Role of RDS and Rhodopsin in Cngb1-Related Retinal Degeneration.

PURPOSE: Rod photoreceptor outer segment (OS) morphogenesis, structural integrity, and proper signal transduction rely on critical proteins found in the different OS membrane domains (e.g., plasma, disc, and disc rim membrane). Among these key elements are retinal degeneration slow (RDS, also known as peripherin-2), rhodopsin, and the beta subunit of the cyclic nucleotide gated channel (CNGB1a), which have been found to interact in a complex. The purpose of this study was to evaluate the potential interplay between these three proteins by examining retinal disease phenotypes in animal models expressing varying amounts of CNGB1a, rhodopsin, and RDS.
METHODS: Outer segment trafficking, retinal function, and photoreceptor structure were evaluated using knockout mouse lines.
RESULTS: Eliminating Cngb1 and reducing RDS leads to additive defects in RDS expression levels and rod electroretinogram (ERG) function, (e.g., Cngb1-/-/rds+/- versus rds+/- or Cngb1-/-) but not to additive defects in rod ultrastructure. These additive effects also manifested in cone function: Photopic ERG responses were significantly lower in the Cngb1-/-/rds+/- versus rds+/- or Cngb1-/-, suggesting that eliminating Cngb1 can accelerate the cone degeneration that usually presents later in the rds+/-. This was not the case with rhodopsin; reducing rhodopsin levels in concert with eliminating CNGB1a did not lead to phenotypes more severe than those observed in the Cngb1 knockout alone.
CONCLUSIONS: These data support a role for RDS as the core component of a multiprotein plasma membrane-rim-disc complex that has both a structural role in photoreceptor OS formation and maintenance and a functional role in orienting proteins for optimal signal transduction.

The rod outer segment (OS) contains a series of tightly packed membranous discs circumscribed by a rim region and enclosed by the plasma membrane. Organization of this structure, and in particular the three membrane domains (the plasma, disc, and rim membranes), is critical for proper phototransduction and vision, and mutations in proteins responsible for organizing/forming this structure lead to inherited retinal degeneration. Here we evaluate the potential interplay and relative contributions of three of the proteins critical for the formation and function of the OS, namely, the visual pigment rhodopsin, the ion channel subunit CNGB1a (cyclic nucleotide gated channel B1a), and the structural protein RDS (retinal degeneration slow, also known as peripherin/rds or peripherin-2).

Mutations in rhodopsin cause autosomal dominant retinitis pigmentosa (ADRP) (http://www.retina-international.org/files/sci-news/rhomut.htm [in the public domain]).[1] Rhodopsin is the predominant component of the disc membrane,[2,3] and in addition to its key role in phototransduction, it also has a structural role in forming and maintaining OSs.[4,5] In the absence of rhodopsin (Rho mice), rod OSs are not elaborated, and rod function is not detectable by ERG.[6,7] However, flattened membranous vesicles reminiscent of nascent discs form inside small sacs of OS plasma membrane, suggesting aborted attempts at OS formation.[7,8] Although rhodopsin is present in the OS in high quantities and is precisely arranged in structurally distinct arrays in the disc membrane, reduction in rhodopsin levels by ∼50% is well tolerated; that is, Rho retinas exhibit well-formed OSs that are quite similar structurally and functionally to those found in wild type (WT).[9,10] Rhodopsin-associated disease is typically dominantly inherited, with dominant-negative, rather than loss-of-function/haploinsufficiency mutations, underlying the majority of cases.

