|Home | About | Journals | Submit | Contact Us | Français|
Author contributions: W.F. and X.W. designed research; W.F., C.G., and X.W. performed research; W.F., C.G., and X.W. analyzed data; W.F. and X.W. wrote the paper.
Photoreceptor-specific transcription of individual genes collectively constitutes the transcriptional profile that orchestrates the structural and functional characteristics of each photoreceptor type. It is challenging, however, to study the transcriptional specificity of individual photoreceptor genes because each gene's distinct spatiotemporal transcription patterns are determined by the unique interactions between a specific set of transcription factors and the gene's own cis-regulatory elements (CREs), which remain unknown for most of the genes. For example, it is unknown what CREs underlie the zebrafish mpp5bponli (ponli) and crumbs2b (crb2b) apical polarity genes' restrictive transcription in the red, green, and blue (RGB) cones in the retina, but not in other retinal cell types. Here we show that the intronic enhancers of both the ponli and crb2b genes are conserved among teleost species and that they share sequence motifs that are critical for RGB cone-specific transcription. Given their similarities in sequences and functions, we name the ponli and crb2b enhancers collectively rainbow enhancers. Rainbow enhancers may represent a cis-regulatory mechanism to turn on a group of genes that are commonly and restrictively expressed in RGB cones, which largely define the beginning of the color vision pathway.
SIGNIFICANCE STATEMENT Dim-light achromatic vision and bright-light color vision are initiated in rod and several types of cone photoreceptors, respectively; these photoreceptors are structurally distinct from each other. In zebrafish, although quite different from rods and UV cones, RGB cones (red, green, and blue cones) are structurally similar and unite into mirror-symmetric pentamers (G-R-B-R-G) by adhesion. This structural commonality and unity suggest that a set of genes is commonly expressed only in RGB cones but not in other cells. Here, we report that the rainbow enhancers activate RGB cone-specific transcription of the ponli and crb2b genes. This study provides a starting point to study how RGB cone-specific transcription defines RGB cones' distinct functions for color vision.
Vertebrate retinal rod photoreceptors sense dim light stimuli for achromatic scotopic vision, whereas various types of cone photoreceptors sense bright light stimuli for color photopic vision. These functional differences result from distinct subcellular structures and molecular compositions of rods and cones, which in turn are determined by cell type-specific transcription profiles (Arshavsky et al., 2002; Pujic et al., 2006; Zhang et al., 2006; Solomon and Lennie, 2007; Jacobs and Nathans, 2009). Clearly, understanding the regulation of photoreceptors' transcriptional specificity is a fundamental aspect of visual science research.
To study transcriptional specificity in photoreceptors, we focus on apical polarity genes in zebrafish. Many polarity genes are expressed in the retina, and they play essential roles in the development and maintenance of zebrafish's rod photoreceptors and four types of cone photoreceptors (UV, red, green, and blue [RGB]) (Tsujikawa and Malicki, 2004; Gosens et al., 2008; Pocha and Knust, 2013). Moreover, in the retina, these polarity genes are expressed in cell type-specific manners. For example, the zebrafish mpp5bponli apical polarity gene (photoreceptor-layer-nok-like, ponli for short) and crumbs2b (crb2b) are not expressed in rods, UV cones, or any other retinal cells; rather, both ponli and crb2b are restrictively expressed in RGB cones (Zou et al., 2010, 2012, 2013), which align mirror-symmetrically into the G-R-B-R-G configuration (Robinson et al., 1993; Raymond et al., 1995) and are further adhered together as pentameric cone units (see Fig. 1A) (Zou et al., 2012). Conversely, the closest homologs of ponli and crb2b, the mpp5anagie oko (nok for short), and crb2a genes are expressed in undifferentiated retinal epithelium cells as well as in all types of photoreceptors and Müller glial cells (see Fig. 1A) (Wei et al., 2002, 2006; Zou et al., 2012). Thus, ponli and crb2b are ideal subjects to study RGB cone-specific transcription.
Like other RNA polymerase II-driven (Pol II) genes (Warner et al., 2008; Yáñez-Cuna et al., 2013), the transcription of ponli and crb2b should be regulated by their cis-regulatory elements (CREs) and trans-regulatory transcription factors. Because the two genes display identical retinal expression patterns, it is possible that they are regulated by similar CREs and transcription factors. However, virtually nothing is known about the mechanisms underlying their distinct transcription patterns.
Cis-regulation of Pol II-driven tissue-specific gene transcription requires both core promoters and enhancers. Core promoters are required for Pol II docking and transcription initiation but not necessarily for the spatiotemporal specificity of transcription; core promoters normally localize within 100 bp around transcription start sites (Lenhard et al., 2012). By contrast, enhancers regulate the spatiotemporal specificity of transcription, and enhancers can localize fairly distally from transcription start sites, either upstream or downstream (Kulaeva et al., 2012; Rouault et al., 2014). Thus, the unique RGB cone-specific transcription of ponli and crb2b may be regulated by novel and similar enhancers in conjunction with core promoters.
