Retinoid-related orphan nuclear receptor RORβ is an early-acting factor in rod photoreceptor development

Abstract

Rods and cones are morphologically and developmentally distinct photoreceptor types with different functions in vision. Cones mediate daylight and color vision and in most mammals express M and S opsin photopigments for sensitivity to medium-long and short light wavelengths, respectively. Rods mediate dim light vision and express rhodopsin photopigment. The transcription factor networks that direct differentiation of each photoreceptor type are incompletely defined. Here, we report that Rorb^−/− mice lacking retinoid-related orphan nuclear receptor β lose rods but overproduce primitive S cones that lack outer segments. The phenotype reflects pronounced plasticity between rod and cone lineages and resembles that described for Nrl^−/− mice lacking neural retina leucine zipper factor. Rorb^−/− mice lack Nrl expression and reexpression of Nrl in Rorb^−/− mice converts cones to rod-like cells. Thus, Rorb directs rod development and does so at least in part by inducing the Nrl-mediated pathway of rod differentiation.

Rods and cones are distinct receptor cell types that mediate dim and bright light vision, respectively. In the mouse, cones are generated between midgestation and birth (1) and subpopulations differentially express M and S opsin photopigments for sensitivity to medium-long and short light wavelengths, respectively (2). Rods express rhodopsin and greatly outnumber cones in mice. Rod generation lags behind that of cones and is more protracted, lasting until about a week after birth. Rods, cones, and other retinal cell types are generated in a stereotypical order from multipotent progenitors, and it has been proposed that a combination of transcription factor activities and external signals at a given developmental stage prompts progenitors to enter specific differentiation pathways (3, 4).

The transcription factors that direct the generation of rod and cone precursors and the terminal differentiation of these cell types are incompletely defined. During terminal differentiation of cones, thyroid hormone receptor TRβ2 is required for M opsin induction such that without TRβ2, cones express only S opsin (5). Factors that promote rod differentiation and survival include leucine zipper protein Nrl (6), orphan nuclear receptor Nr2e3 (7, 8), homeodomain proteins Crx and Otx2 (9–11), and retinoblastoma protein Rb (12, 13). Nrl induces Nr2e3 expression and these two genes define a transcriptional hierarchy for rod differentiation. Nrl^−/− mice overproduce S cones at the expense of rods (6) whereas ectopic Nrl expression converts cones to rods (14). Nr2e3 deficiency causes an enhanced cone phenotype and misexpression of cone genes in rods, suggesting that Nr2e3 represses cone genes to maintain the rod phenotype (15–19). Human NRL and NR2E3 mutations result in retinopathy phenotypes (8, 20). The above findings suggest that rod and cone precursors share a default differentiation program as cones and that rod differentiation requires the action of additional transcription factors.

The Rorb gene encoding retinoid-related orphan nuclear receptor RORβ is expressed in the brain, pineal gland, and retina. Rorb is expressed in all neural retina layers from early stages with a peak at neonatal stages, suggestive of a role in many differentiating retinal cell types including both cones and rods (21–23). We previously reported that Rorb and Crx synergistically induce the S opsin promoter (24), indicating a role for Rorb in cone differentiation. Here, we report a role for Rorb in rods. We found that Rorb^−/− mice overproduce cones at the expense of rods and lack Nrl and Nr2e3 expression. Reexpression of Nrl in these mice converted the excess cones to rod-like cells. Thus, Rorb is critical for rod differentiation and lies upstream of Nrl in the rod transcriptional pathway.

Results

Loss of Nrl-Mediated Rod Differentiation Pathway in Rorb^−/− Mice.

Rorb^−/− mice displayed gross overexpression of S opsin and severe loss of rhodopsin mRNA (Fig. 1A). A small number of rhodopsin-positive (rhodopsin⁺) cells remained around the outer nuclear layer (ONL). Although overexpressed in Rorb^−/− mice, S opsin followed a normal distribution gradient with stronger signals in the inferior (ventral) than superior (dorsal) retina (25). In wild type (+/+) mice, cones represent only 3% of all photoreceptors and reside near the outer edge of the ONL whereas rods represent 97% of photoreceptors and populate the entire depth of the ONL (26). In Rorb^−/− mice at P23, S opsin⁺ cells populated the entire ONL. The extent of S opsin overexpression was emphasized by the fact that Rorb^−/− mice possessed only approximately 40% of the number of photoreceptors found in +/+ mice (see Fig. 3E). A few displaced S opsin⁺ cells were detected in the inner nuclear layer (INL) in Rorb^−/− mice. M opsin expression in Rorb^−/− mice at P23 was slightly elevated in the superior retina but was absent in the inferior reflecting an exaggeration of the normal bias in distribution found in +/+ mice.

