Suppression of retinal degeneration by two novel ERAD ubiquitin E3 ligases SORDD1/2 in Drosophila

Mutations in the gene rhodopsin are one of the major causes of autosomal dominant retinitis pigmentosa (adRP). Mutant forms of Rhodopsin frequently accumulate in the endoplasmic reticulum (ER), cause ER stress, and trigger photoreceptor cell degeneration. Here, we performed a genome-wide screen to identify suppressors of retinal degeneration in a Drosophila model of adRP, carrying a point mutation in the major rhodopsin, Rh1 (Rh1G69D). We identified two novel E3 ubiquitin ligases SORDD1 and SORDD2 that effectively suppressed Rh1G69D-induced photoreceptor dysfunction and retinal degeneration. SORDD1/2 promoted the ubiquitination and degradation of Rh1G69D through VCP (valosin containing protein) and independent of processes reliant on the HRD1 (HMG-CoA reductase degradation protein 1)/HRD3 complex. We further demonstrate that SORDD1/2 and HRD1 function in parallel and in a redundant fashion to maintain rhodopsin homeostasis and integrity of photoreceptor cells. These findings identify a new ER-associated protein degradation (ERAD) pathway and suggest that facilitating SORDD1/2 function may be a therapeutic strategy to treat adRP.


Author summary
Misfolded rhodopsins accumulated in endoplasmic reticulum (ER) could disrupt the homeostasis of the ER and cause ER stress. Chronic ER stress would finally lead to photoreceptor cell death and retinal degeneration. To diminish the stress and sustain homeostasis cells develop alternative strategies to clear the misfolded rhodopsins. Previous studies have suggested that ubiquitin E3 ligase HRD1 is involved in the degradation of misfolded rhodopsins. In this study, we define novel ubiquitin E3 ligase SORDD1/2 based on a genetic screen and demonstrate that SORDD1/2 promotes the degradation of misfolded rhodopsins through ER-associated degradation (ERAD) pathway. Furthermore, we demonstrate that SORDD1/2 function independently of HRD1 in misfolded rhodopsins degradation. We also show SORDD1/2 and HRD1 play redundant roles in rhodopsin homeostasis. Finally, we demonstrate that SORDD1 works well in a Drosophila disease model. Our studies identify a novel ERAD pathway that acts in parallel to HRD1, and suggest that SORDD1 is a good candidate therapeutic target.

Introduction
Misfolded membrane proteins, including G protein-coupled receptors (GPCRs), are often retained in the endoplasmic reticulum (ER) and an overabundance of ER-retained proteins is associated with many diseases. Rhodopsin is a GPCR and dominant mutations in the rhodopsin gene (Rho) are observed in 20-30% of all forms of autosomal dominant retinitis pigmentosa (adRP), the most common form of retinal degeneration [1][2][3]. Most biochemicallycharacterized Rho mutants associated with adRP are likely misfolded and accumulate in the ER of photoreceptor neurons [4]. Indeed, the Rho mutation most commonly associated with adRP is a proline to histidine mutation at position 23 (Rho P23H ), which leads to Rho retention within the ER. This in turn leads to ER stress, activation of the unfolded protein response (UPR), and ultimately photoreceptor degeneration [2,[5][6][7][8]. This is seen in both human patients and animal models of the disease. In most cases, misfolded rhodopsins are inherently unstable and undergo ER-associated protein degradation (ERAD), a process by which substrates are recognized, ubiquitinated by E3 ligases, retro-translocated into the cytosol, and sent to proteasomes for degradation [8][9][10][11]. The UPR induces ERAD to promote the degradation of misfolded rhodopsin and to protect photoreceptor cells against ER stress signals [12][13][14]. Although it has been proposed that ERAD may disrupt Rho homeostasis earlier in development and contribute to retinal degeneration [15], studies in both Drosophila and mouse models suggest that increasing degradation of misfolded rhodopsin through ERAD and the proteasome is protective. Further, decreased efficiency of ERAD/proteasome function leads to the aggregation of rhodopsin and photoreceptor death [16][17][18][19] The E3 ubiquitin-protein ligase HRD1 (HMG-CoA reductase degradation protein 1) plays a central role in ERAD of membrane proteins (ERAD-M) such as rhodopsin. In a fly model of adRP, a glycine is mutated to an aspartate at position 69 within Rh1, the major rhodopsin in flies, which is encoded by the ninaE locus (ninaE G69D ) [20,21]. Importantly, overexpression of HRD1 reduces Rh1 G69D -induced ER stress and alleviates photoreceptor cell loss [19]. Despite the key role played by HRD1 in maintaining rhodopsin homeostasis, flies lacking HRD1 have normal rhodopsin and photoreceptor cell function [16], suggesting that another E3 ubiquitin ligase may help maintain rhodopsin homeostasis.
Here we screened for additional genes that reduce Rh1 G69D -induced photoreceptor cell degeneration, and identified two E3 ubiquitin ligases, SORRD1 and SORRD2, that strongly suppress retinal degeneration in Rh1 G69D -expressing flies. We demonstrate that SORRD1/2 ubiquitinates misfolded Rh1 and promotes degradation of these proteins through valosin containing protein (VCP) and the proteasome, and that this process is completely independent of HRD1. Moreover, although null mutations in either sordd1/2 or hrd1 do not lead to photoreceptor cell dysfunction, animals in which both sordd1/2 and hrd1 are mutated exhibit severe retinal degeneration. Overall, this work identifies an important role for the previously uncharacterized E3 ubiquitin ligases, SORRD1 and SORRD2, in ERAD of misfolded rhodopsin and maintenance of rhodopsin homeostasis, independent of HRD1.