The purpose of this study is to address two questions: First, what are the CREs of the teleost ponli and crb2b genes that are critical for RGB cone-specific transcription in the retina? Second, do ponli and crb2b enhancers bear any similarities that may imply a common cis-regulatory mechanism? By assessing the transcriptional activities of various regions of the ponli and crb2b genes and effects of mutations with transgenic approaches, here we provide affirmative answers to both questions.
Tubingen wild-type zebrafish, Tg(SWS1:GFP) (Takechi et al., 2003), Tg(LCRRH2-RH2-2:GFP)pt115k and Tg(RH2-1:GFP)pt112 fish lines (Fang et al., 2013), as well as medaka fish (obtained from a local pet store) were maintained on a 14 h light/10 h dark cycle. Zebrafish embryos were raised at 28.5°C. Melanin pigmentation was blocked with 0.003% phenylthiourea. In this study, both male and female fish were used for experiments. For larval fish, their sex was undetermined; for founder fish, they were either males or females; and for F1 stable fish, one female and one male were analyzed. Animal care and handling were in accordance with University of Pittsburgh guidelines.
The RH2-1, ponli, and crb2b cis-regulatory genomic DNAs were obtained from the following sources: The 2 kb RH2-1 green opsin gene regulatory DNA (LCRRH2-RH2-1) includes a local control region (LCR), RH2-1 promoter, and 5′-UTR region of zebrafish RH2-1 gene (a gift from Dr. Shoji Kawamura) (Tsujimura et al., 2007); various regions of zebrafish ponli gene, including the 818 bp ponli enhancer (KY379333), were amplified from BAC clone CH211-105D18 (BACPAC Resources Center); the 212 bp tilapia ponli enhancer (KY379335) was amplified from tilapia genomic DNA, which was isolated from a frozen tilapia fish sold in a local supermarket; and the 208 bp medaka ponli (KY379334) and 882 bp crb2b (KY399459) enhancers were amplified from medaka genomic DNA, which was isolated from a whole medaka fish.
To test the transcriptional activities of various cis-regulatory DNAs, we used the I-Sce I meganuclease-based transgenesis system (Thermes et al., 2002) (vector provided by Dr. Michael Tsang). Specifically, each cis-regulatory DNA was inserted between a KpnI and an NheI restriction site and was followed sequentially by a Kozak sequence (GCCACC), the GFP or HA-tagged mCherry reporter ORFs, and an sv40 polyadenylation sequence-containing 3′-UTR (see Fig. 2A, constructs 1–3; Fig. 4A–C, constructs 4–11; Fig. 5A, constructs 12 and 13; Fig. 6D, construct 14; Fig. 9B, construct 15; Table 1). All constructs were verified by restriction digestion and sequencing. After the constructs were made, 20 pg of each construct was coinjected with 0.01 U of I-SceI enzyme into embryos at 1-cell stage. Reporter expression was examined at larval or adult stages either in founder fish or in F1 stable transgenic fish.
For the deletion analysis of the zebrafish ponli intron 1, various internal regions were deleted from construct 4 using the PCR-based Q5 Site-Directed Mutagenesis Kit (NEB, #E0554S) (see Fig. 4B, constructs 5–10), with inverse PCR primers matching to the sequences that flank the deletion regions. Construct 11 was generated by inserting the 531 bp core promoter region, the 818 bp ponli enhancer, and the 3′ end of intron 1, noncoding region of exon 1 between the KpnI and NheI sites.
Constructs 14 and 15 were generated by replacing the 818 bp zebrafish ponli enhancer between the ApaI and Asis I sites of construct 11 with the 212 bp tilapia ponli enhancer and the 882 bp medaka crb2b enhancer, respectively.
For substitution mutation analysis of enhancer motifs, the candidate enhancer motifs in constructs 11 and 14 were substituted with unrelated sequences of the same length (8–18 bp; see Fig. 8), with the center sequence as an AscI site (GGCGCGCC) for construct selection by restriction digestion. Again, these sequence substitutions were performed with the Q5 Site-Directed Mutagenesis Kit.
To compare the mRNA levels of the endogenous ponli and RH2 green opsin genes, we performed real time RT-PCR analysis with the 2−(ΔΔCt) method (Livak and Schmittgen, 2001), and we used an iQ5 real-time PCR detection system and an iQ SYBR Green Supermix kit (Bio-Rad). The experiments were performed with triplicate samples and repeated three times and normalized against the mRNA of the elf2a housekeeping gene. The following PCR primer pairs were used: ponli forward (ACATCGCCTGGTTCTACACC) and ponli reverse (CCAACACGGTACACGTTCAG); RH2-1 forward (GCCGCTCAACTACATTTTGG) and RH2-1 reverse (TGTAGCCCACAAGTGAGCAG); Elf2a forward (TTGAGAAGAAAATCGGTGGTGCTG) and Elf2a reverse (GGAACGGTGTGATTGAGG GAAATTC).
