex. Fluorescence-conjugated (green; Alexa-488; Molecular Probes Inc.) goat anti-mouse was added to the secondary antibody cocktail for double-labeling experiments. After the sections were washed with the TBS-Triton X-100 solution, they were mounted with antifade medium (Mowiol; Calbiochem, La Jolla, CA) and viewed by epifluorescence confocal microscopy at FITC or rhodamine channels. Control slides contained nontransgenic globe sections processed identically in the absence of relevant primary antibodies. 

Results

In earlier work, we have identified a human Rk genomic clone, G1, spanning the first two coding exons: the 5′ portion of the third in addition to nearly 6 kb of DNA extending upstream of the gene. ²² To demonstrate the presence of transcriptionally active sequences within the upstream region and the role of these sequences in driving photoreceptor-specific expression, we analyzed the structure of the immediate 5′ flanking region and examined the activity of this region in human retinoblastoma cells and in transgenic mice. 

As a first step, we established the position of transcription start sites using primer extension and S1 nuclease and RNase protection mapping. The longest Rk cDNA clone, c1, retrieved from a random-primer–generated human retinal cDNA library contained 254 bp of untranslated leader sequence, placing the transcription start site at least 254 bp ahead of the initiation codon. No additional untranslated sequences were found among other Rk cDNA clones generated by the RLM-RACE primer extension approach. The genomic clone G1 also contained exactly the same noncoding sequences as the cDNA clones, further excluding the possibility of additional introns interrupting the 5′ Rk untranslated sequence in the genome. To further verify the position of start sites relative to the genomic sequence, S1 nuclease and RNase mapping were performed with labeled single-stranded probes derived from the ApaI-SmaI genomic fragment (−111 to +87) (Figs. 1A 1B 1C) . After hybridization to total human retinal RNA and digestion of residual single-stranded nucleic acids, the protected hybrids were fractionated by electrophoresis and visualized by autoradiography. A comparable collection of protected bands were observed by both mapping approaches (Figs. 1B 1C ; lanes 1, 5) suggesting heterogeneity of the transcription start site. Position +1 was arbitrarily designated based on the position of the cap site for the most distally initiated transcript detectable on RNase mapping (Fig. 1C) . No protected hybrids were detected with control RNA (Figs. 1B 1C ; lanes 2, 4). All detectable cap sites mapped by these approaches fell within 280 bases of the initiation codon. No Inr sequence was found, consistent with nonspecific transcription initiation near the site of the formation of the photoreceptor-specific transcriptional complex. ³¹ These observations on heterogeneity of the start site are furthermore consistent with the absence of a TATA-motif upstream of the Rk gene. ³²  

Sequence analysis of the 2.0-kb EcoRI-SmaI genomic region revealed the presence of multiple recognizable motifs upstream. A conserved putative homeodomain binding core H1 (TAATC) ³³ was noted, as in most other photoreceptor-specific promoters characterized to date, 28 bases upstream of the start site (Fig. 1D) . ^{19 25} Other partially conserved proximal elements were also noted, including a GC-rich segment (−46) and a GA-box (−77) similar to the sequences found in interphotoreceptor retinoid binding protein (IRBP) ^{34 35} β-phosphodiesterase (β-PDE), ^{36 37 38} and arrestin. ^{39 40 41} An inverted homeodomain binding site H2 (TAATT) ³³ resembling PCE/RET 1-like element ^{39 42} was present at −247. More distally, the genomic segment was composed of a large stretch of alternating, near perfect C₃- and C₄-AG(G/C)TGTG direct repeats (GenBank Acc. No. AY327580). No Nrl-Ap1 sites were found. ⁴³ Further comparison to other TATA-less promoters failed to reveal large stretches of sequence identity by BLAST and BestFit programs, even at low stringency. Comparison to other GRK 5′ flanking sequences was not possible, because these promoters have not yet been characterized. 

