TSC but not PTEN loss in starving cones of retinitis pigmentosa mice leads to an autophagy defect and mTORC1 dissociation from the lysosome

Understanding the mechanisms that contribute to secondary cone photoreceptor loss in retinitis pigmentosa (RP) is critical to devise strategies to prolong vision in this neurodegenerative disease. We previously showed that constitutive activation of the mammalian target of rapamycin complex 1 (mTORC1), by loss of its negative regulator the tuberous sclerosis complex protein 1 (Tsc1; also known as Hamartin), was sufficient to promote robust survival of nutrient-stressed cones in two mouse models of RP by improving glucose uptake and utilization. However, while cone protection remained initially stable for several weeks, eventually cone loss resumed. Here we show that loss of Tsc1 in the cones of RP mice causes a defect in autophagy, leading to the accumulation of ubiquitinated aggregates. We demonstrate that this defect was not due to an inhibition of autophagy initiation, but due to an accumulation of autolysosomes, suggesting a defect in the end-stage of the process causing an amino-acid shortage in cones, thereby hampering long-term cone survival. Because cells with TSC loss fail to completely inhibit mTORC1 and properly activate autophagy in the absence of amino acids, we sporadically administered the mTORC1 inhibitor rapamycin, which was sufficient to correct the defects seen in cones, further enhancing the efficiency of cone survival mediated by Tsc1 loss. Concordantly, activation of mTORC1 by loss of the phosphatase and tensin homolog (Pten) did not affect autophagy and amino-acid metabolism, leading to a more sustained long-term protection of cones. As loss of Pten, which in cones results in less robust mTORC1 activation when compared with loss of Tsc1, still affords long-term cone survival, therapeutic interventions with mTORC1 activators or gene therapy with selected mTORC1 targets that improve glucose metabolism are potential strategies to delay vision loss in patients with RP.

Retinitis pigmentosa (RP) is an untreatable inherited retinal degenerative disease that leads to blindness affecting 1:3500 people in the United States. Most individuals with RP have mutations in genes that are exclusively expressed in the night active rod photoreceptors; however, once rod death has progressed beyond a certain critical threshold, 1 secondary cone death follows leading to loss of daylight, color and highacuity vision. 2 We previously proposed that secondary cone degeneration is the result of inadequate nutrient supply caused by structural changes initiated by the loss of the overabundant rods. 1 Consequently, improving cell metabolism by activation of mammalian target of rapamycin complex 1 (mTORC1), a key kinase that regulates cell metabolism by balancing demand with supply, promoted cone survival, by increased expression of genes involved in glucose uptake, utilization, and retention enabling cones to counteract the conditions of nutrient stress. 1,3,4 Consistent with that loss of mTORC1 activity during disease accelerated cone death as cones failed to balance demand with supply. 3 Cone survival was also dependent on the persistence and strength of mTORC1 activation. Daily systemic injections of insulin, which desensitize the insulin receptor over time and reduce mTORC1 activity due to the feedback loop within the pathway, 4 did only promote cone survival for a period of 4 weeks. 1 In contrast, loss of either of the two negative regulators of mTORC1, Pten (phosphatase and tensin homolog) and Tsc1 (tuberous sclerosis complex protein 1), which circumvent the negative feedback loop, promoted cone survival for a period of up to 8 months. 3 Because loss of the direct upstream negative regulator of mTORC1, Tsc1, led to a more robust activation of mTORC1 in cones than loss of Pten, 5 many retinae with loss of Tsc1 displayed almost a wild-type distribution of cones at 2 months of age. Remarkably, cone death was brought to a halt for at least 1 month 3 in the fast retinal degeneration-1 (rd1) mouse model of RP, which harbors a mutation in the rod-specific phosphodiesterase-6-β gene (Pde6β). 6 As activation of mTORC1 affords a mutation-independent approach to prolong vision, 3 we investigated why in rd1 mice with loss of Tsc1, cone death resumed between 2 and 4 months of age. mTORC1 is a critical negative regulator of macroautophagy, 4 henceforth referred to as autophagy, and a positive regulator of cell growth. Therefore, hypertrophic giant cells and aberrant lysosomes are hallmarks of the human TSC disease pathology. 7 Recently, growth factor-mediated mTORC1 activation by the TSC complex has been shown to require dissociation of the complex from the lysosome, where mTORC1 resides and is activated. 8 Interestingly, lack in amino acids, which leads to inactivation and loss of lysosomal mTORC1, results in lysosomal localization of the TSC complex to inactivate any residual mTORC1. 9 Because of that amino-acid-deprived TSC-null cells fail to completely inactivate mTORC1 and properly activate autophagy, 9 a process important when balancing resources under nutrient stress conditions. During autophagy cytoplasmic cargo is delivered to double-membrane vesicles called autophagosomes, which fuse with the lysosome to form autolysosomes, where content is degraded and amino acids are freed. 10 The process is mediated by a wide range of proteins, some of which are constantly turned over, hence their accumulation is a sign of an impaired process. 11 To study if loss of Tsc1 in rd1 cones, while promoting cone survival through strong activation of mTORC1, may have simultaneously introduced an unwarranted secondary problem by preventing autophagy to occur properly, 9 we analyzed the process of autophagy at 2 months of age, a time point just before when cone death resumes.

