Loss of PRCD alters number and packaging density of rhodopsin in rod photoreceptor disc membranes

Progressive rod-cone degeneration (PRCD) is a small protein localized to photoreceptor outer segment (OS) disc membranes. Several mutations in PRCD are linked to retinitis pigmentosa (RP) in canines and humans, and while recent studies have established that PRCD is required for high fidelity disc morphogenesis, its precise role in this process remains a mystery. To better understand the part which PRCD plays in disease progression as well as its contribution to photoreceptor OS disc morphogenesis, we generated a Prcd-KO animal model using CRISPR/Cas9. Loss of PRCD from the retina results in reduced visual function accompanied by slow rod photoreceptor degeneration. We observed a significant decrease in rhodopsin levels in Prcd-KO retina prior to photoreceptor degeneration. Furthermore, ultrastructural analysis demonstrates that rod photoreceptors lacking PRCD display disoriented and dysmorphic OS disc membranes. Strikingly, atomic force microscopy reveals that many disc membranes in Prcd-KO rod photoreceptor neurons are irregular, containing fewer rhodopsin molecules and decreased rhodopsin packing density compared to wild-type discs. This study strongly suggests an important role for PRCD in regulation of rhodopsin incorporation and packaging density into disc membranes, a process which, when dysregulated, likely gives rise to the visual defects observed in patients with PRCD-associated RP.


Scientific Reports
| (2020) 10:17885 | https://doi.org/10.1038/s41598-020-74628-2 www.nature.com/scientificreports/ to ensure the density of rhodopsin in ROS disc membranes is kept constant, suggesting that regulation of proper rhodopsin density is crucial for photoreceptor disc formation and health 32,33 . Although PRCD's specific role within the photoreceptor remains elusive, recent studies in a PRCD knockout mouse model have identified a requirement for PRCD in high fidelity disc morphogenesis. In this model, it is observed that discs do not flatten or form properly in the absence of PRCD. Consequently, membrane bulges containing rhodopsin appear to separate into extracellular vesicles and accrue within the interphotoreceptor space 34 . A second PRCD knockout mouse model has also shown accumulation of these vesicles and observed a reduced rate of phagocytosis of OS discs by the retinal pigment epithelium (RPE) 35 . Along with evidence of retinal degeneration in these models and structural defects exacerbated by microglia recruitment to clear these vesicles, it is clear that PRCD is required for proper OS maintenance and disc morphogenesis; however further studies are needed to clarify the precise mechanistic role which PRCD plays in these processes.
To further understand the interaction between PRCD and rhodopsin, as well as PRCD's role in photoreceptor OS disc morphogenesis and maintenance, we utilized our Prcd-KO mouse model. In this study, we demonstrate that Prcd-KO retina display dramatic disorganization of the photoreceptor OS. Most importantly, using AFM imaging, we observe that loss of PRCD leads to decreased rhodopsin number and packaging density in ROS disc membranes, likely associated with the observed morphological changes in Prcd-KO retina.

Results
Generation of Prcd-KO mice using CRISPR/Cas9 genome editing. To better understand the importance of PRCD in the maintenance and stability of the photoreceptor outer segment (OS), we generated a global Prcd-KO mouse model using CRISPR/Cas9 genome editing. Among 48 founder mice, we found 12 contained INDEL deletions in exon 1. For this study, we collected data from the 1738 founder, whose Prcd gene contains a single nucleotide deletion in exon 1 at cytosine 61 (c61x), resulting in a frameshift mutation ( Fig. 1A and Supplementary Fig. S1). To rule out off-target effects, additional founder lines were tested, including mice with a 14 bp deletion (1734) and 71 bp deletion in Prcd (1748) (Supplementary Fig. S1). All lines exhibited similar phenotypes, were comparable in weight and development to wild-type (WT) littermates and are fertile with no detectable health issues. Western blot analysis demonstrated that lysate from Prcd-KO retina does not contain PRCD protein (Fig. 1B). Co-immunofluorescent labeling of retinal cross sections with antibodies against PRCD and the OS marker cyclic nucleotide gated channel A (CNGA) revealed expected PRCD OS localization in WT retina, whereas PRCD labeling was absent in Prcd-KO retinal sections (Fig. 1C). Altogether, these data validate the absence of PRCD protein in the Prcd-KO animal model used in this study.
