Nr2e3 is an orphan nuclear receptor expressed specifically by retinal photoreceptor cells. Mutations in Nr2e3 result in syndromes characterized by excess blue cones and loss of rods: enhanced S-cone syndrome (ESCS) in humans and rd7 in mice. Using yeast two-hybrid screens with Nr2e3 as bait, the cone–rod homeobox protein Crx was identified as an interacting partner of Nr2e3. Immunoprecipitation assays confirmed this Nr2e3–Crx interaction and identified the DNA-binding domain of each protein as the interaction motif. Immunohistochemistry demonstrated that Crx and Nr2e3 are co-expressed by rod photoreceptors and their precursors. Chromatin immunoprecipitation assays on mouse retina demonstrated that Nr2e3 and Crx co-occupy the promoter/enhancer region of several rod and cone genes in the rod photoreceptor cells. The promoter/enhancer occupancy of Nr2e3 is Crx-dependent, suggesting that Nr2e3 is associated with photoreceptor gene targets by interacting with Crx. Transient transfection assays in HEK293 cells demonstrated that Nr2e3 enhances rhodopsin, but represses S- or M-cone opsin transcription when interacting with Crx. Quantitative real-time RT–PCR analysis on postnatal day 28 (P28) retina of the rd7 mouse, which lacks Nr2e3 protein, revealed an up-regulation of cone genes, but down-regulation of rod genes. Several mutant forms of human Nr2e3 identified from ESCS patients showed defects in interacting with Crx and/or in transcriptional regulatory function. Altogether, our findings suggest that Nr2e3 is a dual-function transcriptional regulator that acts in concert with Crx to promote and maintain the function of rod photoreceptors.
The vertebrate retina contains two types of visual photoreceptor cells: rods and cones. Rods mediate vision in dim light via the visual pigment rhodopsin, whereas cones mediate color vision in bright light using cone opsins. The ratios of rods versus cones and their spatial distribution in the retina vary greatly among different species. In humans, 5% of the photoreceptors are cones, which occur at highest density in the macula. There are three cone subtypes, which express the short (S, blue), long (L, red) or medium (M, green) wavelength cone opsin (1). In mouse retina, 3% of the photoreceptors are cones with only two subtypes: S (UV) and M (green) cones. They are spatially patterned in the dorsal versus ventral region, although some individual cones in the middle can express both cone opsins (2,3).
During development of the vertebrate retina, the rod and cone photoreceptors and five other major cell types are all derived from multi-potent progenitor cells (4,5). Thymidine birthdating studies in mice showed that cones are born between embryonic day 12.5 (E12.5) and E17.5, whereas rods are generated over a longer period from E13 to postnatal day 10 (P10), peaking at birth (6,7). There is a significant delay between the birth date of photoreceptor precursors and the onset of opsin gene expression (8). Thus, photoreceptor differentiation can be visualized as a process of selectively turning on (or off) the expression of sets of specific genes. This selective gene expression is not only important for development of photoreceptors but also important for maintaining the integrity and function of each photoreceptor subtype, as over-expression or under-expression of a photoreceptor gene, such as rhodopsin, can lead to photoreceptor degeneration (9,10).
Both intrinsic factors and extrinsic cues are important in regulating photoreceptor gene expression and fate specification (11). Some transcription factors and cofactors have recently been identified as intrinsic factors. The homeodomain transcription factor Otx2 (12) and the basic helix–loop–helix protein NeuroD1/Beta2 (13) are involved in specifying photoreceptor lineage. The neuroretina leucine zipper protein Nrl (14) and the retinoblastoma (Rb) protein (15) are critical for the development of rods, whereas the thyroid hormone receptor β2 (Trβ2) (16) is essential for the development of M-cones. The nuclear receptor Nr2e3 (formerly PNR) (17) appears to be involved in the cell fate determination of rods versus S-cones (discussed subsequently). Acting down-stream of Otx2, the cone–rod homeobox (Crx) protein is required for the development and maintenance of photoreceptor function by regulating the transcription of many photoreceptor-specific genes (18–20). Crx plays this role by interacting with other transcription regulators including Nrl (21), Qrx (22), non-histone chromatin protein HMG I (Y) (23), ataxin-7 (24,25) and BAF (26). Not surprisingly, defects of these transcription regulators, such as Nrl (27), Crx (28–31), ataxin-7 (32) and Qrx (22), can lead to transcriptional dysregulation of the target genes, resulting in developmental defects or photoreceptor degeneration. Thus, interactions among a network of photoreceptor transcription factors are required for the development and maintenance of photoreceptors.
Among the members of this network of transcription factors, Nr2e3 is unique in many respects. It was identified as a ligand-dependent nuclear receptor, that is, homologous to an orphan nuclear receptor, Nr2e1 (formerly tailless or TLX), which is involved in the development and maintenance of normal function of the central nervous system (CNS) in Drosophila and mammals (17,33,34). Although Nr2e3 has a DNA-binding domain (DBD) near the N-terminal and can bind to a consensus target sequence of Nr2e1 as a dimer (17), its ability to bind native targets has not been established. Nr2e3 also contains a ligand-binding domain (LBD) in the C-terminal for a ligand not yet identified (17). Nr2e3 is expressed specifically by retinal photoreceptor cells (17) and appears to act down-stream of Nrl to regulate the development of rods versus S-cones (14). Mutations in human Nr2e3 cause an autosomal recessive disease called enhanced S-cone syndrome (ESCS) characterized by hyperfunction of blue cones, but defective function of rods, and blindness in the late stages (35,36). Pathological studies showed excess S-cones in the retina, some of which express both S- and M-cone opsins (37). This phenotype is unique among retinal degenerative diseases, almost all of which show universal loss of rod and cone function. In mice, a 380 bp deletion in the coding region of Nr2e3 is responsible for the phenotype of the Nr2e3rd7/rd7(rd7) mouse (38). The rd7 mouse retina contains whorls and rosettes in the photoreceptor layer at early ages followed by slow photoreceptor degeneration. Similar to ESCS in humans, the rd7 retina has an excess of cones that express mostly S-cone opsin (39), although rod function appears to be normal in young adults by electroretinography (ERG) (38). These results suggest that Nr2e3 is involved in photoreceptor subtype specification and maintenance. More recently, rod photoreceptors have been reported to express Nr2e3 (40,41), which interacts with Nr1d1, another orphan nuclear receptor (40). Co-expression of Nr2e3 and Nr1d1 enhances the potency of Crx and Nrl to activate the rhodopsin promoter in HEK293 cells, suggesting that Nr2e3 plays a role in regulating rod gene expression (40).