Fig. 1.

Enhanced S opsin phenotype in Rorb^−/− mice. (A) Overexpression of S opsin and loss of rhodopsin in Rorb^−/− mice at P23 shown by in situ hybridization in superior (sup) and inferior (inf) retina. At P23, S opsin⁺ cones reside at the outer edge of the ONL (blue bar) in +/+ mice but populate the entire ONL in Rorb^−/− mice. Rorb^−/− mice lack outer segments (OS) and the ONL is adjacent to the retinal pigmented epithelium (RPE, arrowhead). In Rorb^−/− mice, a few (approximately 200) residual rhodopsin⁺ cells were detectable per 10 μm thick cryosection. (B) Late onset of S opsin overexpression in Rorb^−/− mice. Note that in +/+ mice, cones migrate across the nascent ONL toward the inner nuclear layer (INL) then return to the outer edge of the ONL to reach their mature location near weaning age. (C) Western blot showing lack of S opsin at P6 and overexpression at P28 in retina in Rorb^−/− mice.

Fig. 3.

Recovery of rod-like photoreceptors in Rorb^−/−;CrxpNrl mice. (A) Western blot analysis. The CrxpNrl transgene expresses Nrl and reinduces rod markers Nr2e3, rhodopsin, Gnat1, Gnb1,and Pde6a in Rorb^−/− mice at ages indicated. (B) In situ hybridization showing partial rescue of rod gene expression in Rorb^−/−;CrxpNrl mice. In Rorb^−/− mice, the ONL (blue bar) is adjacent to the RPE (arrowhead). (C) In situ hybridization showing suppressed cone gene expression in Rorb^−/−;CrxpNrl mice. (D) Methacrylate sections showing a thin ONL in Rorb^−/− mice at P14. Compared to Rorb^−/− mice, Rorb^−/−;CrxpNrl mice have a more organized ONL, rod-like photoreceptor nuclei and recovery of a small OS layer at P14. OPL, outer plexiform layer, IPL, inner plexiform layer, GCL, ganglion cell layer. (Scale bars, C, D, E, 50 μm.) (E) Rorb^−/− and Rorb^−/−;CrxpNrl mice have approximately 40% of normal photoreceptor numbers. All photoreceptor (cone and rod) nuclei were counted in ONL fields on 3 μm sections at P14. Means ± SD; **, P < 0.001 versus +/+ mice.

The similarity of the above phenotype to that of Nrl^−/− mice suggested that in Rorb^−/− mice, photoreceptor precursors that would normally form rods differentiated instead as S cones. S opsin overexpression in Rorb^−/− mice occurred between P9 and P14 in a large, late-maturing population of photoreceptors (Fig. 1B). Rorb^−/− mice also produced a small, early-appearing cone population like that of +/+ mice that expressed the early cone marker TRβ2 at E18.5 (see Fig. 2C). However, unlike in +/+ mice, this cone population in Rorb^−/− mice failed to express S opsin mRNA or protein at P6 (Fig. 1 B and C) in accord with the previous proposal that Rorb is required with Crx to induce S opsin in normal cone development (24). Somewhat paradoxically, S opsin overexpression at later stages in the excess cones in Rorb^−/− mice was independent of Rorb, suggesting that this is mediated by other means than in normal cone development. The findings indicate that Rorb mediates distinct functions in cone and rod precursors.

Fig. 2.