Candidate CREs of the ponli and crb2b genes were identified by searching homologous sequences among teleost fish genomes with the BLAT searching utility of the UCSC database (http://genome.ucsc.edu/). At the medaka (Oryzias latipes) genome browser, enter ponli or crb2b to locate the genes. The search also automatically align homologous genes of the following species: zebrafish (Danio rerio), stickleback (Gasterosteus aculeatus), fugu (Takifugu rubripes), tetraodon (Tetraodon nigroviridis), and nile tilapia (Oreochromis niloticus). The sequence similarities were calculated with Clustal Omega (http://www.ebi.ac.uk/Tools/msa/, Research resource identifier [RRID]: SCR_001591).
We performed RT-PCR analysis to confirm that the reporter gene was properly transcribed and spliced in Tg(ponli6,092:HA-mCherry)pt118b (pt118b for short). Briefly, we isolated the total RNA from 5 dpf eyes of wild-type and pt118b fish with Trizol (Invitrogen) and then performed reverse transcription using SuperScript III kit (Invitrogen). We then PCR amplified the cDNAs with the following primer pairs to confirm transcription and splicing (see Fig. 2A,B): F1 (GGACCTGATTGGAAACATCG, in exon 1) and R1 (GGCAAAAAGCCAACACTGTACA, in exon 2), which can detect both the endogenous ponli gene and the reporter gene but cannot distinguish between the two; F1 and R2 (GGATGTCGGCGGGGTGCTTCACGTA, in the mCherry coding regions), which can only detect the reporter gene (see Fig. 2A,B).
We performed Western blotting to compare the protein expression levels of reporter genes. Briefly, fish eyes were removed and homogenized on ice in a lysis buffer (1 × PBS buffer with 0.5% Triton X-100, 1 mm PMSF, 1 mm dithiothreitol, and 1 × Roach proteinase inhibitor mixture). The lysates were mixed with 5 × SDS-PAGE loading buffer and heated at 100°C for 5 min. The supernatants of the protein samples were separated in 12% SDS-PAGE gels and transferred to PVDF membranes for Western blotting with anti-mCherry antibody (1:2000; BioVision, catalog #5993-100; RRID:AB_1975001), anti-GFP antibody (1:2000; Sigma-Aldrich, catalog #G6539; RRID:AB_259941), and anti-γ-tubulin antibody (1:3000; Sigma-Aldrich, catalog #T3320; RRID:AB_261655).
We performed immunohistochemical confocal microscopy to examine the cellular expression patterns of reporter proteins in the retina. Briefly, both adult fish and larval eyes were fixed in 4% PFA in 1 × PBS buffer at room temperature for 2 h, infiltrated with 40% sucrose in 1 × PBS overnight, and embedded in tissue freezing media (Tissue-Tek; Sakura Finetek). Eye samples were cryosectioned at 30 μm thickness and immunostained as described previously (Wei et al., 2006). We used the following primary and secondary antibodies: rabbit polyclonal mCherry antibodies (1:300; BioVision, catalog #5993-100; RRID:AB_1975001), mouse monoclonal Zpr1 antibodies (1:300; Zebrafish International Resource Center, catalog #ab174435; RRID:AB_10013803), Cy3-conjugated goat anti-mouse IgG (1:300; Jackson ImmunoResearch Laboratories, catalog #115-165-166; RRID:AB_2338692), goat anti-rabbit IgG (1:300; Jackson ImmunoResearch Laboratories, catalog #111-165-144; RRID:AB_2338006), and Cy5-conjugated donkey anti-rabbit IgG (1:300; Jackson ImmunoResearch Laboratories, catalog #711-175-152; RRID:AB_2340607). AlexaFluor-647 phalloidin (1:300; Thermo Fisher Scientific, catalog #A22287; RRID:AB_2620155) was used for F-actin staining. Images were taken with a Fluoview FV1000 confocal microscope (Olympus).
We took a strategy to first identify ponli CREs, particularly its enhancer; we then used the characteristics of ponli CREs to guide our search for the enhancer of the crb2b gene. Because enhancers can localize distally, either upstream or downstream of transcription sites, and because the transcribed sequence of the ponli gene itself already extends over 75 kb, it was a challenging task to identify the ponli's enhancer region in this broad region and beyond. Therefore, for the sake of practicality and because many enhancers often reside upstream of the transcription start site or within the first intron, we limited our search for ponli enhancer in the 4 kb upstream intergenic region and in the 5.7 kb downstream intron 1 (Fig. 1B).
