Subcellular localization of mutant P23H rhodopsin in an RFP fusion knockin mouse model of retinitis pigmentosa

The P23H mutation in rhodopsin (Rho), the visual pigment protein in rod photoreceptor neurons, is the most common genetic cause of autosomal dominant retinitis pigmentosa (adRP), a retinal disease that causes blindness. Despite multiple studies in animal models, the subcellular details of the fate of misfolded mutant Rho in rod photoreceptors have not been completely defined. We generated a new mouse model of adRP, in which the P23H-Rho mutant allele is fused to the fluorescent protein Tag-RFP-T (P23HhRhoRFP). In heterozygotes, outer segments formed, and WT rhodopsin was properly localized there, but mutant P23H-Rho protein was specifically mislocalized in the inner segments of rods. Despite this cellular phenotype, the P23HhRhoRFP heterozygous mice exhibited only slowly progressing retinal degeneration; in ERG recordings, scotopic a-wave amplitudes were reduced by 24% and 26% at 30 days and 90 days respectively, and the corresponding scotopic b-waves by 18% and 24%. Outer nuclear layer thickness was still 80% of WT at 90 days, but at 364 days had declined to 40% of WT. Transmission electron microscopy revealed greatly expanded membrane lamellae in the inner segment, and by fluorescence imaging, we determined that the mislocalized P23HhRhoRFP was contained in greatly expanded endoplasmic reticulum (ER) membranes. TUNEL staining revealed a slow pace of cell death involving chromosomal endonucleolytic degradation. Quantification of mRNA for markers of ER stress and the unfolded protein response revealed little or no increases in levels of messages encoding the proteins BiP, CHOP, ATF6, XBP1, PERK, Eif2α and Derlin-1, but a decreased level of total Rhodopsin (mouse + human) mRNA levels. The decline in the rate of cell death after an initial burst suggests that P23HhRhoRFP mutant rods undergo an adaptative process that prolongs survival despite gross P23HhRhoRFP protein accumulation in the ER. Because of its slowly progressing nature, and easy visualization of the mutant protein, the P23H-Rho-RFP mouse may represent a useful tool for the future study of the pathology and treatment of P23H-Rho and adRP.

neurons, is the most common genetic cause of autosomal dominant retinitis pigmentosa 23 (adRP), a retinal disease that causes blindness. Despite multiple studies in animal models, the 24 subcellular details of the fate of misfolded mutant Rho in rod photoreceptors have not been 25 completely defined. We generated a new mouse model of adRP, in which the P23H-Rho mutant 26 allele is fused to the fluorescent protein Tag-RFP-T (P23HhRhoRFP). In heterozygotes, outer 27 segments formed, and WT rhodopsin was properly localized there, but mutant P23H-Rho 28 protein was specifically mislocalized in the inner segments of rods. Despite this cellular 29 phenotype, the P23HhRhoRFP heterozygous mice exhibited only slowly progressing retinal 30 degeneration; in ERG recordings, scotopic a-wave amplitudes were reduced by 24% and 26% 31 at 30 days and 90 days respectively, and the corresponding scotopic b-waves by 18% and 24%. 32 Outer nuclear layer thickness was still 80% of WT at 90 days, but at 364 days had declined to 33 40% of WT. Transmission electron microscopy revealed greatly expanded membrane lamellae 34 in the inner segment, and by fluorescence imaging, we determined that the mislocalized 35 P23HhRhoRFP was contained in greatly expanded endoplasmic reticulum (ER) membranes. 36 TUNEL staining revealed a slow pace of cell death involving chromosomal endonucleolytic 37 degradation. Quantification of mRNA for markers of ER stress and the unfolded protein 38 response revealed little or no increases in levels of messages encoding the proteins BiP, 39 CHOP, ATF6, XBP1, PERK, Eif2α and Derlin-1, but a decreased level of total Rhodopsin 40 (mouse + human) mRNA levels. The decline in the rate of cell death after an initial burst 41 suggests that P23HhRhoRFP mutant rods undergo an adaptative process that prolongs survival 42 despite gross P23HhRhoRFP protein accumulation in the ER. Because of its slowly progressing 43 nature, and easy visualization of the mutant protein, the P23H-Rho-RFP mouse may represent 44 a useful tool for the future study of the pathology and treatment of P23H-Rho and adRP. 45 46 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021.

