Xenopus Rhodopsin Promoter IDENTIFICATION OF IMMEDIATE UPSTREAM SEQUENCES NECESSARY FOR HIGH LEVEL, ROD-SPECIFIC TRANSCRIPTION*

,

Phototransduction occurs in the photoreceptor layer of the vertebrate retina, which is composed of distinct cell types: rods and cones (1). These cells express a number of specific proteins that regulate the light-dependent currents mediating vision (2,3). Among these cell-specific proteins are the visual pigments, which combine with 11-cis-retinal to form the light-sensitive component of the transduction cascade. The visual pigments are a large family of genes, which contain rod-specific rhodopsins and at least four classes of cone-specific opsins (4). Rhodopsin, required for nocturnal vision, is the most abundant opsin in many vertebrate retinae by virtue of the size of the rod outer segments, abundance of the rod cells, and the high level of transcription. As such, the regulation of rhodopsin expression has been a focus for understanding mechanisms of cellspecific gene expression in the retina (5).
Transcription initiation has been identified as the major control point for rhodopsin gene expression (6,7). A variety of studies using different approaches have demonstrated that important transcriptional control sequences lie within the 5Ј upstream regions of various rhodopsin genes. Functional assays using transgenic mice have shown that 2-4 kb 1 of upstream sequences from the mouse and bovine rhodopsin genes direct reporter gene expression to the photoreceptor layer (8,9), with sequences proximal to the initiation site (Ϫ500 and Ϫ222) being sufficient for retina-specific expression. The importance of the proximal sequences is highlighted by the high degree of homology found in this region among vertebrate rhodopsins (10). However, the immediate upstream sequences from the mammalian opsins were not able to limit expression of the reporter to rods as expected for rhodopsin (11,12). Expression levels are also regulated by a sequence termed rhodopsin enhancer region, located ϳ2 kb upstream of the initiation site (13). The binding of retina-specific nuclear factors to rhodopsin upstream sequences have been localized in both proximal and distal upstream regions, suggesting a role for these elements in regulating expression (5). However, a complete description of the cis-acting elements that control transcription in rod photoreceptors is not yet available. Transcription factors that potentially regulate gene expression have been identified in the mammalian retina and several have been shown to activate rhodopsin expression in heterologous systems, e.g. Nrl (14,15), Crx (16), and Erx (17). The mechanisms by which the different cis-acting elements in the rhodopsin upstream regions function, either independently or in concert, to produce rod-specific expression are not known.
We have used Xenopus embryos for transient transfection studies and transgenesis to investigate rod-specific transcription. Previously, we found that a 5.5-kb rhodopsin upstream fragment was transcriptionally active, driving the expression of reporter both in Xenopus embryo transfections (10) and in transgenic frogs (18). In this paper, we map the rhodopsin promoter to an upstream region spanning nucleotides Ϫ508 to ϩ41, capable of directing reporter expression to the abundant rod cells as detected by immunolocalization and GFP expression in transgenic Xenopus rod cells. Furthermore, we have used mutational analysis and DNase footprinting to define the functional limits of the promoter to the proximal ϳ233 nucleotides and found multiple regulatory regions that contribute to the transcriptional activity in rod photoreceptors.
5Ј Deletions-The deletion series was constructed using exonuclease III and mung bean nuclease digestion (Stratagene). The plasmid containing a deletion of the TATA box region, pXOP(Ϫ503/Ϫ46)luc, was constructed by digestion of pXOP(Ϫ503/ϩ41) with HindIII and subsequent religation.
Targeted Disruptions-Targeted disruptions of conserved regions in the Xenopus rhodopsin promoter (Ϫ503/ϩ41) were generated using a PCR-based approach using primers (Table I) with a PstI overhang (19). To generate the ⌬(Ϫ84/Ϫ58), the PCR primer (Table I) contained a HindIII overhang, permitting direct cloning of the product into pXOP(Ϫ52/ϩ41)luc. A second construct that disrupted Ϫ84 to Ϫ58 but maintained the spacing of the native promoter was made by cloning a synthetic oligonucleotide (5Ј-CTTGTACGGAGCTCTACTGTGCA-3Ј) into the PstI site in the ⌬(Ϫ84/Ϫ58) construct. PCR was performed using Ultma DNA polymerase (PerkinElmer Life Sciences). The 5Ј products were digested with KpnI-PstI, the 3Ј products were digested with PstI-BamHI, and constructs were generated by three-part ligation of KpnI-PstI product with the PstI-BamHI product. These were cloned directionally into the KpnI-BglII sites of pGL2 (Promega, WI). All mutant promoter constructs were verified by sequencing. The ⌬(Ϫ136/ Ϫ122) contained a single base change (G to C at Ϫ386) in addition to the intended replacement of the Ret-1 site by PstI.
