Adaptation to constant light requires Fic-mediated AMPylation of BiP to protect against reversible photoreceptor degeneration

In response to environmental, developmental, and pathological stressors, cells engage homeostatic pathways to maintain their function. Among these pathways, the Unfolded Protein Response protects cells from the accumulation of misfolded proteins in the ER. Depending on ER stress levels, the ER-resident Fic protein catalyzes AMPylation or de-AMPylation of BiP, the major ER chaperone and regulator of the Unfolded Protein Response. This work elucidates the importance of the reversible AMPylation of BiP in maintaining the Drosophila visual system in response to stress. After 72 hr of constant light, photoreceptors of fic-null and AMPylation-resistant BiPT366A mutants, but not wild-type flies, display loss of synaptic function, disintegration of rhabdomeres, and excessive activation of ER stress reporters. Strikingly, this phenotype is reversible: photoreceptors regain their structure and function within 72 hr once returned to a standard light:dark cycle. These findings show that Fic-mediated AMPylation of BiP is required for neurons to adapt to transient stress demands.


Introduction
Post-translational modifications (PTMs) of proteins are important for rapid responses to environmental challenges of cells. One such PTM is AMPylation, the reversible addition of adenosine monophosphate (AMP) to hydroxyl groups (also known as adenylylation) (Kingdon et al., 1967;Brown et al., 1971;Woolery et al., 2014;Casey and Orth, 2018). AMPylation is catalyzed by at least two protein families, among them the conserved Fic-domain proteins (Harms et al., 2016;Casey and Orth, 2018). Eukaryotic Fic, an ER-resident type-II membrane protein (Rahman et al., 2012), AMPylates BiP (GRP78), a highly conserved and ubiquitous ER chaperone (Ham et al., 2014;Preissler et al., 2015). Working together with a multitude of associated quality control proteins, BiP is critical for the translocation, folding, and secretion of proteins from the ER as well as for aiding in the clearing of misfolded ER aggregates and degradation of membrane-associated proteins (Hendershot et al., 1988;Kozutsumi et al., 1988;Meunier et al., 2002). BiP is both a mediator and transcriptional target of the Unfolded Protein Response (UPR), a coordinated cell signaling pathway that is activated during times of high misfolded protein levels in the ER. Like many protein chaperones, BiP depends on its ATPase activity to undergo a conformational change to bind to its substrates (Gaut and Hendershot, 1993). AMPylation locks BiP into a state resembling the ATP-bound conformation with high substrate off-rates, thereby inhibiting its chaperone function (Preissler et al., 2017b;Wieteska et al., 2017).
In agreement with this PTM's inhibitory role, BiP AMPylation levels are linked to protein homeostasis (Ham et al., 2014). Reduction of ER protein load promotes Fic-mediated AMPylation of BiP, whereas Fic catalyzes the deAMPylation of BiP under elevated ER stress conditions (Ham et al., 2014;Preissler et al., 2015). This switch in Fic's activity is linked to a key regulatory salt bridge in eukaryotic Fic. Mutations in Fic that disrupt this salt bridge result in an overactive AMPylator that lacks deAMPylation activity (Casey et al., 2017;Preissler et al., 2017a). Together, these studies suggest a model in which BiP is AMPylated in times of low ER stress, creating a reserve pool of inactive BiP that can be readily activated to respond to changes of ER homeostasis ( Figure 1A). This reserve pool of BiP is proposed to act as a buffer to attenuate or shorten the need for a more dramatic activation of the transcriptional and translational arms of the UPR (Casey et al., 2017;Preissler et al., 2017a;Wieteska et al., 2017). However, the physiological importance of endogenous Fic-mediated AMPylation remains unclear.
In the fruit fly, Drosophila melanogaster, we previously demonstrated that fic-null mutants harbor a defect in visual signaling, as assessed by electroretinogram (ERG). The well-characterized Drosophila visual system has proven a valuable model for many fields, such as neuroscience (Borycz et al., 2002;Sugie et al., 2015), cell signaling (Dolph et al., 1993;Scott et al., 1995), protein trafficking (Lee et al., 2003;Akbar et al., 2009), and neurodegeneration (Leonard et al., 1992;Johnson et al., 2002;Ryoo et al., 2007). The specialized photoreceptor cells possess tightly packed microvilli-like membranes, termed rhabdomeres, that endow remarkable sensitivity to minute changes in light conditions (Montell, 2012). The ability to maintain this sensitivity is critical for flight behavior, foraging, and escape from predators. Thus, under a wide range of conditions, photoreceptors must maintain their light detection cascade, which requires the constant production, trafficking, and degradation of proteins through the endomembrane system (Colley et al., 1995;Kiselev et al., 2000;Rosenbaum et al., 2006).
We postulated that as a regulator of proteostasis and the UPR, BiP must be tightly regulated for proper maintenance of vision in the fly. Here we demonstrate that mutants lacking AMPylation of BiP at a specific residue, Thr366, possess the same ERG defects as fic-null animals, implicating BiP as the target of Fic necessary for visual signaling. We go on to find that application of an eye-specific stress, constant light, leads to exaggerated electrophysiology defects and the loss of photoreceptor rhabdomeres, akin to neurodegeneration. However, these defects exhibit a remarkable and unique reversibility: structural and functional phenotypes observed in light-stressed fic-null and AMPylationresistant BiP T366A mutants are reversed after the flies are returned to a standard light/dark cycle. In addition, we identify changes in the regulation of UPR during constant light stress in these mutants, implicating dysregulation of ER homeostasis as a probable cause of the inability to adapt to altered light conditions.