Loss of Nrl and Nr2e3 expression in Rorb^−/− mice. (A) Northern blot showing loss of Nrl and rhodopsin and overexpression of S opsin mRNA in Rorb^−/− mice at P14. Numbers below lanes, quantitation relative to +/+ samples, assigned a value 1.0 (normalized to G3PDH); n.d., not detectable. (B) In situ hybridization showing loss of Nrl and Nr2e3 but retention of Crx and Otx2 mRNA expression in Rorb^−/− mice. Expression was analyzed in the first postnatal week (P6) when Nrl, Nr2e3, and Crx approach peak expression. Otx2 was analyzed at E17.5 as peak Otx2 expression in the ONBL occurs in the embryo. (C) Presence of Crx⁺ and TRβ2⁺ photoreceptor precursors in Rorb^−/− embryos. Immunofluorescent detection of RORβ-β-galactosidase (nonnuclear, green cell bodies) expressed from lacZ knocked into Rorb (22) and Crx (red nuclei) a marker for rods and cones or TRβ2 (red nuclei) a marker for cones. (Scale bars, B, 50 μm, C, 20 μm.)

Northern blot, in situ hybridization, and Western blot analyses revealed loss of Nrl and Nr2e3 expression in Rorb^−/− mice (Figs. 2 A and B and 3A). However, Rorb^−/− mice still expressed Crx, which is required for terminal differentiation of both rods and cones (9, 10), and Otx2, which induces Crx expression and participates in rod differentiation (11). The findings indicate that Rorb controls rod differentiation upstream of Nrl and Nr2e3 but independently, or downstream, of Otx2. In situ hybridization and quantitative PCR did show reduced Crx expression in postnatal Rorb^−/− pups (Fig. S1), a decrease that may be explained by the partial loss of photoreceptors, although it is not ruled out that Rorb is also necessary for full expression of Crx.

Immunostaining for Crx as a marker for all photoreceptors (rods and cones) or TRβ2 as a marker for cones indicated that Rorb^−/− embryos produced photoreceptor precursors that would normally form both rods and cones (Fig. 2C). Rorb gene expression, revealed using an Rorb-lacZ knockin allele, was detected in almost all neural retinal cells, including both Crx⁺ and TRβ2⁺ cells. In Rorb^−/− embryos, Crx⁺ cells and TRβ2⁺ cells were dispersed in both outer and inner neuroblastic layers, indicating early abnormalities in migration and retinal organization. Increased numbers of apoptotic cells were detected in the retina of Rorb^−/− embryos and juvenile mice that could account for the ultimate reduction of photoreceptor numbers (Fig. S2).

Nrl Reexpression Partly Rescues Rod Development in Rorb^−/− Mice.

To establish that Rorb acts upstream of Nrl in the same rod differentiation pathway, an Nrl-expressing transgene was introduced into Rorb^−/− mice by crossing Rorb^−/− mice with CrxpNrl transgenic mice. The CrxpNrl transgene carried a Crx promoter that is active in both cone and rod precursors from E12.5 onwards (27, 14). Moreover, expression of Crx, unlike many photoreceptor genes, was retained in Rorb^−/− mice. Rorb^−/− mice carrying the transgene (Rorb^−/−;CrxpNrl) were analyzed from early postnatal stages up to approximately 4 weeks of age, spanning the period of peak Rorb and Nrl expression in the first postnatal week (24, 28) and the subsequent terminal differentiation of rods in normal mice. Given the disorganization of the retina in adult Rorb^−/− mice (22), older mice were not analyzed to avoid the possibility of degeneration obscuring the comparison of Rorb^−/− and Rorb^−/−;CrxpNrl mice.

Rorb^−/−;CrxpNrl expressed Nrl protein at 51% of the level of +/+ mice at P7 as shown by Western blot analysis (Fig. 3A). The lower level may be explained by the lower photoreceptor number in Rorb^−/− mice and by differences in the strength of the transgene promoter relative to the endogenous Nrl promoter. Most rod genes tested, including Nr2e3, Rho, Gnat1, Gnb1 and Pde6a showed a degree of recovered expression in Rorb^−/−;CrxpNrl mice as detected by Western blot, in situ hybridization (Fig. 3B) and microarray (Fig. 5A) analyses. In +/+ mice, the CrxpNrl transgene suppresses cone genes suggesting that Nrl can inhibit all photoreceptor precursors, including those that would normally form cones, from acquiring cone properties (14). In Rorb^−/−;CrxpNrl mice, cone gene expression was strongly suppressed indicating that all cones in Rorb^−/− mice responded to Nrl (Fig. 3C).