6
In addition, we added an additional 1D4 signaling sequence to the carboxy-terminus of 123 the P23H-hRho + Tag-RFP-T fusion allele (Fig 1A-B). Our rationale was that the C-terminal 124 RFP fusion tag in the expressed P23H-hRho-Tag-RFP-T (hereafter P23HhRhoRFP) protein 125 may inhibit the endogenous Rho 1D4 sequence leading to artifacts not attributable to the P23H 126 mutation. The additional 1D4 sequence was also shown to be necessary for normal trafficking of 127 transgenic Rho-Dendra fusion protein in Xenopus laevis rods (40). 128 The P23H-hRho-TagRFP knock-in mice were generated the same way we previously 129 Tyr c-Brd mice (43). Founder mice carrying the HPRT-P23H-hRho-TagRFP allele were crossed to 138 GDF-9-iCre mice (44) to remove the HPRT minigene. P23H-hRho-TagRFP (hereafter "P23H-139 RFP" in reference to the knockin allele) mice were extensively backcrossed to C57BL/6 mice. 140 We validated that the knockin was successful by sequencing genomic DNA from the knockin 141 mouse. We verified expression of the P23HhRhoRFP fusion by fluorescence microscopy of 142 retinas and by immunoblotting (Fig. 1, Fig. 2). 143 We used P23H-RFP/+ heterozygous mouse retinas to examine the expression of the 144 mutant fusion protein alongside the WT mouse Rho protein from the wild-type allele (Fig 1B). 145 Both WT mouse Rho protein and the product of the knockin allele were detected in P23H-RFP/+ 146 retinal lysates by probing with anti-1D4 antibody. The P23HhRhoRFP fusion protein was 147 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint A Human P23H-Rhodopsin-RFP Fusion Gene: Retinal lysates are from a wild-type (+/+) mouse, age P22, and a P23H-RFP/+ mouse, age 45; 100 µg of total protein from each lysate was loaded onto SDS-PAGE gels. Blot membranes were probed with either of the following antibodies: anti-1D4, anti-RFP or anti-beta (ß)-actin (a loading control). Blot scans for the protein ladder were used to mark molecular weight sizes (in kilodaltons, kDa) on the left of each blot image. The ~65 kDa P23HhRhoRFP fusion protein band is present in the P23H-RFP/+ lane in both anti-1D4 and anti-RFP blot scans (magenta arrows). The monomeric mouse Rho protein band is in both lanes in the anti-1D4 blot scan (black arrow). Higher MW species are formed by rhodopsin multimerization.

5' -ctgtacaagacggagacgagccaggtggccccggcctaa L Y K T E T S Q V A P A -
detected as a strong ~65 kDa band specifically in P23H-RFP/+ lysates using both anti-1D4 and 148 anti-RFP antibodies, verifying robust expression. 149 150 P23HhRhoRFP protein was mislocalized in rod photoreceptor neurons. 151 We next tested the fluorescence pattern of the P23HhRhoRFP fusion protein in the 152 retinas of both P23H-RFP/+ heterozygous and P23H-RFP/P23H-RFP homozygous mice. The 153 RFP fluorescence is evident in the outer photoreceptor layers of retinal sections from both 154 heterozygotes and homozygotes at age P30 (Fig 2A). In a confocal z-projection of a retinal 155 section from a P30 P23H-RFP/+ heterozygote, P23HhRhoRFP is most prominently located in 156 brightly fluorescent puncta or "aggregates" within the regions of the inner segments and outer 157 segments of photoreceptors ( Fig 2B). Less prominent but visible in this same section is P23H-158 Rho-RFP fluorescence at the photoreceptor synapses of the outer plexiform layer (OPL) and in 159 the cytoplasm surrounding the photoreceptor nuclei of the outer nuclear layer (ONL) (Fig 2B). 160 Compared to both WT and heterozygotes, the ONL of P30 homozygotes was noticeably thinner 161 based on DAPI+ nuclei staining ( Fig 2C). P23HhRhoRFP aggregates were also clearly visible in 162 the ONL of the homozygous retina. 163 We crossed our new P23H-RFP mouse line with wild-type hRho-GFP mouse lines for 164 dual fluorescent tag multiplex imaging. This approach allowed us to investigate the subcellular 165 dynamics of the mutant P23HhRhoRFP protein relative to hRho-GFP without the P23H 166 mutation. In addition to the hRho-GFP fusion mouse that we previously reported (41), we also 167 generated a new line with an additional 1D4 signal sequence added to the end of the EGFP 168 sequence ("hRho-GFP-1D4"). As heterozygotes, we could not discriminate any phenotypic 169 difference between these GFP fusion mice. 170 Both GFP fusion lines were crossed to our new P23H-RFP knockin mouse. In confocal 171 images of retinal sections from an adult hRho-GFP-1D4/P23H-RFP heterozygote we observed 172 a drastic localization difference between hRho-GFP-1D4, which correctly populated the rod 173 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made photoreceptor OS cilia, and the P23HhRhoRFP aggregates, which were prominently 174 mislocalized in the rod inner segment layer (Fig 3A). We examined retinal sections from this 175 same hRho-GFP-1D4/P23H-RFP heterozygous mouse line with structured illumination 176 microscopy (SIM) superresolution imaging and observed a clear segregation of P23HhRhoRFP 177 from the hRho-GFP-1D4 in the OS cilia ( Fig 3B). Interestingly, in retinal sections from the other 178 heterozygous hRho-GFP/P23H-RFP mice, we found evidence of hRho-GFP mislocalization in 179 the same region as the P23HhRhoRFP aggregates in what appear to be distinct but interwoven 180 membrane compartments (Fig 3C). This result suggests that the added C-terminal 1D4 181 sequence prevents hRho-GFP from being mislocalized with P23HhRhoRFP in the GFP/RFP 182 heterozygous animals. 