ATF6 is required for efficient rhodopsin clearance and retinal homeostasis in the P23H rho retinitis pigmentosa mouse model

Retinitis Pigmentosa (RP) is a blinding disease that arises from loss of rods and subsequently cones. The P23H rhodopsin knock-in (P23H-KI) mouse develops retinal degeneration that mirrors RP phenotype in patients carrying the orthologous variant. Previously, we found that the P23H rhodopsin protein was degraded in P23H-KI retinas, and the Unfolded Protein Response (UPR) promoted P23H rhodopsin degradation in heterologous cells in vitro. Here, we investigated the role of a UPR regulator gene, activating transcription factor 6 (Atf6), in rhodopsin protein homeostasis in heterozygous P23H rhodopsin (Rho+/P23H) mice. Significantly increased rhodopsin protein levels were found in Atf6−/−Rho+/P23H retinas compared to Atf6+/−Rho+/P23H retinas at early ages (~ P12), while rhodopsin mRNA levels were not different. The IRE1 pathway of the UPR was hyper-activated in young Atf6−/−Rho+/P23H retinas, and photoreceptor layer thickness was unchanged at this early age in Rho+/P23H mice lacking Atf6. By contrast, older Atf6−/−Rho+/P23H mice developed significantly increased retinal degeneration in comparison to Atf6+/−Rho+/P23H mice in all retinal layers, accompanied by reduced rhodopsin protein levels. Our findings demonstrate that Atf6 is required for efficient clearance of rhodopsin protein in rod photoreceptors expressing P23H rhodopsin, and that loss of Atf6 ultimately accelerates retinal degeneration in P23H-KI mice.


Results
Loss of Atf6 leads to impaired clearance of rhodopsin protein and hyperactivation of the IRE1-XBP-1s signaling pathway in early age Rho +/P23H mice. Previously, we found that P23H rhodopsin protein was rapidly degraded in photoreceptors of P23H-KI mice at early age (P15) 22 . We also found that chemical-genetic activation of Atf6 signaling pathways promoted P23H rhodopsin protein degradation in heterologous HEK293 cells, while sparing wild-type rhodopsin protein 22,27 . To investigate if Atf6 is important for P23H rhodopsin protein degradation in photoreceptors, we examined retinas of Rho +/P23H mice bred with Atf6 −/− mice.
No gross changes in retinal histology in the absence of Atf6 in young Rho +/P23H retinas. Overexpression of rhodopsin causes photoreceptor cell death and induces retinal degeneration in transgenic animals expressing wild-type rhodopsin or P23H rhodopsin [32][33][34] . With that in mind, we asked if the increased steadystate rhodopsin protein levels in Rho +/P23H mice lacking Atf6 corresponded with photoreceptor cell loss (Fig. 1a). We performed histologic studies to see if photoreceptors or retinal lamination was impacted. A previously published paper has shown that Rho +/P23H retinas have scattered pyknotic nuclei by P15 22 , with Rho +/P23H mice exhibiting a slight reduction in the ONL thickness. Furthermore, the outer segments and inner segments of rod photoreceptors were shorter compared to the wild-type mice. In both Atf6 +/− Rho +/P23H and Atf6 −/− Rho +/P23H mice, we observed scattered and disorganized nuclei in the ONL ofP15 mice. In addition, the thickness of the ONL, outer plexiform layer (OPL), inner nuclear (INL), and inner plexiform layer (IPL) appeared similar (Fig. 2a,b) in retinas (P > 0.05, two-way ANOVA analysis). Therefore, the ~ 2 × increase in rhodopsin protein levels found in the absence of Atf6 did not lead to detectable changes in the thickness of photoreceptor ONL or overall retinal anatomy at this age. These findings demonstrate that Rho +/P23H mice can tolerate increased amounts of rhodopsin protein when Atf6 is lost, at least at this young age.