Fig. 5.

Retinal gene expression profiles. (A) Microarray analysis showing overlap in gene sets with ≥2-fold expression changes in Rorb^−/− and Nrl^−/− mice relative to +/+ mice at P28. For overexpressed genes, 100 are common to Rorb^−/− and Nrl^−/− mice (out of totals 410 and 124 genes, respectively); 104 under-expressed genes are common to Rorb^−/− and Nrl^−/− mice (out of totals 247 and 132, respectively). Unidentified genes were excluded and multiple probes were consolidated for a given gene. (B) Quantitative PCR data showing suppressed rod gene and enhanced cone gene expression in Rorb^−/− mice at P28. CrxpNrl partly restores expression of most rod genes and suppresses cone genes. A few under-expressed genes (Gngt1, Rgs9bp) in Rorb^−/− mice were not substantially reinduced by CrxpNrl. (C) Diagram of RORβ, Nrl and Nr2e3 as a hierarchy of factors required for rod differentiation. Loss of these factors allows precursors that would normally form rods to differentiate instead with cone properties. TRβ2 is required for M opsin induction in cones.

Furthermore, the morphology of photoreceptor nuclei was converted from cone-like in Rorb^−/− mice to rod-like in Rorb^−/−;CrxpNrl mice (discussed below). However, total photoreceptor numbers remained equally reduced in both Rorb^−/− and Rorb^−/−;CrxpNrl mice when counted at P14 (Fig. 3 D and E). It is possible that CrxpNrl expression occurs too late to correct defects in photoreceptor generation or survival in Rorb^−/− mice or that these actions of Rorb are not mediated by Nrl.

Transmission electron micrographs showed in detail the recovery of rod-like cells in Rorb^−/−;CrxpNrl mice (Fig. 4A). In +/+ mice, cones displayed large nuclei with loose chromatin whereas rods displayed smaller and denser nuclei. In Rorb^−/− mice, all photoreceptor nuclei were cone-like whereas in Rorb^−/−;CrxpNrl mice, almost all nuclei had a rod morphology. Photoreceptors in Rorb^−/− mice lacked inner and outer segments (24). Rorb^−/−;CrxpNrl mice possessed many miniature segments (Fig. 4B) indicating that the role of Rorb in segment formation is mediated in part by Nrl.

Fig. 4.

Rod-like morphology of photoreceptors in Rorb^−/−;CrxpNrl mice. (A) Transmission electron micrograph of a +/+ mouse retina at P14 showing many compact rod nuclei with dense chromatin (r) and sparser, larger cone nuclei with loose chromatin (arrowheads). In Rorb^−/− mice, nuclei are cone-like; rod nuclei are absent. In Rorb^−/−;CrxpNrl mice nuclei have a rod morphology. A rare remaining cone-like cell is marked (arrowhead, right panel). (B), Higher power magnification showing long OS containing stacked disc membranes of +/+ mice (Left). Rorb^−/− mice lack inner and outer segments (IS/OS) (Middle) whereas Rorb^−/−;CrxpNrl mice have recovered small IS/OS segments (arrowheads)(Right).

Photoreceptor Gene Expression Profiles.

Microarray analyses showed that most retinal gene expression changes in Nrl^−/− mice were also represented in Rorb^−/− mice at P28, confirming the overlap in Rorb and Nrl functions (Fig. 5A). Approximately 80% (100/124) of the genes with ≥2-fold increased expression in Nrl^−/− mice were similarly represented in Rorb^−/− mice. Almost 80% (104/132) of genes with ≥2-fold decreased expression in Nrl^−/− mice were similarly represented in Rorb^−/− mice. Rorb^−/− mice showed greater total numbers of genes with expression changes than did Nrl^−/− mice in accord with the additional defects in nonphotoreceptor cell types in Rorb^−/− mice. It is also possible that these genes include some photoreceptor genes that are under control of Rorb but not Nrl. Microarray analyses also showed that the CrxpNrl transgene in Rorb^−/− mice partly reversed the expression change of 93% of the genes that were increased in common with Nrl^−/− mice (Tables S1 and S2). The transgene also reversed expression of 45% of the genes that were decreased in common between the two mutant strains. Of the genes that were altered only in Rorb^−/− mice, the transgene reversed the expression pattern of relatively few (<17%), consistent with CrxpNrl influencing primarily photoreceptors but not other retinal cell types.