To assess the transcriptional activity of the 4 kb upstream intergenic DNA, we made transgenic construct 1, which contains the 4 kb intergenic DNA, ponli's 110 bp exon 1, a Kozak sequence, and an mCherry-coding sequence (Fig. 2A, construct 1). We included ponli exon 1 in the construct to best ensure the intactness of the core promoter of the ponli gene. The construct was then injected into embryos at 1-cell stage and monitored for retinal expression at 4 dpf (days post fertilization); we found that none of the 316 injected fish showed mCherry signals in the retina (data not shown). Thus, the 4 kb upstream intergenic DNA does not contain ponli's enhancer and the core promoter region alone is not sufficient to drive retinal transcription.
We next assessed the transcriptional activity of intron 1. To do so, we made transgenic construct 2 by fusing an mCherry reporter gene downstream of a 6102 bp DNA, which covers a 140 bp upstream region, exon 1, intron 1, and the 67 bp noncoding region of exon 2 (Fig. 2A, construct 2). We injected this construct into zebrafish embryos and found that 19 of 208 injected embryos indeed expressed mCherry in the retina at 4 dpf. These mCherry-expressing fish were raised to adulthood, and from their outcross progeny, we identified a stable fish line, which was named Tg(ponli6,102:HA-mCherry)pt118b (pt118b for short). Confocal immunohistochemistry of adult pt118b retina revealed that mCherry is restrictively expressed in RGB cones just as endogenous Ponli protein (Fig. 2C–E) (Zou et al., 2010). In addition, RT-PCR results indicate that the intron 1 DNA of the transgene could be transcribed and subsequently removed from mature mRNA by splicing as the endogenous ponli transcript is processed, although the RT-PCR analysis could not evaluate the efficiency of splicing (Fig. 2B). Interestingly, the mCherry signals in pt118b are ~10 times weaker than those driven by the RH2-1green opsin cis-regulatory DNA in Tg(Rh2-1:HA-mCherry)pt120 (Fig. 2A, construct 3; Figs. 2F, ,33A–C); this weaker transgenic expression echoes the fact that ponli endogenous mRNA is much less abundant than green opsin mRNA (Fig. 3D).
These results indicate that the 6102 bp DNA harbors sufficient CREs for ponli's RGB cone-specific transcription. Considering that construct 2 shares ponli's core promoter region with construct 1, which lacks any transcriptional activity (Fig. 2A), we conclude that the RGB cone-specific transcriptional activity of the 6102 bp DNA must be determined by an enhancer located in ponli's 5785 bp intron 1.
We next sought to narrow down the intronic DNA region that drives RGB cone-specific transcription. To do so, we generated deletion constructs 4–11 and analyzed the transcriptional activities of the intronic DNA sequences (Fig. 4A–C). These constructs all contain ponli's core promoter region (including the 140 bp upstream sequence and 110 bp exon 1), but each includes different subregions of intron 1 (Fig. 4A–C). We injected these constructs into embryos at 1-cell stage and monitored their retinal GFP expression at 4 dpf. We found that, in conjunction with the core promoter region, an 818 bp mid-intronic DNA, but not its flanking 4232 bp intronic DNAs, is sufficient to drive retinal expression (Fig. 4B,C). We believe that this 818 bp intronic DNA regulates retinal expression by promoting transcription but not by ensuring the proper splicing between exon 1 and exon 2 through a splicing branch site. This is because, when exon 1 and exon 2 were fused together by deleting the entire intron 1, retinal expression was inhibited (Fig. 4B, construct 9).
To confirm that the 818 bp DNA drives GFP expression specifically in RGB cones, we next raised two stable fish lines, Tg(ponli1,904:GFP)pt151-#2;#3 (pt151-#2 and pt151-#3 for short; Fig. 4C–G), to reveal the cellular expression patterns of GFP in developed retina. Indeed, in Tg(ponli1,904:GFP)pt151-#3, GFP is restrictively expressed in RGB cones (Fig. 4D–F); however, in Tg(ponli1,904:GFP)pt151-#2, GFP is expressed not only in RGB cones but also in <20% of rods and very sparsely in bipolar cells (Fig. 4G). We also used the same 1904 bp ponli cis-regulatory DNA to drive RGB cone-specific gene expression in 13 other unrelated stable transgenic fish lines; nine of these lines display RGB cone-specific expression, but the remaining four fish lines also display some ectopic GFP expression in 10%–80% of rods and sparsely in bipolar cells, as in Tg(ponli1,904:GFP)pt151-#2 (data not shown). We believe that such ectopic expression in rods and bipolar cells is due to position effects because the ectopic expression does not always happen in every stable fish line. Nevertheless, when ectopic expression does occur, it occurs more often in rods than in other non–RGB-cone cells, suggesting that some rod-specific transcription factors may activate the 1904 bp ponli cis-regulatory DNA under certain position effects.