Results
Impaired autophagy upon loss of Tsc1 in wild-type mice leads to a progressive decline in cone function and conespecific proteins. To study the long-term effect of Tsc1 loss in cones of mice with retinal degeneration, we first analyzed the effect of its loss in wild-type mice. Mice carrying a conditional knockout allele for Tsc1 were crossed to the same cone-specific Cre-driver line 12 used in our previous studies (Tsc1 c/c _M-opsin-Cre, henceforth referred as Tsc1 c/c Creor Tsc1 c/c Cre + ; in all cases Cre + denotes deletion in cones of the gene indicated). Tsc1 loss was verified by increased phosphorylation of the mTORC1 targets ribosomal protein-S6 (p-S6) and the eukaryotic translational initiation factor 4Ebinding protein (p-4EBP1) 4 (Figures 1a and b). Electroretinogram (ERG) recordings on mice with Tsc1 loss showed a strong reduction in cone function over a period of 1 year with a statistically significant decline as early as 3 months of age (Figures 1c and d). We previously showed that the Cre-driver line does not affect cone function up to 1 year of age, indicating that the decline in cone function was due to loss of Tsc1 rather than expression of CRE. 13 Furthermore, rod function at 1 year of age was not affected upon deletion of Tsc1 in cones (Supplementary Figure 1a). In agreement with the decline in cone function, a cone count, using an antibody directed against the cone-specific protein cone arrestin, showed significant loss of cone arrestin-positive cones over time (Figures 1e-g and Supplementary Figure 1b). Impaired autophagy due to constitutive activation of mTORC1 has been shown to affect cellular function and survival in various tissues. 14,15 To examine the status of autophagy, we evaluated the expression of p62 (also known as sequestosome-1), a receptor protein involved in the recognition and targeting of autophagic cargo, such as ubiquitinated proteins, to the lysosome for degradation. 16,17 We found accumulation of both p62 and ubiquitin in cones of Tsc1 c/c Cre + mice, suggesting a defect in the clearance of ubiquitinated proteins (Figures 1h and i). Western blot analyses with retinal protein extracts did not confirm these findings (Supplementary Figure 1c), likely because cones account only for 3% of all retinal cells, complicating the detection of changes in the levels of ubiquitously expressed proteins or proteins that are preferentially expressed in other retinal cells (Supplementary Figure 1d). 18,19 Loss of Tsc1 in cones of RP mice causes an accumulation of autolysosomes. The findings in wild-type mice led us to assesses if loss of Tsc1 also impairs autophagy in RP mice (rd1-Tsc1 c/c Cre + ) where cones are subject to conditions of nutrient deprivation. 1,3 Similar to our observations in wildtype mice, rd1-Tsc1 c/c Cre + mice showed an accumulation of p62 and ubiquitin aggregates in cones (Figure 2a). To test if autophagy initiation was inhibited, we injected rd1-Tsc1 c/c Cre − and Cre + littermates subretinally at birth with a recombinant adeno-associated virus (rAAV9) that expresses a tandem-tagged mCherry-GFP-LC3 gene. 11 LC3 (microtubule-associated protein-light chain 3) is part of the autophagosomal membrane and remains associated with it even after fusion with the lysosome. Therefore, besides assessing for autophagy initiation, the vector also allows monitoring autophagic flux since GFP is pH sensitive and is quenched by the low pH of the lysosome. 