Discussion
Many disease variants in the human RHODOPSIN gene found in RP patients introduce missense mutations in the rhodopsin polypeptide that cause rhodopsin protein misfolding, retention in the ER, and inability to bind to 11-cis-retinal 6,25,37,38 . These molecular defects instigate rod photoreceptor decline by incompletely understood mechanisms and, ultimately, lead to the clinical manifestations of RP. Currently, there is no cure for RP caused by these misfolded rhodopsin proteins. We previously found that chemical-genetic activation of the ATF6 signaling pathway significantly reduced protein levels of several misfolded RP rhodopsin variants, such as T17M, Y178C, C185R, D190G, and K296E rhodopsin, while sparing wild-type rhodopsin when expressed in heterologous HEK293 cells 39 . Furthermore, activation of ATF6 reduced misfolded P23H mutant rhodopsin protein levels (monomer, dimer, and multimers) in HEK293 cell in vitro 39 . Here, we investigated how ATF6 signaling affected P23H mutant rhodopsin protein in photoreceptors in vivo. We examined the steady-state levels of total rhodopsin protein in retinal samples collected from Atf6 +/− Rho +/P23H and Atf6 −/− Rho +/P23H mice. The rhodopsin protein species in these heterozygous Rho +/P23H mice consist of wild-type and P23H rhodopsin. We found significantly more (nearly 2x) total rhodopsin protein in Atf6 −/− Rho +/P23H compared to Atf6 +/− Rho +/P23H while rhodopsin mRNA levels did not significantly change between these strains of mice at 12. These findings provide support that Atf6 is important for rhodopsin protein quality control in rod photoreceptors, because in the Atf6 −/− Rho +/P23H mice, steady state rhodopsin protein levels increased almost 2x. ATF6 signaling likely ensures the efficient degradation of mutant P23H rhodopsin protein through transcriptional induction of factors involved in ER protein folding and ERAD 20,21,40,41 . Therefore, this model demonstrates that loss of Atf6 leads to accumulation of P23H rhodopsin protein that contributes to the ~ 2 × increase in steady-state rhodopsin protein levels at early ages. Our www.nature.com/scientificreports/ findings may provide mechanistic insight into prior studies demonstrating a protective role for ATF6 activity in RP models. For example, in vivo intravitreal AAV injection of one of ATF6's downstream targets, the BiP/Grp78 chaperone, into P23H rhodopsin transgenic rats improved ERG responses 42 . This protective response could arise from increased elimination of the P23H rhodopsin protein through increased BiP/Grp78 chaperone-mediated increase in ERAD. Taken together, these findings underscore the importance of Atf6 plays in rhodopsin protein homeostasis in rods. Variants in the human ATF6 gene cause achromatopsia and cone-rod dystrophy carrying bi-allelic disease alleles 30,[43][44][45][46][47] . Patients with these ATF6 mutations showed malformation of the fovea, dysfunction of photoreceptors, and severe vision loss from infancy 30,45 . Furthermore, it is reported that abnormal retinal vasculature development may lead to malformation of the fovea 48,49 . However, none of these findings are apparent in young Atf6 −/− mice or in young Atf6 −/− Rho +/P23H mice 30 (Supplemental Fig. S1). This difference may reflect a selective function for ATF6 in human cone and/or foveal development. For example, retinal organoids produced from the patients homozygous for ATF6 disease alleles showed significant defects in cone photoreceptor development accompanied by reduction in cone gene expression which included all cone phototransduction genes (CNGB3, CNGA3, PDE6C, PDE6H, and GNAT2) and red and green cone opsin genes 50 . Although cones do not appear to be selectively compromised in Atf6 −/− mice or Atf6 −/− Rho +/P23H mice, the absence of Atf6 does accelerate degeneration throughout the retina in Rho +/P23H mice. This is consistent with Atf6 expression in all retinal cell types, where it likely functions to ensure cell viability in the face of ER stress throughout life 47 . www.nature.com/scientificreports/ In our study, we found that the IRE1 signaling pathway was