In addition to the high reproducibility of RGB cone-specific transgenic expression, the ponli1,904 cis-regulatory DNA also characteristically express GFP at modest levels: Western blotting showed that the protein expression levels in Tg(ponli1,904:GFP)pt151-#2,#3;w1465-#1,w1465-#11 were 50–100 times lower than those driven by opsin regulatory DNAs in Tg(LCRRH2-RH2-2:GFP)pt115-k and Tg(SWS1:GFP) (Takechi et al., 2003; Fang et al., 2013) (Fig. 4H). We also observed modest variations in GFP expression among Tg(ponli1,904:GFP) fish (Fig. 4H).
The high reproducibility of RGB cone-specific transcription in these transgenic fish suggests that 1904 bp DNA contains sufficient basic CREs of the ponli gene for its RGB cone-specific transcription: namely, ponli's core promoter and the enhancer (or at least one of its enhancers). We suspected that the 818 bp intronic DNA harbored the ponli enhancer, which directs ponli's transcriptional specificity; in addition, the 140 bp upstream sequence and exon 1 contain ponli's core promoter, which may be replaced with other generic core promoters without compromising ponli's transcriptional specificity.
To test this idea, we next engineered a chimeric cis-regulatory DNA by combining the 818 bp ponli intronic DNA with the core promoter region of the nok gene, which is expressed in all photoreceptors and Müller cells in developed retina (Wei et al., 2006) (Fig. 1A), and then evaluated its transcriptional activity with a transgenic reporter construct (Fig. 5A, construct 12). Embryonic injection of this construct led to preferred retinal GFP expression at 4 dpf (Fig. 5B,C). By contrast, injection of another construct that only contained nok's core promoter regions produced no retinal expression, even though many muscle cells displayed transient GFP expression (Fig. 5A–C, construct 13).
To evaluate the transcriptional specificity of the GFP reporter in the construct-12-injected fish, we raised these fish to adulthood and then performed confocal microscopy to examine the expression patterns of transgenic clones. We found that GFP was only expressed in the RGB cones in 49 of the 60 transgenic cell clones examined in adult retina (Fig. 5D,E); in the remaining 11 clones, in addition to GFP expression in RGB cones, we also observed ectopic GFP expression in <20% of rods (Fig. 5F,G). These expression patterns are similar to those observed in Tg(ponli1,904:GFP)pt151-#3 and Tg(ponli1,904:GFP)pt151-#2 (Fig. 4E–G), suggesting that the 818 bp intronic DNA underlies the transgenic expression pattern that is characteristic of the ponli gene but not that of the nok gene. Although we cannot exclude the possibility that there is more than one ponli enhancer, our results suggest that the 818 bp intronic DNA contains a ponli enhancer, which when supplied with a basic core promoter (not necessarily ponli's own core promoter), is sufficient to confer modest and RGB cone-specific transcription.
To identify the ponli's CREs in its core promoter and enhancer regions, we next aligned the DNA sequences of the 5′ ends of ponli genes of six teleost fish (medaka, tilapia, fugu, tetraodon, stickleback, and zebrafish). The alignment reveals high sequence conservation near the transcription start site, the 5′ splicing site of intron 1, the middle region of intron 1, and exon 2 (Fig. 6A). In the core promoter regions, we found three conserved core promoter motifs (CPM1–3) in all six fish species. CPM1s locate <20 bp upstream of the transcription start sites; CPM2s cover the transcription start sites (Fig. 6B, arrows); and CPM3 locates 30–70 bp downstream of transcription start sites. The sequence of CPM2s (CCACTTG in zebrafish and CCACGGA in four other fish) is similar to the consensus sequence of human transcription initiators (YYANWYY) (Juven-Gershon et al., 2010), suggesting that CPM2s are the initiator motifs of the teleost ponli genes (Fig. 6B). We could not identify a TATA box, which normally resides between positions −33 and −28 bp upstream of the transcription start site (Carninci et al., 2006; Cooper et al., 2006). Because core promoters normally reside within 50 bp 5′ and 3′ of the transcription start sites (Juven-Gershon et al., 2010; Yáñez-Cuna et al., 2013), these conserved sequence motifs likely constitute the ponli core promoters, which belong to the TATA-less core promoters category (Carninci et al., 2006).
The conserved mid-intronic region overlaps with the 3′ half of the 818 bp zebrafish enhancer-containing intronic DNA (Fig. 6A). Sequence comparison revealed that these mid intronic regions contain five highly conserved DNA sequence motifs, named, enhancer motifs 1–5 (EM1-5). EM1-5s are aligned sequentially and compactly in all teleost fish, except that in zebrafish, they are aligned in the order of EM1, first EM2, second EM2, EM4, EM3, and EM5. Interestingly, zebrafish EM5 and the second EM2 are in reverse orientation compared with EM2 and EM5 in other teleost fish (Fig. 6A,C). Finally, zebrafish EM5 is more separated from the rest of the EMs, being 50 bp downstream of the nearest EM3; by contrast, in other teleost fish, EM5 is only 3 bp downstream of the nearest EM4.