183 To visualize the location of the P23HhRhoRFP inner segment aggregates relative to the 184 connecting cilium (CC) and basal body (BB) we used centrin as an antibody marker for the CC 185 and BB (45, 46) in retinal sections from P23H-RFP/+ mice and imaged by SIM. At age P14, we 186 found many examples of P23HhRhoRFP aggregates that were located just proximal to the BB 187 ( Fig 4A). We observed this same sub-BB localization of the RFP aggregates in retinal sections 188 from age P30 P23H-RFP/+ mice (Fig 4B) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint mouse: hRho-GFP-1D4/P23H-RFP, age 3 weeks In SIM images of retinal sections from an alternate GFP/RFP heterozygote at age P30, in which the wild-type hRho-GFP fusion does not have an additional 1D4 signal peptide, the GFP fluorescence is not exclusively located in the OS layer, but rather is partially mislocalized with P23H-hRho-RFP. In a magnified example, hRho-GFP is colocalized around and within a P23HhRhoRFP aggregate (yellow arrows). In another magnified example, hRho-GFP is wrapped around an P23HhRhoRFP aggregate (white arrow). heterozygous and homozygous mutants (Fig 5A-E). Thereafter, the width of the ONL in the 217 heterozygous retinas declined slowly, approaching a final value of 23 μm with a time constant of 218 56 days, while the ONL in the homozygous mutants declined much more rapidly, with a time 219 constant of 12 days. By age P90 the ONL in P23H-RFP/P23H-RFP homozygous retinas was 220 reduced to a single, disorganized layer of nuclei (See Fig 5A, 5E, & Fig 5G). 221 Although the ONL thickness in P23H-RFP/+ heterozygous retinas were significantly 222 reduced in the inferior retina at P60 and P90 (Fig 5B & 5G), the thickness across all regions of 223 the retina in P23H-RFP/+ mutants was not significantly different from that of +/+ retinas at age 224 P90 ( Fig 5G). Notably, compared to P14 P23H-RFP/+ measurements, the P90 P23H-RFP/+ 225 ONL thickness was significantly reduced (Fig 5D). By age P364 it was evident that the ONL had 226 been severely reduced in P23H-RFP heterozygous retinas due to nuclei loss (Fig 5B, see  227 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint example in Fig 5A); the ONL width at 364 days in the heterozygotes was 40% of that in WT at 230 360 days. 231 232 ERG rod photoreceptor function was moderately diminished in P23H-RFP/+ mice. 233 We next used electroretinogram (ERG) recordings to test and correlate retina visual 234 function to the mild and severe retinal degeneration phenotypes in P23H-RFP/+ mice and 235 P23H-RFP/P23H-RFP mice, respectively. We found generally that ERG waveforms at age P30 236 correlated to our retinal degeneration phenotypes between the mutant P23H-RFP genotypes 237 compared to +/+ control mice ( Fig 6A). In dark-adapted conditions, the scotopic a-wave 238 amplitudes in P23H-RFP/+ mice were significantly reduced compared to +/+ mice at P30 (Fig  239   6B); however, the a-waves stabilized over time and were not further diminished in P23H-RFP/+ 240 mutants at age P90 compared to P30 (Fig 6C). Scotopic b-wave values were not significantly 241 reduced in P30 P23H-RFP/+ mice compared to +/+ (Fig 6D), and like the a-wave, the b-wave 242 values were not diminished in age P90 P23H/+ mice compared to P30 (Fig 6E). We also 243 measured the implicit times from scotopic ERG recordings, which are times from the a-wave 244 deflection to peak b-wave. In P30 P23H-RFP/+ mice compared to +/+ mice, implicit times were 245 higher at moderate flash intensities, but there was not a significant difference over the entire 246 flash range ( Fig 6F). 247 P23H-RFP/P23H-RFP homozygous mice have essentially no scotopic ERG response 248 ( Fig 6A-B and 6D). The minor b-wave response at high flash intensities in homozygotes could 249 be attributed to cones. We also measured the cone ERG response in light-adapted mice from all 250 genotypes at P30. Photopic b-wave amplitudes were recorded in both P23H-RFP/+ and P23H-251 RFP/P23H-RFP mice, and although they appeared slightly lower in amplitude, they were not 252 significantly reduced compared to +/+ mice ( Fig 6G). 253 To determine whether there were any visible alterations of cone morphology in the 254 mutants, we used cone arrestin immunofluorescence labeling in our P23H-RFP mice at P30. 255 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint Cones populated the retinas of P23H-RFP/+ heterozygotes as well as in the retinas of P23H-256 RFP/P23H-RFP homozygotes despite the massive loss of photoreceptor nuclei in the 257 homozygotes ( Fig 7A). At P90 we still observed cones in P23H-RFP/+ heterozygotes, but in 258 P90 homozygotes there were no visible cones remaining, presumably as a result of the nearly 259 complete loss of rods ( Fig 7B). 