We performed a series of experiments aimed at determining the role and mechanism of action of Nr2e3 in photoreceptor subtype specification. Here, we report that Nr2e3 interacts directly with the homeodomain transcription factor Crx on various Crx–target genes in rod photoreceptor cells. The Crx–Nr2e3 complex exerts opposing effects on the transcription of rod versus cone target genes. Some of the Nr2e3 mutations linked to human disease disrupted the interaction with Crx and/or the modulated transcription. Altogether, our results suggest that Nr2e3 promotes the differentiation and maintenance of rod function by differentially regulating the transcription of rod- and cone-specific genes.
Nr2e3 and Crx form a direct interaction through the DBD
To determine the mechanism by which Nr2e3 regulates the development of photoreceptor subtypes, we carried out yeast two-hybrid screens for retinal proteins that interact with Nr2e3. Such an interaction in AH109 yeast cells will activate the His3 reporter, resulting in growth on histidine-deficient selection medium (see Materials and Methods). Yeast cells transformed with only the bait construct Nr2e3-Full, containing a full-length cDNA encoding human Nr2e3, were not able to grow on the selection medium (Fig. 1A). Screening of 2×106 clones from a bovine retinal cDNA library (42) with Nr2e3-Full as bait identified ∼60 positives, six of which encoded Crx. To confirm that Nr2e3 interacts with Crx, we carried out yeast two-hybrid assays on the Crx prey paired with the original and two additional Nr2e3 bait constructs harboring the DBD (Nr2e3-DBD) or the ligand-binding domain (Nr2e3-LBD). As shown in Figure 1A, a positive interaction was detected when Crx was paired with Nr2e3-Full or Nr2e3-DBD, but not with Nr2e3-LBD, suggesting that Nr2e3 interacts with Crx through its DBD.
To determine whether the Nr2e3–Crx interaction in yeast is direct, co-immunoprecipitation assays were performed using in vitro transcribed/translated proteins. An anti-Crx antibody, p261, co-immunoprecipitated the protein produced from Nr2e3-Full in the presence (Fig. 1B, lane 1), but not in the absence (lane 4) of Crx, suggesting that Nr2e3 directly interacts with Crx. In agreement with the yeast two-hybrid assay results, the DBD of Nr2e3 appears to be necessary and sufficient for mediating this interaction, as Nr2e3-DBD but not Nr2e3-LBD co-immunoprecipitated with Crx (lane 3 versus lane 2) under the same conditions. To further confirm this direct interaction between Nr2e3 and Crx, reciprocal experiments were performed. As seen in Figure 1C, a full-length Crx protein (Crx-Full) was immunoprecipitated by an anti-Nr2e3 antibody (p183, discussed subsequently) in an Nr2e3-dependent manner (Fig. 1C, lane 1 versus lane 4). In addition, the N-terminal fragment Crx-1-107, but not the C-terminal fragment Crx-111-299, was co-immunoprecipitated with Nr2e3 (lane 2 versus lane 3). Furthermore, Crx34-107, a peptide containing the homeodomain (HD) and five flanking amino acids on each end (25,43), can also be co-immunoprecipitated with Nr2e3 (data not shown). Therefore, we conclude that the DBD of Nr2e3 interacts directly with the homeodomain of Crx.
Nr2e3 is co-expressed with Crx in mature and developing rod photoreceptors
To investigate whether Nr2e3 interacts with Crx in vivo, we examined the spatial and temporal expression patterns of these two proteins in the retina. For this purpose, we generated rabbit antibodies against two peptides residing in the N-terminal (p20) of mouse Nr2e3 and the middle portion (p183) of human Nr2e3, respectively (Supplementary Material, Figure S1A). We confirmed the specificity of these antibodies for native Nr2e3 in western blots of protein extracts from various tissues and cell lines. Both antibodies recognize a recombinant human Nr2e3 protein expressed by transfected HEK293 cells (Supplementary Material, Figure S1B, lane 13) and a native Nr2e3 protein with an apparent molecular weight of 43 kDa in the rat retina (lane 1). No Nr2e3 reactivity was detected in seven other tissues (lanes 2–8) or three cell lines (lanes 9–11) including two retinoblastoma cell lines, Y79 and Weri-Rb1, which express only a subset of rod photoreceptor genes at low levels. The distribution of protein targets of the two Nr2e3 antibodies was then compared with those of a Crx antibody, on fractionated nuclear and cytoplasmic extracts of wild-type, rd7 and Crx−/− mutant mouse retina (Supplementary Material, Figure S1C). As expected, the two Nr2e3 antibodies (p20 and p183), as well as the anti-Crx p261 antibody, recognize their respective antigens in the retinal nuclear extract of the wild-type mice, but not in that of the corresponding deficient mice, suggesting that the reactivity of these antibodies is specific. These results also demonstrated that Nr2e3 and Crx are nuclear proteins, as expected for transcription factors. Furthermore, no predicted mutant Nr2e3 protein was detected in nuclear or cytoplasmic extracts of rd7 mouse retina by either Nr2e3 antibody, suggesting that the rd7 mouse is a bona fide null mutant of Nr2e3. Interestingly, in the Crx−/− knockout mouse, the protein level of Nr2e3 is much lower than that in the wild-type mouse, whereas the Crx protein level appears unchanged in the rd7 versus the wild-type mouse. These findings are consistent with Nr2e3 being a down-stream target of Crx and suggest that decreased levels of Nr2e3 could contribute to the retinal degeneration phenotype caused by Crx mutations.
To determine which photoreceptor subtype co-expresses Crx and Nr2e3 in the retina, immunofluorescence staining using antibodies specific to Crx (p261) and Nr2e3 (p183) and the cone-specific marker peanut agglutinin (PNA) was performed. Because it is easier to identify cones in primate retina than in mouse retina, adult monkey retinal sections were tested initially. As shown in Figure 2, whereas most of the photoreceptor nuclei in the outer nuclear layer (ONL) reacted with the Nr2e3 antibody, a small number of nuclei located in the outer part of the ONL showed markedly reduced staining (A). Double labeling with PNA identified the Nr2e3-negative cells as cones (A–C), indicating that Nr2e3 is expressed predominantly by rods. In contrast, all photoreceptor nuclei in the ONL, as well as some of the inner nuclear layer (INL) cells stained positively for Crx (D–F). Co-staining for Crx and Nr2e3 (G–I) confirmed that the two proteins are co-localized in most photoreceptor nuclei except for the presumptive cones, which do not express Nr2e3 (arrows). Similar results were obtained with adult mouse retinal sections (data not shown). These findings suggest that Crx and Nr2e3 are co-expressed in rod photoreceptors.
