A Novel, Evolutionarily Conserved Enhancer of Cone Photoreceptor-specific Expression*

The α subunit of cone transducin (TαC) is expressed exclusively in cone photoreceptors of the eye and pineal. TαCisakey phototransduction protein, and inherited mutations in TαC cause total color blindness in humans. We use transgenic zebrafish to identify and characterize cone photoreceptor regulatory element 1 (CPRE-1) a novel 20-bp enhancer element in the TαC promoter (TαCP). CPRE-1 is located ∼2.5 kb upstream of the translation start site and is necessary for strong cone photoreceptor-specific expression in vivo. CPRE-1 comprises of a modular arrangement of two 10-bp elements that have separate, but co-dependent transcriptional activities. In vitro, CPRE-1 specifically binds nuclear factors that are enriched in ocular tissue. Bioinformatic alignments reveal that CPRE-1 sites are evolutionarily conserved in the promoter regions of fish, rodent, and mammalian TαC orthologues and identify a 5′-CTGGAGTG(A/T)TGGA(G/A)GCAGGG(G/C)T-3′ consensus sequence.

The vertebrate retina contains distinct cone and rod photoreceptors that mediate scotopic and photopic vision, respectively (1). Although cones and rods originate from the same population of retinal progenitor cells, they have unique gene expression profiles that account for their differential cell fate, morphology, and signal transduction mechanisms. For example, many components of the G protein-coupled receptor phototransduction cascade are encoded by cone-or rod-specific genes (1). The molecular genetics underpinning cone photoreceptor-specific gene expression remain poorly defined.
Several transcription factors, including Crx, Nr2e3, Nrl, and Tr␤2, regulate photoreceptor-specific gene expression. Cone rod homeobox is expressed in photoreceptors, and cone rod homeobox depletion leads to a developmental loss of both cone and rod photoreceptors in mice and zebrafish and to blindness in humans (2)(3)(4). Nr2e3 and Nrl are expressed in rods where they repress expression of cone-specific genes (5)(6)(7)(8)(9)(10). Muta-tions in Nr2e3 or Nrl lead to rods inappropriately expressing cone-specific markers and to the human retinal disease enhanced S-cone syndrome (5)(6)(7)(8)(9)(10)(11)(12). Thyroid hormone receptor ␤2 is a nuclear receptor required for M-cone development (13,14). Targeted deletion of thyroid hormone receptor ␤2 leads to a loss of M-cone function with an increase in functional S-cones (13,14). Although these transcription factors are known to regulate photoreceptor genes, the molecular regulators of conespecific gene expression are not well defined.
One method of deciphering the molecular mechanisms regulating cone-specific expression is to characterize the cis-elements in a cone-specific promoter. Several studies have identified large promoter regions sufficient to direct transgene expression in cones (15)(16)(17)(18)(19)(20)(21)(22)(23). However, many of these promoter regions have weak activity, exhibit ectopic, non-cone expression, or require a heterologous enhancer element (16 -19, 21-23). With the exception of a cone rod homeobox-binding element, none of these studies have characterized individual cis-elements that direct cone-specific gene expression in vivo (16).
GNAT2 encodes the cone-specific ␣-subunit of cone transducin (T␣C) 3 that is expressed in all cone types and binds to activated cone opsin receptors (24). Missense mutations in the human T␣C gene are linked with achromatopsia, a recessive disease leading to total loss of color vision (24 -26). Similar mutations in the zebrafish T␣C gene result in loss of visual behavior (27). The cone-specific expression of T␣C makes its promoter an ideal candidate for identifying cis-elements controlling cone-specific expression. Previous studies determined that a 277-bp fragment of the human T␣C promoter combined with a heterologous 214-bp interphotoreceptor retinoid binding protein enhancer element is sufficient to drive expression in murine cones (18). However, characterizing functional T␣C cis-elements in mice is limited by the costs and technical expertise associated with transgenics and the limited number of cones in mice. The cost-effective and genetically tractable zebrafish is an attractive alternative model to characterize in vivo, cis-elements controlling cone-specific expression (28,29).