260 To determine the rate of cell death in our P23H-RFP mice, we used terminal 261 deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) fluorescence. We observed 262 TUNEL+ photoreceptor nuclei in the ONL of retinal sections from both P23H-RFP/+ and P23H-263 RFP/P23H-RFP mice ( Fig 7C). We measured the number of TUNEL+ nuclei per mm 2 area of 264 the ONL at multiple timepoints. At P14, the number of TUNEL+ nuclei in P23H-RFP/+ 265 heterozygous retinal sections were comparable to +/+ sections, while the rate was significantly 266 greater in homozygous retinal sections at P14 (Fig 7D). By age P30, the number of TUNEL+ 267 nuclei was significantly greater in P23H-RFP/+ retina compared to +/+ ( Fig 7D); however, at age 268 P90 the TUNEL+ density in heterozygotes was no longer significantly different from that in +/+ 269 mice ( Fig 7D). 270 Overall, we observed that P23H-RFP/+ heterozygous mice have a slow and partial 271 retinal degeneration and mild loss of rod ERG function. At age P90 the heterozygous retinas 272 were still comparable in overall health to control +/+ retinas despite the gross mis-accumulation 273 of P23HhRhoRFP protein in the rod inner segments. The rate of photoreceptor cell death in 274 P23H-RFP/+ retinas indicates that a moderate burst of degeneration begins after P14, which 275 has slowed by age P90. By comparison, the retinal degeneration and ERG phenotypes in 276 homozygous P23H-RFP/P23H-RFP mice were much more severe, such that at P90, nearly all 277 photoreceptor neurons were lost in the retinas of homozygotes. 278 279 280 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Scotopic a-wave
Age P30: * * *  Quantification of select mRNA levels from P23H-RFP/+ mouse retinas by Q-RTPCR. 285 As ER stress and the Unfolded Protein Response (UPR) have been proposed to be 286 important in neurodegeneration due to misfolded proteins in photoreceptors (47-51), we 287 examined mRNA levels for several markers of these pathways, using two different 288 "housekeeping" genes for normalization, RPL19 and HPRT. Genes whose messages we 289 quantified included those encoding BiP, CHOP, ATF6, Eif2α, PERK, DRL1 and XBP1. None 290 showed a statistically significant increase relative to both "housekeeping" genes ( Fig. 8). We 291 also quantified levels of mRNA transcribed from the Rhodopsin locus, using primer pairs that 292 either amplified both human and mouse alleles (mhRho), or the mouse allele only (mRho). Both 293 showed a decrease in message levels in the retinas of heterozygotes relative to those in the 294 WT. The mouse-specific message would be expected to be reduced by ~50%, based strictly on 295 copy number; however, the reduction was about 70%, whereas the total message derived from 296 both alleles was down about 35% suggesting a down-regulation of both Rho mRNAs by 35% 297 relative to WT through either reduced transcription or increased degradation, possibly in 298 response to the presence of aggregated protein. 299 300 P23HhRhoRFP protein was mis-accumulated in the inner segment endoplasmic reticulum. 301 To understand how the mutant P23HhRhoRFP protein is handled by the P23H-RFP/+ 302 rods, we examined the subcellular structures by transmission electron microscopy (TEM) of 303 ultra-thin retina sections following a tannic acid based staining procedure that densely stains 304 internal membranes (8). At age P14, in P23H-RFP/+ heterozygous rods, we observed distinct 305 membranous accumulations in the IS that matched the shape and morphology of the 306 fluorescent RFP+ aggregates (Fig. 9A-B). Compared to +/+ rods, where cytoplasmic 307 membranes are mostly observed in the proximal IS, the membranous accumulations in 308 heterozygous rods were primarily in the form of semi-organized stacks of folded membranes in 309 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint the distal IS. These membranes often filled the mutant IS cytoplasm, apparently distending the 310 width of the IS itself. Indeed, we measured the average maximum IS width of P14 P23H-RFP/+ 311 rods to be significantly greater compared to the WT average maximum width (Fig 9, see  312 legend). Despite these large aberrant IS membranes in P23H-RFP/+ rods, the morphologies of 313 the BB, CC and OS discs were relatively normal and indistinguishable between P23H-RFP/+ 314 315 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint and WT rods at P14, except for the length of the CC, which was significantly longer in mutant 318 rods (Fig 9A-B, see legend). 319 We also used TEM to examine the morphology of the RFP+ IS aggregates in P23H-320 RFP/+ rods at P30, where the RFP granule fluorescence appeared less organized than in P14 321 rods (Fig 4A-B). With TEM, in age P30 P23H-RFP/+ rods, we also observed accumulated IS 322 membranes; however, they did indeed appear more dysmorphic than those at P14, with 323 examples of the membranes wrapping around themselves in whorls (Fig 9C, magenta arrows). 324 As in P14 P23H-RFP/+ retinal sections, the aberrant rod IS membranes at P30 were located 325 among normally formed OS disc stacks, and the CC and BB structures were also 326 morphologically normal at P30 (Fig 9D). Unlike P14 rods, however, we observed some defects 327 in OS membrane morphology at the base of the OS in some P30 P23H-RFP/+ rods. These 328 defects included vesicular and misshapen discs, and an unbound, splayed OS axoneme (Fig  329   9C-D, green arrows). The results indicate that the mutant P23H-Rho-RFP leads to massive 330 alterations of inner segment membrane structure, accompanied by a least partial disruption of 331 outer segment morphology. 