RACK1 modulates polyglutamine-induced neurodegeneration by promoting ERK degradation in Drosophila

Polyglutamine diseases are neurodegenerative diseases caused by the expansion of polyglutamine (polyQ) tracts within different proteins. Although multiple pathways have been found to modulate aggregation of the expanded polyQ proteins, the mechanisms by which polyQ tracts induced neuronal cell death remain unknown. We conducted a genome-wide genetic screen to identify genes that suppress polyQ-induced neurodegeneration when mutated. Loss of the scaffold protein RACK1 alleviated cell death associated with the expression of polyQ tracts alone, as well as in models of Machado-Joseph disease (MJD) and Huntington’s disease (HD), without affecting proteostasis of polyQ proteins. A genome-wide RNAi screen for modifiers of this rack1 suppression phenotype revealed that knockdown of the E3 ubiquitin ligase, POE (Purity of essence), further suppressed polyQ-induced cell death, resulting in nearly wild-type looking eyes. Biochemical analyses demonstrated that RACK1 interacts with POE and ERK to promote ERK degradation. These results suggest that RACK1 plays a key role in polyQ pathogenesis by promoting POE-dependent degradation of ERK, and implicate RACK1/POE/ERK as potent drug targets for treatment of polyQ diseases.

Introduction A class of neurodegenerative disorders are cause by the expansion of CAG repeats within associated genes, resulting in polyglutamine (polyQ) insertions. There are currently nine types of polyQ diseases, including spinocerebellar ataxias (SCAs) and Huntington's disease (HD) [1,2]. One common feature of these polyQ diseases is the formation of polyQ oligomers and aggregates in the cytosol and nucleus [1,3,4]. In both human patients and animal models of these diseases, polyQ structures interrupt a variety of cellular functions, including transcriptional regulation, the endoplasmic reticulum (ER), and synaptic dynamics [5][6][7]. A number of factors were found to reduce the severity of polyQ diseases, primarily by reducing polyQ levels [7][8][9][10][11], but the mechanisms by which polyQ tracts lead to neural degeneration remain unknown.
Receptor for activated C kinase (RACK1) belongs to the WD repeat family of proteins [23] and functions as a scaffold protein to regulate multiple cellular processes, including translation, immunity, apoptosis, and cancer progression [24][25][26][27][28]. In Arabidosis thaliana, RACK1 is involved in regulating MAPK activity [29,30]. In mammalian cells, RACK1 serves as an adaptor to help activate MAPK JNK (c-Jun N-terminal protein kinase) [31]. RACK1 has been shown to function downstreatm of p38b MAPK in helping to clear aggregates of polyubiquitylated proteins from thoracic muscles of aging flies, thereby promoting proteostasis [32]. RACK1 also reduces cellular toxicity associated with protein aggregates. When overexpressed, it localizes to polyQ aggregates and protects cells from polyQ-induced neurodegeneration [33,34]. However, in these latter studies they overexpressed human RACK1 in fly models of MJD rather than the endogenous fly RACK1, and the suppression was weak. Moreover, the mechanisms by which RACK1 regulates polyQ toxicity have not been characterized, and it remains unknown whether MAPK signaling is involved in this process.
Here we established a model of polyQ disease in the Drosophila eye, and conducted a forward genetic screen to investigate the mechanisms involved in polyQ diseases pathogenesis. Strikingly, loss of RACK1 alleviated polyQ-induced cell death without affecting the formation or clearance of protein aggregates. We then performed a genome-wide RNAi screen to identify genes that could modify the polyQ/rack1 RNAi phenotype. Knockdown of the E3 ligases, POE and KCMF1, further suppressed polyQ-induced cell death, resulting in nearly wildtype looking eyes. By contrast, overexpression of POE abolished the suppressive effects of rack1 mutation. Further, RACK1 regulated ERK levels post-translationally, as expected, and ERK was indispensable for the role of RACK1 in polyQ-induced neurodegeneration. Finally, we found that RACK1 regulated ERK level by promoting the interaction between ERK and POE. As previous studies found that the ERK pathway is inhibited in polyQ disease models, we hypothesize that the RACK1/POE/ERK pathway plays an important role in polyQ disease pathogenesis, and that this pathway should be considered a target for treating polyQ diseases.