SORDD1 and SORDD2 are new suppressors of Rh1 G69D
To find factors other than HRD1 that limit retinal degeneration in adRP, we performed a gainof-function screen using a Drosophila model in which Rh1 G69D was ectopically overexpressed using the GMR-Gal4 system (GMR>Rh1 G69D ) (Fig 1A-1D and S1A-S1B Fig). Overexpression of HRD1, the major E3 ubiquitin ligase of the ERAD pathway, ameliorated the glossy eye phenotype of GMR>Rh1 G69D flies and restored ommatidia structures (based on SEM analysis), suggesting the viability of this genetic screen (Fig 1D and 1E, S1B, S1C and S1K Fig and S1 Data).
We screened~3,000 UAS-cDNA libraries from FlyORF [22] and EP collections, and identified three genes that strongly suppressed Rh1 G69D -induced retinal degeneration. These were smg5 (UAS-Smg5.ORF), CG8974 (P[EP]CG8974 G757 ), and CG32581 (P[EPgy2]EY20019). Since Smg5 may affect the stability of rh1 G69D mRNA, we focused on CG8974 and CG32581, which each encode RING-domain E3 ubiquitin ligases. We verified that CG8974 and CG32581 could each suppress Rh1 G69D -induced retinal degeneration by repeating the experiment using UAS-CG8974 or UAS-CG32581 overexpression lines (Fig 1H-1G, S1D, S1E and S1K Fig). We therefore named these proteins SORDD1 (Suppression Of Retinal Degeneration Disease 1 upon overexpression) and SORDD2, respectively. SORDD1 and SORDD2 share 92% protein sequence identity. RNAsequencing and qPCR results indicated that while sordd1 is expressed ubiquitously, sordd2 is not expressed under normal conditions (S2B Fig, S2 Data). The sordd1 and sordd2 gene loci share 95% sequence identity and are adjacent in the genome (S2A Fig). This suggests that sordd2 may have resulted from gene duplication, and therefore may serve a largely, if not entirely, redundant role.
Previous studies reported that HRD1 can also recognize misfolded versions of wild-type Rh1, which are expressed in larval eye imaginal discs and earlier pupa eyes before the expression of chaperons for Rh1, and suppresses ER stress caused by this earlier expression of wildtype Rh1 [19]. We also found that SORDD1 and SORDD2 reduced retinal cell degeneration induced by the expression of wild-type Rh1 in larval eye imaginal discs (S2E Fig). Moreover, we expressed Rh1 P37H (the equivalent of mammalian Rho P23H -the most common genetic mutation associated with ADRP) in larval eye imaginal discs using GMR-Gal4/UAS-rh1 P37H and found that SORDD1 and SORDD2 largely suppressed Rh1 P37H -mediated retinal damage [24] (S2E Fig). These results suggest that SORDD1/2 and HRD1 recognize the folding state of Rh1 rather than specific alterations/mutations in the Rh1 protein sequence.