332 To determine the nature of the membranes in which the RFP fusion protein is located, 333 we used immunofluorescence with antibodies for the endoplasmic reticulum (ER) antigens 334 BiP/GRP78 (binding immunoglobulin protein/glucose-regulated protein 78, an ER lumen 335 chaperone protein) and KDEL (an ER-specific tetrapeptide folding tag), and for the Golgi 336 antigen GM130 (a Golgi-specific membrane marker). In confocal z-projections, we found that 337 BiP ER immunolabeling was co-localized with the RFP+ IS aggregates in rod cells from P14 338 P23H-RFP/+ retinal sections (Fig 10A). Compared to +/+ BiP labeling, which was largely 339 In each example rod, the inner segment (IS) is outlined with yellow lines. The IS max width in P14 P23H-RFP/+ rods is significantly greater than +/+ rods (het: 2.967 µm ± 0.508 µm (standard deviation, sd) (n=14) vs +/+: 1.974 ± 0.481 µm (sd) (n=19), P < 0.0001, unpaired t-test). Ectopic stacks of IS membranes in the swollen P23H-RFP/+ IS are marked with magenta arrows. Double-membraned autophagy compartments are marked with magenta asterisks in the mutant P23H-RFP/+ IS. In magnified views of example connecting cilia (CC) from each genotype, the length of the CC -measured by the densely stained CC membrane -is indicated. In aggregate, the length of the CC in P23H-RFP/+ rods is significantly greater than +/+ CC (het: 1.548 µm ± 0.206 µm (sd) (n=13) vs +/+: 1.27 µm ± 0.247 µm (sd) (n=11), P = 0.0066, unpaired t-test). (C) At age P30, the ectopic IS membranes in P23H/+ rods appear more dysmorphic compared to P14 (magenta arrows). In addition, some outer segments disc membranes within or adjacent to P23H-RFP/+ rods with IS defects are also disrupted and appear dysmorphic (green arrows). (D) In examples of the CC and basal OS regions of P30 P23H/+ rods, the structure of the CC and basal body remain intact despite being adjacent to ectopic IS membranes (magenta arrows); however, there is evidence that basal OS disc morphogenesis is disrupted possibly due to OS axoneme instability (green arrows).
. CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint network was unaffected in P14 P23H-RFP/+ retinas and was not co-localized with RFP+ 348 aggregates (Fig 10B). 349 We used SIM superresolution microscopy to examine the morphology of the ER and 350 Golgi more closely in individual rods in P23H-RFP/+ heterozygous and +/+ retinas at P30. We 351 added centrin immunolabeling to label the CC and BB in these SIM experiments. In P30 P23H-352 RFP/+ retinas, we again observed BiP co-localization with RFP aggregates in the IS (Fig 10 B-353 C, white arrows). The BiP+ ER lumen surrounded and was intercalated with the mislocalized 354 P23HhRhoRFP protein that appeared aggregated within the ER membranes. We also observed 355 BiP+ ER in other regions of the IS including at the BB (Fig 10 B-C, yellow arrows). In control 356 wild-type (+/+) P30 retinas, the BiP-positive ER is located throughout the IS in a reticulated 357 morphology that extends to the BB as well (Fig 10E). We observed similar KDEL+ ER 358 localization in P30 P23H/+ rods: both co-localized within RFP+ aggregates (Fig 10F, white  359 arrows) and in the BB region (Fig 10F, yellow arrows); however, KDEL labeling was more 360 punctate than BiP labeling. In P30 +/+ rods, KDEL was also localized throughout the IS and in 361 the BB region (Fig 10G). Although there is much evidence of co-localization of RFP and ER 362 marker signal, the ER markers are not uniformly distributed throughout the clumps of RFP 363 signal, and there are large sections of RFP-positive aggregates without ER marker signal. 364 We also used SIM to examine the morphology of GM130+ Golgi in P30 P23H-RFP/+ 365 rods. As before we observed the Golgi in the proximal IS of P23H-RFP/+ retinas and 366 segregated from the RFP+ aggregates and the centrin+ CC/BB. In some mutant heterozygous 367 rod IS, however, we observed smaller GM130+ Golgi membranes within the RFP+ aggregates 368 (Fig 10H, white arrows). Also, we found some examples of P30 P23H-RFP/+ rods with more 369 elaborate Golgi that extended into the distal IS and nearby RFP+ aggregates (Fig 10H, yellow  370 arrows). In SIM images of P30+/+ retina immunolabeled for GM130 and centrin, we found most 371 of the Golgi in the proximal IS and dissociated from the centrin+ cilium. Interestingly, however, 372 373 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  P23HhRhoRFP fluorescence is magenta. Co-localized BiP with RFP+ aggregates in the P14 P23H-RFP/+ retina section is marked with white arrows. GM130+ Golgi membranes do not co-localize with RFP+ aggregates and appear unaffected in the P14 P23H-RFP/+ retina compared to +/+ sections. (C) SIM z-projection images of a retina cryosection from an age P30 P23H-RFP/+ mouse that is immunolabeled for BiP (green) and centrin (white), which labels the connecting cilium (CC) and basal body (BB). RFP fluorescence is magenta. In a magnified inner segment (IS), BiP colocalization with a P23HhRhoRFP aggregate is marked with white arrows. BiP+ ER near the cilium that is not colocalized