Mutating the gene rack1 alleviated polyQ-induced cellular toxicity
Several polyQ disease models have been established in flies by expressing disease-associated proteins with expanded polyQ repeat or polyQ chains alone via the GAL4/UAS system [35][36][37][38]. However, to conduct a forward genetic screen to investigate the mechanisms involved in polyQ diseases pathogenesis, we established a simplified model of polyQ disease in the Drosophila eye. In this model, we used the GMR promoter to express 63 CAG repeats tagged with HA (63Q) in the fly eye. The result was loss of pigment and retinal cells immediately after eclosion (S1A and S1B Fig). When 63Q was expressed in third instar eye discs, it initially localized to the cytosol (in anterior cells), but over time gradually formed aggregates (in posterior cells). By the pupal stage, most 63Q signal was detected in aggregates (S1C Fig). By contrast, 31Q localized to the cytosol in both the third instar eye disc and the pupal eye (S1C Fig). We then subjected this model to EMS mutagenesis to screen for genetic modifiers of 63Q-induced photoreceptor cell death. Because genes involved in polyQ pathogenesis could be essential, we combined the polyQ model with the "ey-flp/hid" system to generate flies in which the EMSinduced mutation was homozygous in the eyes [39], but heterozygous in the rest of the animal. Briefly, the ey-flp/hid system combines GMR-hid and the FLP/FRT system. Ey-flp induces mitotic recombination of mutated FRT containing chromosome arms only in the compound eyes. GMR-hid ensures that all eye cells are killed, except for those that are homozygous for your introduced mutation. This allows one to phenotypically screen homozygous loss-of-function mutations in the eye in the F1 generation. Using this strategy, we screened chromosomes 2L, 2R, 3L, and 3R for suppressors of 63Q-induced cell death (Figs 1A, S1D and S8A).
By screening~100,000 flies for each chromosome arm, we identified~30 alleles that strongly suppressed polyQ protein-induced cell death. We then excluded mutants that affected transcription mediated by the GMR promoter by checking if these mutants reduced GFP levels of GMR-GFP flies. The remaining 14 alleles belonged to 4 complementation groups. One complementation group included a single allele, which was homozygous lethal and a strong suppressor of GMR-63Q induced cell death.
Using deficiency mapping and genomic DNA sequencing, we located this mutation to the rack1 genomic locus, which encodes the scaffold protein RACK1 (Receptor for activated C kinase 1). We therefore name this allele rack1 s12 . This allele contains an RNA splicing mutation at the donor (5') site (GT to GA) between the second and third exon (Fig 1B), which disrupts RACK1 expression (S2A Fig). A previously reported null allele of rack1, rack1 1.8 , also suppressed 63Q-induced cell death (Figs 1A and S8A). Moreover, expressing wild-type rack1 via the GMR promoter restored polyQ-induced cell death in rack1 s12 mutants (Figs 1A and S8A). Finally, knocking down rack1 by RNAi suppressed 63Q-induced cell death, although to a lesser extent, whereas expressing rack1 RNAi alone did not affect eye morphology (Figs 1A and S8A). These results confirmed that rack1 mutation alleviated polyQ protein-induced cell death.
To determine whether rack1 modulates specific polyQ disease models, we examined the effect of rack1 in models of Machado-Joseph disease (MJD) and Huntington's disease (HD) [40,41]. Expressing a truncated form of Ataxin-3 with 78 CAG repeats (MJD-Q78) in the fly eye caused a dramatic loss of photoreceptor cells and rhabdomere structures, which are tightly-packed microvilli required for phototransduction [41]. Knocking down rack1 in this context decreased the amount of photoreceptor cells loss [40] (Figs 1C and S2B and S1 Table). To evaluate photoreceptor function, we performed electroretinography (ERG) to measure electrical responses to light stimulation. At day 1, flies expressing MJD-Q78 exhibited decreased amplitude and loss of on-and off-transients, indicating impaired phototransduction and synaptic transmission [42]. Knocking down rack1 partially rescued ERG amplitude and restored ERG transients (Fig 1D and 1E). Similarly, knocking down rack1 alleviated the neurodegeneration associated with mutated Huntington (Htt). Expressing GFP-tagged N-terminal Htt with 96 CAG repeats (Htt96Q) in the eye progressively reduced the number of photoreceptor cells and rhabdomeres. Knocking down rack1 suppressed this retinal degeneration. In 10 day-old flies, GMR>Htt96Q ommatidia generally had only one photoreceptor cell, whereas GMR>Htt96Q ommatidia with rack1 RNAi expression had~4 intact photoreceptor cells (Figs 1F and S2C).
To exclude the possibility that loss of rack1 suppressed polyQ-induced cell death only in photoreceptor neurons, we used elav-GAL4 to express MJD-Q78 in all neurons. This greatly shortened lifespan [40]. elav>MJD-Q78 flies survived longer when rack1 was knocked down, indicating that loss of RACK1 suppressed MJD-Q78-induced neurodegeneration (Fig 1G). We next asked whether loss of rack1 could suppress cell death in models of other neurodegenerative diseases. We modeled retinitis pigmentosa by expressing mutant rhodopsin (Rh1 G69D ) and tauopathy by expressing Tau V337M [43,44]. Importantly, knocking down rack1 did not affect Rh1 G69D -or Tau V337M -mediated retinal cell death (S2D Fig). Taken together, disruption of RACK1 specifically suppressed polyQ-induced neurodegeneration.
RACK1 is a scaffold protein that helps to assemble a variety of signaling molecules involved in multiple processes, including PKC activation, protein translation, and apoptosis [31,32,[45][46][47]. To reveal the mechanism by which RACK1 affects 63Q cellular toxicity, we asked whether RACK1 functions through known pathways in this context. RACK1 promotes the release of eukaryotic translation initiation factor 6 (eIF6) from the 60s ribosome by recruiting PKC (protein kinase C) to eIF6 [45]. However, knocking down, knocking out, or overexpressing pkc53e, which encodes the major fly PKC, failed to suppress 63Q-induced cell death (S3A Fig), indicating that the classic PKC pathways were not involved in suppressing polyQ toxicity in rack1 mutants. RACK1 also regulates ribosomal quality control by interacting with ZNF598 [46]. However, knocking down znf598 did not alleviate 63Q toxicity (S3A Fig). These results indicate that the loss of rack1 did not suppress polyQ toxicity by regulating translation.
RACK1 interacts with the p38/JNK pathway to regulate apoptosis [31,32,47]. However, knocking down or overexpressing JNK, P38a, or P38b did not affect 63Q toxicity or the ability of rack1 RNAi to suppress polyQ toxicity (S3B Fig). Considering that apoptosis is downstream of JNK, we inhibited apoptosis by knocking down apoptosis-related proteins (rpr, hid, and grim) and caspases (dronc, dcp-1, and decay) or by over-expressing Inhibitor of apoptosis (DIAP1). This did not affect 63Q-induced cell death, either with or without rack1 RNAi (S3C Fig). We also assessed the hypoxia pathway and immunity-related proteins, but did not see effects on 63Q toxicity. Therefore, we hypothesized that RACK1 suppresses polyQ-induced neurodegeneration via a novel pathway.