Results
BiP T366A rescues over-expression of constitutively active AMPylating Fic E247G To test the hypothesis that BiP is a critical target of Fic AMPylation in vivo ( Figure 1A), we designed and generated transgenes expressing wild-type and AMPylation-resistant FLAG-tagged BiP proteins under control of the endogenous BiP promoter (Figure 1-figure supplement 1A). BiP null fly mutants die early in development; this lethality is rescued by including a copy of the genomic transgene expressing FLAG-BiP WT or the AMPylation-resistant FLAG-BiP T366A mutant ( Figure 1B). We will refer to these rescued flies as BiP WT or BiP T366A , respectively. In contrast, a BiP transgene mutated at a second reported AMPylation site (Preissler et al., 2015;Casey et al., 2017), BiP T518A , did not rescue BiP -/lethality. As fic null mutants that lack BiP AMPylation are viable, the lethality of the BiP T518A mutant is not likely to be due to the loss of AMPylation (Casey et al., 2017). Instead, these observations indicate an essential role for Thr 518 in BiP's chaperone activity. Consistent with this notion, the equivalent residue, Thr 538 , in the S. cerevisiae BiP homolog Kar2 is required for survival under heat stress even though yeast lack both Fic domain proteins and BiP AMPylation (Figure 1figure supplement 1B).
Previously, we reported that over-expression of the constitutively active AMPylating Fic E247G was lethal in a fic-null fly background (fic 30C ) because it lacks the essential deAMPylation activity (Casey et al., 2017). We tested whether flies expressing the AMPylation-resistant BiP T366A could survive this lethality. Consistent with previous findings, we observe over-expression of the mutant UAS-Fic E247G transgene using the ubiquitous Da-Gal4 driver was lethal in an otherwise fic-null animal ( Figure 1C). Lethality of the constitutively active AMPylating Fic E247G was suppressed in flies expressing the AMPylation-resistant BiP T366A but not BiP WT ( Figure 1C). These results indicate that Thr 366 of BiP is a physiologically relevant AMPylation target in vivo. BiP AMPylation during times of low ER stress reserves a portion of the chaperone to allow for a rapid, deAMPylation-driven, response to high ER stress (Casey et al., 2017;Preissler et al., 2017a). (B) Bar graphs show the percentage of null mutant BiP G0102 /y males rescued by the indicated genomic BiP WT , BiP T366A or BiP T518A genomic transgene (Tg) relative to sibling controls. N = 3 biological replicas. At least 50 flies scored for each replica. Bar graphs show means ± Standard Deviation (SD). (C) Bar graphs show the percentage of viable flies of the indicated wild-type or fic 30C genotypes expressing the overactive Fic E247G under the ubiquitous Da-Gal4 driver relative to sibling controls. Among the indicated genomic BiP transgenes, only BiP T366A provides partial rescue of lethality in the BiP +/+ background and near complete rescue in a BiP G0102 null background. N = 3 biological replicas. At least 100 total flies scored for each replica. Bar graphs show means ± SD. DOI: https://doi.org/10.7554/eLife.38752.002 The following source data and figure supplements are available for figure 1: Source data 1. Relates to Figure 1B  The UPR protects eyes from overactive AMPylation To test the interaction between Fic-mediated AMPylation and the UPR, we employed an eye-specific Fic gain-of-function model. Eye-specific expression of the constitutively active AMPylating UAS-Fic E247G transgene using a LongGMR-Gal4 driver in otherwise fic-null animals results in a severe rough-eye defect (Casey et al., 2017). However, in a fic heterozygous background, eye-specific expression of constitutively active AMPylating Fic E247G yields only a mildly rough eye ( Figure 2A). We used this intermediate phenotype to assess genetic interactions between Fic E247G and components of the UPR with a candidate-based targeted RNAi screen (Figure 2-figure supplement 1). Fic E247G -induced eye roughness was significantly enhanced by knockdown of the key UPR components Perk, Atf4, and Ire1 ( Figure 2C-E and Figure 2-figure supplement 1), but not ATF6 ( Figure 2B). Furthermore, Xbp1 knockdown in conjunction with Fic E247G expression was lethal ( Figure 2F). By contrast, knockdown of these UPR genes in the absence of Fic E247G did not influence eye phenotype or fly survival ( Figure 2A'-F'). These genetic interactions suggest a role for UPR signaling in mitigating cellular stress imposed by expressing the constitutively active AMPylating Fic E247G in the eye.