with RFP is marked with yellow arrows. (D) Within the same SIM images, the BiP+ ER tightly surrounds a P23HhRhoRFP aggregate (white arrows). (E) In age P30 +/+ SIM control images, BiP immunolabeling marks the inner segment ER network that leads to the cilium. (F) SIM z-projections of age P30 P23H-RFP/+ retinal sections immunolabeled for KDEL, another ER lumen marker (green), along with centrin (white) and RFP (magenta). KDEL puncta are more diffuse than BiP, but the localization pattern in the IS is similar; KDEL co-localization with the P23H-RFP aggregates (white arrows) and KDEL+ ER near the basal body that is not associated with RFP (yellow arrows) are labeled. (G) In age P30 +/+ SIM control images, KDEL immunolabeling (green) labels puncta throughout the inner segment. (H) SIM images of a P23H-RFP/+ retina cryosection at age P30 immunolabeled for GM130, a Golgi membrane marker (green), along with centrin (white) and RFP (magenta). Overall, The Golgi is proximally localized to the RFP+ aggregates at P30, but in some examples the Golgi network reaches the centrin+ cilium. On a subcellular scale, the Golgi membranes are generally segregated from RFP aggregates (yellow arrows), but small pieces of Golgi membrane are found co-localized with some RFP aggregates (white arrows). (I) In age P30 +/+ SIM control images of retinal cryosections, the Golgi is predominantly dissociated from the cilia; however, examples of +/+ rods with GM130+ Golgi membrane networks that reach the basal body are shown. Gray dotted arrows throughout mark a region that is magnified from the same image.
. CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint we did observe some examples of +/+ rods in which the Golgi network was more elaborate and 377 extended up to the BB (Fig 10I). 378 Finally, in our SIM images we observed both BiP+ and KDEL+ ER localized with the 379 mislocalized P23HhRhoRFP protein in the region of rod synapses in the OPL (outer plexiform 380 layer) of P30 P23H-RFP/+ retinas (Fig 11A-B). GM130-positive Golgi was not localized in the 381 OPL of P30 P23H-RFP/+ retinas (Fig 11C). 382 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Discussion 387
The P23HhRhoRFP mouse we introduce here is a potentially useful animal model for 388 adRP that can reveal the subcellular and molecular pathology of the misfolding P23H-Rho 389 mutation on the long-term health of mammalian rod neurons. Tag-RFP-T fluorescence in these 390 mice enables both gross and nanoscopic analysis of P23H-Rho ER accumulation in vivo. The 391 WT-hRho-GFP/P23HhRhoRFP dual color heterozygote demonstrates the different ways the cell 392 processes these two proteins with similar fusions, but with and without the P23H rhodopsin 393 mutation (Fig 3). WT-Rho-GFP fusion protein is restricted almost exclusively to the OS in rods 394 of the Rho-GFP heterozygotes (52), whereas the same fusion construct with a P23H mutation 395 was largely confined to the inner segment and ONL (35), as observed in the new model 396 reported here, consistent with the notion that the P23H mutation is responsible for the disruption 397 in normal trafficking, likely as a result of misfolding (18). Furthermore, with superresolution 398 microscopy, we observed P23HhRhoRFP localization proximal to the connecting cilium and 399 basal body in P23H-RFP/+ heterozygous rods and completely excluded from the OS (Figs 2-3). 400 The exclusive IS mislocalization of the P23H-Rho-RFP protein is unlike P23H-Rho localization 401 in other mouse models, in which there is some detectable OS transport despite other 402 mislocalization and degradation phenotypes (25, 27, 31, 34). 403 While both P23H-RFP/+ heterozygous and homozygous retinas had mislocalization of 404 P23H-Rho to the IS, rod neuron degeneration in the homozygotes was more severe. In the 405 heterozygotes, despite dramatic P23H-Rho mislocalization and a burst of rod cell death around 406 age P30, the progression of degeneration in P23H-RFP/+ retinas was only partial and relatively 407 slow. By comparison, in the "untagged" P23H-Rho knockin mouse line, heterozygotes lose 43% 408 of their rod population relative to WT by P63 (28); we do not observe such a severe reduction 409 until age P120 (Fig 5B). binding ER chaperone ERdj5 preserved photoreceptor survival (55). Although we found no 419 evidence for up-regulation of BiP CHOP or PERK at the mRNA level at P30 in P23H-RFP/+ 420 retinas (Fig. 8), we did not measure corresponding protein levels, or mRNA levels at later ages. 421 We did observe widespread BiP/Grp78 protein in the ER throughout the rod cytoplasm, which 422 co-localized with the ER-accumulated P23H-Rho-RFP protein (Fig. 9). 