Expression changes for selected photoreceptor genes were corroborated by quantitative PCR analysis (Fig. 5B). Rorb^−/− mice displayed decreased expression of rod-related genes but enhanced expression of cone-related genes. The expression changes for most genes examined in Rorb^−/− mice showed partial reversal in Rorb^−/−;CrxpNrl mice in accord with the microarray data.

Discussion

This study indicates that Rorb is critical for rod development and that it acts upstream of Nrl and Nr2e3 to control the decision of a photoreceptor precursor to form a rod or cone (Fig. 5C). Mutations in Rorb, Nrl, or Nr2e3 in mice result in enhanced expression of cone genes at the expense of rod genes reflecting an innate potential of rod precursors to differentiate as cones. Mutation in Rorb gives the most severe phenotype of these three genes, with rods being replaced by primitive, nonfunctional cones that lack outer segments. The presence of a small number of rhodopsin⁺ cells in Rorb^−/− mice that has not been reported in Nrl^−/− mice may be explained by possible residual Nrl expression.

We propose that photoreceptor precursors differentiate by default as cones unless Rorb provides an initial impetus to commit to rod differentiation. We suggest that Rorb with other unidentified factors promotes the induction of Nrl. Nrl in turn induces Nr2e3 and the combined function of these genes fixes the precursor in the rod developmental pathway by inducing rod genes and suppressing cone genes (14). It is possible that Rorb also mediates downstream events in rod differentiation, perhaps in cooperation with Nrl, Nr2e3, or Crx. Another newly identified gene in this rod pathway is Pias3 encoding an E3 SUMO ligase that stimulates the cone suppressing activity of Nr2e3 (29). The Nrl gene possesses candidate response elements for nuclear receptors (30), but it is currently unclear if Nrl is directly or indirectly induced by Rorb.

Rorb is widely expressed in the immature retina, and it evidently serves dual roles in promoting both rod and cone differentiation. In contrast, Nrl expression is restricted to rod precursors and, moreover, when ectopically expressed in cones blocks cone differentiation (14). A function for Rorb in cones was proposed in previous studies showing that Rorb and Crx synergistically induce the S opsin gene (Opn1sw) promoter during normal cone development (24). It is likely that these independent activities of Rorb in rod and cone differentiation are determined in cooperation with other currently unidentified rod- and cone-specific factors that direct regulation of distinct target gene networks in each cell type.

The overexpression of Opn1sw in the excess cones in Rorb^−/− mice is seemingly paradoxical because we previously reported that Rorb contributes to Opn1sw induction in normal cone development (24). However, recent studies show that the same response element in the Opn1sw promoter can be activated by RORβ or RORα and that Rora mutant mice have reduced Opn1sw expression (31). Rora may therefore provide a compensatory means of inducing Opn1sw in the excess cones in Rorb^−/− mice. Alternatively, Crx alone may suffice to induce Opn1sw in the excess cones. The evidence that the excess cones in Rorb^−/− mice differ from natural cones in allowing Rorb-independent induction of Opn1sw raises the possibility that the excess and natural cone populations may differ in other subtle ways.

Apart from its role in photoreceptor precursors, Rorb may also act in multipotent progenitors as suggested previously by ectopic Rorb expression in embryonic rat retina that increased cell clone sizes (23). The precise role of Rorb in progenitors is unclear, but it is not essential to generate photoreceptor precursors. However, Rorb does influence the rate of generation or survival of these cells because Rorb^−/− mice ultimately possess only 40% of normal photoreceptor numbers. Like Rorb^−/− mice, Otx2-deficient mice also lack rods but overproduce amacrine-like cells instead of cones (11), revealing a complex interdependence as progenitors are directed toward different retinal cell fates. It is unknown if Otx2 induces Rorb or otherwise promotes rod differentiation. It is possible that the Rorb-Nrl-Nr2e3 gene hierarchy acts in parallel with the Otx2-Crx hierarchy in photoreceptor differentiation, although cross-talk is also likely.