While rhodopsin is a principal disc component, RDS is a tetraspanin membrane protein exclusively localized to the rim region of the disc[11,12] and is required for OS disc morphogenesis and stabilization. Mutations in the RDS gene are associated with ADRP as well as multiple classes of macular degeneration (http://www.retina-international.org/files/sci-news/rdsmut.htm [in the public domain]).[13,14] Retinal degeneration slow function relies on the assembly of a variety of types of homo- and hetero-oligomeric complexes containing RDS and its homologue rod outer segment membrane protein-1 (ROM-1). The C-terminus of RDS promotes membrane fusion in vitro[15] and can initiate membrane curvature,[16] suggesting it may play a role in OS formation. Retinal degeneration slow also is known to interact with several other proteins, often through interactions mediated by the C-terminal. One notable binding partner of RDS is the CNGB1a channel subunit, which we further discuss below.[17,18] More recently, using in vitro analyses, a direct binding of RDS to rhodopsin has been reported.[19] In the rds (actually a naturally occurring null allele [Prph2], here referred to as rds to maintain continuity with previous work) OSs fail to form; the cells instead terminate with the connecting cilia.[8,20] The rds retina exhibits negligible phototransduction activity, and subsequently degenerates.[20-22] In contrast to rhodopsin, lack of a single RDS allele (rds) results in severe abnormalities in OS structure and a reduction in rod function[23]; and consistent with this observation, many RDS ADRP mutations cause disease in patients due to haploinsufficiency.[23,24]

The rod CNG channel consists of three alpha subunits (CNGA1) and one longer beta subunit (CNGB1a).[25] It has an essential role in the phototransduction signaling cascade, and many CNGB1 mutations lead to ADRP.[26-33] The N-terminal of CNGB1a contains a unique proline- and glutamic acid-rich N-terminal extension protein domain called GARP[34]; and in addition to the membrane-bound channel (CNGB1a), two additional cytosolic splice variants called GARP1/2 are expressed from the CNGB1 gene. Though the “free,” that is, nonmembrane-bound GARP isoforms are present throughout the OS,[35,36] the beta subunit (CNGB1a) is present exclusively in the rod OS plasma membrane.

Retinal degeneration slow interactions with CNGB1a and GARP have been hypothesized to stabilize the alignment of OS disc rims with the plasma membrane or to bridge adjacent rims, again to maintain proper OS structure.[17,18] This hypothesis is supported by the observation that, though CNGB1a is not a structural protein per se, in the Cngb1 retina, OS disc alignment and sizing are abnormal.[34,37] Additionally, as noted above, direct binding of RDS to rhodopsin has been shown, and mutations in the fourth transmembrane domain of RDS can interrupt this binding and lead to ADRP.[19] Thus, it has been hypothesized that RDS serves as an organizing component of both adjacent membrane domains (i.e., rhodopsin in the disc membrane and CNGB1a in the plasma membrane) to form functional domains, optimized for efficient signal transduction and structural stability.[19] Here, we evaluate the potential interplay between these three proteins by examining retinal disease phenotypes in animal models expressing varying amounts of CNGB1a, rhodopsin, and RDS.

Materials and Methods

Ethics Statement and Animal Care and Use

Animal use was approved by the Institutional Animal Care and Use Committees at the University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma, and the University of Houston and conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The Cngb1 line, the rds line (originally obtained from Neeraj Agarwal currently at the National Eye Institute, Bethesda, MD, USA), and the Rho line (originally obtained from Janice Lem, Tufts University, Boston, MA, USA) have been described previously.[7,22,34] Cngb1 mice on the rds, rds, Rho, and Rho backgrounds were generated in our facility. Wild-type (i.e., Cngb1/ rds/Rho), rds, rds, Rho, and Rho littermates were used from our colony as controls. All animals were maintained in cyclic light (12 hours light, 12 hours dark, ∼30 lux) and fed standard lab chow.