hyper-activated in Atf6 −/− Rho +/P23H mice when rhodopsin steady-state levels were increased at young ages. We propose that this hyper-activity in IRE1 signaling reflects a compensatory response to loss of Atf6. Specifically, the loss of Atf6 leads to reduced degradation of mutant rhodopsin protein in Rho +/P23H . In turn, this accumulation of misfolded rhodopsin hyper-activates the IRE1 signaling pathway to degrade the increased rhodopsin accumulating in P12 Atf6 −/− Rho +/P23H . Consistent with a role for IRE1 signaling in rhodopsin degradation, we have previously demonstrated that the IRE1 signaling pathway of the UPR is selectively activated in photoreceptors of Rho +/P23H ERAI +/− compared to www.nature.com/scientificreports/ Rho +/+ ERAI +/− mice in a study that used the ERAI mouse GFP reporter line to indicate IRE1-XBP-1 activation 22 . We found that induction of ERAD by IRE1 signaling leads to ubiquitination of P23H rhodopsin in photoreceptors in Rho +/P23H mice 22 . This demonstrated that P23H rhodopsin is rapidly degraded by induction of ERAD in photoreceptors to eliminate misfolded rhodopsin from the ER in vivo. Furthermore, in a 2021 ARVO poster,   51 . Based on our previous and current study and the recent report by Massoudi et al. (2021), both ATF6 and IRE1a protect against ER stress in photoreceptors in Rho +/P23H mice. Our current study showed that the levels of Xbp-1s mRNA, BiP/Grp78 protein, and other transcriptional targets were significantly increased in the retinas of Atf6 −/− Rho +/P23H mice compared to Atf6 +/− Rho +/P23H mice at early age. Many of XBP-1's target genes encode components of the ERAD pathway, and these genes have been found to be upregulated in the retinas of Rho +/P23H mice 12,13,22 . These findings suggest that degradation of P23H rhodopsin via downstream transcriptional activity of the IRE1-XBP-1s pathway and, consequently, ERAD, both work to alleviate ER stress caused by the accumulation of misfolded rhodopsin. Our model is further supported by the findings that E3 ubiquitin ligases, SORDD1/2, was able to facilitate degradation of Rh1 P37H (the Drosophila equivalent of P23H rhodopsin) at larval and earlier stages of growth to allow for development of healthy adult eyes. Furthermore, SORDD1/2 and HRD1/ SYVN1 were also able to prevent retinal degeneration in Drosophila with the G69D (glycine to aspartic acid at amino acid residue 69) rhodopsin mutation 52 . The lack of Atf6 in Rho +/P23H mice may initially increase the ability of E3 ubiquitin ligases downstream of IRE1-XBP-1s-ERAD to target misfolded rhodopsin in early stages of life. In contrast, Chop mRNA levels (an ER stress gene induced by PERK pathway) 22,53 were not affected between Atf6 +/− Rho +/P23H and Atf6 −/− Rho +/P23H mice, which is consistent with previous studies showing that Chop was not induced during retinal degeneration in P23H rhodopsin mice and that the loss of CHOP had no impact on retinal degeneration based on histology or ERG 31,54 . Activation of PERK signaling also did not lead to greater reduction in rhodopsin protein levels in WT or P23H mice 27 .
There are several lines of evidence suggesting alterations of other degradation systems in Atf6 −/− Rho +/P23H mice. We have previously reported in cell culture models that IRE1 relies on functioning proteasomes and lysosomes to degrade the mutated, misfolded rhodopsin 27 . Yao et al. (2018) also reported that P23H mice experience increase in autophagy secondary to ER stress, which leads to proteasome insufficiency and increase retinal degeneration. In contrast, genetic or pharmacologic inhibition of autophagy reduced retinal degeneration and improved proteasome levels 55 . Modulating the ratio between autophagy and proteasome activity (A:P) also helped to improve photoreceptor survival 56 . The authors demonstrated that normalizing the A:P ratio, either by improving folding of P23H rhodopsin or increasing proteasome activity to keep autophagy pathways down, increased photoreceptor survival and preserved retinal function. Taken together, we suggest that autophagy activity is increased as a result of the loss of Atf6.