To study the transcriptional activity of the 5′ flanking region, we generated a series of plasmids carrying the Rk flanking sequences upstream of the luciferase gene and analyzed their activity by transient transfection (Fig. 2) . WERI-RB1 retinoblastoma cells transfected with p-2.0Luc carrying the full 2.0-kb flanking sequence showed luciferase activities on average 50 times higher than those transfected with pLuc. Truncated constructs missing as few as 24 bases from the 3′ end of the 2.0-kb region (p-2.0/0.03Luc) conferred only a fraction of the luciferase activity (0.2×) whereas those without the proximal segment altogether with positions between +63 and −217 missing (p-2.0/0.2Luc) showed no activity at all. In contrast substantial enhancement in activity was observed with truncation from the 5′ direction. Both p-0.5Luc and p-0.11Luc were five to six times as active as p-2.0Luc. The burst in activity was most evident with the elimination of the sequence between positions −556 and −718, with p-0.5Luc showing four times the activity of p-0.75Luc. This abrupt enhancement resulting from the removal of this narrow segment was in contrast to the relatively blunted increase resulting from elimination of nucleotides between −1904 and −719, as evident from comparison of p-2.0-, p-1.5-, p-1.0-, and p-0.75Luc activities. The relatively robust activity of the residual enhancer–promoter residual segment was orientation independent after removal of the positions −1904 to −113, with activity of p-0.11Luc essentially equaling p-0.11Luc(−). The activity, however, appeared to be relatively position sensitive, as judged from the inactivity of p-2.0Luc(−), which in effect carries the highly active, albeit inverted, 0.11-kb region separated from the luciferase gene by more than 1800 bp. Activities of the plasmids were uniformly four- to fivefold lower in Y79 retinoblastoma cells, reflecting the lower transfection efficiency of this cell type by 80% (data not shown). None of the constructs showed any luciferase activity in the nonphotoreceptor Jurkat cell lines, despite the presence of the comparable β-galactosidase activity in both retinoblastoma and Jurkat cells. These results suggest that the transcriptionally active sequences are largely confined to a short proximal enhancer–promoter segment from −112 to +87, with inhibitory sequences located distally and concentrated in the region between −556 and −718. 

The H1 core sequence (TAATC) is a highly conserved motif ³³ located in the midst of the active region. Identical or related sequences are found in most photoreceptor-specific promoters examined to date and are thought to serve as the binding site for photoreceptor-specific homeodomain transcription factor Crx. ^{19 25 26} However, given that the previous data on Crx-deficient mice suggest independence of Rk transcription from Crx function, we wondered whether the H1 site contributes to the activity of the promoter at all. To resolve this question, we mutated the site to eliminate homeodomain binding activity ²⁵ and examined the activity of the resultant plasmids in WERI-RB1 transient transfection assays (Fig. 3) . These plasmids p-2.0h1Luc and p-0.11h1Luc were both lower in activity by 60% compared with the plasmids carrying the corresponding wild-type sequence. The percentage of reduction did not appear to depend on the presence or absence of H2 sequence. This finding suggests that H1 sequence contributes to the activity of the promoter, regardless of the presence of the second putative homeodomain binding site. Mutagenesis did not alter the retinal cell specificity in vitro, as mutant plasmids remained inactive in Jurkat cells (data not shown). 

For further analysis of the function of the immediate upstream sequences in vivo we constructed and examined transgenic mice that carried the 0.11-kb enhancer–promoter sequence linked upstream of the GFP reporter gene. Of the mice generated, four were found to be positive for the transgene by PCR screening of genomic DNAs. Further, RT-PCR analysis demonstrated substantial levels of GFP transcripts in the globes of RK0.11GFP6 (Fig. 4A) . All other tissues tested negative for GFP in this line, despite amplification of comparable actin bands from all samples (Fig. 4B) . The actin band appeared weak in the transgenic eye, presumably suppressed by the robust competing amplification of GFP. A faint GFP band was detected in brain tissue, but only at 35 cycles of amplification consistent with possible expression in the pineal gland. ¹³ No amplification products were detected in the absence of reverse transcriptase. A separate transgenic line showed retina/photoreceptor-specific reporter expression but at lower levels (data not shown). Four others carrying the 2.0-kb promoter region linked to either GFP or LacZ reporter also showed retina-specific expression at substantially lower levels, consistent with the presence of a repressor in this fragment. Among the Rk-LacZ transgenic mice, reporter expression was strongest among the founders and diminished in subsequent generations, consistent with the possibility of epigenetic modification of some LacZ-based transgenes (data not shown). ^{44 45 46}  