423 In contrast, in the untagged P23H-Rho knockin mouse, the inositol-requiring enzyme 1 424 (IRE1) UPR pathway was shown to be activated concurrent with increased ER-associated 425 protein degradation (ERAD) activity, whereas the PERK pathway was not activated in this 426 model (56). Another study supported a key role for protein degradation by demonstrating that 427 P23H-Rho photoreceptors were preserved by genetically over-activating the proteasome by 428 crossing the untagged P23H-Rho knockin mouse with mice constitutively overexpressing either 429 the PA28α or PSMD11 proteasomal cap protein (57). Therefore, the ER membrane expansion 430 we observed in P23H-RFP/+ rods with TEM (Fig 8) could a be compensatory ER stress 431 mechanism in response to an overload of ERAD and proteasome degradation. Such an 432 adaptation warrants further study. In TEM images of P14 P23H-RFP/+ rods, we also observed 433 double membrane autophagosome-like structures adjacent to the ER (Fig 8A), indicating an 434 autophagy component to the P23H-RFP pathology. Similar double membrane vesiculations 435 were observed in transgenic bovine-P23H-Rho Xenopus tadpoles that were exposed to light 436 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint (23), and more recently LC3-positive autophagosomes were localized adjacent to P23H-Rho 437 protein in the inner segments of these same Xenopus rods expressing bovine-P23H-Rho (58). 438 We observed a deterioration in the morphology of the accumulated ER IS membranes in 439 P23H-RFP/+ rods from age P14 to P30 with both RFP fluorescence (Fig 4) and TEM (Fig 8). In 440 addition to large IS membrane whorls in the IS at P30 (Fig 8C), the ER membrane structures 441 were more disorganized than the tight membrane stacks at P14. We also observed ER at the 442 photoreceptor synapse carrying mislocalized P23HhRhoRFP protein in the OPL of P30 P23H-443 RFP/+ retinas (Fig 10). Rhodopsin mislocalization to the synapse layer and in the ONL 444 cytoplasm was previously observed in P23H-Rho transgenic mice (31). This suggests that the 445 ER expands throughout the entire cytoplasmic space in rods by P30 as a broad response to the 446 accumulation of misfolded P23H-Rho protein in the ER. 447 Despite the expanded ER membranes filled with P23HhRhoRFP protein, the 448 morphology of the CC and the OS disc structure was normal in P23H-RFP/+ rods at age P14 449 (Fig 8A-B). At P30, the OS discs is some P23H-RFP/+ rods were slightly dysmorphic, but the 450 CC structure appeared structurally intact, albeit longer than in +/+ rods (Fig 8C-D). The CC 451 elongation P23H-RFP/+ rods is a surprising result. Such an elongation phenotype was also 452 described in knockout mouse models for 2 CC-localized proteins: male germ cell-associated 453 kinase (Mak) and Huntingtin (59, 60). In both knockouts, the elongated CC was accompanied 454 with aberrant OS morphology and Rho mislocalization. One possible cause of the elongated CC 455 in P23H-RFP/+ mutant rods may be an early imbalance of Rho expression during an early 456 ciliogenesis stage due to mutant P23H-Rho-RFP expression. 457 Our analysis of the rate of rod neuronal death using TUNEL staining of the ONL in 458 P23H-RFP mouse retinas is additional evidence that P23H-RFP/+ mutant rods undergo a long-459 term neuroprotective response. Unlike in P23H-RFP/P23H-RFP homozygous retinas where the 460 TUNEL+ rate in the ONL is elevated at P14, the rate of TUNEL+ nuclei spikes at P30 in P23H/+ 461 heterozygous retinas and returns to a level not significantly higher than +/+ retinas by P90 (Fig.  462 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint 7C). Interestingly, a similar spike in TUNEL staining was also observed in both P23H-Rho 463 transgenic rats at age P18 vs age P30 (32) and in the untagged P23H-Rho knockin mouse at 464 age P19 vs P31 (61). 465 In conclusion, our RFP fusion knockin model of adRP retinal degeneration caused by the 466 The P23H-hRho-TagRFP knock-in mice were generated the same way we previously 483 generated P23H-hRho-GFP knock-in mice (35), by gene targeting in the HPRT − embryonic stem 484 (ES) cell line AB2.2 123, which was derived from mouse strain 129SvEv, essentially as 485 described previously (41, 42). The targeting plasmid plasmid contained 5' and 3' sequences 486 identical to flanking sequences of the mouse rhodopsin gene, an intervening sequence 487 corresponding to the human Rhodopsin gene encoding the P23H mutation found in patients, a 488 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint corresponding to the C-terminal 9 residues of human rhodopsin (TETSQVAPA, the epitope for 490 the 1D4 monoclonal antibody and a putative outer segment targeting signal), a STOP codon, 491 and an endogenous polyadenylation signal, followed by an expression cassette (minigene) for 492 human hypoxanthinephosphoribosyltransferase (HPRT). The HPRT minigene was flanked by 493 loxP sites, so it could be looped out in vivo by passing through the germline of GDF-9-iCre mice 494 allele were crossed to GDF-9-iCre mice (44) to remove the HPRT minigene and screened to 506 ensure germline transmission of the correct targeted sequence without HPRT. P23H-hRho-507 TagRFP mice were extensively backcrossed to C57BL/6 mice. We validated that the knockin 508 was successful by sequencing genomic DNA from the knockin mouse. We verified expression 509 of the P23HhRhoRFP fusion by fluorescence microscopy of retinas and by immunoblotting (Fig.  510   1, Fig. 2). 511 The P23H-human-rhodopsin-RFP (P23H-RFP) knockin mice were generated by the BCM 512 Genetically Engineered Mouse core, using a strategy similar to the one previously described for 513 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. The junction between the mouse and human RHO sequences is at the SacI site in the 525 5'UTR between the transcription start sites and the translation start sites. Unlike our P23H-Rho-526 GFP knockin described previously (35), there was no lox site added to the 5'UTR, although 527 there is a lone loxP site remaining in the 3'-end following loop-out of the HPRT minigene. The 528 sequence of the P23H-hRho-TagRFPr allele was verified by Sanger sequencing. We generated 529 a new hRho-EGFP knockin mouse line with an additional C-terminal 1D4 epitope sequence tag 530 by a similar approach, using WT human rhodopsin sequence and EGFP coding sequence 531 instead of Tag-RFP-T. Generation of the original hRho-EGFP knockin line was described 532