AMPylation of BiP is necessary for maintaining vision
The findings that BiP is a target of Fic in vivo and that silencing UPR pathway components enhances the severity of the constitutively active AMPylating Fic E247G -induced rough-eye phenotype prompted us to assay the physiological effects of cellular stress in flies lacking BiP AMPylation. To do this we utilized flies that are either null for fic (fic 30C ) or express the AMPylation-resistant BiP T366A instead of wild-type BiP. By using this strategy, we are able to discern BiP AMPylation-specific changes from other potential changes that are due to as-yet-unknown targets of Fic AMPylation.
As previously reported in ERG recordings, fic-null flies display a reduction of the initial response (termed the ON Transient, Figure 3A) to a light pulse compared to wild-type controls. Interestingly, BiP T366A , but not BiP WT flies, exhibited the same defect in ON Transients as fic 30C mutants, consistent with BiP being the primary target of Fic AMPylation required for proper visual neurotransmission ( Figure 1-figure supplement 2). Of note, we used an eye-specific RNAi construct against white to minimize any effect of the mini-white gene used as a marker in these BiP transgenes. When we compared ERG traces of fic 30C and BiP T366A flies in white+ (red eyed) backgrounds, the   reductions in ON transients were no longer detectable ( Figure 1-figure supplement 3). This is likely due to the previously established protective effect provided by the red pigment in white + flies. Indeed, a similar white-dependent phenotype has been reported for photoreceptor synaptic plasticity after prolonged light exposure (Sugie et al., 2015;Damulewicz et al., 2017), consistent with previous observations that a functional white gene masks some degenerative phenotypes in the retina (Lee and Montell, 2004;Soukup et al., 2013). Therefore, we used white-eyed flies to test whether AMPylation may play a role in this type of photoreceptor plasticity, which is induced by rearing flies in uninterrupted low light for 72 hr (Sugie et al., 2015;Damulewicz et al., 2017).
We conducted ERG recordings under different light conditions with four fly lines, w 1118 , fic 30C , BiP WT and BiP T366A ( Figure 3B). Compared to age-matched siblings reared under the standard 12 hr Light:12 hr Dark (LD) treatment, fic 30C and BiP T366A flies reared for three days under continuous light (LL) at 500 lux exhibited severe ERG defects. This included reduction in the sustained negative potential (SNP), a measure of photoreceptor activation, and loss of ON and OFF transients, which reflect synaptic transmission to downstream L1/L2 lamina neurons ( Figure 3C and D). Wild-type controls maintained healthy OFF transients following LL, but ON transients were reduced, reflecting the sensitivity of this component to various light conditions ( Figure 3D). To test for behavioral consequences, we assayed w 1118 and fic 30C flies after 72 hr of LD or LL treatment for light-induced startle behavior using single-fly activity chambers (Ni et al., 2017). Following a 5 min light pulse, LD-reared fic 30C flies exhibited a startle response indistinguishable from control w 1118 flies, whereas fic 30C flies, but not w 1118 flies, failed to respond to the light pulse after 72 hr of LL (Figure 3-figure supplement 1). Thus, Fic-mediated AMPylation is required to maintain vision acuity under LL conditions. We next designed experiments to test whether these functional ERG changes in flies lacking AMPylation reflected light-induced neurodegeneration or a failure to adapt to constant stimulation. First, we asked if the LL-induced ERG defects of BiP T366A and fic 30C flies were reversible. We reared mutant and control flies for 72 hr in LL followed by 72 hr of recovery in LD (referred to as 'Rec'; Figure 3B). This recovery period was sufficient to restore both healthy OFF transients and SNPs in BiP T366A and fic 30C flies ( Figure 3C and D). Second, we asked if the intensity of the light would exaggerate the defects of BiP T366A and fic 30C flies. Exposure of mutant or control flies with 5000 lux, instead of 500 lux, did not alter the severity of ERG defects, indicating the changes were not simply a reflection of the increased amount of total light exposure during LL treatment (Figure 3-figure supplement 2). Third, we asked if prolonging the LL stress would alter the reversibility of these defects. Mutant flies reared under LL for ten days retained the capability to recover healthy ERG traces after only three days on LD, indicating that photoreceptors are not dying but maintained during prolonged light stress ( Figure 3-figure supplement 3). Together, these data support a model for a dysregulated adaptive response, rather than phototoxicity, inducing the reversible loss of OFF transients and reduced SNPs.