11 At 2 months of age, we found that while the number of autophagosomes was similar between Cre − and Cre + littermates, there was a significant increase in the number of autolysosomes in rd1-Tsc1 c/c Cre + cones (Figures 2b and c). This suggests that loss of Tsc1 caused a defect in the clearance of autolysosomes, as autophagy was functional up to the stage of autolysosome formation. While the data contradict the notion that activated mTORC1 inhibits autophagy initiation through direct phosphorylation of Ser757 on Unc-51-like autophagyactivating kinase-1 (ULK1), a protein that is part of the autophagy initiation complex, 20,21 they are in agreement with a recent report that showed that unlike proliferating cells, postmitotic neurons with TSC loss maintain autophagy through an adenosine monophosphate-activated protein kinase (AMPK)-dependent phosphorylation 22 of ULK1 at Ser555. Consistent with that we observe increased phosphorylation in rd1-Tsc1 c/c Cre + cones of both ULK1 sites, Ser757 and Ser555, as well as increased phosphorylation of AMPK (Figure 2d), suggesting that the mTORC1-dependent inhibition of ULK1 was overridden by AMPK-dependent activation of ULK1.
In agreement with an increase in the number of autolysosomes, other proteins required for autophagosome formation such as autophagy gene 12 (ATG12), and autophagy-linked FYVE protein (ALFY), a scaffold protein implicated in the selective degradation of ubiquitinated proteins, 23,24 were also  (Figures 3b and c), corroborating the increase in the number of autolysosomes. This increase in the number of lysosomes coincided with increased nuclear expression of the forkhead box protein O3 (FOXO3A), a transcription factor that regulates the transcription of various autophagy-related genes ( Figure 3d). 25,26 The upregulation of ATG12, ALFY and accumulation of p62 as well as increased expression of FOXO3A was lost upon concurrent removal of Tsc1 and Raptor (regulatory-associated protein of mTOR), suggesting that the increase was mTORC1 dependent (Supplementary Figure 2).
In summary, our data suggest that the defect in autophagy does not arise from autophagy initiation, autophagosome maturation or fusion with the lysosome. As mTORC1 is a negative regulator of the transcription factor EB (Tfeb), a master regulator of lysosomal enzymes, the increased accumulation of autolysosomes could result from a deficiency of lysosomal enzymes due to increased mTORC1 activity. 22,27 Tsc1 loss induces a shortage of free amino acids in cones. The autophagy defect in rd1-Tsc1 c/c Cre + cones could cause an imbalance in cellular homeostasis depleting the cell of free amino acids. To test this notion, we exploited the mechanism of amino-acid sensing by mTORC1. Under aminoacid-replete conditions, mTORC1 is recruited by the Rasrelated GTPases to LAMP2-containing compartments where it encounters Ras homolog enriched in brain (Rheb) for activation. 28 Therefore, colocalization of mTORC1 with LAMP2 can be used to assess if the cell has a sufficiency of amino acids. We observed at 2 months of age in Cre − cones distinct focal mTOR staining that was strongly associated with LAMP2. However, in age-matched rd1-Tsc1 c/c Cre + cones mTOR staining was diffuse showing almost no colocalization with LAMP2 ( Figure 4a). Because mTOR/LAMP2 colocalization was lost in rd1-Raptor c/c Cre + retinae, but maintained in rd1-Rictor c/c Cre + retinae, the colocalization was reflective of mTORC1 staining at the lysosome (Figure 4b).