To determine whether this co-expression occurs in the developing photoreceptor cells, we double labeled eye sections of developing mice for Crx and Nr2e3. Crx is first detectable at embryonic day 14.5 (E14.5) (data not shown). At E18.5, Crx staining is seen in a subset of nuclei located in the ventricular zone of the outer retina, whereas no Nr2e3 staining is visible (Fig. 3 A–C). Nr2e3-expressing cells are not detected until after birth, consistent with the observation that Nr2e3 acts down-stream of Crx. At postnatal day 0.5 (P0.5) and P3.5, Nr2e3 reactivity is seen in Crx-positive nuclei (D–I). Beginning at P5.5, a few Crx-positive nuclei become negative for Nr2e3 (arrows), although most cells stain for both proteins (J–L). This pattern of Crx and Nr2e3 expression is even more apparent at P7.5 (M–O), when rhodopsin expression begins to peak, and persists thereafter (P–R) (data not shown). The Nr2e3-negative cells in the ONL of P7.5 and P12.5 retinal sections were confirmed as cones by co-staining for Nr2e3 and cone arrestin (data not shown). These findings suggest that Nr2e3 is co-expressed with Crx in developing and mature rods, implicating Nr2e3 in the development and maintenance of rod photoreceptors.
Crx and Nr2e3 co-occupy the regulatory region of both rod and cone genes in rod photoreceptors
To determine whether the Crx–Nr2e3 interaction occurs on target gene chromatin in vivo, chromatin immunoprecipitation (ChIP) assays were performed on mouse retina using the anti-Crx p261 and anti-Nr2e3 p183 antibodies. ChIP assays have previously demonstrated that the promoter/enhancer regions of several photoreceptor genes are direct targets of Crx (24). Here, we examined whether Nr2e3 is associated with the chromatin of these Crx targets. To distinguish target occupancy in individual photoreceptor subtypes, we tested two mutant mouse strains, Red-DT(A) and C57BL/6J-Pdebrd1le/Pdebrd1le (rd1), in parallel with wild-type mice. Red-DT(A) is a transgenic mouse line that carries diphtheria toxin-A (DT-A) chain driven by a 6.5 kb L-cone opsin promoter, so adults of this strain lack cone photoreceptors (44). Red-DT(A) thus serves as ‘coneless’ mouse in our study. The rd1 strain is congenic with C57BL/6J but carries a retinal degeneration mutation in the Pde6b gene encoding the beta subunit of cGMP phosphodiesterase (45). The vast majority of rods in rd1 mouse retina (98%) degenerate by 17 days of age, and no rods are left after the third or fourth postnatal week. Rod death is followed by loss of cones (46). By 80 days of age, the rd1 mouse has no rods and <1.5% of cones left in the retina, and therefore it represents a rodless and practically coneless mouse. Figure 4A shows that in the wild-type mouse retina, Crx and Nr2e3 co-occupy the promoter and enhancer region of three opsin genes, as well as the promoter region (but not the 3′ region; see Supplementary Material, Figure S2) of several other photoreceptor genes. Crx and Nr2e3 do not bind to the promoter region of a liver-specific gene Albumin (Alb), although this sequence contains several consensus Crx binding sites. The cell type in which Crx and Nr2e3 associate with their target promoters appears to be the rod photoreceptor cells, as a similar pattern is seen with the coneless mice, but not with the rodless/coneless mice. Thus, Crx and Nr2e3 co-occupy the regulatory region of both rod and cone genes in rod photoreceptors as would be expected if the Nr2e3–Crx interaction regulates photoreceptor gene expression in vivo.
ChIP assays were carried out using retinas from P14 Crx knockout (Crx−/−) mice (before photoreceptor degeneration occurs) to determine whether target occupancy by Nr2e3 is dependent on Crx. As shown in Figure 4B, Nr2e3 occupancy of Crx-dependent genes rhodopsin, Pde6b, M-cone opsin and S-cone opsin is abolished in the absence of Crx. This loss of Nr2e3–target interaction in Crx−/− mice is unlikely due solely to a reduction in the Nr2e3 protein level, because Nr2e3 is still associated with the promoter of a Crx-independent gene, Irbp (arrow), which is expressed normally in Crx−/− mice (20). These results suggest that Nr2e3 recognition of the chromatin targets of Crx-regulated genes depends on Crx in vivo. In contrast, binding of Crx to its targets appears not to be affected by Nr2e3 deficiency in P14 rd7 mice (Fig. 4B). Altogether, these observations suggest that Nr2e3 associates with Crx-dependent targets in vivo by interacting with Crx, whereas Crx binding is independent of Nr2e3.
Nr2e3–Crx interaction enhances the transcription of rhodopsin, but represses that of cone opsins in transfected cells
To determine the physiological relevance of Nr2e3–Crx interaction on target gene expression, we measured the ability of Nr2e3 to regulate the transcription of a luciferase reporter controlled by the rhodopsin promoter (BR225-Luc) in the presence of Crx and/or Nrl using transient transfection assays in HEK293 cells. We previously established that Crx and Nrl synergistically activate the activity of the rhodopsin promoter in transfection assays (18). Here, an Nr2e3 expression plasmid was used, alone or in combinations with Crx and/or Nrl expression plasmids. As shown in Figure 5A, Nr2e3 alone doubles rhodopsin promoter activity (lane 3) when compared with a 3-fold increase with Crx (lane 5) and 9-fold with Nrl (lane 7) at the same DNA concentration (100 ng). When Nr2e3 was paired with either Crx (lane 6) or Nrl (lane 8), promoter activation increased. When all three factors were present, the rhodopsin promoter activity increased 175-fold on average, significantly higher than 70-fold achieved in the presence of Crx+Nrl (lane 10 versus lane 9). Thus, Nr2e3 enhances the synergistic activation of the rhodopsin promoter by Crx and Nrl together.