Although the photoreceptor phenotype shows substantial overlap in Rorb^−/− and Nrl^−/− mice, subtle distinctions also suggest some different functions for Rorb. An example is that overexpressed S opsin retains a normal distribution gradient over the inferior-superior axis of the retina in Rorb^−/− mice whereas in Nrl^−/− mice, S opsin is more evenly expressed in all retinal regions (6). A final observation is that Nrl^−/− and Nr2e3 mutant mice (7, 6) but not Rorb^−/− mice form prominent hyperplastic folds in the photoreceptor layer, suggesting that changes in Rorb^−/− mice preclude the events that cause folding. A possible explanation is that the lower photoreceptor density in Rorb^−/− mice alleviates folding (32).

Materials and Methods

Details of cDNAs, primers and antibodies are given in Tables S3, S4, and S5.

In Situ Hybridization and Northern Blot Analysis.

Eyes were fixed in 4% paraformaldehyde and cryosections were hybridized with digoxigenin-labeled riboprobes as described (33). For Northern blots, 15 μg samples of eye total RNA (7 mice/group) were analyzed using ³²P-labeled probes (5).

Western Blot Analysis, Antibodies and Immunohistochemistry.

Ten microgram protein samples of nuclear extracts (for Nrl, Nr2e3) or whole cell extracts (other proteins) or 25 μg of whole cell extract (S opsin) were analyzed by Western blot. Bands were quantified by densitometry and normalized to actin (33). Samples were heated at 95 °C before gel-loading only for analysis of Nrl and Nr2e3. Rabbit antiserum was raised against Crx residues 261–274 YSPVDSLEFKDPTG (17) (Covance). For immunofluorescence, 10 μm cryosections were incubated with rabbit anti-TRβ2 or anti-Crx (1:2,500) and Alexa Fluor 568 goat anti-rabbit (1:500), monoclonal anti-βGal (1:200), and Alexa Fluor 488 goat anti-mouse (1:500) antibodies.

Mouse Strains.

Rorb^−/− mice carrying a lacZ knockin (22) on a C3H/HeN background were backcrossed for 2 generations onto a C57BL/6J background. The rd1 mutation was bred out of the C3H background using PCR genotyping (34) with a modified primer (RD6 5′-TACCCACCCTTCCTAATTTTTCTCACGC-3′). Rorb^+/− mice were crossed to generate +/+, +/− and −/− littermates for analysis. CrxpNrl transgenic mice (14) were rederived onto a C57BL/6 × 129/Sv background then crossed with Rorb^−/− mice. Animal studies followed approved institutional protocols.

Histology and Transmission Electron Microscopy.

Eyes were fixed in 3% glutaraldehyde/2% paraformaldehyde. Three micrometer methacrylate sections were stained with hematoxylin and eosin (24). Cells were counted on 160 μm lengths of ONL in the central retina on 3 sections each from 5, 5, and 4 eyes from 3 +/+, 3 Rorb^−/− and 2 Rorb^−/−;CrxpNrl mice, respectively at P14. Electron microscopy was performed on 2 mice/group (JFE Enterprises) (24).

Microarray and Quantitative PCR Analysis.

Microarray data are available in the National Center for Biotechnology Information Gene Expression Omnibus under accession no. GSE16585 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE16585). Microarray analysis was performed on Affymetrix gene chips on 4 independent samples each representing an individual mouse. The experiment was repeated at P28 and P14. Data were compared with data of P28 Nrl^−/− mice, using a false discovery rate with confidence interval <0.5 as described (35). Real-time PCR was performed twice on triplicate samples of cDNA made from 2.5 μg RNA samples (≥3 mice/group) using SybrGreen I (Molecular Probes) and iCycler IQ PCR detection (Bio-Rad). Average threshold cycle (Ct) differences were normalized to β-actin control. To assess reversal of gene expression in Rorb^−/−;CrxpNrl mice, for overexpressed genes in Rorb^−/− mice, genes were scored when the ratio of Rorb^−/− to Rorb^−/−;CrxpNrl was >1.5-fold. For underexpressed genes, genes were scored when the ratio of Rorb^−/−;CrxpNrl to Rorb^−/− was >1.5 fold.