GFP Constructs-The XOP basal region (Ϫ52/ϩ41) was cloned as a HindIII-BamHI fragment into complementary sites in pEGFP (Ϫ) (18) to generate EGFP-XOP basal. This plasmid was digested with KpnI and HindIII and gel-purified. Rhodopsin-targeted deletion luciferase constructs were digested with KpnI/HindIII and cloned into the KpnI-HindIII sites of pEGFP XOP basal to generate rhodopsin-targeted deletion constructs in EGFP. The ⌬(Ϫ52/Ϫ38) KpnI-HindIII fragment was cloned directly into EGFP digested with the same enzymes. The clones were verified by restriction digestion and sequencing. Plasmid preparations were performed using the Qiagen protocol (Chatsworth, CA).
Basal Element Replacement Constructs-Wild type Xenopus rhodopsin promoter luciferase construct (Ϫ503/ϩ41) was digested with HindIII to liberate the basal region (Ϫ52/ϩ41). The 93-bp basal region was replaced by synthetic oligonucleotides encoding Xenopus rhodopsin basal region (Ϫ52/Ϫ37) and Xenopus cardiac actin (CAR) basal region (Ϫ36/ϩ36) or with the CMV promoter basal region (Ϫ36/ϩ28) with HindIII linkers on either end. All constructs were confirmed by sequencing, and plasmid DNA preparations were performed using the Qiagen protocol.
Cardiac Actin Luciferase Constructs-The cardiac actin promoter (Ϫ3270/ϩ23; Ref. 20) was liberated as a KpnI-HindIII fragment and cloned directionally into KpnI-HindIII sites in pGL2 (Promega, WI). The clones were verified by restriction analysis and sequencing of junctions to confirm the orientations. Luciferase constructs containing Xenopus cardiac actin TATA and ϩ1 (pCARTATA) were generated by removing the XOP (Ϫ503/Ϫ52) region from XOP-CAR using HindIII and SmaI. The ends were filled in using Klenow and religated. This construct contains XOP (Ϫ52/Ϫ37 and the cardiac actin basal region (Ϫ36/ϩ36) driving the expression of luciferase.
Artificial Promoter Construct-Artificial promoter was made by design of synthetic oligonucleotides (IDT, Coralville, IA). p(Ϫ96/ϩ41)GFP contained wild type sequence of the rhodopsin proximal promoter from Ϫ96 to ϩ41. Oligonucleotides were designed with overhanging KpnI and HindIII sites, ligated into the same sites in the XOP basal construct, and confirmed by sequencing.

TABLE I
List of primer sequences used in generating targeted deletion and artificial constructs The gene-specific forward primers were used in PCR reaction with GL2 3Ј primer (5Ј-CGGGATCCAAGCTT-ACCAACAGTACCGGAATGCC-3Ј). The gene-specific reverse primers were used in conjunction with the GL2 5Ј primer (5Ј-GGGGTACCTGTATCTTATGGTACTGTAACTGA-3Ј).

Construct
Sense primer (5Ј-3Ј) Antisense primer (5Ј-3Ј) CACTAGGATTAATAGTGCGCTAAATCCTTTGTTC GTGACGCTGGGGGTTGCA AGCTTGCAACCCCCAGCGTCACGAACAAAGG ATTTAGCGCACATATTAATCCTAGTGGTAC a No sense primer was necessary for this construct (see "Experimental Procedures"). and 1:3, respectively. DNA used for transfections were either purified by twice banding in CsCl (24) or by the Qiagen protocol. At least two different preparations were tested for each plasmid. Groupwise comparisons of mean activities (RLU/head) generated in individual transfection trials were performed using single-factor analysis of variance (␣ ϭ 0.05 for planned comparisons; Ref. 25). Transfection of trunks and whole embryos were done as controls. For luciferase assay, the heads (6 -10) were homogenized and aliquots were assayed (using a protocol supplied by Promega) to determine the picograms of luciferase produced per well. The luciferase assay gave 50,000 -100,000 relative light units/pg of purified luciferase (Sigma)/30 s using the single-channel luminometer (Berthold). Protein estimations were done using the Bradford reagent (Bio-Rad) and bovine serum albumin standards, and ranged from 10 to 20 g/head. Luciferase activity is reported as the mean (including all trials, n given in figure legend) Ϯ S.E.