To test the effect of Nr2e3 on cone gene transcription, similar transfection assays were performed using Sop600-Luc, a S-cone opsin promoter/luciferase reporter. Because the effects of these transcription factors on this promoter had not been established, we examined various concentrations of each expression vector in these assays. As seen in Figure 5B, the transactivating activity of Crx is concentration-dependent (lanes 5–7), while neither Nrl nor Nr2e3 demonstrated much transactivating activity under the same conditions (lanes 11–13 and lanes 2–4). Co-expression of both Crx and Nrl showed essentially an additive effect on the S-cone opsin promoter (lane 17 versus lanes 6 and 12), in contrast to the results seen with the rhodopsin promoter. However, Nr2e3 represses, in a dose-dependent manner, the ability of Crx alone (lanes 8–10 versus lane 6) or Crx+Nrl (lanes 18–20 versus lane 17) to activate the S-cone opsin promoter, but has little effect on the activity of Nrl alone (lanes 14–16 versus lane 12). These results suggest that Crx (and possibly Nrl) trans-activates the S-cone opsin promoter, whereas Nr2e3 is a repressor for this promoter by antagonizing the activity of Crx.
To determine whether this suppressive effect of Nr2e3 on Crx-mediated transcriptional activation applies to other cone genes, we performed similar transfection assays with Mop250-Luc, an M-cone opsin promoter/luciferase reporter. As observed with Sop600-Luc, Crx and/or Nrl transactivated the promoter activity of M-cone opsin, whereas Nr2e3 repressed this activity in a concentration-dependent manner (Fig. 5C). Altogether, our results suggest that Crx and Nrl are transactivators of both rod and cone genes, whereas Nr2e3 is a dual regulator that interacts with Crx to enhance the expression of rod genes but repress that of cone genes.
The Nr2e3 mutation in the rd7 mouse has opposing effects on the transcription of rod versus cone genes
To determine whether the dual regulatory function of Nr2e3 detected in the transfection assays actually occurs in vivo, quantitative real-time RT–PCR analysis was used to compare transcriptional levels of selected rod and cone genes in the retinas of P28 rd7 and wild-type mice. At this age, rod and cone function are normal by ERG in the rd7 mouse (38). We analyzed three rod genes, rhodopsin, Pde6b and Pde6a, and three cone genes, S-cone opsin, M-cone opsin and Arr3 (cone arrestin), which are direct targets of Crx and Nr2e3 (Fig. 4). As shown in Figure 6A, the transcription levels of the rod genes in the rd7 retina were reduced by 13% (rhodopsin), 16% (Pde6a) and 16% (Pde6b), respectively, compared with those in the wild-type retina. In contrast, transcription of cone genes was elevated in the rd7 retina: 18% increase for M-cone opsin, 24% for S-cone opsin and 19% for Arr3. Transcription of Irbp, a gene expressed independently of Crx in both rods and cones, reached comparable levels in wild-type and rd7 retinas (data not shown), suggesting that only Crx-dependent transcriptional regulation is altered in the absence of Nr2e3. These results confirm our previous observation that Nr2e3 differentially regulates transcription of rod and cone genes by modulating the activity of Crx.
To determine whether this transcriptional dysregulation in the rd7 retina occurs during development, we examined changes in RNA levels for the earlier mentioned six genes in both rd7 and wild-type mouse retina at P7 and P14, using the P28 RNA levels as references (Fig. 6B and C). In rd7 mice, the transcription of cone genes appears to increase faster and reach higher levels than that in wild-type mice. In contrast, the transcription of rod genes peaks at P14 in both rd7 and wild-type mice. However, in rd7 mice, transcription of rod genes does not persist at these high levels and decreases after P14. As a control, the expression of the Irbp gene peaks before P7 and is maintained at a steady level in the period of P7–P28 in both wild-type and rd7 mice (data not shown), suggesting that Irbp transcription is independent of Nr2e3, as well as Crx. The opposing effects of Nr2e3 deficiency on rod versus cone gene expression further support our observation that Nr2e3 acts as a dual transcription regulator for Crx-dependent rod and cone genes.
Missense Nr2e3 mutations associated with ESCS disrupt the interaction with Crx and/or the regulation of target gene transcription
To elucidate the mechanisms by which missense Nr2e3 mutations cause ESCS in humans, we performed functional analysis on some of the genetically identified human mutations (35). We chose four mutations, R97H and R76W located in the DBD and W234S and R311Q located in the LBD, for our analysis. Each of these four mutations was introduced into the Nr2e3-Full mammalian expression vector by using site-directed mutagenesis. To see whether these mutations affect the ability of Nr2e3 to interact with Crx, we generated 35S-labeled Nr2e3 proteins using in vitro transcription/translation from these mutated constructs. We then tested the ability of these mutant proteins to be co-immunoprecipitated with Crx by the anti-Crx antibody using the wild-type Nr2e3 as a control. Figure 7A shows that R97H, one of the DBD mutations, dramatically reduced the ability of Nr2e3 to be co-immunoprecipitated with Crx (lane 5 versus lane 1). In contrast, no apparent changes were observed for the other three Nr2e3 mutants under the same conditions (lanes 2–4 versus lane 1). These results suggest that R97 in the DBD provides an important residue for interacting with Crx.
Next, we carried out reciprocal experiments to test whether any of the five Crx homeodomain mutations identified genetically, R41Q, R41W, A56T, E80A and R90W (43), has any effect on the ability of Crx to interact with Nr2e3. 35S-labeled mutant forms of Crx were tested for co-immunoprecipitation with Nr2e3 using an anti-Nr2e3 antibody (p183). Figure 7B shows that E80A significantly reduced the ability of Crx to be co-immunoprecipitated with Nr2e3 (lane 3 versus lane 1). No changes in Nr2e3 binding were observed for the other four mutants when compared with the wild-type Crx (lanes 2, 4–6 versus lane 1). These results suggest that E80 is an important residue in Crx for interacting with Nr2e3. Interestingly, E80A is known to have no effect on the DNA-binding activity of Crx, whereas R41Q, R41W and R90W all cause a reduction in Crx's DNA-binding activity (43). Thus, different residues in the Crx homeodomain could play distinct roles in binding to DNA or interacting with Nr2e3. Taken together, the earlier results suggest that a mutation in the DBD of either Nr2e3 or Crx could lead to an aberrant Crx–Nr2e3 interaction, which may contribute to the disease phenotype associated with either gene.