To test if the absence of mTOR/LAMP2 colocalization in rd1-Tsc1 c/c Cre + cones was caused by a shortage of amino acids, we explanted retinae from rd1-Tsc1 c/c Cre + mice and incubated them in glucose-rich and glucose-free Dulbecco's modified Eagle's media (DMEM). In both cases, short-time  Figure 4c). Similarly, a single intraperitoneal injection of rapamycin, an allosteric mTORC1 inhibitor, was also able to restore mTOR/LAMP2 colocalization in rd1-Tsc1 c/c Cre + cones, but not in rd1-Tsc1 c/c Raptor c/c Cre + cones (Figure 4d). Taken together, the data suggest that loss of Tsc1 in cones of RP mice induces an imbalance in the supply and demand of amino acids, which could contribute to the demise of cones see between 2 and 4 months of age.
Rapamycin reverses the autophagy defect and improves cone survival upon Tsc1 loss. The restoration of colocalization between mTOR and LAMP2 in rd1-Tsc1 c/c Cre + cones upon rapamycin administration (Figure 4c) indicates that autophagy was restored resulting in the release of free amino acids. To test whether the p62 and ubiquitin aggregates or the lack of free amino acids contribute to cone death in rd1-Tsc1 c/c Cre + mice, we performed a long-term rapamycin treatment. Our previous study showed that rd1-Tsc1 c/c Cre + mice display a decline in cone survival between 2 and 4 months of age. 3 Similarly, removal of Tsc1 in cones of wildtype mice (Tsc1 c/c Cre + ) causes loss of cone arrestin expression by 4 months of age. Because loss of mTORC1 activity did not affect cones in a wild-type background but accelerated cone death during disease, 3,13 we first administered rapamycin to Tsc1 c/c Cre + mice, where repeated injections between 1 and 4 months of age (18 in total) did clear p62 aggregates and also restored cone arrestin expression (Figures 5a and b).
In rd1-Tsc1 c/c Cre + mice where mTORC1 activity is critical to promote cone survival, we first tested the effect of rapamycin between 1 and 2 moths of age, a time window in which cone survival remains stable at around 78%. 3 While one injection of rapamycin at 2 months of age causes an increase in free amino acids, it was not able to clear p62 and ubiquitin aggregates (Supplementary Figure 3a). This is because accumulation of p62 and ubiquitin occurs over time because of buildup of autophagic cargo and may require more sustained induction of autophagy for complete clearance. In this regard, repeated injections between 1 and 2 month of age at an interval of 5 days were sufficient to clear p62 and ubiquitin aggregates without affecting cone survival ( Supplementary Figures 3a and b). The 5-day interval period was based on the time window it took for p-S6 levels to almost fully recover in cones of rd1-Tsc1 c/c Cre + mice after one rapamycin injection ( Supplementary Figures 3c and d). Extended injections of rapamycin (18 in total) up to 4 months of age, however, led to a drop in cone survival when compared with vehicle-injected mice (Figure 5c), suggesting that mTORC1 was inhibited too often over the 3 months time period. We therefore performed two additional treatment regimens reducing the frequency of administration by threefold each time. This led to a dose-dependent increase in cone survival with six injections being at par with vehicle-treated mice and two injections showing a significant improvement in cone survival by 4 months of age (Figures 5c and d). Interestingly, none of the treatment regimens had a negative effect on cone survival in Cremice (Supplementary Figure 3e), suggesting that the amount of mTORC1 activity maintained was sufficient to not alter the natural course of the disease. Postinjection analysis of p62 aggregates at 4 months showed that neither two nor six injections of rapamycin were not sufficient to clear p62 aggregates ( Supplementary  Figure 4a), suggesting that clearance of p62 is not required to improve cone survival. Consistent with that one injection of rapamycin was sufficient to facilitate autophagy and release free amino acids that directed mTOR to the lysosome for at least 14 days (Supplementary Figure 4b).