Like humans, zebrafish have unique spectral cone subtypes with distinct cytology and synaptic connectivity that mediate color vision (28,30). Zebrafish cones are divided into three morphological types; the UV-sensitive long single cones, the blue-sensitive short single cones, and the red-sensitive and green-sensitive double cones (30,31). In vivo studies in zebrafish have identified large regulatory regions that direct cone photoreceptor gene expression (15,19,32). Transient transgenic studies identified a 105-bp cis-region of the zebrafish UV opsin promoter that can alter a rod-specific promoter to express in rod and UV cones (15). Recently a 500-bp region directing M-cone specific expression was identified that is sufficient to induce a non-retinal keratin 8 promoter to express in zebrafish M-cones (32). Previously we reported that a 3.2-kb promoter fragment of the zebrafish T␣C gene is sufficient to drive strong EGFP expression in zebrafish cone photoreceptors and that the most distal 1.2-kb sequence contains an enhancer-like activity (19). A 55-bp T␣C promoter fragment enhances EGFP expression in cone photoreceptors. A, schematic of the promoter-reporter portion of the 5Ј zebrafish T␣CP deletion constructs and corresponding activity in vivo. The number of transgenic fish scored for each construct is indicated at the right of the histograms. Enhancer-1 activity localizes to a 55-bp region (highlighted in red) between Ϫ2521 and Ϫ2466 bp of T␣CP. B, examples of the activity levels used to determine the activity of 5Ј zebrafish T␣CP deletion constructs. C-G, confocal micrographs of retinal sections from 5-dpf embryos injected with Ϫ2521 T␣CP. C-E, optical sections demonstrate that the EGFP-expressing cells generated co-localize with double cones (Zpr-1), UV (UV opsin), and blue cone (blue opsin) photoreceptors (highlighted by white arrows). F, EGFP-expressing cells do not co-localize with rod photoreceptors (rhodopsin). G, all EGFP-positive cells are situated in the outer nuclear layer (ONL). INL, inner nuclear layer; GCL, ganglion cell layer.
To identify this enhancer of cone-specific expression, we performed an in vivo analysis of the 3.2-kb zebrafish T␣C promoter using deletion, mutant and chimeric reporter constructs, and in vitro characterization of binding to eye nuclear protein.
Here we describe a novel, evolutionarily conserved 20-bp enhancer of the zebrafish T␣C promoter that preferentially binds eye nuclear protein, is active upstream of a heterologous promoter, and is necessary for driving strong cone-specific expression.

EXPERIMENTAL PROCEDURES
Generation of Reporter Constructs-The Ϫ3173-bp T␣CP-EGFP-1 plasmid based on the pEGFP-1 plasmid vector (Clontech) was used as the initial template for all constructs. The Ϫ3173(⌬Ϫ2698/Ϫ2300) deletion was constructed by removing a 397-bp ScaI-BmgBI fragment and re-ligation. The 5Ј deletions, internal mutants, Ϫ2510/Ϫ2501 "RR" repeat and "LR" inversion were generated by introducing PCR fragments into Ϫ3173 T␣CP-EGFP-1 digested with PstI and EcoRI. The UV zebrafish opsin promoter (zUVOP) was PCR-amplified from a zebrafish genomic library clone (Stratagene) and inserted into the pEGFP-1 vector. Chimeric constructs were engineered by introducing PCR fragments upstream of the Ϫ897-bp zUVOP minimal promoter. The 3 and 6 copy repeats of "LR" were generated by cloning synthetic oligonucleotides upstream of the Ϫ897-bp zUVOP. A spacer sequence was introduced by PCRamplifying the kanamycin cassette from the EGFP-1 vector and cloning between the Ϫ897-bp zUVOP and enhancer repeat sequence using SpeI restriction sites. The (Ϫ2521/Ϫ890-bp zT␣CP)-(Ϫ836-bp human T␣C promoter)-EGFP-1 construct was generated by digesting (Ϫ2521/Ϫ890-bp zT␣CP)-(Ϫ897-bp zUVOP)-EGFP-1 with SpeI and BamHI and cloning a Ϫ836-bp human T␣C promoter fragment PCR-amplified from genomic clone pCYPAC2 (RZPD GmBH). All constructs were confirmed by DNA sequencing.