Finally, we examined whether any of the four Nr2e3 mutations affect regulation of Crx (+/− Nrl) transactivation in transient transfection assays. C-terminal (DBD) and N-terminal (LBD) deletion constructs were also tested to map the domains responsible. As shown in Figure 7C, both deletion constructs DBD and LBD significantly reduced activation of the rhodopsin promoter/luciferase (BR225-Luc) by Crx (+/− Nrl). This suggests that both DBD and LBD are required for Nr2e3 transactivating function. Interestingly, expression of LBD also represses the activity of Crx (+/− Nrl) as compared to the samples without Nr2e3. This could be due to an over-expression artifact associated with Nr2e3-mediated transactivation, such as competition for limiting amounts of co-activators. Two missense mutations in the DBD, R76W and R97H (35), significantly reduced Nr2e3-mediated enhancement of Crx (+/− Nrl) transactivation, consistent with the role of the DBD in interacting with Crx. In contrast, two missense mutations in LBD, R311Q and W234S (35), had little or no effect on Nr2e3-mediated potentiation of Crx (+/− Nrl) activation of the rhodopsin promoter.
Similar transfection assays using the M-cone opsin promoter/luciferase reporter (Mop250-Luc) were performed. As shown in Figure 7D, the DBD protein significantly reduced repression of Crx (+/− Nrl) on this promoter. In contrast, the LBD protein significantly increased repression, likely due to the over-expression effect seen with the rhodopsin promoter. Each of the four missense mutations significantly reduced Nr2e3's repression of the M-cone opsin promoter. Altogether, our results suggest that both DBD and LBD are important for Nr2e3 trans-regulatory function and that missense mutations of Nr2e3 can disrupt transcriptional regulation of the photoreceptor genes, leading to photoreceptor disease.
Crx–Nr2e3 interaction on target gene chromatin
This study provides several lines of evidence suggesting that Nr2e3 interacts directly with Crx. This was detected in yeast and confirmed by in vitro co-immunoprecipitation assays using antibodies specific to each protein. Analyses of mutant forms of Nr2e3 and Crx further support this interpretation. In contrast, no interaction between Nr2e3 and Nrl was detected under similar conditions (data not shown), although Nr2e3 also appeared to increase the activity of Nrl on the rhodopsin promoter in transfection assays. Recently another orphan nuclear receptor, Nr1d1, was identified as a Nr2e3 interacting partner, which forms a multi-protein complex with Nr2e3, Nrl and Crx, which can be co-immunoprecipitated from bovine retinal nuclear extracts (40). Because Crx is also known to interact with Nrl (21), our results suggest that Crx may serve as a bridge protein between Nrl and Nr2e3, providing insight into how such complexes are assembled.
Subsequently, ChIP assays showed that Nr2e3 and Crx co-occupy the promoter/enhancer region of both rod and cone genes in rod photoreceptor cells. Nr2e3 was reported to bind to DNA on a dimeric direct repeat containing two AAGTCA half-sites in vitro (17). However, no such dimeric site was found in the promoter of rhodopsin, S-cone opsins or M-cone opsins. Similar half-site(s) with AGGTCA or GAGTCA are present, but Nr2e3 did not recognize these in in vitro protein-DNA-binding assays (Chen, unpublished data). Although we cannot rule out the possibility that Nr2e3 can bind to a target DNA sequence(s) on its own, our combined results from the protein–protein interaction and ChIP assays suggest that Nr2e3 occupies these promoters in vivo by interacting with Crx (and possibly other DNA-binding proteins) in the complex of photoreceptor transcription factors. This is further supported by the finding that Nr2e3 does not occupy Crx-dependent gene targets in Crx−/− mice. Interestingly, Nr2e3 is found in association with the Irbp promoter in Crx−/− mice. The mechanism involved in this phenomenon requires further investigation, but is likely mediated by Otx2, a close family member of Crx, because Otx2 (1) has a homeodomain highly homologous to that of Crx (18), (2) binds to and regulates the Irbp promoter (47,48) and (3) occupies the Irbp promoter in the retina of Crx−/− mice (Peng and Chen, unpublished data). Nevertheless, our finding that Nr2e3 associates with Crx–target chromatin suggests a direct involvement of Nr2e3 in regulating the expression of these photoreceptor genes. This is supported by the results of the co-transfection assays, which demonstrated that Crx and Nr2e3 functionally interact to regulate the promoter activity of specific target genes. This conclusion was further confirmed by the transcriptional dysregulation of these specific target genes in rd7 retina, which lacks Nr2e3 protein. Altogether, our results strongly suggest that Crx and Nr2e3 interact in vivo and that this interaction directly regulates the transcription of both rod and cone genes.
Nr2e3 functions in rod differentiation and maintenance at the cellular level
Consistent with two recent reports (40,41), our immunohistochemistry studies with monkey and mouse retinal sections demonstrated that Nr2e3 is expressed by rod photoreceptors. We also showed that during mouse development, Nr2e3 protein is initially expressed by Crx-expressing cells in the first 3 days after birth (P0.5–P3.5). These Crx positive cells represent postmitotic photoreceptor precursors (49). However, during the period of peak expression for both Nr2e3 and Crx between P5.5 and P12.5, Nr2e3 expression decreases in a subset of Crx positive cells as terminal differentiation of photoreceptor subtypes occurs. This event could be important for establishing rod (Crx+/Nr2e3+) versus cone (Crx+/Nr2e3−) identity, as demonstrated by the fact that lack of a functional Nr2e3 leads to the development of more cones in mice and humans. Nrl is likely to be the key regulator for this event, because the expression of Nrl precedes that of Nr2e3 (but lags behind Crx) and is required for Nr2e3 expression (14,41). Nrl and Nr2e3 are also co-expressed in the rod photoreceptor cells of the human retina (41). Interestingly, Nrl was reported to be present in the nucleus of rods, but in the cytoplasm of cones (50). This distinct subcelluar localization could also account for the apparent decrease in Nr2e3 expression in cone cells in the studies presented here. These immunohistochemistry findings combined with our biochemical data strongly support the notion that Nr2e3 is mainly involved in the differentiation and maintenance of rods, rather than promoting the proliferation of cones as previously hypothesized (39).