Immunofluorescence Labeling and Protein Chemistry

Western blot and velocity sedimentation were performed as described previously.[38,39] Experiments were repeated three to seven times, and densitometry was done using the Image Lab Software (Bio-Rad, Temecula, CA, USA). Eyes for immunofluorescence were harvested, fixed, sectioned, and immunolabeled as previously described.[38,40] Primary antibodies are described in the Table,[38,39,41-43] and anti-mouse, anti-rabbit, or anti-goat AlexaFluor conjugated secondary antibodies (Life Technologies, Grand Island, NY, USA) were used. Images were captured on an Olympus BX-62 microscope (Center Valley, PA, USA) equipped with a spinning disc confocal unit and analyzed as described previously.[41,44]

Table

Antibodies Used for Immunofluorescence and Western Blotting

Light and Transmission Electron Microscopy

Tissue collection, processing, and plastic embedding and transmission electron microscopy (TEM) have been described previously.[45,46] Light microscopy images were captured at ×40 using a Zeiss (Peabody, MA, USA) universal microscope. To evaluate outer nuclear layer (ONL) and OS layer thickness, images were captured from central superior and inferior regions containing the optic nerve head, and at least three retinal sections from two to four eyes/genotype were used. Layer thickness was measured using Adobe Photoshop CS5 (Adobe Systems, Inc., San Jose, CA, USA).

Electroretinography

Electroretinography (ERG) was performed as previously described.[45,46] Assessment of rod photoreceptor function (scotopic ERG) was performed on dark-adapted animals with a strobe flash stimulus of 157 cd-s/m2 intensity while photopic responses were recorded using a similar stimulus from light-adapted animals. At least five or six mice per genotype were analyzed.

Statistical Analysis

Graphs are presented as mean ± SEM. Differences between genotypes were assessed by 1-way or 2-way ANOVA with Bonferroni's post hoc pairwise comparisons. P < 0.05 was considered significant (P < 0.05, P < 0.01, and P < 0.001).

Results

Levels of Rhodopsin, RDS, and ROM-1 Are Altered by Co-elimination of CNGB1 and RDS/Rhodopsin

Previously we evaluated the interplay between RDS and rhodopsin in OS morphogenesis,[8] and here we wanted to understand the effects of eliminating CNGB1 at the same time as either rhodopsin or RDS. We crossed the Cngb1 line, which lacks all Cngb1 gene products including CNGB1a and GARP1/2,[34] onto the rds, rds, Rho, and Rho backgrounds (which were all used as control lines for the purpose of comparison throughout the study). At postnatal day (P) 30 we assessed levels of rhodopsin, RDS, and ROM-1 by reducing SDS-PAGE/Western blot. Data are presented as percentage of WT after normalization to actin (a loading control) (Figs. 1A–C) and are plotted as means ± SEM (P < 0.05, P < 0.01, and P < 0.001 by 1-way ANOVA with Bonferroni's post hoc comparison). Though rhodopsin levels were virtually normal in the Cngb1 and the rds (Fig. 1A), they were reduced in Cngb1/rds to ∼56% of levels in the rds. Rhodopsin levels in the Cngb1/Rho were ∼64% those of the Rho. Levels of rhodopsin in the rds were reduced by 84% and were not further reduced by ablating Cngb1.

Figure 1

Levels of OS proteins are altered when Cngb1 is absent. Retinal extracts were isolated from P30 eyes of the indicated genotypes and were analyzed by reducing SDS-PAGE/Western blot. The blots were probed with (A) mAB 1D4 against rhodopsin, (B) anti-RDS-CT, and (C) anti-ROM-1-CT. Blots were also labeled with actin as a loading control. Protein was quantified densitometrically, normalized to actin, and plotted as a percentage of WT (shown are means ± SEM, *P < 0.05, **P < 0.01, and ***P < 0.001 by 1-way ANOVA with Bonferroni's post hoc comparison). Measurements were performed in 3 to 7 retinas per genotype. Nonreducing sucrose gradient velocity sedimentation (D, E) was performed on P30 retinal extracts from WT and Cngb1. Fractions from gradients (1–12) were further separated by SDS-PAGE/reducing Western blot. Resulting blots were probed with anti-RDS-CT (D) and anti-ROM-1-CT (E). The relative percent of total RDS or ROM-1 in each fraction was assessed densitometrically and plotted as mean ± SEM from 3 or 4 independent experiments.