The E3 ubiquitin ligases SORDD1/2 promote the degradation of Rh1 G69D
Overexpression SORDD1/2 associate with the membrane fraction of S2 cells and colocalize with the ER protein Calnexin (S2C and S2D Fig). Moreover, RNF185, the human homolog of SORDD1/2 was found to be an ER resident E3 ligase and function in the degradation of transmembrane proteins in ER, suggesting that SORDD1/2 may be directly involved in degrading misfolded Rh1 G69D in ER [25,26]. To test this, we expressed a Myc-tagged version of Rh1 G69D in the eye (GMR>Rh1 G69D -Myc), and found that overexpression of SORDD1 or SORDD2 could lower the level of Rh1 G69D protein in both eye imaginal discs and pupa eyes (Fig 2A-2D, S3 Data and S4 Data). Accumulation of Rh1 G69D induced ER stress and activated the UPR, as both Xbp1-EGFP and ATF4-mCherry, two independent UPR reporters, were activated in eye imaginal discs of GMR>Rh1 G69D animals [19,27]. However, Rh1 G69D failed to activate Xbp1-EGFP and ATF4-mCherry when SORDD1 or SORDD2, but not CG8141, was overexpressed (Fig 2A and 2B and S3 Data). Thus, like HRD1, SORDD1/2 facilitates the degradation of misfolded Rh1 G69D , suppressing Rh1 G69D -induced ER stress and retinal degeneration.
Knocking-down the proteasome component Rpn12, or the VCP components TER94/VCP or UFD1 (Ubiquitin Fusion-Degradation 1), which are required to extract substrate from the ER into the cytosol, caused eye damage in GMR>Rh1 G69D sordd1 flies, but did not cause eye damage when knocked down alone ( Fig 3B). Moreover, disruption of both VCP and the proteasome increased Rh1 G69D protein levels in GMR>Rh1 G69D , sordd1 retinae (S3B and S3C Fig  and S6 Data). Based on these results, we conclude that SORDD1-mediated degradation of Rh1 G69D is dependent on the VCP/proteasome system.
Degrading misfolded ER membrane proteins, which is called ERAD-M, requires proteins to be retro-translocated into the cytosol and subsequently poly-ubiquitinated [28]. HRD1, in addition to functioning as an E3 ubiquitin ligase, also complexes with the ER luminal protein HRD3 to form a retro-translocation channel for the movement and degradation of misfolded ER proteins. [29,30]. However, knocking down hrd1 or hrd3 did not affect the eyes of GMR>Rh1 G69D sordd1 flies, raising the possibility that SORDD1/2-dependent ERAD does not rely on retro-translocation of substrates by the HRD1/3 complex (Fig 3C). To confirm the dispensable role of the HRD1/3 complex in SORDD1-mediated protein degradation, we used the "ey-flp/hid" system to generated flies with eyes that were homozygous mutant for either hrd1 or hrd3 [16]. SORDD1 continued to suppress Rh1 G69D -induced retinal degeneration in the absence of HRD1 or HRD3 (Fig 3C and 3D and S7 Data). Taken together, we conclude that SORDD1-mediated Rh1 G69D degradation depends on VCP and the proteasome, but is independent of the HRD1/3 complex.
To further explore the physiological roles of SORDD1/2 and HRD1 in maintaining photoreceptor cell homeostasis, we examined photoreceptor cell integrity in sordd1/2 1 and hrd1 1 mutants that expressed wild-type rhodopsin. We first used deep pseudopupil analysis, which reflects the compact structure of photoreceptor cells, assess levels or retinal degeneration [31]. Animals lacking sordd1/2 1 or hrd1 1 had normal retinal cytoarchitecture across a broad range of ages, whereas double mutants lacking both sordd1/2 1 and hrd1 1 exhibited a gradual loss of the deep pseudopupil, indicating age-dependent retinal degeneration (Fig 4C and S9 Data).
We examined the function of these photoreceptor neurons using an electroretinogram (ERG) assay, which is an extracellular recording that measures the summed light responses of photoreceptor cells. The ERG exhibits two primary features: a rapid corneal negative response, reflecting phototransduction; and on-and off-transients, which reflect synaptic transmission from photoreceptor cells [32]. Twenty-five-day-old ninaE G69D /+ flies exhibited small ERG response amplitudes and a complete loss of transients, suggesting disruption of several neural functions (Fig 5F and 5G and S12 Data). Consistent with the results obtained via TEM, GMR-hrd1; ninaE G69D /+ flies showed significant increased ERG amplitude at the age of 25 days relative to ninaE G69D /+ flies, and expression of SORDD1 fully restored the ERG amplitude and on/off transients, indicating a complete rescue of photoreceptor function (Fig 5F and 5G and S12 Data). Furthermore, at 32 days, while ninaE G69D flies had no ERG response, ERG responses and transients were detected in ninaE G69D /GMR-sordd1 flies (Fig 5F).