Embryo Microinjections-Covalently closed circular plasmids were suspended in 0.1 M Tris (pH 7.9) plus 0.1% Phenol FIGURE 2. Enhancer 1 preferentially binds eye nuclear protein. A, schematic of overlapping 42-bp probes within the 55-bp enhancer-1 sequence. B, electrophoretic mobility shift assay demonstrates that Probes 1 and 2 bind nuclear protein isolated from adult zebrafish eyes. C, electrophoretic mobility shift assay with probe 1 shows a larger quantity of radiolabeled probe 1 is bound by eye nuclear protein (ENP) than body nuclear extract (BNP) suggesting that binding of the trans-factor(s) is more highly expressed in ocular tissue. This interaction is specific, because unlabeled probe 1 at 50-or 100-fold molar excess is able to compete away binding of eye nuclear protein to labeled probe 1, whereas a non-related cold competitor for clusterin at 50-or 100-fold molar excess is not. Eye nuclear protein (ENP), body nuclear protein (BNP), cold competitor (CC).

TABLE 1 Number of EGFP؉ cells from transient transgenic fish sections that label with rod or cone opsin antibodies
Parentheses indicated number of EGFPϩ cells and the number that labeled with cone or rod opsin antibodies is highlighted in bold.

Enhancer of Cone-specific Expression
Red (Sigma) at a final concentration of 25 ng/l. The pressure injector was regulated so that ϳ300 pl (ϳ7.5 pg) of each reporter construct was injected per embryo. The covalently closed plasmids were injected into the cytoplasm of 1-2 cell stage zebrafish embryos using an air pressure-regulated Pico Pump (World Precision Instruments) attached to a Narishige micromanipulator and needle holder. Injected embryos were transferred to water containing 0. Screening for EGFP Expression-Injected embryos were anesthetized and analyzed by fluorescence microscopy using a Zeiss Axioplan 2 fluorescence microscope at 5 days post fertilization (dpf). Construct activity was scored on the basis of a semiquantitative assay, where one of three ordinal activity levels was assigned to an embryo based on the number of fluorescent cells within the eye. Embryos with Ͼ50 GFPϩ cells in the eye were assigned an activity level of ϩϩϩ, fish with 5-50 GFPϩ cells were assigned as ϩϩ, 1-5 fluorescent cells as ϩ, and fish with no GFPϩ cells were assigned an activity level of Ϫ (15). The sum of each ordinal level was divided by the total number of embryos to give relative activity. Statistical analysis was performed using Student's t tests, and p values of Յ0.05 were considered significant.