previously (41) 533
Both P23H-RFP and hRho-EGFP-1D4 mice were extensively backcrossed to C57BL/6 534 (>10 generations). Wild-type (+/+) C57BL/6 littermates were used as controls throughout this 535 study. The following genotyping PCR primers were used for the P23HhRhoRFP knockin allele 536 (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint eye cups were cryoprotected in 30% sucrose before mounting in OCT in plastic molds and flash 565 freezing. 8 µm -10 µm cryosections were collected on poly-L-lysine coated glass slides (EMS), 566 and unfixed eye cups sections were immediately fixed with 2% PFA for 2 minutes. This light 567 fixation method was used for all immunolabeling experiments that included anti-centrin cilia 568 immunolabeling. Superior-inferior positions were marked in eyes to be used for retina thickness 569 measurements prior to enucleation to maintain proper orientation throughout fixation and 570 sectioning. 571 For immunohistochemistry, sections were blocked with either 2% normal goat serum 572 (NGS) (Fitzgerald), 2% bovine serum albumin (BSA) (Sigma), 2% fish skin gelatin (FSG) 573 (Sigma), 0.2% saponin diluted in 1x PBS, or with SUPER block: 15% NGS, 5% BSA, 5% BSA-c 574 (Aurion), 5% FSG, 0.2% saponin in 1xPBS (sections from Fig 7C-D, Fig 9, Fig 10). Blocked 575 sections were probed with 0.5 µg -2 µg of primary antibodies in the same blocking buffer 576 overnight at room temperature, protected from light. The following antibodies were used: mouse  Widefield imaging was performed on an inverted Nikon Eclipse TE2000U microscope 589 with mercury lamp excitation, a 10x objective (Nikon, Plan Fluor 10x) and imaging via a 590 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint Photometrics CoolSnap cf Photometrics digital camera (Roper Scientific) and excitation via a 591 mercury lamp using dichroic mirrors and filters for excitation and emission wavelength selection. 592 Full retina sections were generated by merging overlapping captures in Fiji/ImageJ using the 593 "Stitching" plugin (74, 75). ONL thickness was measured from these full retina section files in 594 All images were pseudo-colored and processed for clarity in Fiji/ImageJ; minimum and 614 maximum input values were adjusted maintaining a linear slope. Magnified images of rod cilia 615 were digitally straightened with the Straighten tool in Fiji/ImageJ. 616 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Mice were dark adapted overnight and anesthetized with 90 mg/kg ketamine + 14 mg/kg 619 xylazine. 0.5% tropicamide was added as a mydriatic to both eyes with 2.5% phenylephrine 620 hydrochloride and 0.5% proparacaine hydrochloride for analgesia/anesthesia. 2.5% 621 methylcellulose was used to maintain conductivity and for corneal hydration. A ground electrode 622 was inserted into the mouse forehead, and wire electrode loops were placed over each eye. We The following TEM preparation, based on (8), enhances staining and contrast of internal 641 cell membranes in rod photoreceptors. Mouse eyes for TEM were cornea punctured in the 642 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made and statistical significance analysis of calculated RQ values were carried out using GraphPad 700 Prism 8.4.3. A 95% confidence interval for a two-tailed t-Test was used as the criterion for 701 statistical significance. P values for the two-tailed t-tests are reported. 702 703 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint  . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint 43 707

Statistical analysis 708
Unpaired Student's t-tests, Two-way ANOVA with Šídák multiple comparison tests, and 709 cubic spline curve fittings were performed in GraphPad Prism® software. Each decline in ONL 710 width with age was fit to a single exponential decay as W(t) = (W(0) -W(∞))exp-(t/τ) + W(∞) 711 Where W(t) is the measured width at age, W(∞) is the plateau value to which it declines, W(0) 712 is an initial value, and τ is the time constant. The data were fit to the equation with W(0), W(∞) 713 and τ as floating fit parameters in a Marquardt-Levenberg least-squares fitting algorithm using 714 Prism®. 715 716 . CC-BY-NC 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint EY01173, R01EY026545 and R01EY031949, core grants P30EY002520 and P30CA125123, 720 MAR was supported by a grant from the Knights Templar Eye Foundation and NIH grant F32-EY027171. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted October 11, 2021. ; https://doi.org/10.1101/2021.10.11.463949 doi: bioRxiv preprint