RACK1 genetically interacted with the E3 ligases POE and KCMF1
To determine the pathway through which rack1 modulated polyQ pathology, we conducted a genome-wide RNAi screen to identify genes that modified the polyQ/rack1 RNAi phenotype. We expressed~6,000 RNAi lines individually in compound eyes that also expressed rack1 RNAi and 63Q. We identified more than 300 genes that restored 63Q toxicity when knocked down, but most also induced cell death when expressed alone or combined with 63Q, suggesting effects independent of RACK1. We also identified 14 RNAi lines that further suppressed 63Qinduced cell death in combination with rack1 RNAi (Table 1). Among these 14 were two E3 ligases, POE (pushover or purity of essence) and KCMF1 (zinc-finger protein Potassium Channel Modulatory Factor 1), which each further suppressed polyQ-induced cell death in combination with rack1 RNAi , but only slightly suppressed 63Q toxicity when knocked down alone. In contrast, no cell death was detected when poe RNAi or kcmf1 RNAi were expressed alone (Figs 3A and S8B and S1 Table).
Previous studies have shown that POE and KCMF1, together with another E3 ligase UFD4 (Ubiquitin fusion-degradation 4-like), counteract the deubiquitinase USP47 (Ubiquitin Specific Peptidase 47) to destabilize ERK [22]. However, knocking down ufd4 did not affect 63Q toxicity. This may be because UFD4 had a much weaker effect on ERK levels compared with POE or KCMF1 (Figs 3A and S8B and S1 Table). We then overexpressed USP47 in the compound eyes of GMR-63Q flies and found that USP47 suppressed polyQ-induced cell death. Moreover, the suppression of 63Q toxicity mediated by rack1 knock down was greatly enhanced by overexpressing USP47 (Figs 3A and S8B). This indicated that rack1 loss suppressed polyQ toxicity through pathways involving POE/KCMF1/USP47. Given that rack1 mutants suppressed polyQ-induced cell death independent of aggregate clearance, we next asked whether knocking down poe/kcmf1 or over-expressing USP47 reduced polyQ aggregates. We generated clones of cells expressing poe RNAi , kcmf1 RNAi , or USP47 in 63Q-expressing eyes with the "flip-out" method [48]. We found that knocking down poe/kcmf1 or over-expressing USP47 did not affect 63Q aggregate formation, whereas over-expressing the chaperone, DnaJ, significantly reduced polyQ aggregates (S4A and S4B Fig) [36,49].
We further confirmed the genetic interaction between RACK1 and POE/KCMF1/USP47 in the MJD model. As seen with 63Q, knocking down poe or kcmf1 (or overexpressing USP47) alleviated MJD-Q78-induced photoreceptor cells loss. Knocking down ufd4 had no effect ( Fig  3B and 3C). Moreover, we found that either knocking down poe or kcmf1 or overexpressing USP47 in the rack1 RNAi background largely promoted MJD-Q78 photoreceptor cell survival compared with the expression of rack1 RNAi , poe RNAi , kcmf1 RNAi , or USP47 alone (Fig 3B and  3C). We further evaluated photoreceptor function via ERG recordings. Consistent with the degeneration results, knocking down poe or kcmf1 (or overexpressing USP47) increased ERG amplitude for MJD-Q78 flies, whereas co-expression of rack1 RNAi with poe RNAi , kcmf1 RNAi , or USP47 completely rescued ERG amplitude and transients for MJD-Q78 flies (Fig 4A-4C). These results suggest that rack1 alleviated polyQ toxicity through the POE/KCMF1 ubiquitin ligase systems.