Nr2e3 promotes rod differentiation and maintenance via differential regulation of rod and cone gene expression
One important finding from our transient transfection assays is the observation that Nr2e3 activates the expression of rod genes but represses cone genes when interacting with Crx (+/− Nrl). This dual regulatory function is not unique for Nr2e3. Several other nuclear receptors have been reported to interact with both co-activator and co-repressor complexes and to up- or down-regulate transcription, depending on the promoter and cell context. As an example, the thyroid hormone receptors (TRs) repress the transcription of target genes by recruiting co-repressors when the ligand is not available (51), but activate transcription by recruiting co-activators in the presence of the ligand (52). This dual regulatory function plays a critical role in the transformation of the tadpole to the adult frog. Furthermore, Nr2e1 (tailless or TLX), a close family member of Nr2e3, is capable of activating and repressing the transcription of different target genes in Drosophila and mammals (53–55). Interestingly, it has been reported that Otd, a member of the Otx/Crx family, plays a dual role in regulating the specification of ommatidia subtypes in the Drosophila retina by enhancing the transcription of two specific rhodopsin genes, rh5 (blue) or rh3 (UV), but repressing the transcription of rh6 (green) (56). This dual regulatory role of Otd might be analogous to the Crx–Nr2e3 interaction in mammals. The mechanisms by which Nr2e3 activates or represses transcription require further investigation. However, the LBD might play an important role, because deleting or mutating this domain alters the dual regulatory function of Nr2e3. Obviously, the promoter context and the interplay among Nr2e3, Crx, Nrl and other regulatory proteins on various promoter elements must also play critical roles in mediating this dual regulatory function.
The dual regulatory role of Nr2e3 is supported by the results of the quantitative RT–PCR analysis of rod and cone genes in the retina of rd7 versus wild-type mice. The transcription of three cone genes, S-cone opsin, M-cone opsin and Arr3 (cone arrestin), is significantly up-regulated, similar to observations in the Nrl−/− mice (14). However, in contrast to the Nrl−/− mice, which do not express rhodopsin, the rd7 mice produce transcripts for three rod-specific genes, rhodopsin, Pde6a and Pde6b, although their expression levels are reduced when compared with those of the wild-type mice. This is consistent with normal rod function as measured by ERG in young adult rd7 mice. Thus, Nr2e3 is not required for specifying the rod cell lineage initially. Instead, it is required for terminal differentiation of rods by preventing immature rods from adopting a cone cell fate. Our results suggest that repressing the expression of cone genes and enhancing the expression of rod genes is the mechanism involved. In the rd7 retina, rods also undergo slow degeneration with time, which could be due to reduced expression levels of rod-specific genes. Thus, Nr2e3 is also required for maintenance of rod function. The changes in the temporal expression patterns of rod and cone genes in the developing rd7 retina as demonstrated in Figure 6B and C further support the role of Nr2e3 in both development and maintenance of rods. The role of Nr2e3 is summarized by the model shown in Figure 8. The expression of Crx commits a population of progenitor cells to become photoreceptor precursors. Some of these precursor cells become rods by expressing first Nrl and subsequently Nr2e3. In these premature rods, Nr2e3 interacts with Crx and Nrl to enhance their synergistic activity in activating rod-specific genes. At the same time, Nr2e3 also interacts with Crx to suppress cone gene expression. As a result, these Nrl+, Nr2e3+ cells differentiate into mature rods. Precursor cells that are Nrl−, Nr2e3− fail to express rod-specific genes, and therefore, will develop into cones, depending on the action of other transcription factors (‘X’ in Figure 8), such as Trβ2, which is required for M-cone development (16). In the presence of a defective Nr2e3 gene (represented by ΔNr2e3), some Nrl+ immature rods will adopt a cone cell fate (S-cones) due to derepression of cone genes. In the mean time, other rods that have completed terminal differentiation will degenerate because of insufficient levels of rod-specific proteins.
The mechanisms by which missense mutations of human Nr2e3 cause disease
On the basis of the earlier model, we predicted that Nr2e3 mutations associated with disease would show defects in regulatory function for rod and/or cone gene expression. This was confirmed by functional analysis of four missense Nr2e3 mutations, all of which are transversions. Among the four, R97H has been shown to co-segregate with the disease, and R311Q is the most common mutation found in ESCS patients (35). All four mutations appear to alter the regulatory function of Nr2e3 in various ways. R97H reduces the ability of Nr2e3 to interact with Crx both physically and functionally. The other three mutations cause defects in regulatory function, although the gene products appear to interact with Crx normally in vitro. Furthermore, each mutation alters the regulatory function in a different way, depending on which domain is affected. For example, two mutations in the LBD, R311Q and R234S have little effect on transactivating the rhodopsin promoter, but reduce repression of cone promoters. In contrast, two mutations in the DBD, R76W and R97H affect both the transactivating and repressing activities. These results suggest that the phenotypes associated with different mutations reflect changes in some, but not all interactions between Nr2e3 and its partners (co-activators or co-repressors). Future studies to clarify these mechanisms may help in elucidating a phenotype–genotype correlation.
We have also shown that a missense mutation in the homeodomain of human Crx, E80A, diminished the ability of Crx to interact with Nr2e3. The E80A mutation is associated with an autosomal dominant cone–rod dystrophy (28), although it does not impair the ability of Crx to bind DNA (43). The finding that this mutated version of Crx demonstrates an altered affinity for Nr2e3 further supports our model, which postulates that the interaction between Crx and Nr2e3 is crucial in establishing and maintaining the rod photoreceptor phenotype by regulating target gene expression. The Crx homeodomain, therefore, has two distinct functions: binding to DNA regulatory elements and interacting with other regulatory proteins. Defects in either of the two functions could cause disease by mechanisms involving both Crx and Nr2e3. For example, reduced Crx binding to targets caused by R90W, R41Q and R41W mutations could also affect the ability of Nr2e3 to regulate Crx–target genes, as demonstrated by our ChIP assays in Crx−/− mice. Our studies on the Nr2e3–Crx interaction have provided new insights into mechanisms involved in photoreceptor development and disease pathogenesis.