Constant light induces severe but reversible morphological defects in AMPylation mutants
To determine if the underlying eye substructures were being altered in these AMPylation deficient mutants, we performed TEM on ultrathin transverse eye sections. Under LD conditions, fic 30C and BiP T366A mutant and wild-type controls appeared indistinguishable ( Figure 4A). However, following 72 hr of LL (500 lux), fic 30C and BiP T366A mutants, but not w 1118 and BiP WT controls, displayed severe defects in the integrity of rhabdomeres, the microvilli-like membrane structures that house the phototransduction cascade ( Figure 4B). After a three-day recovery at LD, the rhabdomeres were nearly restored in both AMPylation-deficient mutants ( Figure 4C). To quantify these structural changes in large cohorts of flies, we assessed flies for the presence of wild-type 'deep pseudopupils' (DPP) ( Figure 4D). Visualization of the DPP affords an assessment of rhabdomere structural integrity in living flies (Franceschini and Kirschfeld, 1971). Consistent with the TEM data, 3 days of LL caused loss of DPP in fic 30C and BiP T366A , and DPPs returned after a 3 day recovery ( Figure 4D). This suggests that proper regulation of BiP through AMPylation is required for maintaining both function and structure of photoreceptor cells. Fic is required for ER homeostasis in the visual system during constant light Given the unique role of Fic in both AMPylating and deAMPylating BiP to modulate its chaperone activity and maintaining ER homeostasis, we evaluated fic 30C flies for changes in the UPR under LD, LL, and Rec conditions. First, we performed immunohistochemistry against BiP, a transcriptional target of the UPR, which is upregulated during states of ER stress (Gardner et al., 2013;Ham et al., 2014). After 3 days of LL, sections of fic 30C revealed increased levels of BiP in retinas and in the lamina neuropils where photoreceptor axons form synapses onto lamina neurons. BiP levels returned to control levels following three days of recovery ( Figure 5A and B). To further assess UPR signaling in these tissues, we utilized a sensor for Ire1-mediated Xbp1 splicing (Sone et al., 2013) and an Atf4 translational reporter which serves as a proxy for Perk-mediated phosphorylation of eIF-2a (Kang et al., 2015). In wild-type flies, Xbp1-GFP was slightly induced in the lamina after 24 hr of LL in wild-type flies and the signal decreased over time ( Figure 5C, top row, and Figure 5D). However, in fic 30C flies, the Xbp1-GFP signal in the lamina continued to increase after 48 hr of LL and remained elevated after 72 hr (Figure bottom row, and Figure 5D). In the retina, control flies showed little to no increase of Xbp1-GFP levels, while fic 30C flies showed a significant transient increase after one and two days LL. With the Atf4-DsRed reporter, we observed a significant increase of signal in both the retina and lamina of wild-type flies after one day, but no difference in fic 30C mutants at one or two days LL when compared to LD controls ( Figure 5E and F). However, by three days of LL, Atf4-DsRed reporter activity in the wild-type flies returned to control levels, while the fic 30C mutants showed a significant increase in both the retina and lamina neuropil ( Figure 5E and F). These changes in UPR signaling were reversible as each of the reporters returned to near control intensity after 72 hr of LD recovery ( Figure 5A,C and E, last columns). The elevated UPR response in fic 30C mutants correlated with the timing of the observed severe defects in the integrity of rhabdomeres ( Figure 4B). Together, these data identify a role for Fic-mediated BiP AMPylation in regulating ER stress during homeostatic responses of the visual system.