This more pronounced effect of Cngb1 ablation in the rds compared to the Rho continued when we examined RDS levels (Fig. 1B). In Cngb1−/−/rds retinas, RDS levels were reduced to ∼54% of those in the rds while RDS levels in Cngb1/Rho remained at ∼87% of those in the Rho. Rod outer segment membrane protein-1 levels were not altered in either the Cngb1/rds versus the rds or the Cngb1/Rho versus the Rho (Fig. 1C). We also observed that RDS and ROM-1 are severely reduced by the coablation of Cngb1 and rhodopsin (Cngb1/Rho had RDS/ROM-1 levels that were 8.8% and 7.9% those of Rho, respectively).

Cyclic nucleotide gated channel B1a is known to bind to RDS in the OS,[17,18] and proper binding and complex assembly of RDS/ROM-1 are necessary for OS formation,[38] so we asked whether eliminating Cngb1 altered RDS/ROM-1 oligomerization. Retinal extracts from WT and Cngb1 animals were fractionated on 5% to 20% nonreducing sucrose gradients and then separated on reducing SDS-PAGE. In the WT retina, RDS is present as tetramers (fractions 6–8), octamers (fractions 4–5), and higher-order oligomers (fractions 1–3) while ROM-1 is detected only in tetrameric and octameric fractions. We did not observe significant alteration in RDS (Fig. 1D) or ROM-1 (Fig. 1E) complex formation in Cngb1 retinas, suggesting that RDS/ROM-1 oligomerization is not tied to CNGB1.

Previously we observed that CNGB1a and GARP2 protein levels are reduced by ∼50% in the rds and are undetectable in the rds.[47] When combined with our data here showing that RDS levels are decreased in the Cngb1/rds compared to the rds, these data suggest that stability of CNGB1a and RDS may be interrelated.

Trafficking of OS Protein Is Largely Unaffected by Removing Cngb1

Cyclic nucleotide gated channel B1a and RDS binding occurs during the OS targeting process prior to arrival in the OS,[18] so we next used immunofluorescence labeling to ask whether the targeting of RDS, ROM-1, rhodopsin, or CNGB1a/GARP1/2 is affected by co-elimination of Cngb1, rhodopsin, or Rds. Retinal degeneration slow and ROM-1 labeling was properly restricted to the OS layer in the WT, Cngb1, rds, Cngb1/Rho, Rho, and Cngb1/Rho retinas (Fig. 2A, green). Similarly, rhodopsin localization is not affected in most of these genotypes; however, some accumulation of rhodopsin in the ONL is observed in the Cngb1/Rho (Fig. 2A, red, arrow). In the rds, large amounts of rhodopsin accumulate in the inner segment and ONL,[8] and this is recapitulated in the Cngb1/rds (Fig. 2B, red). In contrast, in the Rho and Cngb1/Rho retinas, the majority of RDS is not found in the inner segment, but is properly restricted to the distal tips of the photoreceptors, consistent with our previous observation showing that RDS and ROM-1 are found in the small nascent OSs of the Rho.[8] However, in the Rho and Cngb1/Rho retinas, some mild mislocalization of RDS (but not ROM-1) to the inner segment, ONL, and outer plexiform layer is seen (arrows, Fig. 2C). Combined, these data suggest that while rhodopsin is more susceptible to mislocalization than RDS, CNGB1 ablation has little to no effect on the OS targeting of RDS, ROM-1, or rhodopsin.

Figure 2

Eliminating Cngb1 does not affect targeting of RDS or rhodopsin. Paraffin-embedded retinal sections from P30 animals of the indicated genotypes were labeled (A) with antibodies against RDS (RDS-CT, green), rhodopsin (mAB-1D4, red), or ROM-1 (ROM1-CT, green); (B) mAB 1D4 (red) or ROM1-CT (green); (C) RDS-CT or ROM1-CT (green), or (D) Cngb1 gene products (GARP-4B1, red). In all cases nuclei were counterstained with DAPI (4′,6-Diamidino-2-Phenylindole; blue). Arrows indicate accumulation of protein outside the OS. Scale bar: 20 μm. OS, outer segment; ONL, outer nuclear layer; IS, inner segment; OPL, outer plexiform layer.