Nuclear Protein Extract Preparation and Electrophoretic
Mobility Shift Assay-Nuclear protein extracts were prepared from adult zebrafish eyes. Adult zebrafish eyes were dissected and homogenized in 4 ml of homogenization buffer (50 mM Hepes/pH 7.9, 25 mM KCl/5 mM MgCl 2 /200 mM sucrose/0.5 mM dithiothreitol/0.25 mM phenylmethylsulfonyl fluoride (Sigma)), centrifuged at 2000 ϫ g for 10 min at 4°C. The pellet was re-suspended in 600 l of Chelsky buffer (10 mM Tris, pH 7.0/10 mM NaCl/3 mM MgCl 2 /30 mM sucrose/1% Nonidet P-40), incubated on ice for 10 min, and centrifuged at 3000 ϫ g for 10 min at 4°C. Pellets were re-suspended in 600 l of Chelsky buffer plus 10 mM CaCl 2 , layered on sucrose cushion (50 mM Hepes, pH 7.9/25 mM KCl/5 mM MgCl 2 /876 mM sucrose/0.2 mM EDTA/0.5 mM dithiothreitol/0.25 mM phenylmethylsulfonyl fluoride), and centrifuged at 3000 ϫ g for 10 min at 4°C. Nuclei were re-suspended in Chelsky buffer c-low salt (20% glycerol/20 mM Tris-Cl pH 7.9/25 mM KCl/0.2 mM EDTA/0.5 mM dithiothreitol/0.25 mM phenylmethylsulfonyl fluoride), lysed by the dropwise addition of a high salt Chelsky buffer c-low salt plus 1.2 M KCl, and homogenized. Samples were incubated on ice for 30 min and centrifuged at 13,000 -25,000 ϫ g for 10 min at 4°C. Supernatant was removed and concentrated by ultra centrifugal filter devices (Amicon ultra 4). Protein content was determined by using the BCA Protein Assay Reagent (Pierce). Oligonucleotide probes (supplemental Table S1) were synthesized with 5Ј overhangs and radioactively labeled with [␣-32 P]dCTP using Klenow Exo-(New England Biolabs). Nuclear extracts were incubated with a 200-fold molar excess of poly(dI-dC) at 4°C for 10 min. Labeled probe was added, and the mixture was incubated at room temperature for 30 min. Reactions were stopped and loaded on 5% non-denaturing polyacrylamide gels, followed by autoradiography. For the competition assay, cold competitor was added prior to addition of labeled probe, in 50-, 100-, and 200-fold molar excess.
Sequence Analysis-The 10-kb 5Ј promoter sequence of GNAT2 genes were obtained from ENSEMBL and aligned with the 20-bp CPRE-1 sequence using ClustalW (35). Only sequences with a pairwise alignment score greater than 0.61 were considered significant. The consensus CPRE-1 sequences was generated using the online Weblogo server (36).

A 55-bp Distal Promoter Region Is Necessary for Strong EGFP Expression in Cone
Photoreceptors-In transient transgenic assays, the Ϫ3173-bp fragment of the zebrafish T␣C promoter (T␣CP) region drives robust EGFP expression specifically in retinal cone photoreceptors and in the pineal (Fig. 1, data not shown). The expression pattern is identical to that observed in FIGURE 3. Refinement of the enhancer 1 sequence to CPRE-1 a 20-bp region containing separate, but dependent, functional domains. A, probes containing 10-bp mutations in probe 1 were designed to further narrow the trans-factor(s) binding site. Introduced mutations are highlighted by boxes, and the relative position of the deletion constructs are illustrated by arrows. B, electrophoretic mobility shift assay of probes 1, 1M L, 1M R, and 1M D with eye nuclear extract. All three mutant probes are unable to form a complex with eye nuclear extract, and probes 1M L and 1M R are unable to compete away binding of eye nuclear extract to labeled probe 1. C, to test the functional significance of the 10-bp mutations on CPRE-1 activity in vivo, constructs containing corresponding 10-bp deletions were tested. In agreement with the in vitro assay deletions of 10-bp regions from Ϫ2521 to Ϫ2510 bp (L) and Ϫ2510 to Ϫ2501 bp (R) of T␣CP resulted in a significant loss of functional activity in vivo. A significant reduction in EGFP expression levels is also observed when R is mutated. When L is mutated a statistically significant reduction in EGFP expression is observed (C and D). A duplicate RR combination is insufficient to recapitulate activity to that of the wild type LR arrangement. The enhancer activity is also partially dependent on orientation, because when the LR was inverted and reversed there is a significant loss of EGFP expression. (All constructs were injected as covalently closed plasmids.) D, p values for comparisons of construct activities (*, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.0005). NSS, no statistical significance.
the Tg(3.2T␣CP-EGFP-1) transgenic line generated using the same promoter fragment (19). To identify cis-elements controlling this cone photoreceptor-specific expression, we tested the reporter activity of a series of 5Ј deletions of the Ϫ3173-bp T␣CP in vivo.