Discussion
Here we demonstrate that BiP is a critical in-vivo target of Fic-mediated AMPylation, as an AMPylation-resistant BiP blocks lethality caused by over-expressed constitutively active AMPylating Fic E247G and recapitulates fic-null vision defects. This work also sheds light on the physiological connection between AMPylation/deAMPylation of BiP and the UPR. We observe genetic interactions with the constitutively active AMPylating Fic E247G and the UPR sensors Ire1 and Perk as well as their effectors, perhaps due to the critical role of BiP as both a regulator (Bertolotti et al., 2000;Shen et al., 2005;Carrara et al., 2015;Amin-Wetzel et al., 2017) and downstream transcriptional target of the UPR (Kozutsumi et al., 1988;Ham et al., 2014). Indeed, we hypothesize that unregulated Fic E247G , in the absence of deAMPylation activity, AMPylates excess BiP, excluding it from its normal chaperone role and leading to cell death (Casey et al., 2017;Truttmann et al., 2017). In support of this hypothesis, the developmental defects due to excessive unregulated AMPylation are suppressed by AMPylation-resistant BiP T366A . Furthermore, the enhancement of the rough-eye Fic E247G phenotype by knockdown of the Ire1 and PERK pathways suggest a protective role for the UPR, perhaps through the known effects on translation by Ire1-mediated decay of mRNA, Xbp1-driven transcription or Perk-mediated phosphorylation of eIF-2a (Gardner et al., 2013).
Our work supports an in-vivo requirement for Fic-mediated AMPylation of BiP T366 in the context of long-term adaptation to prolonged light exposure. BiP T366A replacement mutants phenocopy ficnull flies in both the light-induced blindness and the unexpected recovery from this phenotype. We observe changes in the ON/OFF transients reflecting defects in transmission to the downstream L1/ L2 neurons and also a decreased magnitude in the SNP of photoreceptors. Functional changes in the SNP component are mirrored in the structural changes of photoreceptor rhabdomeres. Rhabdomere appearance of AMPylation-deficient flies was reminiscent of retinal degeneration mutants (Smith et al., 1991;Ryoo et al., 2007), however the remarkable recovery of the tissue structure in three days of LD is unlike any reported retinal degeneration phenotype. It is worth emphasizing that both rhabdomere structure and synaptic transmission recover, as these are not necessarily linked. For example, some mutant alleles of ninaE, encoding the R1-6 rhodopsin, or rdgC, encoding a rhodopsin phosphatase, maintain synaptic transmission despite substantial rhabdomere degeneration  (Webel et al., 2000;Lee and Montell, 2001). Together, our findings demonstrate a seminal role for Fic-mediated AMPylation of BiP in vivo: enabling photoreceptors to adapt and maintain both structural and functional integrity during periods of prolonged stress due to uninterrupted light stimulation. The exact mechanism through which these defects in fic mutants arise remains undetermined, but previous studies have demonstrated a requirement for maintaining normal ER folding and trafficking of transmembrane visual signaling components such as Rhodopsin (Colley et al., 1995;Rosenbaum et al., 2006). We hypothesize this intense demand for proper ER stress regulation sensitizes the eye to otherwise mild defects in fic mutants, and the additional demands posed by the stress of constant light stimulation.