Conversely, to see whether reduction/elimination of RDS or rhodopsin affected CNGB1a/GARP1/2 targeting, we labeled with GARP-4B1, which recognizes all Cngb1 gene products (Fig. 2D, red). In WT, rds, and Rho retinas, CNGB1a/GARP1/2 properly localized to the OS layer. This observation was replicated in the Rho retina; CNGB1a/GARP1/2 properly localized to the OS layer, even though the Rho does not form fully elaborated OSs. In striking contrast, and consistent with our Western blot data,[47] in the absence of RDS (rds), CNGB1a and GARP1/2 are virtually undetectable in the retina. This decrease in Cngb1 gene products in the absence of RDS is clearly not due to a requirement for proper OS formation, since the OSs in the Rho retina, while more formed than in the rds, are still just tiny sacs of vesicle-containing membrane at the tip of the OS,[8] yet the Rho model still expresses CNGB1A/GARP1/2. Rather it suggests that the stability or transport of CNGB1a and/or GARP 1/2 are dependent directly on RDS or on the initiation/formation of OSs by RDS (which can be seen in nascent form even in the Rho).

Elimination of CNGB1 Exacerbates OS Shortening When RDS Is Also Reduced

We next conducted histologic and morphometric analysis at P30 (Fig. 3A). As expected at early time points, ONL thickness is not changed in rds or Rho versus WT (Figs. 3B, 3C). However, though the differences do not attain statistical significance, mean ONL thickness in the Cngb1, Cngb1/rds, and Cngb1/Rho was decreased by 10% to 15% compared to their respective controls (WT, rds, and Rho), suggesting that degeneration in the Cngb134,48 starts earlier than that in the rds and Rho. There is no difference in ONL thickness in Rho versus Cngb1/Rho or rds versus Cngb1/rds, suggesting that the photoreceptor degeneration in the rds and Rho is not additive with that in Cngb1.

Figure 3

Outer segment length is reduced in the absence of Cngb1. Representative light micrographs (LM) were captured from plastic-embedded retinal sections harvested at P30 from the indicated genotypes. Scale bar for LM images is 20 μm. RPE, retinal pigment epithelium; OS, outer segment; IS, inner segment; ONL, outer nuclear layer. (B, C) Outer nuclear layer and OS layer thicknesses were measured in sections captured from the central superior and central inferior retina from 2 to 4 eyes per genotype. Plotted are means ± SEM, *P < 0.05, **P < 0.01, and ***P < 0.001 by 2-way ANOVA with Bonferroni's post hoc comparison.

It was previously shown that OSs are shorter and more disorganized in the Cngb1 and rds retinas compared to WT,[23,34] and our data confirm this (Fig. 3C). Interestingly, this effect was additive: In the Cngb1/rds, mean OS length was 41% of WT, compared to 52% and 73% of WT for rds and Cngb1, respectively (Fig. 3C). As expected, at P30, OSs in the Rho are similar to those in WT, but we do see OS shortening in the Cngb1/Rho due to the absence of CNGB1a; OSs in the Rho are 89% of WT, while those in the Cngb1/Rho are 59% of WT. Again, the effects of eliminating CNGB1a from the rds retina are more severe than in the Rho retina; for example, OSs in the Cngb1/rds are decreased to 56% of those in Cngb1 (P < 0.05), while the OSs of the Cngb1/Rho are 81% of Cngb1 (not significant).