To determine whether the conserved intronic regions of other teleost fish also drive RGB cone-specific transcription, we replaced the zebrafish 818 bp enhancer region in construct 11 with the 212 bp conserved tilapia intronic DNA and analyzed, in zebrafish, the GFP expression patterns of the resulting chimeric transgenic construct 14 (Fig. 6D). As expected, the 212 bp tilapia intronic DNAs indeed drove GFP expression in zebrafish RGB cones in two stable fish lines Tg(ponlitilapia:GFP)pt152-#14,#D (Fig. 6E). These results suggest that the mid intronic regions of the teleost ponli genes are conserved not only in sequences but also in RGB cone-specific transcriptional activity.
To determine whether or not ponli EM1-5 and their nearby sequences are required for RGB cone-specific transcription, we mutated these motifs individually and then examined the resulting effects on retinal expression patterns. Because such analysis was laborious and because it is impractical to generate and analyze a stable fish line for each mutation, we devised a larval immunohistological microscopic method to evaluate the effects directly in injected fish. The essence of this method is to inject the mutation constructs in Tg(Rh2-1:HA-mCherry)pt120 embryos at the 1-cell stage and then at 9 dpf, to determine the types of GFP-expressing cells according to the photoreceptors' morphologies, Zpr1 immunoreactivity, and mCherry signals (Fig. 7).
The substitution mutations were generated for both tilapia and zebrafish ponli enhancer regions by replacing these short sequences individually with unrelated sequences of the same lengths (~10 bp): a total of 14 substitution mutations for tilapia ponli enhancer and 18 substitution mutations for zebrafish ponli enhancer (Fig. 8A–C).
These substitution mutation analyses revealed that mutation of EM2 (GAACAGATGG), which is conserved in all six fish species, abolished retinal expression, suggesting that EM2 plays an activation role in RGB cone-specific transcription (Fig. 8B,C). In addition, the sequences around EM2 also appear to be essential, even though there is no clear similarity conservation between zebrafish and the other five fish species in these nearby sequences (Fig. 6C). However, mutation of the second zebrafish EM2, which is in reverse orientation and 10 bp downstream of the first EM2, did not affect retinal expression (Fig. 8C).
Surprisingly, unlike EM2, loss of EM1, EM3, EM4, and EM5 did not appear to affect retinal expression, except when EM1 was next to EM2 in tilapia (Fig. 8B,C). Nevertheless, the cell type-specific transcription can be slightly compromised. For example, mutation of tilapia motif 13 resulted in more ectopic expression in non–RGB-cone cells that form vertical clones, suggesting that ectopic expression might be activated in retinal progenitors; this ectopic expression might be augmented by position effect because different transgenic clones in the same retina displayed a different level of specificity (Fig. 8D). Interestingly, even though these mutant constructs resulted in ectopic expression in rods, bipolar cells, amacrine cells, and ganglion cells, we very rarely observed any expression in UV cones. And the ratio of GFP expression in RGB cones remains at an ~2:2:1 ratio (Fig. 8B,C). Together, the mutation analyses suggest that EM2 and its close neighboring sequences are essential for retinal expression in an all-or-none fashion and that other conserved motifs may collectively influence cell type specificity.
The RGB cones are very similar in their morphologies and functions; such structural and functional similarities must be based on the similarity in their gene expression profiling. This made us wonder whether similar enhancers are used to govern other RGB cone-specific genes, such as crumbs2b (crb2b), which is expressed with the same spatiotemporal patterns as the ponli gene in the retina, and whose protein product colocalizes with Ponli (Zou et al., 2010, 2012).
To explore this possibility, we next compared the first intron of the crb2b genes among the five fish species (medaka, fugu, tetraodon, stickleback, and zebrafish) and found four stretches of sequence that are conserved in the first introns (C1–C4; Fig. 9A). Strikingly, the conserved stretch C3 also contains three motifs that are identical or very similar to the EM2, EM3, and EM4 of the medaka ponli enhancer (Fig. 9A), implying that this stretch of DNA might function as a crb2b enhancer. To test this hypothesis, we replaced the ponli enhancer of the transgenic reporter assay construct 11 with an 882 bp C3-containing DNA fragment (Fig. 9B); the resulting construct 15 was injected into zebrafish embryos, and the transcriptional activity in larvae was examined. We found that this 882 bp DNA indeed induced specific RGB cone expression (Fig. 9C,E). As for ponli enhancers, loss of EM2 completely abolished the 882 bp DNA's retinal transcriptional activity (Fig. 9D). However, mutations of EM3 and EM4 did not abolish retinal expression; rather, they slightly increased nonphotoreceptor expression and also slightly altered RGB cone ratios (Fig. 9D,E). These results suggest that the 882 bp DNA fragment of crb2b intron 1 contains a crb2b enhancer that is sufficient to drive RGB cone-specific transcription in conjunction with a core promoter.