Discussion
This work suggests a new ERAD-M pathway that depends on the E3 ubiquitin ligase SORDD1/2 (Fig 4G). Misfolded membrane proteins such as Rh1 G69D may be ubiquitinated by  two independent ER enzymes, HRD1 and SORDD1/2. The ubiquitylated protein is then transported out of the ER membrane via the VCP complex and degradation in cytosolic proteasomes. Although HRD1 also functions (in complex with HRD3) as a retro-translocation channel to facilitate removal of substrates from the ER [29,30], HRD1 and HRD3 are dispensable for SORDD1-mediated degradation of Rh1 G69D ; thus, SORDD1 suppresses Rh1 G69Dinduced ER stress and cell damage independent of HRD1 and HRD3. Although we demonstrated that VCP complex components are essential for SORDD1-mediated degradation of Rh1 G69D , we did not identify a candidate retro-translocation channel. One possibility is that Rh1 G69D ubiquitinated by SORDD1 is removed directly from the ER membrane by the VCP complex without channels due to the plasticity of the ER membrane [33]. Alternatively, key translocation channels, such as Sec61, may serve dual functional roles and be essential for cell viability [34]. Disrupting rhodopsin homeostasis-the most abundant protein in photoreceptors-induces severe stress and frequently results in the degeneration of photoreceptor neurons [35]. Degradation of misfolded rhodopsin via ERAD is one of several mechanisms that maintains rhodopsin homeostasis and normal photoreceptor function and viability. Despite HRD1 serving as a key E3 ligase in ERAD, null hrd1 mutants exhibit normal rhodopsin levels and light responses [16,19]. One possibility is that, in addition to being degraded by ERAD, misfolded rhodopsin might also be targeted by autophagy pathways [36,37]. However, recent reports argue against a complementary function of autophagy in degrading misfolded rhodopsin-inhibiting autophagy reduced retinal degeneration in Rho P23H mice by promoting proteasomal degradation of misfolded mutant rhodopsin [17,18,38].
Here we present data suggesting that SORDD1/2 functions in a similar, and independent, manner to degrade mutant Rh1 G69D . While knocking out either hrd1 or sordd1/2 did not prevent the degradation of Rh1 G69D protein, Rh1 G69D accumulated in photoreceptors that lacked both E3 ubiquitin ligases. Moreover, flies lacking both hrd1 and sordd1/2 exhibit severe retinal degeneration, whereas photoreceptor cells remain intact in flies lacking either hrd1 or sordd1/ 2. Therefore, SORDD1 functions redundantly with HRD1 in ERAD of misfolded rhodopsin, thereby helping to maintain rhodopsin homeostasis. Two redundant ERAD pathways might be beneficial for the maximum clearance of misfolded rhodopsin and to maintain photoreceptor cell function.