rack1 mutations suppressed polyQ toxicity through ERK
Since ERK was the substrate of the POE/KCMF1 E3 ubiquitin ligase, we speculated that loss of rack1 suppresses polyQ toxicity by regulating levels of ERK. We first measured ERK levels in adult flies and found that overexpressing USP47 or knocking down rack1 increased ERK levels by two-fold (Fig 5A and 5B). We then used the "flip-out" system to express rack1 RNAi in GFPpositive cells in eye imaginal discs. Comparing ERK levels between rack1 RNAi cells and neighboring wild-type cells, we further confirmed that down-regulating rack1 increased ERK levels (Fig 5C and 5D). By contrast, rack1 RNAi and USP47 overexpression did not impact erk mRNA levels (S6A Fig). Moreover, we found that knocking down rack1 also increased ERK levels when 63Q was expressed. Similar results were seen when poe or kcmf1 was knocked down, but a UAS-RNAi lines were crossed with 63Q;GMR>rack1 RNAi to identify genes that modified the polyQ/rack1 RNAi phenotype. "yes" indicates genes that when knockeddown enhanced the ability of rack1 RNAi to suppress polyQ-induced cell death.
b UAS-RNAi lines were crossed with 63Q;GMR-GAL4 to identify those that suppressed polyQ-induced cell death. "yes" indicates genes that when knocked-down suppressed polyQ-induced cell death. We also found that knocking down rack1 in S2 cells decreased ERK ubiquitination level (Fig 6A and 6B). Finally, knocking down rack1 using engrailed-GAL4 (en-GAL4) induced extra wing vein material within the posterior compartment, a phenotype consistent with erk activation (Fig 6C) [50]. Thus, RACK1 negatively regulated ERK protein levels post-translationally. Consistent with the up-regulation of ERK by loss of rack1, overexpressing erk suppressed both 63Q-and MJD-Q87-associated toxicities (Figs 6D, 6E and S8C). To further prove that the suppression of polyQ toxicity by rack1 loss depended on ERK, we knocked down erk using erk RNAi and found that erk RNAi prevented rack1 RNAi from suppressing both 63Q-and MJD-Q87-induced cell death (Figs 6D, 6E and S8C). Finally, we found that the knock down or overexpression of erk did not affect polyQ levels (S4A and S4B Fig). We also assessed ERK levels in polyQ disease models and found that expressing 63Q or MJD-78Q did not affect ERK levels (S6C and S6D Fig). These results strongly indicate that the ERK pathway is involved in rack1-mediated suppression of polyQ toxicity.