Loss of PRCD leads to progressive loss of visual function and slow photoreceptor degeneration. To evaluate the visual response of rod and cone photoreceptor neurons, we assessed scotopic (rod) and photopic (cone) responses in Prcd-KO and WT littermates by electroretinography (ERG). A light-evoked ERG response is comprised of an "a-wave" (photoreceptor) and a "b-wave" (inner retinal neurons), which result (B) Immunoblot analysis shows loss of PRCD protein from Prcd-KO retinal lysate at postnatal (P) 30 and P100, whereas wild-type (WT) retinal lysates demonstrate PRCD immunoreactivity (n = 4). Please note that immunoblot data is cropped; full-length, raw data is available in Supplementary Fig. S2. (C) Immunofluorescent staining of P30 retinal cross-sections from WT and Prcd-KO mice, probing with antibodies against PRCD (red) and the OS marker cyclic nucleotide gated channel alpha-1 and alpha-3 (CNGA 1/3; green) to demonstrate proper localization of PRCD to the OS in WT retina and loss of PRCD from Prcd-KO retina (n = 3). Scale bar = 20 μm. All experiments were conducted with littermate controls. www.nature.com/scientificreports/ from hyperpolarization of photoreceptors and depolarization of downstream retinal neurons, respectively 36 . At postnatal day (P) 30, Prcd-KO mice present with an approximate 30% reduction in rod response compared to WT littermate controls in bright light conditions ( Fig. 2A). At low light intensities (0.00025 and 0.001 cd*s/ m 2 ), we were unable to identify any discernable differences in scotopic rod response between WT and Prcd-KO animals. However, at brighter flash intensities 0.158 (p = 0.009424), 0.995 (p = 0.001771), and 2.5 cd*s/m 2 (p = 0.001363), rod a-waves were significantly reduced in Prcd-KO animals compared to WT littermate controls (Fig. 2B). Furthermore, we observed a slow decrease in visual response over time at various ages (P30, P60, P100, and P200); by P200, Prcd-KO animals demonstrate an approximate 50% reduction in rod photoresponse compared to WT littermate controls (Fig. 2C). In contrast, Prcd-KO photopic ERGs demonstrated no noticeable changes compared to WT responses at any tested ages or flash intensities (Fig. 2D). The evident impairment and gradual reduction in rod visual function of Prcd-KO animals suggests a possible defect in the morphology or survival of photoreceptor cells. We utilized hematoxylin and eosin (H&E) staining of WT and Prcd-KO retina at P30, P120, and P250 in order to analyze retinal morphology. No significant defects in retinal lamination or morphology were observed in P30 Prcd-KO mice (Fig. 2E). To determine if retinal degeneration was evident at P30, where we observed significant loss of rod function in bright light conditions, we counted photoreceptor nuclei at 10 different locations in the same H&E sections. No significant reduction in photoreceptor nuclei was apparent in retina lacking PRCD until after P120. At P250, we observed a significant reduction in photoreceptor nuclei, with approximately 23.8% fewer nuclei in Prcd-KO retina compared to WT (p = 0.001332; Fig. 2F). Based on these data, the early rod ERG defects observed in bright light conditions were likely not due to loss of rod photoreceptors.

Prcd-KO retina exhibit reduced levels of rhodopsin and increased structural disorganization of rod OS as disease progresses.
As Prcd-KO retina do not undergo degeneration until after P120, it is possible that the observed defect in visual function is due to changes in major photoreceptor or disc-resident proteins, as PRCD is localized exclusively to the OS disc membrane. Therefore, we analyzed retinal lysate from WT and Prcd-KO mice at P30 and P100 by western blot and found no significant changes in any tested proteins ( Supplementary Fig. S3). However, due to the reported interaction between rhodopsin and PRCD, we measured the concentration of rhodopsin in WT and Prcd-KO retina with spectrophotometric analysis, a technique which has been widely used to measure and calculate picomolar concentrations of rhodopsin in the retina 37,38 . At P30, total rhodopsin concentration was similar between Prcd-KO and WT retina (Fig. 3A). However, by P120, we observed a 20.4% decrease (p = 0.000072) in the amount of rhodopsin per Prcd-KO retina compared to age-matched WT controls (Fig. 3B). Despite the observed decrease in rhodopsin protein at P120, rhodopsin mRNA levels were comparable between WT and Prcd-KO animals at P30 and P120 ( Supplementary Fig. S4). Furthermore, no mislocalization of rhodopsin is evident at P120 in Prcd-KO retinal sections when analyzed by immunofluorescent staining (Supplementary Fig. S5).