Because increasing the efficiency of protein folding in the ER by expressing BiP prevents retinal degeneration in Rho P23H models, chemical chaperones capable of reducing Rho misfolding are promising therapeutic interventions [5,[39][40][41]. However, the escape of mutant forms of rhodopsin from the ER might contribute to defects in the morphogenesis of rod photoreceptor cell outer segments [42]. Recently, it was shown that increasing proteasome-mediated degradation of misfolded rhodopsin protects against adRP [17,18,38]. However, induction of proteasomal activity might be problematic, since it impairs general cellular proteostasis. In a fly model of adRP, inhibition of VCP or the proteasome alleviated retinal degeneration caused by Rh1 P37H , despite the fact the clearance of misfolded RH1 P37H was disrupted [24]. This might be because inhibition of the VCP or the proteasome may suppress a cell death pathway that acts downstream of misfolded rhodopsin. Indeed, VCP is required for autophagy flux, and blocking autophagy suppresses retinal degeneration in Rho P23H mice [38,43,44]. In flies co-expressing SORDD1 and Rh1 G69D , disruption of VCP or the proteasome dramatically blocked SORDD1-accelerated degradation of Rh1 G69D , thus reversing the suppression of Rh1 G69D -induced eye damage by SORDD1. Given that E3 ubiquitin ligases are the rate-limiting step of ERAD, inducing ERAD by overexpression of HRD1 effectively reduces levels of misfolded Rh1 G69D protein and suppresses retinal degeneration in this fly model of adRP. This raises the possibility that key ERAD components may be a promising therapeutic target for treating adRP. However, because ERAD regulates the turn-over of a variety of very important proteins [40,45,46], hrd1 mutations are lethal, as is the widespread overexpression of HRD1 (e.g., via the tubulin or actin promoter).
Unlike HRD1, flies that lack both sordd1 and sordd2, or flies that express both SORDD1 and SORDD2 ubiquitously, do not exhibit obvious defects, suggesting that SORDD1 is more specific to misfolded proteins. Moreover, overexpression of SORDD1 efficiently degraded Rh1 G69D and, therefore, better suppressed retinal degeneration and restored light response than HRD1. Furthermore, ERAD of misfolded proteins via SORDD1 is likely conserved in mammals, as RNF5 and RNF185 also suppressed Rh1 G69D -induced retinal degeneration. Indeed, both E3 ligases have been implicated in ERAD of a misfolded version of the cystic fibrosis transmembrane conductance regulator, CFTR ΔF508 , which is associated with cystic fibrosis [25,47]. However, unlike SORDD1, which suppressed Rh1 G69D -associated retinal degeneration, genetic and pharmacological inhibition of RNF5 in vivo attenuates intestinal pathological phenotypes in a mouse model of cystic fibrosis without blocking degradation of CFTR ΔF508 [48,49]. This might be because RNF5 and RNF185 function redundantly to degrade CFTR ΔF508 [25]. Moreover, RNF185 has recently been shown to target ER membrane proteins, including CYP51A1 and TMUB2 [26]. Since SORDD1 is ubiquitously expressed, it might also be involved in the degradation and turnover of a subset of membrane proteins. All the evidence strongly suggests that SORDD1 and its human homologs play roles in the quality control of membrane proteins. Overall, these data suggest that the facilitation of SORDD1-dependent ERAD of misfolded rhodopsin is a promising treatment for adRP.