Deletion of the region between Ϫ3173 and Ϫ2749-bp of zebrafish T␣CP results in a significant reduction of reporter activity, revealing a potential enhancer element, designated "enhancer region 2," within this 424-bp region (Fig. 1A). However, deletion of the 283-bp region between Ϫ2749 and Ϫ2466-bp of the T␣CP, designated "enhancer region 1," practically abolishes EGFP expression (Fig. 1A). An internal deletion of enhancer region 1, within the Ϫ3173 bp T␣CP construct, results in minimal reporter activity (Fig. 1A). These results suggest that enhancer region 1 is necessary for high level cone-specific expression and that enhancer region 2 is not sufficient to compensate for loss of enhancer region 1.
We focused on refining the enhancer element (enhancer-1) in enhancer region 1. Analyses of additional 5Ј deletions initially narrowed enhancer-1 to the 55-bp region, between Ϫ2521 and Ϫ2466 bp of T␣CP (Fig. 1A). The Ϫ2521-bp T␣CP construct results in robust expression, with Ͼ34% of injected embryos expressing EGFP in the retina. The Ϫ2521-bp T␣CP-EGFP construct confines transgene expression to all cone photoreceptor types based on expression in the outer nuclear layer and co-immunolocalization of EGFP with markers for double, short single, and long single cones, but not with a rod photoreceptor marker (Fig. 1, C-F, and Table 1). Deletion of the 55 bp between Ϫ2521 and Ϫ2466 bp of T␣CP results in a significant loss of EGFP expression, with Ͻ10% of injected embryos expressing EGFP in only in one to five cells per eye (Fig. 1A).
Eye Nuclear Protein Binds Specifically to Enhancer-1-The ability of enhancer-1 to bind eye nuclear protein was evaluated by testing three overlapping oligonucleotide probes that span the 55-bp sequence of Ϫ2521 to Ϫ2466 bp of T␣CP ( Fig. 2A). Probe 1 and Probe 2 form distinct complexes with eye nuclear protein, but Probe 1 became the focus of our studies as it generates a more intense complex (Fig. 2B).
The specificity of eye nuclear protein binding to Probe 1 was analyzed. Addition of Probe 1 as a cold competitor at 50 and 100 M excess demonstrates a specific ability of unlabeled Probe 1 to compete for nuclear proteins binding to Probe 1 in a concentration-dependent manner (Fig. 2C). An unrelated competitor is unable to compete away binding of nuclear protein to Probe 1 (Fig. 2C). In addition, non-retinal nuclear protein isolated from zebrafish body is capable of binding to Probe 1, although at a much reduced intensity compared with eye nuclear protein (Fig. 2C). This suggests an enrichment of trans-factors in the eye that specifically bind to enhancer-1.