The presence of the GFP in the RK0.11GFP6 transgenic retinal tissue was further verified by examination of freshly dissected retinas under blue light. Transgenic retinas appeared bright green compared with nontransgenic retinas (Figs. 4C 4D) and the fluorescence density was uniform throughout the transgenic retina, paralleling the expected distribution of Rk. Confocal microscopy of retinal wholemounts suggested differentially higher fluorescence in the optical sections of outer retina, as opposed to inner layers (data not shown). The results of these studies confirmed the presence of transcriptionally active elements within the 0.11-kb Rk flanking sequence capable of driving topographically accurate expression. 

Confocal microscopy was used to further examine the cellular localization of GFP fluorescence. Eyecups were briefly fixed and immediately viewed after frozen sectioning to minimize autofluorescence. The confocal micrograph in Figure 5A shows bright fluorescence confined to the outer layers of the retina occupied by photoreceptors and their processes, including the outer and inner segments. A lower level of fluorescence was also seen in outer nuclear layer and synaptic terminals. This pattern is consistent with primarily cytosolic distribution of GFP throughout the photoreceptors. No background was visible on nontransgenic sections (Fig. 5B) , except for a faint yellowish fluorescence that was barely detectable in enhanced micrographs. 

To further verify photoreceptor-restricted expression of GFP and to exclude any interference from yellowish green autofluorescent artifacts near the GFP detection wavelength, fully fixed transgenic globes were sectioned and immunostained with an anti-GFP polyclonal antibody (kindly provided by W. Clay Smith) in conjunction with a secondary red fluorescent antibody (Alexa Fluor, Molecular Probes Inc.). The micrographs in Figures 5C and 5E show that robust red fluorescence in transgenic globes was exclusively confined to the outer retinal layer and overlapped the location of the photoreceptors and their processes. The cellular distribution of red fluorescence (Fig. 5C) essentially matched the distribution of the GFP direct fluorescence in eyecups (Fig. 5A) except that outer nuclei were better outlined because of the added sensitivity of the indirect immunofluorescent approach. No other foci within the transgenic globe showed any red fluorescence. Only a very low-level nonspecific red fluorescence pattern was seen in the micrographs of nontransgenic eyes stained in the absence of primary antibody, even after prolonged exposure (Figs. 5D 5F) . None of the GFP-expressing mice showed any aberrations in photoreceptor morphology or reduction in the number of rows of outer nuclei compared with C57BL/6 mice (Figs. 5E 5F) , even up to the age of 1 year, consistent with the absence of any detectable GFP photoreceptor toxicity. 

To demonstrate colocalization of the endogenous Rk with GFP, we stained the transgenic globes with a combination of monoclonal anti-Rk and polyclonal anti-GFP antibodies. Secondary green and red antibodies were used to stain Rk green and GFP red. The direct green fluorescence from GFP was largely obscured by background autofluorescence in the fully fixed sections and did not interfere with the high quantum yield fluorescent signal expected from Rk bound to a green fluorescence-conjugated antibody. Double-labeled sections show uniform distribution of GFP epitopes stained red in cytosol (Fig. 5G , rhodamine channel) complemented with prominent specific green fluorescence from Rk immunoreaction concentrated in the outer segment (Fig. 5H , FITC channel). The preferential localization of Rk signal in the outer segments was anticipated reflecting the interaction of Rk with its membrane-bound light-activated rhodopsin substrates, ⁴⁷ whereas GFP was expected to be more uniformly distributed without any membrane localization signal. Nevertheless, the colocalization of the fluorescence is evident from uniformly yellow photoreceptor outer segments in the composite confocal micrograph (Fig. 5I) and provides compelling evidence for the photoreceptor-specific activity of the human 0.11-kb promoter, paralleling the specificity of the endogenous mouse Rk promoter. 