Eliminating Cngb1 Severely Affects Rod Function

We performed full-field scotopic ERG at P30 to evaluate rod function. Figure 4A shows representative waveforms, and maximum saturating a- and b-wave amplitudes are plotted in Figure 4B (mean ± SEM, P < 0.001 by 1-way ANOVA with Bonferroni's post hoc comparison). Our results for control animals (Cngb1, rds, and rho) are consistent with previous publications[8,34] (Fig. 4B). Again, elimination of Cngb1 had a much more severe effect in the rds background than it did in the Rho background. Scotopic a-wave values in the Cngb1/Rho were not significantly reduced compared to the Cngb1, suggesting that all of the functional decrease in the Cngb1/Rho is due to the absence of Cngb1. In contrast, in the Cngb1/rds, scotopic ERG responses were severely reduced compared to both Cngb1 and rds (Fig. 4B). This effect was seen in both the scotopic a- and b-waves, and scotopic a-waves in the Cngb1/rds were virtually undetectable. Scotopic ERG function is undetectable in the rds and Rho retina at P30, so we did not perform ERGs on animals in those backgrounds. The additive ERG defect in the Cngb1/rds could be due to reductions in CNGA1; however, levels of CNGA1 in Cngb1 are virtually undetectable,[34,49] and this finding is recapitulated in Cngb1/rds (Supplementary Fig. S1). Thus while it is possible that further reduction in CNGA1 levels contributes to the additive defects seen in the Cngb1/rds, we cannot conclusively assess the possibility given that the levels of CNGA1 are below the limit of detection.

Figure 4

Absence of Cngb1 significantly affects rod structure and function. Full-field scotopic ERG responses were obtained at P30. (A) Representative waveforms; (B) plots of mean amplitude ± SEM from 4 to 7 mice per genotype (*P < 0.05, **P < 0.01, and ***P < 0.001 by 1-way ANOVA with Bonferroni's post hoc comparison). (C) Representative TEM images of the photoreceptor-RPE interface at P30. Scale bars: 10 μm for low-magnification EM images and 500 nm for higher-magnification EM images. RPE, retinal pigment epithelium; OS, outer segment; CC, connecting cilium.

We conducted TEM to determine whether the severe effects of simultaneously eliminating Cngb1 and reducing rhodopsin/Rds led to defects in OS ultrastructure (Fig. 4C with additional examples in Supplementary Fig. S2). The OS shortening we quantified on the light microscopy level in the Cngb1 is also evident by low-magnification EM (×3000, top, Fig. 4C). By high-magnification EM (×20,000, bottom, Fig. 4C), we find that the disc diameter and often the disc alignment are abnormal in the Cngb1−/− retina. In some cases, new discs grew abnormally parallel to the plasma membrane (arrow, Fig. 4C, with additional examples in Supplementary Fig. S2), as previously reported.[34] Outer segments in the rds were characterized by abnormal disc size and the formation of highly dysmorphic whorl structures, but interestingly, this phenotype was not worsened in the Cngb1/rds retina. Outer segment structure in the Cngb1/Rho was worse than in the Rho (which is largely normal). The phenotype in the Cngb1/Rho was of the same type seen in the Cngb1, namely, abnormalities in disc size and alignment but not whorl formation; however, the defects appeared to be more severe in the Cngb1/Rho compared to the Cngb1. The rds retina forms no OS at the tip of the connecting cilia (Fig. 5, CC marks connecting cilia), and the Cngb1/rds is not different from the rds. In contrast, the Rho forms well-characterized[7,8] tiny nascent OSs (arrows, Fig. 5) distal to the connecting cilia with enclosed small flattened vesicular structures. Similar structures at the distal tip of the connecting cilium were seen in the Cngb1/Rho (arrows, Fig. 5).

Figure 5

Outer segment ultrastructure in the rds and Rho is not affected by the ablation of CNGB1. Shown are representative TEM images captured from the indicated genotypes at P30. CC, connecting cilium. Arrows indicate nascent OS structures formed in retinas of animals on the Rho background. Scale bars: 10 μm for low-magnification EM images (top) and 500 nm for higher-magnification EM images (bottom).