Increased mTORC1 activity by loss of Pten is more beneficial for long-term cone survival. The TSC complex has recently been shown to be required for full inactivation of mTORC1 under amino-acid-deprived conditions. 9 Because loss of Pten also promoted cone survival 3 through constitutively activated mTORC1, albeit to a lesser extent than loss of Tsc1 3 due to less activation of mTORC1 in cones, 5 we investigated the effect of Pten loss on autophagy, mTORC1 localization and long-term cone survival. Interestingly, in rd1-Pten c/c Cre + retinae we found no accumulation of p62 or ubiquitin in cones at 2 months of age (Figure 6a). Concordantly, mTOR/LAMP2 colocalization in cones was similar between Cre − and Cre + littermates (Figure 6b), with a less (c) Representative immunofluorescence on retinal flat mounts for LAMP1 and LAMP2 (red signal). Cones were detected by cone ARRESTIN staining (green signal). (d) Immunofluorescence analysis of FOXO3A (red signal). Cones were detected by SW OPSIN staining (green signal). In all panels blue signal is nuclear DAPI. Scale bars: 20 μm strong increase in lysosomes per cone when compared with loss of Tsc1, as assessed by LAMP1-positive punctae (Figure 6c). In agreement with these findings, we did not observe an apparent increase in ULK1 phosphorylation at both sites nor in AMPK phosphorylation (Supplementary Figure 5a) when compared with loss of Tsc1, suggesting that autophagy is maintained at a steady state.
In wild-type mice, we observed only occasional accumulation of p62 in cones of Pten c/c Cre + mice when compared with mice where Tsc1 was removed in cones (Supplementary Figure 5b). This accumulation was not detrimental to cone function and survival up to 1 year of age ( Supplementary Figures 5c and d). Finally, evaluation of long-term cone survival at 1 year of age revealed a twofold increase in the number of cones in rd1-Pten c/c Cre + mice when compared with Crelittermates.
Cone survival was also significantly higher when compared with loss of Tsc1 (Figures 6d and e), which did not show any protective effect at 1 year of age in our previous study. 3 A linear regression analysis comparing the decline of the protective effect between loss of Tsc1 and loss of Pten showed that while the initial effect upon loss of Pten is less robust, the decline over time is not as steep when compared with loss of Tsc1 (Figure 6f). In summary the data indicate that any increase in mTORC1 activity is beneficial for long-term cone survival as long as autophagy is allowed to progress normally.

Discussion
Here we set out to understand why loss of Tsc1 in cones of rd1 mice promotes initially robust cone survival that remains stable for over a month and then leads to a decline in the number of surviving cones. We found that loss of Tsc1 results in the accumulation of autolysosomes causing a shortage of free amino acids in cones, which could have ultimately contributed to cone death. Similar to findings in other neurons, 13,22 cones with loss of TSC use an AMPKdependent mechanism to maintain autophagy flux, despite mTORC1-mediated inhibition of autophagy initiation. This contrasts the finding in non-neuronal cells where loss of TSC completely inhibits autophagy. 22 To compensate for the lack of autophagy and keep up with the increased demand for protein degradation due to the increase in protein synthesis, nonneuronal TSC-null cells appear to increase proteasomal activity. 29 Consequently, non-neuronal TSC-null cells in which the proteasome is inhibited suffer from an intracellular shortage of free amino acids. While we did not analyze the status of the proteasome, cones in rd1 mice with Tsc1 loss increase autophagy similar to other neurons likely to compensate for the increase in protein sysnthesis. 22 Consistent with an increase in autophagy in cones, we found increased nuclear localization of FOXO3A, a transcription factor that regulates the expression of autophagy and lysosomal-related genes.
The accumulation of autolysosomes seen in rd1-Tsc1 c/c Cre + cones indicates a failure to digest the autolysosomal content, which could be caused by a lack of sufficient lysosomal enzymes. However, it remains to be determined if this was due to mTORC1-mediated inhibition of Tfeb. 27,30 Nonetheless, the accumulation of autolysosomes and p62-positive aggregates may provide an explanation for the observed neuronal dysfunction in TSC. It may also explain why these neurons are more sensitive to stress, as they cannot dilute undigested aggregates through cell division. 7,31,32 In our case, the added stress condition to the loss of TSC in cones was the disease condition itself, which leads to a nutrient shortage in cones. Thus, lack of amino acids in cones of rd1 mice upon loss of Tsc1 may be a diseasespecific condition not seen in other neurons with TSC loss. As two administrations of rapamycin were sufficient to further improve cone survival without clearing p62 aggregates, our experiments suggest that the lack of amino acids may have been a more critical factor for cone death than the accumulation of p62 aggregates.