CPRE-1 Comprises Two 10-bp Functional cis-Elements-
To narrow the minimal sequence of enhancer-1, three mutant probes containing sequential 10-bp mutations were analyzed by electrophoretic mobility shift assay (Fig. 3A). The three mutant probes fail to form a complex with eye nuclear protein (Fig. 3B). However, when used as cold competitors, Probe 1M R is unable to compete, Probe 1M L partially competes, and Probe 1M D efficiently competes away binding of eye nuclear protein to wild-type Probe 1 (Fig. 3B). Thus, the 20 bp spanning Probes 1M L and 1M R, designated "cone photoreceptor regulatory element 1" (CPRE-1) appear essential for trans-factor(s) binding to enhancer-1 (Fig. 3A). To test the significance of CPRE-1 for transcriptional activity in vivo, constructs containing complete or partial deletions of CPRE-1 were tested (Fig. 3C). The Ϫ2501-bp T␣CP-EGFP construct had limited activity compared with the Ϫ2521 bp T␣CP-EGFP construct demonstrating that CPRE-1 is necessary for robust cone-specific expression (Fig. 3C). 5Ј deletion of the sequence from Ϫ2521 and Ϫ2510 bp of T␣CP resulted in a significant loss of EGFP expression, and deletion of the sequence from Ϫ2510 and Ϫ2501 bp of T␣CP results in a further statistically significant loss of EGFP expression (Fig. 3, C and D). The in vitro and in vivo assays suggest that CPRE-1 comprises two separate cis-elements, "R" located between Ϫ2521 and Ϫ2510 bp of T␣CP and "L" located between Ϫ2510 and Ϫ2501 bp of T␣CP. To identify the functional importance of the L and R cis-elements, they were mutated independently in the Ϫ2521-bp background and tested in vivo. Mutation of either L or R results in a significant loss of transcriptional activity compared with the Ϫ2521-bp T␣CP construct, with mutation of R having a more pronounced affect (Fig. 3, C and D). However, duplication of R is insufficient to recapitulate the transgene expression levels of the L plus R arrangement, demonstrating a co-dependence of the modules for enhanced transcriptional activation (Fig. 3, C and D). The ability of CPRE-1 to increase transcription appears orientationdependent as inverting and reversing the L and R modules results in a significant reduction in transgene expression, compared with their wild-type orientation (Fig. 3, C and D).
Identification of L and R Sequences Essential for trans-Factor(s) Binding-To identify the base pairs necessary for binding of trans-factor(s) to CPRE-1, the ability of L and R probes, with consecutive 3-bp mutations, to bind eye nuclear protein was tested (Fig. 4, A and B). The 6-bp sequence from Ϫ2518 to Ϫ2512 bp (5Ј-GAGTGT) of L is essential for binding, because mutation of these residues in Probes 1M L.2 and 1M L.3 abolishes the ability to form a complex with eye nuclear protein and to compete with binding to wild-type Probe 1 (Fig. 4C and Table 2). For R, only the 3-bp sequence from Ϫ2507 to Ϫ2504 bp (5Ј-GGC), mutated in Probe 1M R.2, is essential for binding ( Fig. 4D and Table 2). The sequences mutated in Probes 1M L.4 and 1M D.1, although not essential for binding, result in less intense complexes suggesting a role in stabilizing the complex. A significant role for the thymidine nucleotide at position Ϫ2511 bp is demonstrated as Probe 1M L.4 forms complexes much less efficiently than overlapping Probe 1M R.1, and this nucleotide is the distinguishing mutation (Fig. 4, C and D, and Table 2).

CPRE-1 Is Necessary but Not Sufficient to Drive Expression of a Heterologous Promoter in Vivo-
The ability of CPRE-1 to drive the expression of a heterologous zebrafish UV opsin promoter (zUVOP) was characterized (Fig. 5A). We identified a Ϫ800-bp fragment of the zebrafish UV opsin proximal promoter region that directs low but specific EGFP expression in UV cones (Fig. 5A, data not shown). Chimeric constructs comprising of the heterologous UV opsin promoter with either the Ϫ3173/Ϫ890-bp or Ϫ2521/Ϫ890-bp regions of the distal T␣CP upstream of the UV opsin promoter are sufficient to enhance levels of EGFP expression (Fig. 5, A and B-F). Although the Ϫ2521/Ϫ890-bp T␣CP fragment originates from a promoter expressing in all cones, in the chimera it is not sufficient to override the UV cone specificity of the heterologous promoter nor to direct expression of the ubiquitous SV40 minimal promoter to cone photoreceptors (Fig. 5, A and B). However, when CPRE-1 is removed (Ϫ2501/Ϫ890 bp) from the chimeric construct a significant loss in transgene expression was observed, confirming that CPRE-1 is necessary to drive expression from the heterologous zUVOP promoter (Fig. 5A). CPRE-1 is not sufficient to enhance expression from the heterologous promoter, because single or multiple copies cloned upstream of the minimal UV opsin promoter do not increase expression from the minimal UV opsin promoter (Fig. 5A). To determine if this insufficiency relates to spatial context, six repeats of the 20-bp "LR" CPRE-1 sequence were placed 1721 bp upstream of the UV opsin promoter by introducing an unrelated spacer sequence. No increase in expression from the heterologous promoter was observed (Fig. 5A), suggesting that the LR CPRE-1 sequence requires interaction with specific cis-ele-

TABLE 2 The position of 3-bp mutations on mutant probes in relation to enhancer-1 sequence
In Probe 1 the L cis-element is in bold, and the R is in italic. Altered bases base pairs are underlined; engineered overhang sequences are shown in lowercase italic letters.