To gain further insight into the observed reduction in rod photoreceptor function at P30, we investigated photoreceptor ultrastructure at various ages (P30, P60, and P120) using transmission electron microscopy (TEM). At the earliest tested age (P30), many Prcd-KO rod photoreceptors contained discs that were disoriented and vertically aligned, although rod OS had an overall normal shape ( Fig. 3D; arrowhead). Ultrastructural analysis of Prcd-KO rods at P60 showed the OS layer to be disorganized compared to WT. Outer segments had irregular diameters, included frequent examples of overgrown and misoriented disc membranes, and occasionally shed OS fragments and whorls of membrane packets into the OS layer. Closer inspection revealed a distinct population of Prcd-KO rods containing small packets of overgrown discs (10-40 discs), which were aligned abnormally, having grown along the long axis of the photoreceptors ( Fig. 3E and Supplementary Fig. S6B-E). Although these defects were observed in Prcd-KO rods, there were still many rods which were comparable to WT, with no apparent differences in gross OS ultrastructure and contained nicely stacked, correctly sized, and properly aligned disc membranes ( Supplementary Fig. S6A,B,F). Additionally, our evaluation of Prcd-KO retina ultrastructure found no apparent difference in the connecting cilium, inner segment (IS), mitochondria, or adjacent RPE layer compared to WT littermates. With increased age, Prcd-KO mice showed an increased frequency of defective photoreceptor cells with incorrectly stacked discs, further disorganization of ROS, and more space between discs at P120 compared to age-matched WT (Fig. 3C,F).

Disc membranes in Prcd-KO rod photoreceptors contain reduced packaging density and number of rhodopsin molecules. To investigate the interplay between rhodopsin and the observed defects
in the Prcd-KO ROS ultrastructure, we evaluated ROS disc membranes isolated from Prcd-KO mice and age matched WT controls by atomic force microscopy (AFM). Previous studies have successfully used AFM to visualize rhodopsin nanodomains in the disc membranes 39 . Rhodopsin nanodomains are formed by oligomers of rhodopsin of various size and are not an artifact of phase separation of lipids or those related to the AFM procedure 28,30,40 . To further our confidence in this detection method, ROS disc membranes from WT mice at P30 and P120 exhibited properties similar to those reported previously 30,32 (Fig. 4A-C). In these discs, rhodopsin was arranged into nanodomains and the nanodomains were largely densely packed within the disc ( Fig. 4A-C). Prcd-KO mice at P30 and P120 exhibited some ROS disc membranes which resembled those from WT mice with rhodopsin nanodomains densely packed within the membrane (Fig. 4D,G); however, they also exhibited ROS disc membranes that were irregular. Irregular ROS disc membranes from Prcd-KO discs had fewer rhodopsin nanodomains with larger areas devoid of rhodopsin nanodomains (Fig. 4E,F,H, I; asterisks). Irregular ROS disc membranes were also observed for WT mice; however, they were only sporadically observed, representing 6% or less of the ROS disc membranes analyzed. The proportion of regular and irregular ROS disc membranes were quantified for P30 and P120 Prcd-KO mice. In P30 KO mice, about one-third of the ROS disc membranes were Scientific Reports | (2020) 10:17885 | https://doi.org/10.1038/s41598-020-74628-2 www.nature.com/scientificreports/  www.nature.com/scientificreports/ irregular, and this increased for P120 mice, where about half of the ROS disc membranes were irregular (Fig. 5A, Table 1). Thus, the proportion of irregular ROS disc membranes increased with age for Prcd-KO mice. Since ROS discs from photoreceptor cells are heterogeneous, quantitative analysis is required 32 . The number of rhodopsin contained within a ROS disc membrane was estimated from the size of nanodomains and the density of rhodopsin was computed presuming single rhodopsin molecules homogenously distribute throughout the area of the lamellar region of the discs 39 . The number and density of rhodopsin nanodomains and molecules of rhodopsin in the irregular ROS disc membranes of both P30 and P120 mice were lower compared to that of regular ROS disc membranes (Fig. 5B-E). The regular ROS disc membranes of both P30 and P120 Prcd-KO mice contained a similar number and density of rhodopsin as that of WT ROS disc membranes (Table 2). Altogether, these data demonstrate that Prcd-KO mice have an impaired ability to form proper ROS disc membranes with sufficient rhodopsin number and packaging density.