Transgenic Xenopus
DNA was digested with XhoI to linearize the plasmid and purified after digestion (High Pure PCR Product Purification Kit, Roche Molecular Biochemicals), with final elution in water. Transgenic Xenopus embryos were produced using restriction enzyme-mediated integration as described (23). Normal embryos were selected and cultured in 0.1ϫ MMR at 18 -20°C until approximately stage 41 (3-4 days), at which point non-mosaic green fluorescence above the weak background fluorescence from the yolk could be observed. Normal tadpoles were maintained in 0.1ϫ MMR and used for analysis. GFP-positive transgenic tadpoles were fixed overnight in phosphate-buffered 4% paraformaldehyde. Frozen sections (10 -15 m thickness) were photographed under bright field and fluorescent lighting conditions (18). Photographic slides were scanned and figures produced using Photoshop (Adobe). The brightness and contrast of the images in sections from the various ⌬ construct transgenic tadpoles were adjusted in order to visualize the cell type expressing the GFP reporter. At least three or four independent transgenic lines for each construct were sectioned and analyzed. The pattern and intensity of GFP expression between the various lines for any one construct was consistent.

DNase I Footprinting
Nuclear proteins were extracted from adult Xenopus retinae as described (26). Glutathione S-transferase-tagged bovine Crx homeodomain and flanking regions (residues 37-107, GST-CrxHD) and a hexahistidine-tagged mouse Nrl (residues 16 -237, His-Nrl) were overexpressed and purified from Escherichia coli as described previously (16,27). 32 P-End-labeled DNA fragments of the Xenopus rhodopsin proximal promoter (Ϫ279/ϩ40) was generated by PCR amplification, and footprinting reactions were carried out as described previously (16). Varying concentrations of purified transcription factors (as indicated in Fig. 2 legend) were used in the footprinting reactions. 1 g of poly(dI-dC) was included only in the reactions containing retinal extracts.

Vertebrate Rhodopsin Upstream Sequence Comparisons-
Vertebrate rhodopsin 5Ј-flanking regions possess significant interspecies homology, with many short stretches of sequence identity in the 250-bp region immediately upstream of the TATA box/initiation site (Fig. 1). The overall sequence identity in this region ranged from 45% to 55% over a 200-bp stretch of sequence between Xenopus rhodopsin and other vertebrate rhodopsin sequences, suggesting a potential role for these conserved sequences in rhodopsin gene regulation. Several of the conserved regions in XOP are similar to binding sites for transcription factors in mammalian promoters: Ret4-like (Ϫ53/Ϫ38), which is a binding site for Crx (16,28); NRE-like (Ϫ121/Ϫ110 and Ϫ84/Ϫ58), which is a binding site for a retina specific leucine zipper protein, Nrl (15,29); the GATTA repeat (Ϫ106/Ϫ91), which is a binding site for Crx (16) and potentially Rx1; the photoreceptor cell element PCE I/Ret1 (Ϫ133/Ϫ127), found in many retina-specific genes, which is present immediately upstream of the GATTA sequence in rhodopsin genes (30,31); and the Ret2/E box region (Ϫ158/Ϫ188) (13,32). Such high proximal promoter sequence homology highlights a potential conservation of the mechanism regulating rhodopsin gene expression in vertebrate photoreceptors.