Recent studies have shown that under various stress conditions, TSC is required to localize mTORC1 away from the lysosome for complete shut off. 9,33 Thus, in proliferating cells lacking TSC, mTORC1 is still localized at the lysosome and activity is maintained even under conditions of amino-acid deprivation. 9 Consistent with that, we found 20% of mTORC1 colocalized with LAMP2 in rd1-Tsc1 c/c Cre + cones at 2 months of age. The difference in the extent of mTORC1 localization at the lysosome between the studies may be attributed to differences in experimental time windows, in vivo versus in vitro, and/or differences in mTOR regulation between proliferating cells and neurons. 22 While in our study the reduced mTORC1/LAMP2 colocalization indicates a shortage of amino acids, the 20% of active mTORC1 localized at the lysosome is likely causing the autophagy defect. In contrast to Tsc1-null cells, Pten-null cells are able to turn off the mTORC1 activity, 9 because of the presence of a functional TSC. This explains why Pten loss in cones results in a more sustained long-term survival effect. Consistent with that in our previous study, we found fewer p-S6-positive cones upon loss of Pten than upon loss of Tsc1. 3 Thus, rd1-Pten c/c Cre + cones, where S6 phosphorylation was not detected, may be the ones where the mTORC1 activity was turned off intermittently to induce autophagy.
In summary, our data show that mTORC1-mediated cone survival is directly dependent on the magnitude of mTORC1 activation. Thus, treatments with mTORC1 activators that do not interfere with TSC or gene therapy with mTORC1 target genes should be beneficial for long-term cone survival as even moderate activation of mTORC1 by loss of Pten affords longterm protection.  Animal Care and Use Committees of the University of Massachusetts Medical School. Animals were maintained on a 12-h light/12-h dark cycle with unrestricted access to food and water. Lighting conditions were kept constant in all cages, with illumination ranging between 10 and 15 lux. The Pten c/c , Tsc1 c/c , Raptor c/c , Rictor c/c mice and the cone-specific Cre line have all been described previously. 12,[34][35][36][37] Genotyping was performed as described in the original publications. In all instances, Cre + and Cre − littermates were used for analysis. All mice were genotyped for the absence of the rd8 allele with mutation in the Crumbs 1 gene 38 and none of the mice analyzed were albino.
Rapamycin administration. Rapamycin was diluted in 50% ethanol to 10 mg/ml. Before injection, the solution was diluted to a concentration of 2 mg/ml in 50% ethanol. All mice were treated with 2 mg/kg body weight of rapamycin by intraperitoneal injections. Mice were either injected at 5-day intervals beginning at P28 until 2 or 4 months of age or at P28, P35, P49, P63, P77 and P105 for the sixinjection regimen, or at P28 and P90 for the two-injection regimen. Fifty percent ethanol was used as the vehicle control.
Retinal explant cultures. Retina was dissected free from other ocular tissue in PBS and then incubated either in regular DMEM media or glucose-free DMEM media. Fetal calf serum was added at 10% (vol/vol) in both cases. Incubation was performed for 2 h at 37°C and 5% CO 2 . Thereafter, retinae were fixed and processed for antibody staining as described.
Autophagy flux, mTOR and LAMP quantifications. The mCherry-GFP-LC3 vector 39 (Plasmid: Claudio Hetz; packaging: UMass Vector Core, rAAV9) was injected subretinally at birth (1 μl of 1 × 10 13 gc/ml) as described previously. 40 Retinae were harvested at 2 months of age and processed as described. Images from Cre − and Cre + retinae were acquired (100 × ) at the same exposure and manual counting of the GFP and mCherry spots was performed in a blind manner. The same method was used to assess mTOR and LAMP2 colocalization as well as for LAMP1/2 counting. All images were acquired on a Leica DM5500 fluorescence microscope (Leica, Wetzlar, Germany).