RACK1 functioned as an adaptor protein to promote the degradation of ERK by POE
Because RACK1 is a scaffold protein and regulated ERK levels post-translationally, we hypothesized that RACK1 may recruit E3 ligases to ERK to promote its degradation. To test this hypothesis, we expressed GFP-tagged RACK1 and FLAG-tagged ERK in S2 cells, and found that ERK could be co-immunoprecipitated with RACK1 ( Fig 7A). We next assessed the interaction between RACK1 and the E3 ligases, POE/KCMF1/UFD4. Although KCMF1 and UFD4 could not be co-immunoprecipitated with RACK1 ( Fig 7B and 7C), POE interacted with RACK1 through multiple binding sites, including the UBR4 domain and the region between the UBR and UBR4 domains (Figs 7D, S7A and S7B). Moreover, as with RACK1, POE coimmunoprecipitated with ERK via multiple binding sites. The interaction between POE and ERK depended on RACK1, since knocking down rack1 by double-stand RNA (dsRNA) weakened the interaction between POE and ERK ( Fig 7E). We also found that overexpressing RACK1 did not enhance the interaction between POE and ERK (S7C Fig). We then overexpressed poe using a CRISPR/Cas9-based transcriptional activation system [51] and found that POE abolished the suppression of polyQ-associated cell death by rack1 RNAi . By contrast overexpressing kcmf1 had no effect (Figs 7F and S8D). Finally, overexpression of poe reverted ERK levels back to normal in the rack1 RNAi background, while overexpressing kcmf1 had no effect (S7D and S7E Fig). We conclude that RACK1 functions to stabilize the interaction between ERK and the E3 ligase POE, and that mutations in rack1 disrupt POE-dependent ERK degradation. These increased levels of ERK then suppress polyQ-associated neurodegeneration.