The functional and sequence similarities between ponli and crb2b enhancers suggest that they constitute a conserved cis-regulatory mechanism that governs RGB cone-specific transcription of ponli and crb2b. We thus name these enhancers collectively rainbow enhancers to honor their transcriptional activity in RGB cones, which largely constitute the beginning of the color vision pathway.
In this study, we set out to identify CREs of the ponli and crb2b genes that regulate RGB cone-specific transcription and to determine whether or not their cis-regulatory mechanisms are conserved. By assessing the transcriptional activities of various DNA regions of teleost ponli and crb2b genes with homology comparison and transgenic reporter assays, we found that the ponli core promoter is TATA-less and contains three conserved core promoter motifs (CPM1-3); in addition, the intronic ponli enhancers, containing five conserved enhancer motifs (EM1-5), determine RGB cone-specific transcription. We further showed that crb2b enhancers are also intronic and that they share EMs with ponli enhancers, suggesting that crb2b and ponli enhancers, collectively called rainbow enhancers, underlie a conserved cis-regulatory mechanism.
The sequence and functional conservation among ponli and crb2b enhancers suggest that rainbow enhancers are constructed according to a common “code.” Indeed, the concept of “enhancer cis-regulatory codes” is gaining support from growing evidence. According to this concept, a specific code governs how and what enhancer motifs are organized in an enhancer to confer its tissue-specific transcriptional activity (Yáñez-Cuna et al., 2013). However, it remains challenging to interpret cis-regulatory codes because enhancer motifs are often short and degenerate; in addition, enhancers are often studied individually, making it more difficult to reveal the underlying general principles. Thus, the identification of rainbow enhancers for two different genes offers a good opportunity to use homology and functional conservation to crack the cis-regulatory code that governs RGB cone-specific transcription. Although a complete understanding of rainbow enhancer code has yet to be reached, our current study revealed two prominent features of the rainbow enhancers.
First, the transcriptional activity of rainbow enhancers depends on EM2 in an all-or-none fashion because mutation of the EM2s in the zebrafish ponli enhancer, tilapia ponli enhancer, and medaka crb2b enhancer completely abolished their transcriptional activities (Figs. 8B,C, ,99D,E). The consensus sequence of EM2 (GAACAGATGG) and its surrounding sequences do not resemble the consensus sequences of the U2 and U12 splicing branch sites TTCTCATTC and TTTTCCTTAACTTTT, which localize most commonly 20–40 bp upstream of the 3′ splicing sites (Yeo et al., 2004; Turunen et al., 2013), suggesting that loss of EM2 is unlikely to block splicing; rather, the transcriptional dependency on EM2 would suggest that EM2 plays an essential role in promoting transcription, possibly by recruiting a transcription factor that has yet to be identified. Supporting this notion, eliminating the need for splicing by fusing exon 1 and exon 2 blocked retinal expression completely (Fig. 4B, construct 9). The EM2-binding transcription factor may itself be expressed in an RGB cone-specific manor. In such a case, the enhancer of an EM-2-binding transcription factor gene may also be a rainbow enhancer, and it may augment its own transcription in a positive-feedback loop to increase the robustness of RGB cone-specific transcription. Alternatively, the EM2-binding transcription factor may not be restrictively expressed in RGB cones; instead, it may promote nonspecific transcription. In such a case, other enhancer motifs may restrict rainbow enhancers' transcriptional activity to RGB cones. Interestingly, the sequences immediately next to EM2 are also important for transcriptional activity, even though they do not have to be as highly conserved as EM2 itself (Figs. 6, Fig. 8). Such sequence divergence of essential motifs is not unique; another example is the critical cis-regulatory elements of the fly sparkling enhancer, which has rapidly changed over a short period of time during evolution (Swanson et al., 2011).
Second, in addition to EM2, rainbow enhancers also contain several other highly conserved motifs, which are concentrated around EM2 in a range of ~200 bp (Figs. 6C, ,99A), a typical size for many known enhancers (Yáñez-Cuna et al., 2013). However, these conserved motifs are not absolutely required because some of them, such as EM5, are even absent or highly degenerated and could not be identified in medaka crb2b enhancer (Fig. 9A); in addition, mutations of these motifs did not abolish retinal expression as EM2 mutations, even though these mutations slightly compromised cell type specificity of transgenic expression. In addition to sequence variations, the distance between highly conserved motifs can vary between species and between genes (Fig. 8B,C and Fig. 9A, respectively). Moreover, the orientation of zebrafish ponli enhancer's EM5 and the second duplicated EM2 are even in a reverse orientation compared with the same motifs in tilapia ponli enhancer (Fig. 6C). Despite these variations among rainbow enhancers, both tilapia ponli enhancer and medaka crb2b enhancer can still activate RGB cone-specific transcription in zebrafish, suggesting that the transcription factors are not required to bind to rainbow enhancers in a strict spatial order in order for them to restrict transcriptional activation in RGB cones. This functional conservation is surprising because it is thought that the composition, orientation, spacing, and alignment of enhancer motifs play critical roles in transcriptional specificity by creating unique and continuous recognition surfaces for transcription factor binding and activation in a distinct combinatory fashion, as exemplified by the paradigmatic interferon-β enhanceosome (Panne et al., 2007; Panne, 2008; Levine, 2010; Swanson et al., 2010). The coexistence of both organizational variations of rainbow enhancer motifs and functional conservation of rainbow enhancers raises an interesting question: How can enhancers be structurally varied on the one hand and still maintain their strict transcriptional specificity on the other hand? Perhaps loose cis-regulatory codes confer the freedom that transcription factors need to assemble in distinct configurations and to paradoxically enforce strict transcriptional specificity.