Discussion
In the current study, we sought to further understand the role of PRCD, a small disc-resident protein, in photoreceptor disc morphogenesis. Mutations in PRCD are associated with dramatic misalignment of disc membranes and disorganization of the OS, followed by slow retinal degeneration. The most common PRCD mutation (C2Y) has been extensively characterized in several dog breeds 16 . Recent studies in additional mouse models have shown that high fidelity disc morphogenesis is compromised when PRCD is not present 34,35 . However, the precise role PRCD plays in this process remains unclear. In this study, we demonstrate that loss of PRCD disrupts rhodopsin packaging into ROS disc membranes, leading to defects in visual function and compromised OS structure, phenotypes consistent with clinical findings in humans and dogs with a PRCD-C2Y mutation 10 .
Previous studies in canines demonstrate that PRCD-associated disease is heterogeneous, with spatial and temporal differences in disease onset and progression between dog breeds 14,15,41 In these models, it is reported that changes in ERG response correspond to the presence of morphological defects in affected animals 16 . In our Prcd-KO mouse model, retina appear to develop normally, exhibiting proper lamination. Our ERG studies demonstrate that loss of PRCD impairs mass rod photoreceptor response in bright light conditions as early as P30. At P30, ultrastructural analysis reveals Prcd-KO rod photoreceptors with defects in OS morphology, including focal misalignment of discs and imperfect disc stacking. As mice age, ERG response continues to decline and more pronounced morphological defects are observed. Importantly, all tested disc-resident and phototransduction proteins appear normal in Prcd-KO retina at P30, suggesting the observed reduction in Prcd-KO rod response at this age is due to morphological defects rather than loss of crucial phototransduction proteins 34 .
Similar to two recently published mouse models lacking PRCD, photoreceptor degeneration in our Prcd-KO mice occurs well after disease onset, as we only see significant degeneration after P120. However, it seems that disease onset occurs earlier in our mouse model compared to others. In one model, Spencer et al. report that single cell recording of rods shows no differences in visual response compared to WT at P40-P45. Additionally, Prcd −/− rods appear relatively normal compared to WT at 2 months, with only a few deformed ROS 34 . In another model, Allon et al. report that mass rod response was not reduced until after 20 weeks of age and do not mention disorganized disc membranes until 30 weeks 35 . We believe these discrepancies can be explained by mouse strain differences and genetic heterogeneity, as our Prcd-KO model is in a 129/SV-E background and the other two are in a C57BL/6J background. If the characterized correlation between ERG response and morphology in canines with a PRCD-C2Y mutation holds true in Prcd-KO animals, it explains why we see reductions in ERG response at P30 when we also observe morphological defects. Additionally, it may explain why ERG response is not affected in these other knockout models, when ROS appear normal.