Discussion

The human Rk promoter represents the first of the GRK promoters to be characterized to date. Our initial structural and in vitro functional studies of the 2.0-kb 5′ flanking region suggested a clustering of transcriptionally active elements within the short 0.11-kb segment immediately upstream of the start site. Included in this narrow region was a highly conserved putative homeodomain binding site that contributed prominently to the overall transcriptional activity. The more distal sequences were largely composed of C₃ and C₄ repeats that appeared largely inhibitory. Additional studies in transgenic mice confirmed the photoreceptor-specific promoter activity of the short 0.11-kb region in vivo. The studies provided strong evidence for the transcriptional basis of photoreceptor-specific expression mediated through a short proximal region immediately upstream of the Rk gene. 

Among photoreceptor-specific genes characterized to date, Rk generally resembles Irbp, ^{34 35} arrestin, ^{39 40} and β-PDE ^{36 37 38} in promoter architecture and function. Comparable cis-acting elements are found in all these TATA-less promoters, including key homeodomain binding sites, GC-rich sequences, and GA-boxes concentrated within the vicinity of the start site immediately upstream of the gene. ⁴⁸ As in the Rk promoter, sequences tend to be most active proximally and the distal regions generally inhibit the overall promoter activity, both in vitro ^{38 48 49} and in transgenic mouse or Xenopus models. ^{37 41 50} Such parallels in architecture and function could mean that this group of relatively simple TATA-less promoters is coregulated by a set of mechanisms fundamentally distinct from those that govern specialized photoreceptor-specific genes, such as opsins with complex TATA-containing promoters linked to photoreceptor terminal differentiation. Subtle differences in expression profiles among Rk-like genes including rod-restricted expression of arrestin and β-PDE in contrast to expression of Rk and Irbp in both rods and cones could be explained by the divergence in the sequence in the proximal or distal promoters. 

The conserved proximally located H1 sequence (TCTAATCG) appears to be a major contributor to the overall basal transcriptional activity of the Rk promoter. This sequence is nearly identical with putative homeodomain binding sites found upstream of nearly all photoreceptor-specific promoters characterized to date. ²⁶ A cone–rod photoreceptor-specific homeodomain protein Crx, ^{25 26} responsible for terminal differentiation of the photoreceptors recognizes a consensus H1-related sequence (C/TTAATCC) ²⁶ making Crx a logical first candidate cognate for this site. However, recent studies on Crx^−/− mice suggest that transcriptional activity of Rk, arrestin, and Irbp may not be effected greatly by the absence of Crx activity, despite a profound decline in the transcription of opsin and opsinlike genes tied to the terminal differentiation of photoreceptors. ¹⁸ Otx2 has been proposed as the alternative transcription factor capable of driving some photoreceptor-specific genes in the presence or absence of Crx, as it has been retrieved by yeast one hybrid using the bait from the Irbp promoter sequence, even in the presence of Crx. ^{34 49} The role of Otx2 alone, however, remains controversial, given its more generalized distribution in the eye beyond the photoreceptor. ^{51 52} The decline in activity of the promoter with the mutagenesis of H1 sequence clearly suggests a functional role for this sequence in Rk promoter. Whether this sequence interacts at least in part with Crx, Otx2, or other homeodomain factors remains to be elucidated through future biochemical studies of the promoter. 