Generation of transgenic flies
The cDNA sequences of Rh1, sordd1, sordd2, hrd1 and CG32847 were amplified from RH01460, GH14055, RE35552, GH11117 and FI06431 of DGRC gold cDNA collections, respectively (Drosophila Genomics Resource Center). cDNA of CG8141 was amplified from the RT-PCR products of w 1118 flies. The cDNA sequences of human genes HsSYVN1, HsRNF5 and HsRNF185 were amplified from ORF Clones IOH21699, IOH3743 and IOH14506 obtained from Thermo Fisher Scientific. These cDNA sequences were then subcloned into the pUAST-attB or GMR-attB vectors. Rh1 G69D was mutated from the ninaE cDNA and subcloned into the pninaE-attB vector with a c-terminal GFP-tag. The constructs were injected into M To generate UAS-Rh1 G69D or UAS-Rh1 G69D -Myc flies, the Rh1 G69D cDNA was subcloned to pUAST vector with or without Myc tag, and the constructs were injected into w 1118 embryos and transformants with random insertions were identified based on eye color. The 2 nd chromosome insertions with weak Rh1 G69D expressing were used for experiments.

RNA extraction and quantitative real-time PCR analysis
Larvae or adult flies were dissected to obtain the different tissues and tissue-specific RNA was extracted using TRIzol Reagent (Invitrogen) according to the manufacturer's protocol. Total RNA was reverse-transcribed using PrimeScript RT-PCR kit (Takara). qRT-PCR was performed using SYBR Premix Ex Taq (Takara) on a CFX96 real time PCR detection system (Bio-Rad). The average threshold cycle value (CT) was calculated from at least three replicates per sample. Expression of sordd1 and sordd2 was standardized relative to rp49.Relative expression values were determined by the 44CT method. Primers: sordd1-f: CTAATCGGGGAACAAGGCAATACG, sordd1-r: CTATGCATAGAACAGCCACAATAT; sordd2-f: ATATTAGAAGGCTACACGAAGATG, sordd2-r: TTATCGGGATGATCCGTGCACTAT; rp49-f: CAGTCGGATCGATATGCTAAGCTG, rp49-r: TAACCGATGTTGGGCATCAGATAC
To examine the cellular localization of SORDD1, S2 cells were seeded onto coverslips coated with concanavalin A and transformed with Myc-tagged SORDD1 plasmid. After 36 h of transformation, the cells were fixed using 4% paraformaldehyde in PBS for 0.5 h, followed by incubation with primary antibodies including mouse anti-Myc (1:200; Santa Cruz) and anti-Calnexin (1:100; Developmental Studies Hybridoma Bank) for 1.5 h at room temperature. Samples were incubated with Alexa Fluor 488-conjugated, Alexa Fluor 568-conjugated secondary antibodies (1:500; Invitrogen) with 20 nM DAPI (Invitrogen) for 1 h at room temperature. Fluorescence images were acquired at room temperature with an A1 confocal laser scanning microscope (Nikon) using a 60x objective.

Fractionation assay
Fractionation assays were performed with a Membrane and Cytosol Protein Extraction Kit (Beyotime Biotechnology) according to the manufacturer's protocol. Briefly, S2 cells were grown in 6 cm dishes and transformed with Myc-tagged SORDD1 plasmid. After 48 h of transformation, the cells were harvested and lysed with Buffer A, followed by centrifugation at 14,000g for 30 min at 4˚C. The supernatant (S) was collected, and the pellet was lysed again by Buffer B, followed by centrifugation at 14,000g for 5 min at 4˚C to collect the pellet (P). S and P fractions were analyzed by western blot.

Scanning electron microscopy
Dissected adult fly eyes were fixed in 2.5% glutaraldehyde at 4˚C overnight followed by incubation in 1% osmium tetroxide for 1 h at room temperature. Samples were dehydrated in a series of ethanol dilutions (20,35,50,70, and 100% ethanol), and substituted by liquid carbon dioxide. After the liquid carbon dioxide gasified, samples were scanned using a ZEISS EVOL LS10 scanning electron microscope at room temperature. The number of ommatidia in the middle 3 rows in each sample was counted to assess the damage degree, and images from 5 flies were used for quantification.

Transmission electron microscopy
TEM was performed with standard methods as described [52]. Dissected adult fly eyes were fixed in 4% paraformaldehyde and 2.5% glutaraldehyde at 4˚C overnight followed by incubation in 1% osmium tetroxide for 1 h at room temperature. Then samples were dehydrated in a series of ethanol dilutions (20,35,50,70,80,90, and 100% ethanol) and embedded in LR White resin (Polysciences, Inc.). Thin sections (80 nm) were stained with 0.06% uranyl acetate and 0.1% lead citrate (Sigma, St. Louis, MO) and examined using a JEM-1400 transmission electron microscope (JEOL) at room temperature. The images were acquired using a Gatan camera (model 794; Gatan, Inc.).

Electroretinogram recordings
Electroretinogram (ERG) recordings were performed as described [53]. Microelectrodes filled with Ringer's solution were placed onto the surface of the compound eye and the thorax of a fixed fly. A Newport light projector (model 765) was used for stimulation. After 1 min of dark adaptation, red-eyed flies were exposed to a 5-s pulse of~2000 lux orange light (source light was filtered using a FSR-OG550 filter, Newport). ERG signals were amplified with a Warner electrometer IE-210 and recorded with a MacLab/4 s analogue-to-digital converter and the clampelx10.2 program (Warner Instruments, Hamden, USA). All the flies used in ERG assay were raised under constant white light of~700 Lux at 25˚C.

Statistics
Statistical results were generated by GraphPad Prism 8 and statistical significance was assessed through Ordinary one-way ANOVA, Sidak's multiple comparisons test analyses. All error bars represent S.E.M.