Electroretinography. ERG was performed using the Espion E3 console in conjunction with the ColorDome (Diagnosys LLC, Lowell, MA, USA) as described previously. 13 Histological methods. Antibody staining on retinal cryosections and retinal flat mounts were performed as described previously, 3,40 except that Triton was replaced by 0.1% saponin for staining procedures involving autophagy and lysosomal proteins. All antibody stainings were performed in PBS buffer. The following primary antibodies and concentrations were used: rabbit α-p-S6 (Ser240) (1:300; catalog no. 5364); rabbit α-p-4EBP1 (1:300; catalog no. 2855); rabbit α-FOXO3A (1:300; catalog no. 12829); rabbit α-mTOR (1:500; catalog no. 2983); rabbit α-p-ULK1 (Ser757) (1:300; catalog no. 6888); rabbit α-p-ULK1 (Ser555) (1:300; catalog no. 5869); rabbit α-p-AMPKβ1 (Ser182) (1:300; catalog no. 4186); rabbit α-ATG12 Quantification of cone survival. Retinal flat mount images for the cone arrestin expression analyses were acquired by tiling individual images taken at × 16 over the entire retinal surface area with an automated scanning stage. It must be noted that dormant cones that no longer express cone arrestin are not detected by this method and hence cone arrestin only serves as a proxy to estimate the number of healthy cones. Quantification in wild-type mice: retinae were divided into two sectors with radii of 1 and 2 mm, respectively (Figure 1e). Cones were counted manually in four squares per sector, each square measuring 40 000 μm 2 , to determine the average cone density per sector and genotype (cones per mm 2 ). Quantification in rd1-mutant mice: cone survival was evaluated by our previously described method by calculating the surface area of the retina that is covered by the cone arrestin signal to extrapolate the percentage of cone survival. 1,3,5 Colocalization between the cone arrestin staining and the retinal surface area was calculated using the CoLocalizer Pro software (Colocalization Research Software, Switzerland, Japan). 41 We have previously validated the cone survival quantification obtained by this method with actual cone counting and found it to be a reliable estimate of cone survival during degeneration. 3 Western blot analysis. Western blot analysis was performed as described previously 3 with the following modifications. For the western blots depicted in Supplementary Figure 1, the membrane was blocked for 1 h at room temperature in 5% fat-free dry milk powder, incubated with primary antibody overnight at 4°C, washed three times for 20 min each wash at room temperature, incubated with an HRP-coupled secondary antibody (1:10 000; Santa Cruz Biotechnology Inc.) for 2 h, and washed three times for 20 min (each wash at room temperature). The signal was detected with SuperSignal West Dura (Pierce Biotechnology, catalogue no. 35076, Thermo Fisher Scientific, Waltham, MA, USA). All incubations were performed in PBS in the presence of 0.1% Tween-20 and 5% fat-free dry milk powder. The following primary antibodies were used: rabbit α-p-S6 (1:1000; catalog no. 5364), rabbit α-p-4EBP1 (1:1000; catalog no. 2855) and rabbit α-S6 (1:1000; catalog no. 2217) from Cell Signaling Technology; guinea-pig α-p62 (1:500; catalog no. GP62-C) from Progen; rabbit α-CRE (1:2000; catalog no. 69050-3) from Novagen (EMD Millipore); mouse α-β-tubulin (1:2,000; catalog no. 8328) from Sigma-Aldrich; and mouse α-ubiquitin HRP conjugate (1:10 000; catalog no. 14049) from Cell Signaling Technology. For the western blots depicted in Supplementary  Figure 3, instead of fat-free milk powder, Odyssey Blocking Buffer (LI-COR) (catalog no. 927-40000) from LI-COR (Lincoln, NE, USA) was used for blocking and antibody incubations. No detergent was added during blocking and antibody incubations. Primary antibodies used were same as above, while secondary antibodies were infrared dye-conjugated (1:10 000; LI-COR). Membrane was scanned on an Odyssey Infrared Scanner from LI-COR to detect signal.
Statistics. The Student's t-test was used for statistical analyses. P-values o0.05 were considered statistically significant. All error bars represent the S.E.M.