Elements other than H1 could participate in regulating the activity of the Rk promoter. The GC-rich and GA sequences may interact with Sp proteins, as previously documented for several other photoreceptor-specific proteins, including Irbp ⁵⁰ and β-PDE promoters. ^{37 38} More distally, the PCE-like sequence H2 could participate as a developmentally sensitive element by interacting with Rx/rax an antennapedia-type homeodomain factor expressed most prominently in retina and anterior brain during early embryonic development. ^{53 54 55} At this point, we do not have evidence supporting transcriptional activity of H2 in adult mice, but the possibility of its activity during embryonic phase cannot be excluded based on our data. The physiologic role of the more distal region containing the C₃ and C₄ repeats remains uncertain at this point; however, it is possible that this region may act as a regulatory silencer or shield adjacent genes against the effect of a powerful proximal promoter/enhancer segment. A major contribution from another key photoreceptor transcription factor Nrl is unlikely given that there are no identifiable sites in the Rk promoter so far examined and that the targeted disruption of Nrl leads to only a relatively small change in the expression of Rk. ²⁰ The possibility that this or other cognates may nevertheless act through remotely located sequences to regulate the expression in photoreceptors or even pineal ¹³ or cone photoreceptors ^{8 10} cannot be excluded in light of the current data. The precise overlap of GFP (red) and endogenous Rk (green) immunoreactivities in confocal micrographs (Fig. 5I) , evident from the uniformly yellow outer segments uninterrupted by any discernible green skip areas, suggests that the key elements governing expression in all photoreceptors, including cones, are contained within the 0.11-kb Rk promoter segment. Additional evidence in favor of the presence of both cone- and rod-responsive elements comes from the demonstration of the comparable in vitro activity of proximal promoter–enhancer in both conelikeWERI-RB1 and rodlike Y79 cell lines (data not shown). 

Rk plays an important role in protecting from light-induced apoptosis, and alterations in Rk levels could modify the susceptibility of photoreceptors to light. ⁵ Absence of Rk leads to light-induced retinal degeneration in mouse models, even under conditions of relatively low light, especially when compounded by additional mutations in other photoreceptor-specific genes. ⁷ Whether changes in Rk promoter activity could ultimately compensate for or exacerbate genetic deficits affecting photoreceptors and help avert or expedite the appearance of a clinically detectable retinal disease remains an open question. Previous studies designed to examine the role of Rk mutations in causing human retinal disease have primarily focused on alterations of the coding sequence. ^{10 11 56} Availability of the additional information from the present study opens the way to assessing the potential pathogenic impact of Rk promoter mutations in initiation and progression of retinal degenerative disease. 

 Supported by awards from Alexandrine & Alexander Sinsheimer (SCK); Knights-Templar Eye Foundations (SCK); National Eye Institute R01 HL48459 (KWG); and in part by Roswell Park Cancer Institute’s National Cancer Institute–funded Cancer Center Support Grant CA16056.

 Submitted for publication February 26, 2003; revised April 14, 2003; accepted April 15, 2003.

 Disclosure: J.E. Young, None; T. Vogt, None; K.W. Gross, None; S.C. Khani, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Structure of the 5′ flanking sequence upstream of the human Rk gene. (A) Partial restriction map and architecture of the region upstream of Rk gene. The 16-kb genomic insert from previously isolated clone G1 containing the immediate first two coding exons of the Rk gene was mapped by partial-restriction digestion followed by sequence analysis of the 2.0-kb EcoRI-SmaI segment upstream of the gene. Horizontal lines: the 2.0- and 0.11-kb regions most closely studied. The position of the key restriction enzymes and the conserved motifs (H1, H2, GC, and GA) and the repeat (C₃ and C₄) elements are shown. The enzyme restriction sites are designated E, EcoRI; A, ApaI; S, SmaI. The translation initiation codon is indicated by ATG. (B) S1 mapping of the transcription start site. Gel purified ³²P end-labeled single-stranded probe (10⁵ cpm) from the ApaI-SmaI genomic region was hybridized to human retinal RNA (35 μg, lane 1) or control tRNA (35 μg, lane 2) and digested with (lanes 1, 2) or without (lane 3) S1 nuclease, followed by denaturing gel electrophoresis. The start site region is indicated on the autoradiogram by parentheses. The sequencing ladder was generated with the same end-labeled oligonucleotide used to synthesize the single-stranded probe. Arrowhead: position of the undigested probe. (C) RNase mapping of the start site. Uniformly labeled antisense RNA (10⁵ cpm) derived from the ApaI-SmaI fragment containing 0.11-kb of the Rk upstream region was hybridized to tRNA (lane 4) or total human retinal RNA (lane 5) and digested with RNase A/T1 followed by electrophoresis. Arrowhead: position of the most distal start site designated +1 on the autoradiogram. No stable hybrids were detected with the ³²P-labeled sense RNA as the probe (data not shown). (D) Sequence of the region surrounding the transcription start sites. The position of key potentially active elements, including the bicoid-type homeodomain response element (H1) and the two elements found in other TATA-less genes, including the GC-rich and GA sequences, are underscored. Long, left-pointing arrow (D): position of the oligonucleotide used for generation of sequencing ladder and S1 probe; short, right-pointing arrows (A): location of direct repeats. Right-angled arrows (A, D): the distal start site. The coding sequence including the initiation codon are boxed.