These models also show that retina lacking PRCD develop an abundance of extracellular vesicles (EV) containing rhodopsin, which phagocytic microglia cannot manage to clear properly 34,35 . Spencer et al. demonstrates that discs lacking PRCD do not form properly and that prior to flattening, membrane bulges containing rhodopsin separate into EV in the interphotoreceptor space 34 . We were able to identify several examples of EV in our model, but they often appear in the same areas as fixation artifacts and were also observed in WT retina. These newly formed EV are likely labile structures which require cardiac perfusion to retain their ultrastructure. It is important to note that the presence of vesicular profiles is also a hallmark of PRCD-associated disease in miniature poodles 14 . However, these vesicular profiles were reported to be missing or sparsely distributed in retina of English cocker spaniels with PRCD-C2Y mutations 16 . Therefore, while the traditional TEM fixation protocols used in this study may not have captured these vesicles properly, it is also possible that formation of these vesicles is just another phenotypic variation that can be attributed to genetic differences in mouse strain, as demonstrated in canine models. We plan to investigate this possibility in our future studies. In general, the variable phenotypes observed in canines and mice with PRCD-associated disease underscores the importance of continued study of this protein in multiple model systems.
In contrast to other published studies, we furthered our analysis of Prcd-KO disc membranes using AFM. PRCD is identified as one of 11 proteins exclusively present in the disc membrane and has been shown to interact with rhodopsin 7,18 . As mentioned, rhodopsin is organized into nanoscale domains and AFM has been extensively used to visualize these nanodomains in various studies 28,30,32,42,43 . Rhodopsin protein is reduced by half in Rho+/− retina and AFM analysis of Rho+/− discs demonstrates that although disc size decreases in response to availability of rhodopsin, at 6 weeks of age, Rho+/− disc membranes contain a density of rhodopsin comparable to WT 22,32 . With an average density of about 20,000 μm 2 in murine discs, rhodopsin is likely maintained at this density in order for rods to retain high sensitivity to light 23,30,33,39,44,45 . In fact, high rhodopsin packaging density www.nature.com/scientificreports/ appears very important for proper visual function, as mice housed in constant dark display increased rhodopsin packing density within ROS disc membranes and show improved visual function compared to mice housed in cyclic light conditions 33 . In contrast to tight regulation of rhodopsin packaging density, our Prcd-KO mice demonstrate a population (~ 33%) of "irregular" discs at P30 which contain both fewer molecules of rhodopsin and decreased density of rhodopsin in comparison to WT "regular" discs. By P120, the relative proportion of these irregular discs increases to ~ 50%, corresponding with worsening retinal defects as assessed by ERG and TEM. Thus, if there do exist mechanisms to maintain consistent rhodopsin packaging density in ROS disc membranes, our data suggest that PRCD may be involved in regulation of this process. Although rhodopsin levels and availability appear relatively normal in Prcd-KO retina at P30, there is an apparent inability for rod photoreceptor neurons lacking PRCD to maintain the proper density of rhodopsin within ROS disc membranes. Furthermore, the presence of these irregular discs strongly suggests a possible explanation for the observed reduction in Prcd-KO rod ERG response and ROS ultrastructural defects at P30 and subsequent ages. By P120, we detect an approximate 20.4% reduction in total rhodopsin protein. We speculate that this decrease is related to the disrupted incorporation of rhodopsin into these aforementioned irregular discs.
In all PRCD knockout models to date, cones appear unaffected until rods degenerate. As cones rely on opsins other than rhodopsin for phototransduction, the specific role for PRCD in rod photoreceptors through packaging of rhodopsin into disc membranes is compelling. Considering PRCD's small size, one hypothesis is that PRCD works by anchoring nascent discs in place to ensure that the proper density of rhodopsin is packed into each disc before flattening and enclosure. Investigation of this possibility, as well as determining why Prcd-KO rods are still able to form a population of phenotypically normal discs, are topics of future study.
Altogether, our data provide further insight into PRCD's involvement in proper rhodopsin packaging density in ROS disc membranes, a process which is crucial for OS maintenance and disc morphogenesis.