DNase Footprinting-In order to locate binding sites for potential transcription factors, DNase footprinting was performed on the rhodopsin proximal promoter. Using an adult Xenopus retinal extract, numerous extended regions were protected from digestion, including those predicted from sequence comparisons in the proximal promoter ( Fig. 2A). At present, no Xenopus homologues of Crx, Nrl, or other rod-specific transcription factors have been identified. Therefore, footprinting of the proximal promoter was done using recombinant mammalian proteins purified from E. coli: the bovine Crx homeodomain (GST-CrxHD; Ref. 16) and the DNA binding domain and surrounding bZIP regions of murine Nrl containing a hexahistidine tag (His-Nrl; Ref. 27). Both proteins protected the rhodopsin proximal region (Fig. 2, B and C). GST-CrxHD produced an extended footprint encompassing the Crx consensus sites from Ϫ153/Ϫ73 on the sense strand and Ϫ70/Ϫ156 on the antisense strand, respectively. Nrl protected a number of regions (Ϫ81/ Ϫ56 on the sense strand and Ϫ57/Ϫ84 on the antisense strand) that included an AP1 consensus region and NRE, adjacent to two Crx sites. These results strongly suggest that the sequences in the Xenopus rhodopsin proximal promoter contain binding sites for members of the Otx2-related and neural retinal leucine zipper families.
Xenopus Rhodopsin Proximal Promoter: Analysis by Transfection of Xenopus Embryos-A comparison of promoter activity was performed using an improved transfection protocol (23), which allows direct comparison between the mean activities observed for different DNA constructs, without normalization with a second reporter plasmid. Initial experiments were per- formed using three upstream fragments: (Ϫ5500/ϩ41, Ϫ1300/ ϩ41, and Ϫ508/ϩ41), which were all capable of driving tissuespecific expression in embryos (Table II). Comparative transfections performed at equimolar concentrations showed that the activity from the Ϫ5500/ϩ41 fragment was ϳ35% lower compared with the other two promoters, which did not differ, although this difference was not significant at p Ͻ 0.05. Size of the plasmid used in the transfection appeared to have little influence on reporter gene expression, as two derivatives containing additional sequences produced similar results (Table II). One of the derivatives contained the Xenopus rhodopsin gene and 3Ј sequences, suggesting no significant transcriptional control regions downstream of the initiation region. Unlike that observed in heads, reporter expression from the upstream sequences in trunks was Ͻ2-fold over that obtained using GL2, which produced a relatively consistent but very low level of activity compared with heads transfected without DNA. We conclude that the major regulatory elements are located in the 508-bp proximal region.
To further characterize the transcriptional control sequences in the proximal promoter, transgenic tadpoles were generated (18,33) using XOP(Ϫ508/ϩ41) driving expression of GFP. GFP expression was found only in the eye and, transiently, in the pineal. In fixed sections of transgenic retina, GFP expression was observed only in rods (see below). Each rhodopsin-positive cell also expressed GFP, with expression apparent by stage 40 (data not shown). The levels of GFP expression for the three large upstream constructs were qualitatively very similar, confirming the results from the transfections: that there is apparently no significant difference between these constructs. Taken together, these results demonstrate that the immediate upstream rhodopsin sequences direct expression to the rod cells.
Mapping the Rhodopsin Proximal Promoter-To identify cisacting elements in the XOP proximal promoter, a series of mutations in pXOP(Ϫ503/ϩ41)luc were generated by either selective removal of nucleotides or by sequential DNA deletions from the 5Ј end (Fig. 3). The 92-bp encompassing the transcription initiation site and including the TATA region were essential for activity, as deletion of this region from the 503-bp upstream fragment decreased luciferase expression 190-fold (Fig. 3). Activity from the Ϫ503/Ϫ46 fragment was equivalent to the promoterless control. However, the 92-bp region (Ϫ46/ ϩ41) in either orientation or as a tandem repeat gave background levels of expression, indicating that these nucleotides, although necessary, were not sufficient for transcription of reporter plasmids.