MATERIALS AND METHODS
Nr2e3 expression vectors and other plasmids
The full-length human Nr2e3 coding region (Full, amino acids 1–410) and its deletions and DBD (amino acids 1–141) and LBD (amino acids 221–410) were cloned in frame with the His6-tag in a mammalian expression vector pcDNA3.1/HisC (Invitrogen) at the BamHI and EcoRI sites. This was achieved using PCR with a human EST clone (GenBank accession no. AI591043, from ATCC) as template and primer pairs specific to the respective regions as follows: Full: F1, 5′-ATAGATCTATGGAGACCAGACCAACA-3′ (with BglII site at 5′) and R410, 5′-AGAATTCCTAGTTTTTGAACATATCAC-3′ (with EcoRI site at 5′); DBD: F1 and R141, 5′-TGAATTCAGGACTCAGTGTTGGACTCC-3′; LBD: F221, 5′-ATAGATCTTCCATGAGACCTCGGCTCG-3′ and R410. The same DNA fragments were also cloned in frame with the Gal4 DBD in the yeast bait vector pGBKT7 (BD Biosciences Clontech). Bovine or human Crx and their deleted and mutated forms in mammalian and yeast expression vectors were described previously (25,43). pMT-Nrl and the rhodopsin luciferase reporter BR225-Luc were also reported previously (57). Mop250-Luc contains a human M(green)-cone opsin promoter (−253 to +38 bp, relative to the transcription start site) up-stream of the luciferase coding sequence. This promoter region was cloned in the SmaI site of the pGL2-basic vector (Promega) by PCR, using human genomic DNA as template and a pair of primers specific to the M-cone opsin promoter (5′-TCCGCCTCCCAGATTCAAG-3′+ 5′-TATGGAAAGCCCTGTCCCC-3′). Sop600-Luc containing a mouse S-cone opsin promoter (−504 to +58 bp relative to the transcriptional start site) up-stream of the luciferase coding region in the pGL3-basic vector (Promega) was kindly provided by Dr Douglas Forrest (Mount Sinai School of Medicine, New York, NY, USA). The insert of all the constructs described earlier was sequence approved for the absence of mutations in the respective gene.
Four Nr2e3 mutations identified genetically were introduced into the full-length Nr2e3-pcDNA3.1/HisC mammalian expression vector using GeneEditor in vitro site-directed mutagenesis kit (Promega) with the following mutagenesis oligonucleotides: R76W, 5′-AAGAGGAGCGTAtGGCGGAGGCTCATC-3′; R97H, 5′-ACAAGGCCCACCaCAACCAGTGCCAG-3′; R311Q, 5′-ATCTCTCGGTTCCaGGCATTGGCGGTG-3′ and W234S, 5′-ATGGCCGTCAAGTcGGCCAAGAACCTG-3′. The presence of a specific mutation in each construct was confirmed by sequencing.
Yeast two-hybrid screening and interaction assays
Yeast two-hybrid screening and interaction assays were performed essentially as described previously (26) with minor modifications. pGBKT7 in the Matchmaker Two-Hybrid System 3 (BD Biosciences Clontech) was used as the parental bait vector instead of pAS2. Nr2e3-Full-pGBKT7 containing the full-length human Nr2e3 was used as bait for screening a bovine retinal cDNA library made in the prey vector pACTII (42) in the yeast strain AH109 grown on the selection medium SD-Trp− Leu− His− with 15 mM 3-AT (3-amino-1,2,4-triazole, Sigma). Similar conditions were used for the yeast two-hybrid interaction assays with Nr2e3-Full-pGBKT7, Nr2e3-DBD-pGBKT7 and Nr2e3-DBD-pGBKT7 as bait and Crx-pACTII (25) as prey.
Recombinant Nr2e3 and Crx proteins with or without 35S-labeling were made using TnT T7 Quick Coupled Transcription/Translation Kit (Promega) with the corresponding mammalian expression vectors. Co-immunoprecipitation assays were performed as described previously (25) with minor modifications. The anti-Crx antibody p261 or anti-Nr2e3 antibody p183 was used to immunoprecipitate the respective antigen and its interacting protein. A modified washing buffer containing 50 mM Tris–Cl (pH7.5), 400 mM NaCl and 0.5% Triton X-100 was used in the wash procedure.
Western blot assays
Two anti-Nr2e3 antibodies were generated in rabbits using the synthetic peptides corresponding to amino acids 20–39 (p20) of mouse Nr2e3 and amino acids 183–202 (p183) of human Nr2e3 and affinity purified (Proteintech Group Inc., Chicago, IL, USA). Western blot assays using whole cell extracts of various rat tissues and cultured human cell lines were performed as described previously (26), except ECL Plus western blotting detection system and ECL Hyperfilm (Amersham Biosciences) were used for detection. The nuclear and cytoplasmic extracts of mouse retina were made using the NE-PER nuclear and cytoplasmic extraction reagents (Pierce, Rockford, IL, USA) following manufacturer's instructions. The mouse strains used in these assays include C57BL/6J (wild-type, The Jackson Lab), Nr2e3rd7/rd7(rd7, The Jackson Lab), a mutant mouse homozygous for a deletion mutation of Nr2e3 (38,39), and a Crx knockout mouse (Crx−/−, kindly provided by Dr Constance Cepko, Harvard Medical School, Boston, MA, USA) (20). The primary antibodies used in the western blots include rabbit polyclonal antibodies anti-Nr2e3-p20 (1 : 500), anti-Nr2e3-p183 (1 : 2000), anti-Crx-p261 (1 : 100) (25) anti-BAF (1 : 1000) (26) and the monoclonal antibody anti-β-actin (1 : 5000, Sigma) as loading controls. The primary antibodies were visualized with horseradish peroxidase coupled anti-mouse IgG (for β-actin) or anti-rabbit IgG antibodies (Santa Cruz) at a 1 : 10,000 dilution.
Eye sections from adult and embryonic mice were prepared and immunostained with specific antibodies as described previously (26) with some modifications. C57BL/6J (wild-type) mice were used for these studies, and Nr2e3rd7/rd7 (rd7) mice were used as negative controls in Nr2e3 staining. Paraffin sections of the retina from a 8-year-old female Rhesus monkey (Macaca mullata) were provided by Dr Rosario Hernandez (Washington University, St. Louis, MO, USA). Goat anti-rabbit IgG antibodies coupled to Alex Fluor A488 or Alex Fluor A568 (Molecular Probes) were used as secondary antibodies at a 1 : 400 dilution. In addition, a fluorescein-labeled a fluorescein-labeled peanut agglutinin (PNA) (Vector Laboratories) at a 1 : 500 dilution was used to stain the cone outer segments. The stained slides were coverslipped with Vectashield mounting medium with DAPI (Vector Laboratories) and visualized using either a fluorescence light microscope (Olympus BX51) with a Spot Cooled Color Digital Camera (Diagnostic Instrument Inc.) or a scanning confocal microscope (Model 510, Carl Zeiss). For double-labeling with Crx and Nr2e3, sequential reactions with two rabbit primary antibodies were performed. First, Nr2e3 was labeled using either anti-Nr2e3-p20 (1 : 200) or p183 (1 : 1000) as the primary antibody and goat anti-rabbit IgG-A568 (1 : 400) as the secondary, followed by labeling Crx using anti-Crx-p261(1 : 100) as the primary and goat anti-rabbit IgG-A488 (1 : 400) as the secondary. An extra fixation step (10% formalin for 1 h) was performed between the two incubations to minimize cross-reactions. Omitting the primary antibody (anti-Crx-p261) in the second labeling was used as a negative control. Antigen retrieval was performed in each single staining, as described previously (26). Markers used in labeling cones include fluorescein PNA (Vector Laboratories, 1 : 400) and a rabbit anti-mCAR (cone arrestin) antibody (1 : 1000) (58) kindly provided by Dr Cheryl Craft, University of Southern California.