Elimination of Cngb1 Severely Affects Cone Function in the rds Retina

Though Cngb1 is expressed only in rods, secondary cone loss can occur after rod defects. Therefore, we evaluated cone function in our models. Figure 6A shows representative photopic ERG waveforms while Figure 6B plots mean photopic b-wave amplitudes (mean ± SEM, P < 0.01 by 1-way ANOVA with Bonferroni's post hoc comparison). Consistent with previous reports, at this time point significant decreases in cone function are not detected in the rds or Cngb1.[23,48] Likewise, no significant reductions in cone function were noticed in Cngb1/Rho+/− versus Rho or WT. However, a striking ∼75% reduction in cone function was observed in Cngb1/rds compared to WT, rds, and Cngb1. To identify whether this defect was associated with loss of cone cells, M-opsin-positive cones were counted in a region in the superior central retina while S-opsin-positive cones were counted in the inferior central portion of the retina (representative images in Fig. 6C; Fig. 6D shows mean ± SEM). Consistent with the dramatic decrease in photopic ERG response in the Cngb1/rds compared to the Cngb1 or the rds, we observed a 50% reduction in S-cones in the Cngb1/rds compared to Cngb1, rds, or WT controls (P < 0.001 by 1-way ANOVA).

Figure 6

Cone function is significantly reduced in the Cngb1/rds retina. Photopic ERGs were recorded from WT, Cngb1, rds, Cngb1/rds, rho, and Cngb1/Rho at P30. (A) Representative waveforms; (B) plots of the mean maximum photopic b-wave amplitude. n = 4 to 7 mice per genotype (*P < 0.05, **P < 0.01, and ***P < 0.001 by 1-way ANOVA with Bonferroni's post hoc comparison). (C) Paraffin-embedded retinal sections from P30 animals were labeled with anti S-opsin (green, top) and anti-M-opsin (green, bottom) and were all counterstained with DAPI (blue). Scale bar: 20 μm. OS, outer segment; ONL, outer nuclear layer; OPL, outer plexiform layer. (D) Cells positive for cone opsins were counted in a 300-μm region of the central retina either inferior (S-opsin-positive cells) or superior (M-opsin-positive cells) of the optic nerve. Plotted are means ± SEM.

Discussion

In this study, our goal was to understand how RDS (a rim protein), rhodopsin (primarily a disc protein), and CNGB1a (a plasma membrane protein) are interrelated in their ability to support OS structure and function. Recent work showed that RDS can bind rhodopsin and suggested that a complex composed of CNGB1a, RDS, and rhodopsin may play a structural role in anchoring the rim to the plasma membrane (via RDS-CNGB1a interactions), and a functional role ensuring that proteins are properly localized for optimal phototransduction.[19] When combined with other work showing that CNGB1a/GARP interactions with phosphodiesterase 6 may provide a role for GARP in cGMP turnover,[17,50-52] these observations suggest that RDS may organize the rim region into a hotspot area or environment for highly regulated phototransduction. This hypothesis is consistent with our results here showing that CNGB1a and RDS can act additively, exacerbating RDS haploinsufficiency, and resulting in (1) severely decreased rod function in Cngb1/rds versus Cngb1 or rds and (2) modest additional reductions in the thickness of the OS layer and the ONL in the Cngb1/rds versus the rds and Cngb1. Critically, however, these additive effects occur without any difference in rod ultrastructure in Cngb1/rds versus rds. This suggests that defects in rod function and overall retinal structure cannot be completely attributed to a role for CNGB1/RDS interactions in maintaining rim structure, but rather that these interactions also organize a functional environment critical for proper signal transduction. This role for RDS as an interactor/organizer of many proteins (i.e., not just ROM-1) is consistent with the observation that RDS also interacts with rhodopsin[19] as well as the known broader role of tetraspanins, which often act as organizers of membrane domains containing a wide variety of different proteins.[53]