Analysis of luciferase activity from 5Ј promoter deletions identified other sequences required for promoter function. All of the deletion constructs, except XOP(Ϫ44/ϩ41), yielded luciferase activity significantly above that obtained using GL2, in transfections of embryo heads (Fig. 3). No significant differences in luciferase activity were observed when nucleotides Ϫ508 to Ϫ234 were deleted (Fig. 3), indicating that these nu- a Activity measured as RLU/head is presented as: mean Ϯ S.E. and data analyzed using analysis of variance (single-factor; ␣ ϭ 0.05) yielded p Ͻ 0.13 (*) and p Ͻ 0.11 (**) relative to Ϫ5500/ϩ41. cleotides do not contribute significantly to the head-specific transcription from the Ϫ503/ϩ41 rhodopsin promoter region. Deletion of the region spanning position Ϫ233 to Ϫ203 enhanced luciferase activity 2.4 -2.8-fold over that observed using any of the four larger constructs (Fig. 3). No significant changes in activity compared with Ϫ202/ϩ41 were obtained when nucleotides Ϫ202 to Ϫ171 were deleted. A drop in activity was observed with the deletion of Ϫ170 to Ϫ145, with activity comparable to that obtained using the full fragment, Ϫ508/ ϩ41. Thus, these results are consistent with a second regulatory region from Ϫ170 to Ϫ146. A further decrease in luciferase activity (4.5-fold) compared with XOP(Ϫ145/ϩ41) was found when Ϫ145 to Ϫ128 was deleted. A 40-fold stimulation of luciferase expression from the Ϫ127/ϩ41 fragment relative to that obtained using XOP (Ϫ44/ϩ41) indicated that the sequence spanning positions Ϫ127 to Ϫ46 is the smallest fragment sufficient for promoter activity.
To test for tissue specificity, embryonic trunks were transfected using the various deletion constructs and activity was compared with the corresponding activity in head by normalization to total protein. When compared with the activity measured in heads, all trunk activities except that of XOP(Ϫ127/ ϩ41) were less than 4% of the head activity. Trunk activity from all the deletion constructs was slightly elevated compared with Ϫ503/ϩ41 and were also higher than GL2, suggesting that some nonspecific, low level transcription may occur when nucleotides Ϫ503 to Ϫ330 are deleted. However, these results show predominant head-specific expression driven by sequences further upstream of proximal promoter sequences.
To further characterize the Xenopus rhodopsin proximal promoter, targeted disruptions were created to replace putative regulatory elements with a short linker sequence (Table I). To test if the selective disruption resulted in expression in nonretinal cells, the mutant constructs were simultaneously analyzed by measuring reporter activity in transfected heads and trunks. A disruption of the region from Ϫ233 to Ϫ203 of the Xenopus rhodopsin upstream sequence (XOP 4) resulted in a 3-fold increase in reporter gene expression in heads (Fig. 4). This increase in activity in heads did not result in any significant change in reporter activity measured in transfected trunks. These observations are consistent with the increase in activity associated with the 5Ј deletion construct lacking the Ϫ233/Ϫ203 region. The removal of Ϫ170 to Ϫ146 (XOP 3) did not alter the expression levels in transfected heads or trunks (Fig. 4). Selective disruption of the Ret1-like core region and flanking nucleotides (Ϫ136 to Ϫ122) resulted in a 2-3-fold increase in luciferase expression in heads as compared with the wild type promoter. This increase in expression was not accompanied by any significant change in the levels of reporter expression in trunks. Alteration of one of the NRE/AP1-like sites (Ϫ120 to Ϫ109) did not significantly affect the head-specific expression of the rhodopsin promoter or lead to any change in activity in trunks. However, disruption of the second NRE-like site (Ϫ84 to Ϫ58) dramatically reduced expression (Fig. 4). This was not caused by a change in spacing since an additional construct in which the spacing of the wild type rhodopsin promoter was maintained by addition of a short insert of random sequence was also significantly lower than the Ϫ503/ϩ41 promoter construct. These results indicate an essential role for this NRE-like element in maintenance of high level expression of rhodopsin in Xenopus rod cells. A change in either one (Ϫ98 to Ϫ91) or both (Ϫ107/Ϫ91) of the GATTA sequences caused a small reduction in expression of the transgene, 20% or 32%, respectively, that was not statistically significant. Finally, replacement of nucleotides Ϫ52 to Ϫ38, corresponding to the conserved Ret4 region in the bovine rhodopsin upstream sequence, did not change either the reporter luciferase levels or head-specific expression from the wild type promoter.