Methods
Generation of Prcd knockout model. Prcd-KO mice were generated using CRISPR/Cas9 genome editing at the transgenic core facility at West Virginia University. Small guide RNA (sgRNA) were designed as described in earlier studies 46,47 . Two sgRNA target sequences were created for exon 1 of Prcd, both upstream (CCG TAG ATT TAC CAA CCG AGTGG ) and downstream (CCA CTC GGT TGG TAA ATC TACGG ) of the 5′-NGG "PAM" (protospacer adjacent motif). These sequences were annealed and ligated into a pX330 vector (pSpCas9). Using PCR amplification, a T7 promoter (TTA ATA CGA CTC ACT ATA GGG) was added to the sgRNA template and purified using an RNA purification kit from Ambion. After confirming the specificity and efficacy of the sgRNA/ Cas9 cutting by in-vitro assays, both sgRNA (17 ng/μl) and Cas9 (34 ng/μl) mRNA (Invitrogen) were injected into the pronuclei of FVB blastocysts. The correctly targeted founder mice were identified by sequencing and backcrossed with 129/SV-E mice (Charles River Laboratories) to eliminate the rd1 allele, which naturally occurs in FVB mice. Furthermore, Prcd-KO mice were confirmed to lack the rd8 mutation and were extensively backcrossed for more than 6 generations with 129/SV-E mice to rule out "off-target" effects sometimes associated with CRISPR/Cas9 gene editing. Genotyping was performed by PCR amplification followed by sequencing. All experimental procedures involving animals in this study were approved and conducted in strict accordance with relevant guidelines and regulations by the Institutional Animal Care and Use Committee at West Virginia University.
Electroretinography. Prior to measuring light dependent electrical response, mice were dark-adapted overnight before testing and ERGs were performed under dim red light. Animals were anesthetized with isoflurane (5% in 2.5% oxygen) and eyes were dilated (1:1 Phenylephrine : Tropicamide) for 10 min and placed on a stage heated to 37 °C with a nose cone supplying isoflurane (1.5% in 2.5% oxygen) for testing. After lubrication with GenTeal gel (0.3% Hypromellose), silver electrodes were placed above the cornea to measure ERG response. A reference electrode was placed between the ears on top of the animal's head. ERG recordings were conducted using the UTAS Visual Diagnostic System with BigShot Ganzfeld, UBA-4200 amplifier and interface, and EMWIN software (version 9.0.0, LKC Technologies, Gaithersburg, MD, USA). Scotopic (rod) ERGs were performed in the dark with flashes of white light at increasing intensities. Photopic (cone) responses were performed with flashes of increasing intensities of white light with a 30 cd/m 2 white background light for 10 min to saturate rods.
Immunoblotting. To evaluate protein expression by immunoblot, two retinas were homogenized by sonication (Microson Ultrasonic cell disruptor, 3 pulses 5 s at power setting 10) in 200 μl of urea sample buffer (USB) containing 6 M, 4% SDS, and 125 mM Tris-HCl, pH 6.8 in a 1.5-ml microcentrifuge tube on ice. Protein concentration was determined using a NanoDrop spectrophotometer (ND-1000, Thermo Fisher Scientific). Samples were normalized to 5 mg/ml with additional USB and 5% 2-mercaptoethanol and bromophenol blue were added prior to SDS-PAGE. Equal concentration of each sample was loaded and separated by SDS-PAGE in a 4-15% tris-glycine precast gel (4-20% Mini-PROTEAN TGX, Bio-Rad, Hercules, CA, USA) and subsequently transferred to an Immobilon-FL PVDF membrane (Immobilon-FL, Millipore, Burlington, MA, USA). After blocking membranes with blocking buffer (Odyssey Blocking Buffer; LI-COR Biosciences, Lincoln, NE, USA) for 1 h at room temperature (RT) and incubated with primary antibodies for 2 h at RT or overnight at 4 °C on a bidirectional rotator. After primary antibody incubation (see Table 3), immunoblots were washed three times in 1X PBST (1X PBS/0.1% Tween-20) for 5 min each at RT and incubated with secondary antibody, goat anti-rabbit Alexa Fluor 680, goat anti-mouse DyLight 800 or donkey anti-sheep Alexa Fluor 680 (Thermo Fisher Scientific) for 30 min at RT. After washing three additional times in 1X PBST, membranes were scanned using an Odyssey Immunohistochemistry. Whole eyes from P30, P120, and P250 were enucleated, immersed, and fixed in Excalibur's Z-fix. Embedding, sectioning, and hematoxylin and eosin (H&E) staining was carried out on WT and Prcd-KO eyes by Excalibur Pathology, Inc. (Norman, OK, USA). Images were collected using a Nikon C2 confocal microscope. Five points were chosen to count the number of photoreceptor nuclei in the outer nuclear layer (every 350 μm) above and below the optic nerve. Counts were completed for three independent eyes (n = 3) and averaged to determine the amount of photoreceptor degeneration at each age. Images were processed using ImageJ software along with the Bio-Formats plugin 48,49 .