Discussion
The fact that polyQ repeats aggregate and are toxic by themselves suggest that they are the critical driver of neurodegenerative disease pathogenesis, rather than the proteins they are  associated with in different diseases. High-throughput screens have been performed in multiple systems to identify factors that modify polyQ aggregation [52][53][54], with most of these modifiers affecting the accumulation or the physical properties of polyQ aggregates [55]. In this study, we generated a simple model of polyQ disease by expressing 63 CAG repeats in the fly eye and conducted a genome-wide, loss-of-function screen to identify suppressors of polyQinduced cell death. We found that a loss-of-function mutation in rack1 alleviated the cytotoxicity of polyQ proteins. Importantly, loss of rack1 did not affect the proteostatic control of polyQ proteins, indicating that RACK1 instead affected cytotoxicity downstream of aggregate formation. Chaperones, including HSP40/DNAJ, protect cells in the context of different polyQ disease models, again indicating that the toxicity of polyQ proteins is derived specifically from the polyQ tracts, not from the unique features of the pathogenic proteins [10,36,38,56]. However, since chaperons reduce levels of both the polyQ tracts and the associated protein, this result is not definitive proof that the polyQ tracts are the primary culprit in mediated neuronal cell death. It has also been proposed that unique features of polyQ proteins also modulate pathogenesis of special polyQ diseases [57,58]. Our results demonstrate that mutating rack1 prevented neurodegeneration in a general model of polyQ disease, as well as two models of specific polyQ diseases, namely MJD and HD. This indicates that different polyQ diseases share common pathogenic features.
RACK1 is a seven WD repeat domain protein, and provides a platform for protein-protein interactions [23]. RACK1 helps to regulate stress responses and apoptosis through scaffolding signaling proteins such as PKC, ZNF598, and JNK [31,45,46]. However, all these signaling pathways are not required for RACK1 function in the context of polyQ pathogenesis. Here, we demonstrate that RACK1 physically interacted with ERK and negatively regulated ERK levels post-translationally. We provide further genetic evidence that the suppression of polyQ toxicity by rack1 mutation depended on ERK stabilization. Knocking down erk abolished the suppressing effects of rack1 on polyQ-induced neurodegeneration. Therefore, we present both biochemical and genetic evidence that loss of RACK1 suppressed polyQ-induced cytotoxicity by stabilizing ERK.
POE is an N-terminal UBR box E3 ubiquitin ligase, and regulates ERK stability together with another E3 ligase, KCMF1 [22]. In a genome-wide RNAi screen for modifiers of the rack1 phenotype, we found that knockdown of POE or KCMF1 further suppressed polyQ toxicity when expressed with rack1 RNAi . In contrast, overexpressing their counterpart, the deubiquitinase USP47, suppressed polyQ-induced cell death, and this suppression was improved in combination with rack1 mutation. It has been suggested that ERK is a substrate for POE, and indeed we found that ERK interacted with POE. Further, both POE and ERK bound RACK1, and interaction between POE and ERK was disrupted by knocking down rack1, suggesting that the interaction between POE and ERK depends on their ability to bind RACK1.  Furthermore, rack1 mutation did not suppress polyQ-induced cell death when POE was overexpressed. In contrast, KCMF1 did not interact with RACK1, and, consistent with this, overexpressing kcmf1 did not reverse the suppression effects of rack1 mutants on polyQ cytotoxicity. This suggests that KCMF1 may not be the key RACK1 cofactor. Instead, our data indicate that RACK1 plays an essential role in promoting the degradation of ERK by bringing together ERK and the E3 ligase, POE. These results further validate our finding that overexpressing rack1 does not enhance polyQ toxicity, since excessive scaffolding proteins may not enhance the interaction between POE and ERK.
The ERK pathway is involved in promoting cell survival and the clearance of protein aggregates in some neurodegenerative diseases. While ERK signaling is upregulated in HD cell lines and a mouse model of SCA-17 [14,17], studies in a fly model of HD and patient fibroblasts indicate that polyQ proteins inhibit ERK signaling [16,59]. Despite the fact that ERK regulation by polyQ is controversial, a number of studies agree that activating the ERK signaling pathway protects against polyQ toxicity [14,16,17,59]. Further studies utilizing the SCA1 model suggest that ERK may help prevent polyQ aggregation [60]. Consistent with the role of ERK in cell survival, we found that ERK signaling protected against polyQ toxicity, both in models of general polyQ disease and models of specific polyQ diseases, namely MJD and HD. This suggests that ERK protects against polyQ disease pathogenesis. Moreover, upregulating levels of ERK protein through the genetic manipulation of POE, USP47, or RACK1 prevented polyQ-induced cell death in MJD and HD models. Although the mechanisms by which ERK signaling regulates polyQ toxicity remain unknown, our results suggest that ERK suppresses polyQ-induced cell death downstream of aggregate formation. Moreover, ERK levels was not reduced in polyQ-expressing cells, indicating that inhibition of ERK pathway polyQ is not directly involved in the pathogenesis of polyQ toxicity.
Because the abnormal expansion of polyQ repeats plays a pivotal role in pathogenesis of polyQ diseases, clearing polyQ aggregates via molecular chaperones has been extensively studied as a strategy for treating polyQ diseases. However, these types of treatment regimens could impair general cellular proteostasis, as many polyQ containing proteins play essential roles. Here we show that the RACK1/POE/ERK signaling pathway does not clear polyQ protein aggregates, but instead specifically ameliorates late pathogenic events associated with polyQ diseases. Our data thus indicate that the RACK1/POE/ERK pathway is a potential therapeutic target for treating polyQ diseases.