Role of the Xenopus Rhodopsin Basal Element-To address the contribution of the basal region in determining the specificity of expression from the Xenopus rhodopsin promoter, the sequences encoding the TATA region, initiation site, and sur- rounding nucleotides were tested. The XOP basal region comprising nucleotides Ϫ36/ϩ41 was replaced by basal element regions from two non-retina-specific promoters, Xenopus cardiac actin (Ϫ36/ϩ36; Ref. 20) and CMV (Ϫ36/ϩ25) promoters (Fig. 5A). These basal regions do not possess any obvious sequence homology to the XOP basal region (Fig. 5B). Replacement of the XOP basal element with either cardiac actin or CMV basal region did not significantly affect the expression levels of the hybrid constructs as compared with wild type rhodopsin in transfected heads or trunks (Fig. 5C). These results show that the heterologous basal regions can functionally substitute for the XOP basal region and that the upstream cis-acting sequences are capable of controlling the cell specificity of rhodopsin expression.
Transgenic Xenopus-To investigate the cellular expression patterns from the altered rhodopsin promoters, transgenic Xenopus were generated with the various deletion constructs driving expression of GFP. Several transgenic lines were generated for each construct, and GFP expression was analyzed in tadpoles. Transgenic animals harboring deletions of (Ϫ233/ Ϫ203) and (Ϫ136/Ϫ122) exhibited significant enhancement of GFP fluorescence, which was restricted to the eye. Animals transgenic for XOP promoter deletion of (Ϫ52/Ϫ38) showed GFP expression predominantly restricted to the eye, with weak fluorescence in pineal and anterior head structures in a few animals. The intensity of fluorescence in the eye was comparable to that of transgenic tadpoles generated with the wild type (Ϫ508/ϩ41) rhodopsin construct. Deletion of Ϫ170 to Ϫ146 produced animals that exhibited much lower levels of GFP expression, which was localized to the eye. Transgenics with deletions of (Ϫ120/Ϫ109), (Ϫ98/Ϫ91) or (Ϫ107/Ϫ91) had essentially wild type expression levels. Deletions of the second NRElike site (Ϫ84/Ϫ58) resulted in animals with extremely reduced levels of fluorescence in the eye, with expression fading to undetectable in some animals. However, the expression from these four constructs was strictly limited to the eye. To identify the cell type expressing the reporter, transverse sections of retina were analyzed by bright field and fluorescence microscopy. Within the eye, the expression of GFP from all the targeted deletion constructs was confined to the rod photoreceptor cells in the retina irrespective of the differences in fluorescence levels (Fig. 6). These experiments show that the mutant rhodopsin promoters are capable of directing rod-specific expression in vivo. This extends and confirms the observed head specificity of reporter expression in the transfection approach using the various constructs and demonstrates a redundancy in regulatory element function in targeting the expression of rhodopsin to the rod photoreceptors.
Artificial Promoter Constructs-To test the role of the GATTA and NRE-like element in targeting expression to rods, an artificial promoter construct containing the Ϫ96/ϩ41 region upstream of GFP was used to generate transgenic animals. (Fig. 7). GFP expression was seen in the eyes as well as the brain and portions of the spinal cord extending from the base of the brain to varying distances along the tail. Later in development, GFP expression became more cranially restricted and eventually faded to undetectable levels. These results suggest that the GATTA region and the NRE-like element can support reporter gene expression in the central nervous system during the early stages of development. However, high level rod-specific expression requires the presence of other sequences in the Xenopus rhodopsin proximal promoter. DISCUSSION To study the mechanisms of rod-specific transcription in the vertebrate retina, we have focused on the analysis of the Xenopus rhodopsin promoter. Using a dual approach of embryo transfections and transgenesis, in vivo functional characterization of a rod-specific promoter has been performed in photoreceptors. This environment is permissive for photoreceptor cellspecific transcription, even though during transfection partial disruption of cell contacts prevents normal rod outer segment formation (34). The transfected heads are harvested at the onset of rhodopsin expression and photoreceptors express endogenous rhodopsin to levels similar to untreated heads at the same stage. 2 The levels of luciferase with the complete promoter (Ϫ503/ϩ41) are Ͼ100-fold higher than the promoterless control, and direct comparison between various constructs allowed us to quantitate the effects of removing individual cisacting elements. Transgenic approaches in Xenopus have proven to be a powerful complementary approach for the analysis of cell-specific promoters (18,33), permitting the in vivo analysis of transcription in chromatin (35,36). In the restriction enzyme-mediated integration method, multiple copies of the plasmid integrate into random loci in the genome (18,33). include the conserved Ret 4-core sequence that was included while synthesizing the XOP-CAR and XOP-CMV constructs. C, luciferase activity normalized to the wild type promoter (WT) from four to six independent transfection experiments using heads or trunks was plotted. Activity levels from the wild type cardiac actin construct and CAR and CMV promoter in heads and trunks are shown for comparison.