Compared with other cone-specific CREs, rainbow enhancers display three prominent features. First, rainbow enhancers differ from others in the cellular patterns of transcriptional activities. To date, a limited number of cone-specific CREs have been identified; they are either active in one distinct type of cone photoreceptor or active in all types of cone photoreceptor. Examples of the former are the CREs of various cone opsin genes (Takechi et al., 2003, 2008; Luo et al., 2004; Tsujimura et al., 2010, 2015); examples of the latter are the CREs of cone arrestin and cone transducin α subunit (Pickrell et al., 2004; Smyth et al., 2008). Unlike these two classes of cone-specific CREs, rainbow enhancers' unique RGB cone-specific expression implies that the underlying mechanism of rainbow enhancers is distinct from others. Moreover, because the ponli and crb2b genes are unlikely to be the only genes that define RGB cones' structural and functional characteristics, it is tempting to speculate that other RGB cone-specific genes exist and that their transcription is regulated by a family of rainbow enhancers.
Second, like most other identified cone-specific CREs, all rainbow enhancers contain a sequence motif that is either identical or very similar to CRX's (cone rod homeobox) consensus target sequence TTAATCC (Chen et al., 1997). These CRX sites localize 5–12 bp upstream of EM2 and constitute 70% of the EM1 motifs (Figs. 6C, ,99A). Because retinal expression was abolished by substitution mutations 4t and d, which removed the tilapia ponli CRX site entirely and three sevenths of the zebrafish CRX site, respectively, it is tempting to speculate that CRX recognizes EM1. However, it is puzzling that the removal of the 5′ four-sevenths of the zebrafish CRX site or 60% of EM1 (substitution mutation 4z) did not affect retinal expression (Figs. 6C, ,88C). Regardless, because CRX is also expressed in rods and regulates rod opsin by interacting with rod-specific Nrl (neural retina-specific leucine zipper protein) (Rehemtulla et al., 1996; Babu et al., 2006; Peng and Chen, 2007; Reks et al., 2014), it is unlikely that CRX itself determines rainbow enhancers' RGB cone-specific transcription.
Finally, rainbow enhancers' most critical and conserved motif, EM2 (GAACAGATGG), does not appear to show significant similarity to other cone CRE motifs, including the CPRE-1 motif (cone photoreceptor regulatory element 1) of the α subunit of cone transducin, which is expressed in all four zebrafish cone types (Smyth et al., 2008). This uniqueness of EM2 highlights the importance of identifying an EM2 transcription factor for understanding rainbow enhancer activation. Perhaps as Nrl drives photoreceptors to become rods (Oh et al., 2007), the yet to be identified EM2-binding transcription factor may drive photoreceptor to become RGB cones.
Two useful expression features of rainbow enhancers make them excellent tools for transgenic research of teleost photoreceptors. First, unlike opsin cis-regulatory DNAs (Hamaoka et al., 2002; Takechi et al., 2003; Tsujimura et al., 2007), rainbow enhancers can be used to drive simultaneous transcription in zebrafish RGB cones; this expression pattern is particularly desirable either for studying cell-nonautonomous gene functions because transgenic manipulation of such genes in one member of the RGB cones may not result in apparent phenotypes due to rescuing effects by neighboring cells or for therapeutic gene expression in multiple types of cones. Second, to avoid potential adverse consequences of transgenic overexpression by opsin cis-regulatory DNAs, rainbow enhancers can be used to drive much more modest transcription (Figs. 3, ,44D,H).
In conclusion, we have identified two teleost rainbow enhancers, the ponli and crb2b enhancers; to our knowledge, this is the first report of such enhancers with RGB cone-specific transcriptional activity in teleost retina. Our work provides a starting point from which to study rainbow enhancer-mediated transcription of ponli and crb2b and possibly other RGB cone-specific genes; these genes may collectively define RGB cones' distinct transcriptional profile and consequently their functions in initiating color vision.
This work was supported by the National Institutes of Health Grants P30EY008098, EY016099, EY025638, and R21EY023665, Eye and Ear Foundation of Pittsburgh, and Research to Prevent Blindness. We thank Lynne Sunderman for proofreading the manuscript.
The authors declare no competing financial interests.