Generation of plasmid constructs
The rack1, erk, and kcmf1 cDNAs were amplified from the cDNA clones RE74715, RE08694, and LP17815, respectively, which were obtained from the Drosophila Genomic Resource Center (DGRC). The ufd4 and truncated poe cDNAs were reverse transcribed and amplified from total RNA extracted from w 1118 flies. To construct pIB-rack1-GFP, the entire coding sequence of rack1 was subcloned into the pIB-cGFP vector between the EcoRI and NotI sites. To construct pIB-erk-FLAG, the 3XFLAG tag was subcloned into the pIB vector between the BamHI and XbaI sites, followed by subcloning erk cDNA between the EcoRI and NotI sites. To construct pIB-poe.tr-mCherry, the mCherry tag was subcloned into the pIB vector between the BamHI and XbaI sites, and the specific truncated coding sequences of poe were subcloned into the pIB-mCherry vector between the EcoRI and NotI sites. To construct pIB-kcmf1-HA, the 3XHA tag was subcloned into the pIB vector between the BamHI and XbaI sites, followed by subcloning the entire coding sequence of kcmf1 between the EcoRI and NotI sites. To construct pIB-ufd4-Myc, the Myc tag was subcloned into the pIB vector between the BamHI and XbaI sites, and the entire coding sequence of ufd4 were subcloned into pIB-Myc vector between the EcoRI and NotI sites.

Generation of transgenic flies
To generate the pGMR-63Q-HA construct, a HA tag was added to the C-terminus of the 63Q cDNA and subsequently subcloned into the pGMR vector [69]. The construct was injected into w 1118 embryos, and transgenic insertions on the X-chromosome were maintained. To The ufd4 RNAi line was generated as described [70]. Briefly, a designed two 23nt short hairpin RNA sequences (GCTGTGCTGCTAGATATTTGT) targeting the coding region of ufd4 was cloned into a VALIUM20 vector. The plasmids were then injected into M{vas-int.Dm}ZH-2A;M {3xP3-RFP.attP}ZH-86Fb embryos, and transformants were identified on the basis of eye color.
poe was overexpressed using the flySAM system [51]. Briefly, one sgRNA (TGGCTCCAGCTTGACGTCGG) targeting 350 base pairs upstream of the transcription starting site of poe was cloned into the flySAM vector. Plasmids were then injected into M{vasint.Dm}ZH-2A;M{3xP3-RFP.attP}ZH-86Fb embryos, and transformants were identified on the basis of eye color, and further crossed with GMR-GAL4 flies to overexpress poe.

Objective criteria for scoring retinal phenotypes
All genotypes presented here exhibited highly uniform retinal phenotypes. We examined eye phenotypes of �100 flies per genotype, and phenotypes shown in the images (obtained via light microscopy) are present in 100% of the animals. For genetically identical flies, there was no appreciable phenotypic variability. In cases where enhancement or suppression is reported, this phenotype was present in 100% of the animals. Nevertheless, to apply quantitative analyses, we scored the severity of eye phenotypes, as previously reported [71].

EMS screening
The EMS screening were performed as described [72]. Briefly, the second chromosome of

Electroretinogram recordings
The electroretinogram experiments were performed as described [69]. Briefly, two glass microelectrodes were filled with Ringer's solution; one was placed on the surface of the compound eye, and one was placed on the thorax. The light source intensity was~2000 lux, and the light color was orange (source light was filtered using a FSR-OG550 filter). ERG signals were amplified with a Warner electrometer IE-210, and recorded with a MacLab/4 s A/D converter and the clampelx 10.2 program (Warner Instruments, Hamden, USA).

Transmission electron microscopy (TEM) and retinal degeneration analysis
Transmission electron microscopy was performed as described [73]. Briefly, adult fly heads were dissected, fixed, dehydrated, and embedded in LR White resin. Thin sections (80 nm) were prepared at a depth of 30-40 μm, then stained with uranyl acetate and lead citrate (Sigma, St. Louis, MO) and examined using a JEOL JEM-1400 transmission electron microscope (JEOL Ltd., Tokyo, JAPAN). For retinal degeneration analysis, flies were raised at 25˚C in a 12 h light/12 h dark cycle, and collected at indicated ages. More than 20 ommatidia were counted for each eye section, and more than three flies per genotype were scored.

Lifespan assay
Lifespan assay were performed as described [74]. Briefly,~200 newly enclosed female flies of each genotype were collected and placed in groups of 20 individuals. Flies were kept at 25˚C with 65% humidity under 12 h light/12 h dark cycles. Flies were transferred to new vials every 2 days, and the number of dead flies was counted every day.