Figure 2.

 

Transcriptional activity of the human Rk promoter segments. Fragments of the Rk 5′ flanking sequence derived from the 2.0-kb EcoRI-SmaI region by restriction or exonuclease III/S1 digestion were linked upstream of firefly luciferase gene in the forward or reverse (−) orientation in promoterless pXP plasmids (pLuc). The constructs (2.5 μg) were transfected into (A) WERI-RB1 or (B) Jurkat-B cells, together with pCMV-LacF (0.5 μg). The luciferase and nuclear–cytosolic β-galactosidase activities were measured fluorometrically, and the normalized luciferase activities were determined relative to β-galactosidase activity and promoterless plasmid pLuc. The results, averaged from at least three different transfections, are shown in the histogram, together with the SEM. The numbers in italics represent the exact position of the ends of the fragments relative to the start site.

Figure 3.

 

H1-dependent transcriptional activity of the promoter region. p-2.0h1Luc and p-0.11h1Luc were constructed by oligonucleotide-mediated site-directed mutagenesis from p-2.0Luc and p-0.11Luc, respectively. The sequence of the mutagenized region and relative activities of the constructs in WERI retinoblastoma cells are shown.

Figure 4.

 

Retina-specific expression of GFP in a transgenic mouse line carrying the 0.11-kb Rk-GFP transgene. Lines carrying the transgene were screened for (A) GFP transcripts or (B) GFP and actin transcripts, both by PCR amplification of the reverse transcription product from total RNA (3 μg) from various tissues. The amplified products are separated on a 1% agarose gel, and the positions of GFP (131-bp; filled arrowhead) and actin bands (117-bp; open arrowhead) are shown. Freshly dissected retinas from the GFP-expressing RK0.11GFP6 (+) line and nontransgenic C57BL/6 mice (−) were compared under incident blue (C) and white light (D), and the gray-scale images were recorded with and without the GFP filter.

Figure 5.

 

Cellular localization of GFP in photoreceptors of the Rk-GFP transgenic line. Micrographs show green fluorescence emission recorded from briefly fixed frozen sections of transgenic RK0.11GFP6 (A, +) and control (B, −) eyecups after excitation at 418 nm by a blue-green argon source. Immunofluorescent detection of GFP in whole globes. Sections of transgenic GFP-positive (+) (C, E) and control (−) (D, F) globes were incubated with anti-GFP antibody (1:500) followed by (1:2000) red fluorescence-conjugated secondary antibody, mounted in antifade medium, and examined by epifluorescence microscopy. Fluorescent (C, D) and phase contrast with fluorescent overlay (E, F) images were recorded at 20× magnification to maximize the field. Double immunofluorescence labeling with polyclonal anti-GFP (1:500; red) and anti-Rk monoclonal D11 (1:1000; green) antibody after incubation with the primary and secondary antibody cocktails. I represents G and H images merged. OS, outer segment; IS, inner segment; ON, outer nuclear layer; ST, synaptic terminals.

The authors thank Mary K. Ellsworth and Colleen Kane for expert and indispensable technical help in generation, maintenance, and propagation of transgenic lines; Joanne Ballard for help in propagation and maintenance of the cell lines; Richard Palmiter (University of Washington, Seattle, WA) for providing the plasmid pCMV-LacF; Nicholas Muzyczka (University of Florida, Gainesville, FL) for providing the plasmid pTR-UF5; and W. Clay Smith (University of Florida, Gainesville, FL) for the gift of anti-GFP antibody and Krzysztof Palczewski (University of Washington, Seattle, WA) for the anti-Rk antibody.