We also observed that loss of BiP AMPylation deregulates, but does not block, the UPR to this physiological stress. This observation supports previous claims that AMPylation and deAMPylation of BiP aids in maintaining ER homeostasis ( Figure 1A) by establishing a reserve pool of BiP that can readily be deAMPylated in response to acute ER insults (Ham et al., 2014;Casey et al., 2017;Preissler et al., 2017a). This regulation would allow for fine-tuning of the UPR response under specific contexts, comparable to findings in C. elegans in which fic-1 and hsp3 (a BiP homologue) mutants are sensitive to bacterial infection (Truttmann et al., 2016). We speculate that the eye requires tight control of BiP activity, to facilitate adaptation of the vision signaling cascade. Under standard LD conditions, only slight differences are observed, presumably because ER stress is low ( Figure 5C and D, first column). However, exposure to constant light results in ER stress and UPR signaling which wild-type flies can clear, presumably because there is a reserve pool of AMPylated BiP to rapidly respond to the stress. In fic mutants, we speculate, loss of the reserve BiP results in the UPR dysregulation revealed by the IreI and Perk activity reporters ( Figure 5C and D) as a larger proportion of BiP would be previously engaged and not able to respond to the extra stress. Additionally, the prolonged UPR response in the eyes with dysregulated AMPylation may result in increased expression of UPR-regulated proteins while blocking production of the visual signaling components necessary for adapting to transient stress. It is unknown whether the ratio of AMPylated to unmodified BiP directly regulates activation of the UPR sensors, given the multiple levels of feedback on the system (Gardner et al., 2013). It remains to be determined whether the effects on UPR signaling are the direct cause of the visual system defects or if the phenotypes in fic and BiP T366A mutants are primarily due to changes in protein folding and secretion. We also cannot rule out the influence of AMPylation on other ER processes that involve BiP, such as ER-associated degradation (Hegde et al., 2006).
BiP expression is subject to multiple levels of feedback regulation and is induced when the UPR is activated (Kozutsumi et al., 1988;Ma and Hendershot, 2003). However, in a negative feedback loop, BiP also inhibits activation of the UPR sensors Ire1, Perk, and Atf6, through direct binding (Bertolotti et al., 2000;Shen et al., 2005). It remains unknown how AMPylation affects the interactions of BiP with these proteins in vivo; however, in-vitro work suggests that AMPylation of BiP abolishes its inhibitory effect on Ire1 dimerization and activation (Amin-Wetzel et al., 2017). We speculate that the loss of BiP AMPylation/deAMPylation cycle in a fic null affects both the ability of BiP to quickly respond to misfolded protein aggregates and to regulate UPR activation. Future studies on the dynamic role of reversible BiP AMPylation and its interaction with downstream UPR sensors should provide unique insight into neuronal plasticity and regeneration. intensity, normalized to wild-type LD controls, in the lamina neuropil and retina from two independent experiments. (E) Representative images of an Atf4-dsRed reporter in either a wild-type or fic 30C background following LD, 1 day LL, 2 day LL, 3 day LL, and Recovery conditions. (F) Quantification of Atf4-dsRed intensity, normalized to wild-type LD controls, in the lamina neuropil and retina from two independent experiments. For all experiments, n = 8 flies per genotype/condition, with exceptions of outliers falling three standard deviations outside the mean. Bar graphs show means ± SD. For all experiments, significance is indicated for treatment compared to the LD condition for the corresponding genotype. ****p<0.0001; ***p<0.001; **p<0.01; *p<0.05. All scale bars: 50 mM. DOI: https://doi.org/10.7554/eLife.38752.021 The following source data is available for figure 5: Source data 1. Relates to Figures 5B, D