We found increased retinal degeneration and diminished rhodopsin protein levels in P60 Atf6 −/− Rho +/P23H retinas compared to P60 Atf6 +/− Rho +/P23H . Furthermore, we found that the thickness of retinal layers including ONL, OPL, INL, and IPL were also significantly lower in the ventral part of the Atf6 −/− Rho +/P23H retina compared to Atf6 +/− Rho +/P23H . The reduction of ONL in the ventral part of the retina is consistent with previous histological data but the thickness of other retinal layers was not measured previously [22][23][24] . In Rho +/P23H mice, approximately half of the rod photoreceptor cells had disappeared between P14-P40 when compared to Rho +/+ retina, which showed no reduction of rod photoreceptors between P40 and P63 as described in previous studies [22][23][24] . Our data demonstrate that by P60, loss of Atf6 accelerates retinal degeneration in Rho +/P23H mice. Why does loss of Atf6 increase retinal degeneration in Rho +/P23H mice at P60, while not affecting younger animals? Our previous study showed that early wave of photoreceptor cell death and peak induction of the IRE1 reporter occur during the first postnatal month in Rho +/P23H mice 22 . The activation of IRE1 in Rho +/P23H is maintained throughout life to regulate proteostatic balance to remove P23H rhodopsin 22 . We propose that hyperactivation of IRE1 (as seen in the younger animals) restored rhodopsin protein homeostasis in the absence of Atf6 beginning at P12 (i.e., early stage), so that P30 Rho +/P23H retinas looked indistinguishable. Why can't IRE1 hyperactivation keep rhodopsin and retina healthy at P60? We propose that the capacity of IRE1 to support ER homeostasis may ultimately be overwhelmed in the absence of Atf6, leading to increased rod photoreceptor cell death, and reduction of rhodopsin protein levels at later ages of Atf6 −/− Rho +/P23H mice 35,36 . The ongoing photoreceptor cell death in RP retina likely causes widespread ER stress from oxidative damage, mitochondrial dysfunction, and other metabolic degenerative mechanisms 57,58 . Other sources of ER stress arising in the degenerating retina include damaged lipids, proteins, carbohydrates, enzymes, and DNA in photoreceptor cells, which ultimately results in further photoreceptor cell death through lipid peroxidation 59 . Thus, P23H rhodopsin-induced cell damage in addition to P23H rhodopsin protein itself could elicit too much ER stress, overwhelming the proteostatic balance maintained by the IRE1 in in the Atf6 −/− Rho +/P23H mice.
Here, we observed no detectable difference in the function of rods and cones between P60 Atf6 +/− Rho +/P23H and P60 Atf6 −/− Rho +/P23H mice. Although, we observed a reduction of ONL and other retinal layers in Atf6 −/− Rho +/P23H mice compared to Atf6 +/− Rho +/P23H mice, no significant difference was noted in amplitude of either the scotopic or photopic b-wave in the strains of mice. Why did the reduction of retinal layers in Atf6 −/− Rho +/P23H mice compared to Atf6 +/− Rho +/P23H mice show no functional changes? We propose that the full-field flash ERG is relatively insensitive to detect smaller defects 60,61 because it represents the global retinal function via summed electrical response of the whole retina excited by a flash of light 62 . Thus, the reduction of retinal layers in ventral retina observed in P60 in the absence of Atf6 in Rho +/P23H retina was likely not detected with full-field ERG due to preservation of the dorsal ONL layer.
In recent years, numerous small molecules have been identified that activate or inhibit ATF6 or IRE1 27,63-68 . Agonists of IRE1 signaling include Type 1 IRE1 kinase inhibitors, which allosterically activate the RNAse function of IRE1, and IRE1 activators, which activate both RNAse and kinase function; however, these small molecule candidates (e.g. 474, IXA4, and IXA6), albeit showing no activation of IRE1-dependent cell death pathways, have yet to be fully tested for rhodopsin proteostatic properties 67,69 . By contrast, ATF6 agonists (e.g., AA147 and AA263) are effective in vivo and may have significant implications for amyloid related diseases and retinal development through ATF6 activation 53 www.nature.com/scientificreports/ BiP/Grp78, a prominent target of ATF6 upon ER stress, alleviates P23H RP symptoms 42 . We propose that ATF6 and IRE1-XBP-1 small molecule agonists are promising agents for further RP clinical studies if their rhodopsin proteostatic properties can be shown in vivo. There are no reported differences in the phenotypes between Atf6 +/+ and Atf6 +/− mice 30,41,42,44 . Injection of Atf6 +/+ , Atf6 +/− , and Atf6 −/− mice with tunicamycin led to kidney and liver toxicity only in Atf6 −/− animals but not in Atf6 +/+ or Atf6 +/− mice; in addition, our previous study has shown normal morphology and normal rhodopsin expression when comparing Atf6 +/+ to Atf6 +/− mice 30 . All experiments used female or male Atf6 −/− Rho +/P23H mice in comparison to control littermates Atf6 +/− Rho +/P23H , at the postnatal (P) days 12, 15, 30, and 60 (number (n) = 3 ~ 6 respectively for each stage). For retinal vasculature assessment in Atf6 −/− mice, female and male P30 Atf6 +/+ and P30 Atf6 −/− mice (n = 3 animals per group) on a C57BL/6 J background were used as described in previous studies 30,41 . For all experiments, animals were kept in cyclic 12-h light/dark conditions with free access to food and water. All mouse care and experimental procedures in this study were approved and conducted in strict accordance with relevant guidelines and regulations by the Institutional Animal Care and Use Committee at the Stanford University and in compliance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and the ARRIVE (Animal Research: Reporting of in Vivo Experiments) guidelines.