Western blot analysis
Western blotting was performed as described [39]. Briefly, 20 fly eyes were dissected and homogenized in SDS sample buffer with a Pellet Pestle (Kimble/Kontes). The proteins were fractionated by SDS-PAGE and transferred to Immobilon-P transfer membranes (Millipore) in Tris-glycine buffer. The blots were probed with Mouse Tubulin primary antibodies (1:2000 dilution, Developmental Studies Hybridoma Bank), Rabbit ERK (1:1000, Cell Signaling Technology), Mouse β-Actin primary antibodies (1:1000 dilution, Santa Cruz Biotechnology) and Rabbit RACK1 antibodies (1:1000 dilution, Dr. J. Kadrmas lab) followed by IRDye 680 goat anti-Rabbit IgG (LI-COR) and IRDye 800 goat anti-Mouse IgG (LI-COR) as the secondary antibodies. The signals were detected with the Odyssey infrared imaging system (LI-COR). Cell culture, dsRNA synthesis, and immunoprecipitation S2 cells were grown at 26˚C in Schneider's Drosophila medium (Sigma-Aldrich) with 10% fetal bovine serum (Gibco BRL), dsRNA was synthesized in vitro and transfected as reported [75]. Plasmids were transfected using Vigofect reagent (Vigorous Biotechnology), and cells were collected and lysed with 10 mM Tris-HCl lysis buffer (pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.5% NP-40, 25 mM NaF, and 1 mM Na 3 VO 4 with 1× proteinase inhibitor cocktail [Sigma-Aldrich]). S2 cells were pre-treated with dsRNA against GFP or rack1 for four days, and then transfected with plasmids for two days. To inhibit proteasome activity, cells were treated with 5 μM MG132 for two days. Immunoprecipitations were performed with mCherry beads (Chromotek) and GFP beads (Chromotek). The bound proteins were analyzed by western blotting against Rabbit GFP antibodies (1:1000 dilution, Torrey Pines Biolabs), Mouse

Quantitative RT-PCR
Quantitative RT-PCR was performed as described [76]. Total RNA was prepared with Trizol reagent (Invitrogen) from dissected fly eyes, and was treated with TURBO DNase (Thermo-Fisher). The cDNA was synthesized with RT master mix (RR036A-1; Takara). An iQ SYBR green supermix (Bio-Rad) was used for real-time PCR, and results were analyzed with a CFX96 Real-Time PCR Detection System (Bio-Rad).

Polysome fractionation
Polysome fractionation was performed as described [77]. S2 cells were cultured in dsRNA for four days and then treated with cycloheximide (CHX) at 0.1 mg/ml for 30 min. For each RNAi, 10 8 cells were pelleted and washed once with cold PBS containing 0.1 mg/ml CHX. The cells were then lysed in polysome lysis buffer (PLB; 20 mM Tris-Cl, pH 7.5, 250 mM KCl, 10 mM MgCl 2 , 1% Triton X-100, 1 mM DTT, 0.1 mg/ml CHX, 1X proteinase inhibitor cocktail (Sigma-Aldrich), and 1 mM PMSF). The lysates were centrifuged at 12,000 x g for 10 min at 4˚C and 1/10 of the supernatant was removed as input sample. The remainder of each supernatant was loaded on a 20%-50% sucrose gradient prepared in PLB (without Triton X-100) and resolved by centrifugation in an SW41 rotor (Beckman) at 38,000 rpm for 3 h at 4˚C. Fractions were collected while monitoring the absorbance at 254 nm. RNA was prepared from each fraction and assayed by quantitative RT-PCR.

Quantification and statistical analysis
Statistical details for each experiment, including n numbers and the statistical test performed, can be found in the corresponding figure legend. Data are presented as mean ± standard error of the mean (SEM). Statistical analysis was performed using Prism 6 software. Student's t test was used for datasets with a normal distribution and a single intervention. One-way ANOVA was performed with Tukey's post hoc test for multiple comparisons, and two-way ANOVA with Sidak's post hoc test was used to compare ribosome loading of erk mRNA. p < 0.05 was considered statistically significant. � p < 0.05, �� p < 0.01, ��� p < 0.001, ���� p < 0.0001.  Table. qPCR validation of indicated RNAi lines. UAS-RNAi lines were crossed with GMR-GAL4, and adult eyes of F1 flies were dissected for RNA extraction and subsequent qPCR. Samples were prepared in triplicate. All values were normalized to GMR-GAL4 samples. RpL32 was used as the reference gene for standardization. The p-values were obtained using a Student's T-tests against the GMR-GAL4 samples. (XLSX) S1 Data. Original data statistics. (XLSX)