An interesting outcome from our work is that cone ERG function is significantly reduced in the Cngb1/rds compared to controls. Cngb1 is expressed only in rods, and cone function in Cngb1 animals is normal at P3048 and starts decreasing only later, likely a secondary phenomenon due to rod loss. Similarly, while RDS is expressed in rods and cones, cones can better tolerate reduced RDS levels than rods, so cone function in the rds is also preserved until approximately 4 months of age.[23] Yet here we see cone degeneration and a dramatic reduction in cone function in Cngb1/rds as early as P30. This likely represents a rapid acceleration of RDS haploinsufficiency in cones; however, why elimination of CNGB1 in rods should accelerate RDS-associated cone degeneration is not clear. This cone phenotype is not completely due to overt rod loss since ONL thickness is only mildly affected in the Cngb1/rds versus the rds. There may be several underlying causes, including overall alterations in retinal homeostasis, altered secretion of signaling factors from rods needed for cone health, or alterations in other cell types such as horizontal cells and retinal glia that interact with both rods and cones. Further exploration of the mechanisms underlying this cone defect are of great interest given the proliferation of retinal disease wherein secondary cone loss following rod-specific defects leads to worsened patient visual outcomes.

Outer segment trafficking has recently been a field of particular interest. The trafficking pathways for rhodopsin have been extensively studied,[54] and RDS and rhodopsin are thought to traffic separately.[55] Recent work has shown that in contrast to rhodopsin, the majority of RDS does not traffic through the trans-Golgi, but uses an unconventional secretory pathway during OS targeting.[56,57] Little is known about the regulation of CNGB1a/GARP1/2 trafficking; however, recent work from Goldberg's group[18] shows that RDS/CNGB1a interactions first occur in the inner segment, raising the intriguing possibility that CNGB1a may traffic with RDS. However, whether the CNGB1a would traffic with conventionally or unconventionally processed RDS remains to be further explored. Here we find that in general, trafficking of RDS, ROM-1, and rhodopsin was not affected by Cngb1 ablation. However, the reverse is more difficult to assess. The presence of a clear region of CNGB1a/GARP1/2 immunoreactivity in the distal inner segment region, likely inside the small abnormally formed OS nubs, suggests that overall, CNGB1a/GARP1/2 targeting is not affected by elimination of rhodopsin; however, we cannot assess the role of RDS in CNGB1a/GARP1/2 trafficking because CNGB1a/GARP1/2 are not detectable in rds mice.[47] Thus, it is not possible to differentiate whether CNGB1a/GARP1/2 trafficking to the OS requires RDS, whether CNGB1a protein stability requires RDS, or whether CNGB1a trafficking/stability require the RDS-mediated assembly of OSs.

Interestingly, in some cases eliminating the CNGB1 channel can be beneficial in preventing or retarding retinal degeneration. For example, the removal of Cngb1 significantly rescued the rapid degenerative phenotypes in the rd1 mouse.[58] Degeneration in the rd1 was hypothesized to be tied to observed elevated cGMP levels.[59] However, the removal of CNG channels led to rescue of rod photoreceptors without ameliorating cGMP levels,[58] suggesting that removing cGMP-mediated signaling was beneficial rather than removing cGMP per se. As a parallel, eliminating RDS in the rd1 retina (rds/Pde6b) also delayed retinal degeneration regardless of the massive accumulation of cGMP.[60] Our observation that CNGB1a levels are drastically reduced in the rds suggests that attenuation of rd1-associated retinal degeneration in the rd1 rds double knockout is likely due to an elimination of the CNG channel/CNGB1a rather than to the loss of RDS per se.

In conclusion, we here present data showing that elimination of CNGB1 exacerbates RDS-associated functional but not structural haploinsufficiency. These observations suggest that RDS/CNGB1 interactions have a role in OS function in addition to structure and are consistent with the hypothesis that RDS functions as a component of a multiprotein plasma membrane-rim-disc complex critical for OSs.

Click here for additional data file.