The eyes were enucleated for collection of retinal tissue. For secondary method, we performed cervical dislocation. The lens and the anterior segment were removed, and the eyecups were further dissected to collect whole retinal lysate for biochemistry or molecular biology, or eyecups were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB), for 60 min at 4 °C. After fixation, the eyecups were processed for hematoxylin and eosin (H&E) staining 71 and cryostat sectioning. For cryostat sectioning, eyecups were transferred from 10% for 1 h to 20% for 1 h to 30% sucrose overnight at 4 °C, then eyecups were embedded in Optimal Cutting Temperature (OCT) medium (Tissue-Tek, Elkhart, IN), frozen in liquid nitrogen and subsequently vertically sectioned on a Leica cryostat (Leica Biosystems Inc, Buffalo Grove, IL) at a thickness of 20 μm. For wholemount retinal preparation, the retinas were isolated from the eyecups and dissected as wholemounts.
H & E staining. The detail protocols for H & E staining in retinal layer was performed as previously published 71 . Three to five left eyecups from three to five animals (n = 3-5) were sectioned along the vertical meridian on a cryostat at a thickness of 20 μm. Sections were then collected on gelatin-coated slides for H&E staining. Slides were dipped in Harris hematoxylin for 1 min then they were washed in tap water and dehydrated in alcohol. Slides were then dipped in Eosin-Phloxyine for 30 s, then dehydrated in a series of 95% ethanol and 100% ethanol followed by 5 min in xylene, and mounted in Vectashield mounting medium (Vector Labs, Burlingame, CA).
ERGs (Diagnosys LLC, Lowell, MA) were recorded from both eyes of Atf6 −/− Rho +/P23H (n = 6) mice and compared to ERGs from eyes of Atf6 +/− Rho +/P23H control littermates (n = 6) at P60 as described previously 30 . Mice were anaesthetized using a combination of ketamine (20 mg/kg; KETASET, Fort Dodge, IA, USA) and xylazine (5 mg/kg, X-Ject SA; Butler, Dublin, OH, USA) using similar procedures as our published protocols 30 . Under a dim red light, the pupils were dilated with Atropine sulfate ophthalmic solution 1% (Akorn Inc, Lake Forest, IL, USA). The recording electrodes attached to two gold wire rings were placed on the cornea of both eyes. The eye lubricant hypromellose ophthalmic gel, USP 2.5% (HUB pharmaceuticals, LLC, Rancho Cucamonga, CA, USA) was applied to keep the hydration and conductivity between the cornea and recording electrodes. The ground and reference electrodes were placed at the tail and tongue, respectively. The eyes were then given scotopic ERG responses (a series of white light flashes varying from -1.5 to 2 log cd s/m 2 ). After 10 min of light adaptation, photopic ERG responses of -0.31 to 2.81 log cd s/m 2 were recorded. The amplitudes for the resulting b-wave responses at the series of light flash intensity were plotted.
Retinal vasculature staining in Atf6 −/− mice. For wholemount immunohistochemical staining, the same procedures described in our previous studies were used 30,71 . Three right retinas from three animals (n = 3) were used for wholemount staining. Wholemounts were treated with 1% Triton X-100 in 0. Statistical analysis. All the statistics were expressed as mean ± standard error of the mean (SEM). Student's t-test was used for comparison. Two-way ANOVA and Fisher's least significant difference procedure (LSD test) were used to examine the differences among the group of means. All the statistical tests were performed using GraphPad Prism Version 8.3.1. The difference between the means of separate experimental groups was considered statistically significant at P < 0.05. www.nature.com/scientificreports/