To control for potential variation due to copy number and sites of integration, independent transgenic lines were studied. Variation due to the site of integration and/or cosuppression of supernumerary copies of exogenous transgenes prevented direct comparison of expression between individual animals (37,38). Qualitative comparison of relative expression levels for all animals made with the targeted deletion and intact GFP reporter constructs in transgenic experiments was comparable to the quantitative results found in transfection experiments with the luciferase reporter constructs, except in the case of ⌬XOP 3 (see below). The pattern and temporal onset of transgene expression in native photoreceptors was reproducible among animals generated with the same construct. Therefore, transgenic animals permitted a clear morphological identification of cellspecific expression, in a retina with roughly equal numbers of rods and cones (39). The combined approaches of transient transfection and transgenic analysis permitted a detailed examination of cis-acting elements in the Xenopus opsin promoter.
We have found four cis-acting regulatory elements in the Xenopus rhodopsin proximal promoter, two of which function as positive regulatory elements and two as negative regulatory elements. The sequences between Ϫ84 and Ϫ58, which contain a match to the binding site TG(N 6 -8 )GC(A/C/T) for Nrl (class V binding site; Ref. 27), are required for high level transcription in both transgenic animals and transfected retina. Our data suggest that these nucleotides contain an Nrl-like binding site, which in mammalian promoters is located in the same position relative to the transcriptional start site (15).
The Xenopus rhodopsin promoter contains a number of potential Otx2-like binding sites (TAAT or ATTA) between Ϫ154 to Ϫ72. A GATTA repeat, previously identified as BAT-1, shows a high degree of conservation among the opsin promoters (40) as well as other photoreceptor-specific genes (41). The BAT1 region in the bovine opsin promoter has been shown to be a binding site for HMG I(Y) (29,42) and to contribute to the promoter activities in transfected cells, but not in in vitro transcription assays (28). Surprisingly, targeted disruption of this region in the Xenopus promoter caused no significant change in either transcriptional levels or pattern of expression (Figs. 4 and 6). Moreover, the presence of a GATTA sequence upstream of the basal promoter was not sufficient to support a high level of rod-specific expression in transgenic Xenopus (Fig.  7). The GATTA repeat may serve different functions in the different photoreceptor gene promoters. In the case of the IRBP promoter, which drives expression in both rods and cones, deletion of the GATTA element dramatically reduced transcription in transfected chick retinal cultures (43) and transgenic mice (44). In this case, the Ret-1 GAATTA site was not sufficient to overcome the mutations in the GATTA sequence. Due to the high degree of sequence conservation across species, it is tempting to propose that the GATTA region in XOP does have a functional role. Our inability to detect a significant effect in the BAT-1 mutation experiments (Figs. 4 and 6) could be explained by the presence of redundant elements in the proximal promoter.
Targeted disruption of the conserved cis-acting element, Ret-1/PCE I in the Xenopus opsin promoter resulted in a 2-fold increase in transcriptional activity in transfected retina and increased GFP expression in transgenic animals. The discrepancy between the results obtained in the 5Ј deletion and targeted disruptions indicates that the regulatory properties of the Ret-1 element may be determined by upstream cis-acting sequences, perhaps through interactions with transcription factors bound to these elements. This supports our results obtained using synthetic promoter constructs, containing one or more copies of the Ret-1 region (Ϫ144 to Ϫ120) upstream of the rhodopsin basal region (Ϫ51/ϩ41), which contribute to 0.1% or less of the Xenopus rhodopsin promoter's transcriptional activity in transient transfections (data not shown). Several proteins have been shown to bind to the Ret-1/PCE I sequence including Crx (16), Rx (31), and Erx (17), each of which functions as a weak transcriptional activator in transfection assays. However, these proteins most likely interact with other retinal transcription factors, as seen in the case of a direct interaction between Nrl and Crx (16,45), resulting in synergistic activation of the bovine opsin promoter. It is unclear if the spacing of the Ret-1 sequence relative to the transcriptional start site affects its regulatory role. The targeted disruptions may have resulted in a promoter conformation more favorable to the assembly of transcription factors, thereby stimulating the transcriptional activity.