Chromatin immunoprecipitation assays
These were performed using six pooled mouse retinas, as described previously (24). The mouse strains used in ChIP assays include C57BL/6J (wild-type, 6-8-weeks-old or P14 during development), a mutant strain C57BL/6J-Pdebrd1le/Pdebrd1le (rd1, The Jackson Lab) referred as rodless/coneless (at 80 days) (46), a transgenic line, Red-DT(A) referred as coneless (at 48 days, a gift form Dr Jeremy Nathans, Johns Hopkins University, Baltimore, MD, USA) (44), Crx−/− (at 14 days) and Nr2e3rd7/rd7 (rd7) (at 14 days). The chromatin DNA immunoprecipited by anti-Crx (p261), anti-Nr2e3 (p183) and normal rabbit IgG (Santa Cruz, a negative control), as well as input controls (without immunoprecipitation), was analyzed by PCR, using primers specific to the promoter or enhancer region of selected Crx target genes. In addition to those primers described earlier (24), new pairs of primers were included to detect Pde6a 5′ promoter region (200 nt, 5′-GTATGGATGTCTCCTCTTGGC-3′+5′-GGTTCTCTGGGTCTTGTCTGC-3′), Pde6a 3′ region (230 nt, 5′-ACCATCAACATCCCCGCTG-3′ + 5′-CAGAACCAACCCCCCAATC-3′), Arr3 (cone arrestin) 5′ promoter region (196 nt, 5′-CAACCCAGCACAGGATAATG-3′ + 5′-TGATTGTTACTGAGGAGGTAGGC-3′), Arr3 3′ region (200 nt, 5′-ATCTCTCTGCTTCAGGCACC-3′ + 5′-TCCTCTCTGGGGTTTTTCAC-3′), Irbp 5′ promoter region (123 nt, 5′-CCTCACATCTAACTCCCACATTG-3′ + 5′-CCTTGGCTCCTGGATAAGAG-3′) and SER (S-cone opsin enhancer region) (199 nt, 5′-CCAGCCTGATTTCTCTTACACC-3′ + 5′-TGCTTCTCCCAGACTATGTGAG-3′).
Transient transfection and dual luciferase assays
These were performed in HEK293T cells cultured in 35 mm plates, as described previously (25). A 2 µg of luciferase reporters, BR225-Luc, Sop600-Luc or Mop250-Luc, 1 ng of the internal control pRL-CMV and 100–400 ng of each mammalian expression vector were transfected into 293 cells and analyzed using the dual luciferase assays. At least three independent trials were performed for each experiment. The P-values were calculated by Student's t-test assuming two samples with equal variance.
Quantitative real-time RT–PCR analysis
Total RNA was isolated from four retinas dissected from the rd7 and wild-type (C57BL/6J) mice at the ages of 7, 14 and 28 days, respectively, using Versagene RNA Tissue Kit (Gentra Systems, Inc. Minneapolis, MN, USA). One microgram of total RNA was reverse transcribed in independent triplicates, using Transcriptor First Strand cDNA Synthesis Kit (Roche Applied Science, Indianapolis, IN, USA) and anchored-oligo-(dT)18 primer. The cDNA was diluted 10-fold and quantified by real-time PCR analysis in triplicates, using SYBR Green JumpStart ReadyMix (Sigma) and iCycler PCR machine (Bio-Rad). Primers pairs used in the PCR reactions include Rhodopsin (190 nt, 5′-GCTTCCCTACGCCAGTGTG-3′ + 5′-CAGTGGATTCTTGCCGCAG-3′), Pde6b (216 nt, 5′GTGCTGCTGTGGTCGGCCAAC-3′+5′-CCGGCCATCAGGCGTGCGTGG-3′), Pde6a (274 nt, 5′-CACTCCTGAGAGATGAGAGCC-3′ + 5′-CAGGGTTTGGTGATGGCTG-3′), S-cone opsin (252 nt, 5′-GCTGGACTTACGGCTTGTCACC-3′ + 5′-TGTGGCGTTGTGTTTGCTGC-3′), M-cone opsin (246 nt, 5′-GGTGGTGATGGTCTTCGCATAC-3′ + 5′-TTGGAGGTGCTGGAAAGTTCAG-3′), Arr3 (163 nt, 5′-AAGTTTTCCATCTACCTGGGG-3′ + 5′-TCACATCCAAGTCATCACGG-3′), Irbp (Rbp3, 176 nt, 5′-CTCGGTCAGCGAACTTTGG-3′ + 5′-GGCAACCTCCACTTGTCACTTC-3′) and β-Actin (215 nt, 5′-CCAACTGGGACGACATGGAG-3′ + 5′-TGGTACGACCAGAGGCATACAG-3′) as a loading control. Relative expression levels were determined as described previously (25) with calibration curves generated for each transcript using a serial dilution of reference retinal cDNA from adult C57BL/6J mice. Linear relationships between the amounts of sample and cycle threshold (Ct) values were obtained for all standard curves. Relative expression levels were determined by comparing the amounts of cDNA in individual samples with the standard curve and then normalizing to the amount of β-Actin cDNA.
We wish to thank Dr Connie Cepko for providing Crx knockout mouse, Dr Jeremy Nathans for Red-DT(A) mouse, Dr Douglas Forrest for Sop600-luc plasmid, Dr Ching-Hwa Sung for the bovine retinal cDNA library, Dr Rosario Hernandez for providing monkey retinal sections, Dr Cheryl Craft for mCAR antibody, Dr Xiaojie Yao for constructing the full-length Nr2e3 mammalian expression vector, Siqun Xu, Qingsheng Xu, Belinda McMahan and Steve Turney for technical assistance and Dr Anne Hennig, Dr David Beebe and Dr Russell van Gelder for critical reading of the manuscript. This work was supported by grants from the National Institutes of Health (EY12543 to S.C., EY02687 to WU-DOVS), Research to Prevent Blindness (RPB) and Knights Templar Foundation (to G.P.). S.C. is a recipient of Sybil B. Harrington Scholar Award from RPB.
Supplementary Material is available at HMG Online.