Immunofluorescent staining. After enucleation, mouse eyes were prepared as described previously 17 .
Using a Leica CM1850 Cyrostat, 16 μM cross-sections were cut and placed on Superfrost plus slides (Fisher Scientific). Slides were washed with 1X PBS to hydrate tissue and to remove OCT. To prevent non-specific antibody labeling, slides were blocked for 1 h. After blocking, primary antibodies were added at a 1:1000 dilution and incubated overnight at 4 °C (see Table 3). After primary antibody incubation, sections were washed in 1X PBST (1X PBS/0.1% Triton X-100) three times and incubated with the nuclear stain DAPI (1:5,000; Molecular Probes) and secondary antibody (Thermo Fisher Scientific goat anti-mouse Alexa Fluor-488 or goat anti-rabbit Alexa Fluor-568) at a 1:1,000 dilution for 1 h at RT. Slides were mounted with Prolong Gold antifade reagent (Thermo Fisher Scientific) and coverslipped (1 mm). Slides were imaged using a Nikon C2 confocal microscope. Images were processed using ImageJ software along with the Bio-Formats plugin 48,49 .
Transmission electron microscopy. Enucleated eyes were placed in freshly prepared buffer containing 2% paraformaldehyde, 2.5% glutaraldehyde, and 0.1 M cacodylate buffer, pH 7.5. After 30 min of fixation, cornea and lens were removed, and eyecups were fixed for an additional 48 h at RT with constant rotation on a nutator. Dissection, embedding, and transmission electron microscopy were performed as described previously 50 .

Rod outer segment and disc membrane isolation. Retina from 15 age-matched WT (129/SV-E) and
Prcd-KO mice were collected. Retinal tissues were suspended in 300 μl of 8% (vol/vol) Optiprep (Sigma-Aldrich, St. Louis, MO, USA) in Ringer's buffer (10 mM HEPES, 130 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl 2 , 1.2 mM CaCl 2 , and 0.02 mM EDTA) and vortexed for 1 min. Next, samples were centrifuged at 238 × g for 1 min at 4ºC and the supernatant was collected into a clean tube. This was repeated four additional times and the subsequent supernatant was pooled and placed on top of a 10-30% Optiprep in Ringer's buffer gradient. The gradient containing the sample was then centrifuged at 26,500 × g for 50 min at 4ºC with no brakes (Beckman Coulter Optima LE-80 K; SW-41Ti). A dense band approximately two-thirds from the top of the gradient was collected and diluted fourfold in Ringer's buffer. After a 3 min centrifugation at 627 × g at 4ºC, the supernatant was next transferred to a high-speed centrifuge tube and centrifuged at 26,500 × g for 30 min at 4ºC (Beckman Coulter Optima TLX; rotor-TLA55). The resulting pellet was then resuspended in 2 mM Tris-HCl and used for AFM studies.
Atomic force microscopy. All experimental procedures were conducted under dim red light conditions to avoid photobleaching. ROS disc membranes were diluted in Ringer's buffer (10 mM HEPES, 130 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl 2 , 1.2 mM CaCl 2 , and 0.02 mM EDTA, pH 7.4), immobilized on freshly cleaved mica and imaged by AFM in imaging buffer (20 mM Tris, 150 mM KCl, 25 mM MgCl 2 , pH 7.8), as described previously 32 . Contact mode AFM imaging was performed using Bruker Multimode II atomic force microscope equipped with an E scanner (13 μm scan size) and silicon nitride cantilevers with a nominal spring constant of 0.06 N/m (DNP-S, Bruker, Santa Barbara, CA). Deflection images were analyzed using SPIP (version 6.5, Image Metrology A/S, Hørsholm, Denmark) to determine the number and density of rhodopsin, as described previously 32,39 . Graphs and statistical analyses were done using Prism 7 (GraphPad Software, San Diego, CA). www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.