The region between Ϫ170 to Ϫ146 (XOP 3) and Ϫ233 and Ϫ203 (XOP 4) contains sequences that were footprinted with adult Xenopus nuclear extracts ( Fig. 2A). In retinal transfections, deletion of Ϫ233 to Ϫ203 caused a 3-fold increase in transcription that was also seen in transgenic animals, suggesting that this region contains a negative regulatory element. The sequences in the XOP 4 region do not possess any sequence similarity to the mammalian opsin promoters or a recognizable consensus transcription factor binding site. Although there is evidence for a negative regulatory region in mammalian opsin FIG. 6. Rod-specific expression of GFP reporter in transgenic Xenopus generated with different targeted disruption constructs. Bright field images of cross sections of representative transgenic tadpoles generated using wild type (Ϫ503/ϩ41) or targeted disruption constructs showing the pigment epithelium (PE) and different cellular layers (photoreceptors (PR), inner nuclear layer (INL) in the retina. The lens (L) and optic nerve (ON) are seen only in a few sections. Fluorescence image of the same sections show GFP expression limited to rods. Wild type section magnification, ϫ20; mutant magnification, ϫ40. In order to identify the cell type expressing GFP photographs were taken with varying exposure times, as a result fluorescence intensities do not represent actual levels of expression seen in individual tadpoles. promoters, the sequences do not share any similarity with XOP 4. There is limited sequence similarity between the Xenopus and mammalian promoters in the XOP 3, only encompassing an E-box motif, CANNTG at nucleotide Ϫ163 to Ϫ158. In the mouse opsin promoter, the equivalent E-box sequences were shown to bind MASH-1 using mobility shift assays (46). The Xenopus homologues of MASH proteins (XASH-1 and XASH-3) are only expressed in the ciliary margins in the laminated retina (47) and not in the photoreceptors. Therefore, if XOP 3 function is regulated by its conserved E-box motif, then either a different class of bHLH proteins bind to this region in the Xenopus opsin promoter, or the effect of XOP 3 on opsin expression is exerted at a step prior to photoreceptor differentiation. Targeted disruption of these sequences, however, showed no change in transfected retina but greatly reduced levels of reporter expression in transgenic animals. The discrepancy between the two assays for this construct suggests that protein binding to XOP 3 deletion may be sensitive to the chromatin environment or the proteins that bind to XOP 3 may be part of the chromatin-remodeling complex. This region contains an AT-rich sequences (at least five consecutive A/T at Ϫ152 to Ϫ147). Further experiments are needed to determine if HMG I(Y) actually binds to the Xenopus proximal promoter and activates transcription via any of these sequences. FIG. 7. The BAT 1 and NRE sequences in an artificial XOP promoter construct drives GFP expression in the developing central nervous system. A, schematic representation of an artificial promoter construct containing wild type XOP promoter sequences from Ϫ96 to ϩ41 driving expression of the GFP reporter gene. The construct includes the 3Ј GATTAATA sequence of the BAT1 region (Ϫ96/Ϫ89), the NRE consensus sequence TG(N) 8 GC (Ϫ73/Ϫ62) denoted by boxes, and the basal promoter region (Ϫ61/ϩ41). The sequences upstream of Ϫ96 include a short (6 bp) random linker used for cloning. B, developmental time course of GFP expression in transgenic tadpoles generated using the Ϫ96/ϩ41 GFP construct. Transgenic tadpoles were produced and analyzed for GFP expression from 3 to 7 days post injection (dpi). Bright field and fluorescent images of tadpoles generated with Ϫ96/ϩ41GFP (top panel) showing GFP expression in eye, brain, and spinal cord 4 -7 days after injection. The background auto fluorescence from the yolk is seen in developing tadpoles injected with either Ϫ96/ϩ41 XOP-GFP or sperm nuclei alone (nontransgenic controls, lower panels).