Enhancer of Cone-specific Expression
ment(s) in the Ϫ2501/Ϫ890-bp T␣C distal promoter. These dependent elements are not functionally or spatially conserved in the human T␣C proximal promoter, because replacement of the spacer sequence with the Ϫ2453/Ϫ847 bp of the human T␣C proximal promoter did not increase transgene expression (Fig. 5A). Thus, CPRE-1 is necessary, in the context of the Ϫ2521/Ϫ890 distal zebrafish T␣CP-UV opsin chimeric construct, to enhance transgene expression from a heterologous promoter. However, it is not sufficient to enhance transcription from the heterologous promoter nor to overwrite the UV cone specificity of the minimal promoter by itself.
CPRE-1 Is Conserved in T␣CP Orthologues-Alignment of the CPRE-1 sequence with orthologous T␣C promoters reveals a conserved copy within divergent vertebrate species (Fig. 6A). The location of the CPRE-1 site is also conserved among more closely related species, e.g. primates and rodents. Bioinformatics reveal a CPRE-1 consensus sequence of 5Ј-CTGGAGTG(A/ T)TGGA(G/A)GCAGGG(G/C)T-3Ј (Fig. 6B), which does not match any known transcription factor site in the Transfac data base (37). We searched for CPRE-1 consensus sites in the promoter regions of other zebrafish cone-specific (arr3, cnga3, gngt2, pde6C, and pde6h), rod-specific (rho, pde6a, and gnat1), and housekeeping genes (tubulin and ␤-actin). A significantly matching consensus site was identified in the cone-specific arrestin (arr3) and cone cyclicgated nucleotide channel ␣3 subunit (cnga3) genes. Thus, CPRE-1 is an evolutionarily conserved enhancer element in a cohort of cone-specific genes.

DISCUSSION
Cone photoreceptors are highly specialized sensory neurons that enable color vision and visual acuity. Severe forms of human blindness associated with cone defects include rare genetic diseases, e.g. achromatopsia, and prevalent complex diseases, e.g. age-related macular degeneration. To develop our understanding of the molecular genetics of cones we focused on identifying cis-elements in the zebrafish T␣C promoter region.
This study identifies and characterizes cone photoreceptor regulatory element-1 (CPRE-1) a 20-bp enhancer of cone-specific expression. CPRE-1 is located in the distal zebrafish T␣C promoter region, ϳ2.5 kb upstream of the coding sequence. CPRE-1 is necessary for robust, cone-specific expression, and regulates expression in all cone photoreceptor types. CPRE-1 comprises a modular arrangement of two 10-bp cis-elements. These L and R elements have separate but co-dependent transcriptional activities that are required to drive high levels of cone-specific expression. The activity of CPRE-1 appears orientation-dependent as inversion and reversion of LR result in significantly less activity than the wildtype orientation. CPRE-1 is not sufficient to drive expression from heterologous promoters. However,  a larger distal fragment of the zebrafish T␣CP is sufficient to enhance expression from a heterologous UV cone-specific promoter, and CPRE-1 is required for this enhanced activity. Notably, CPRE-1 does not override the UV specificity of the heterologous promoter. Thus, CPRE-1 is an enhancer module that increases the rate of transcriptional activity, but not the tissue specificity, of cone-specific promoters in vivo.
Enhancement of transcriptional activity by CPRE-1 may result from direct binding of transcriptional activators on site or by modification of the chromatin architecture to enable access of transcription factors to the promoter region. However, in the transient transgenesis assay, the reporter constructs are microinjected as closed-circular DNA, which does not integrate into the zebrafish genome, but remains as episomal DNA (38). Chromatin assembly on episomal DNA is a "time-dependent" process with the majority of episomal DNA not bound by chromatin until ϳ5 weeks after DNA administration (39). At 1 and 7 days post DNA administration there is ϳ5to 10-fold less episomal DNA bound by chromatin than at 5 weeks (39). Because our reporter constructs are assayed for activity at 5 days post microinjection, it is unlikely that regulation of chromatin accessibility is a significant determinant of reporter activity. Thus, we conclude that CPRE-1 most likely enhances transcriptional activity by assembly of transcriptional activators.
Bioinformatic analyses suggest that CPRE-1 sequences are evolutionarily conserved in the promoter regions of T␣C orthologues from zebrafish to humans and identify a novel consensus sequence of 5Ј-CTGGAGTG(A/T)TGGA(G/A)GCAGGG(G/C)T-3Ј. In vitro experiments confirm that DNA probes encompassing the CPRE-1 element preferentially bind eye nuclear protein in a dose-dependent manner, and CPRE-1 is required for these interactions. Within the CPRE-1 of zebrafish T␣CP we identify nine central nucleotides essential for binding eye nuclear factors (Ϫ2518/Ϫ2512 and Ϫ2507/Ϫ2504 bp), and four nucleotides regulating the efficacy of binding (Ϫ2511 and Ϫ2501/ Ϫ2499 bp). These sites include four of the eight most conserved bases identified in the CPRE-1 consensus sequence.
This study advances our understanding of the transcriptional regulation of the cone-specific GNAT2 gene family. Previous studies of the human promoter reported several silencer sequences that bind nuclear protein in vitro, but their functional importance in vivo has not been reported (40). A transgenic line consisting of a 277-bp proximal region of the human T␣C promoter directs expression in murine cones when a heterologous enhancer from the human interphotoreceptor retinoid-binding protein gene is placed upstream (18), but a native human T␣C enhancer has not been identified. Chimeric reporter constructs with a distal fragment of the zebrafish T␣C promoter region containing CPRE-1 cannot regulate a proximal human T␣C promoter, and vice versa, suggesting speciesspecific differences in promoter organization or transcription factor binding.
Our proposed model of transcriptional regulation of the zebrafish T␣C gene is depicted in Fig. 7. The proximal 0.7 kb of promoter has weak transcriptional activity but controls the "all" cone specificity of expression (19). In the distal promoter, enhancer activities are located between Ϫ3.2/Ϫ2.8 kb (enhancer region 2) and Ϫ2.8/Ϫ2.5 kb (enhancer region 1). A novel 20-bp enhancer, CPRE-1 is located within enhancer region 1. CPRE-1 comprises of separate but co-dependent L and R modules necessary for robust, cone-specific expression. CPRE-1 enhances the rate of transcriptional activity, but not the tissue specificity. The region from Ϫ2.5/Ϫ0.9 kb, bridging the tissue specificity and enhancer domains, is essential for the enhancer activity of CPRE-1. This dependence is not a trivial spatial dependence, because neither a generic spacer nor an equivalent region of the human T␣C distal promoter can substitute in chimeric reporter assays. We speculate that transcription factors binding to these regulatory DNA regions are brought into contact in a complex resulting in high level conespecific expression (Fig. 7B). Our future studies are focused on identifying these factors and characterizing their expression and function.