Generation of genomic BiP transgenes
PCR-amplified BiP genomic sequences were cloned into a pAttB vector and a 3X-FLAG tag was inserted after the N-terminal signal sequence. To create the T366A and T518A mutations, gBlocks (IDT, Coralville, IA) for the mutant sequences were synthesized and subcloned into the pAttB_genomic BiP vector via NEB HiFi Assembly Kit (NEB, Ipswich, MA). These constructs were sequenceverified and injected into embryos (BestGene, Chino Hills, CA) for insertion at the 89E11 landing site. Expression levels of FLAG-BiP transgenes were determined with western blotting. In brief, fly heads were homogenized in lysis buffer (10% SDS, 6M urea, and 50 mM Tris-HCl, pH 6.8 + 10% DTT), sonicated for 5 min, boiled for 2 min, and centrifuged for 10 min at 10,000 g to remove debris.

Generation of GMR_dsRNA white transgenes
To make the eye-targeted dsRNA constructs against the white gene, the dsRNA sequence was obtained from a pAttb-UAS S.cer -dsRNA white vector (Kalidas and Smith, 2002) (a gift from Dr. Dean Smith, UT Southwestern Medical Center, TX) and the UAS S.cer -Hsp40 promotor sequence was replaced with a 5X-GMR promotor sequence, synthesized as a gBlock (IDT) and cloned with NEB HiFi Assembly Kit (NEB).

Fly rearing conditions
All flies were reared on standard molasses fly food, under room temperature conditions. For light treatments, flies were collected within one to two days of enclosing, and placed in 5 cm diameter vials containing normal food, with no more than 25 flies, and placed at either LD (lights ON 8am/ lights OFF 8pm) or LL. ERGs, head dissections and behavior assays were performed between 1pm and 4pm. The same intensity white LED light source was used for both conditions and flies were kept the same distance away from the light source, which amounted to approximately 500 lux. LD and LL treatments were done at 25˚C. For the UPR and Fic E247G rough-eye interaction experiments, all flies were raised at 28˚C.

Electroretinograms
ERGs were recorded as previously described (Montell, 2012). Glass electrodes filled with 2M NaCl were placed in the fly thorax and surface of the corneal lens (recording). A computer-controlled LED light source (MC1500; Schott, Mainz, Germany) was pulsed for 1 s at 4 s intervals. The resulting ERG traces were collected by an electrometer (IE-210; Warner Instruments, Hamden, CT), digitized with a Digidata 1440A and MiniDigi 1B system (Molecular Devices, San Jose, CA), and recorded using Clampex 10.2 (Molecular Devices) and quantified with Clampfit software (Molecular Devices). Flies were assayed in batches of eight to ten, and resulting quantifications are pooled from three independent biological repeats.

Deep pseudopupil analysis
Flies were anesthetized on CO 2 and aligned with one eye facing up. Using a stereoscopic dissection microscope, each fly was scored for presence or loss of the deep pseudopupil (Franceschini and Kirschfeld, 1971), and the percentage of flies with intact pseudopupils was calculated. For each genotype/treatment, over 50 flies were scored per replica and three biological replicas were performed (n = 3).

Light-startle behavior assay
Assay was adapted from a previously described method (Ni et al., 2017

Scanning electron microscopy
SEMs of fly eyes were obtained as previously described (Wolff, 2011). Eyes were fixed in 2% paraformaldehyde, 2% glutaraldehyde, 0.2% Tween 20, and 0.1 M cacodylate buffer, pH 7.4, for 2 hr. Samples were washed four times with increasing ethanol (25-100%) for 12 hr each followed by a series of hexamethyldisilazane washes (25-100% in ethanol) for one hour each. Flies were air dried for 24 hr, mounted on SEM stubs, and the bodies were coated in fast-drying silver paint. Flies were sputter coated with a gold/pallidum mixture for 60 s and imaged at 900X magnification, with extra high tension set at 3.0 kV on a scanning electron microscope (Sigma SEM; Carl Zeiss, Germany). Ten flies per genotype were mounted and three were imaged (n = 3). Blinding of the samples' identity to the user acquiring the images was performed.

Transmission electron microscopy
Quantification of fluorescence staining Fluorescence images were quantified using ImageJ (NIH) adapting previous methods (Nandi et al., 2017). For each antibody, a threshold was determined, removing the lowest 10% of signal in LD control samples (to reduce variation from low level background signals). This same threshold was applied, and a mask was created for every image in a batch of staining. Within a 1 mm optical slice, the retina and lamina regions were selected manually using an Actin stain and assigned as Regions of Interest. The integrated pixel intensity per unit area was measured within this selected area, redirecting to the threshold mask. In each fly, four sequential optical slices were quantified and averaged. For each genotype and treatment, four flies were quantified from two independent biological replicas for a total of eight flies. Data was normalized to the wild-type LD control for each replica. Outliers of greater than three standard deviations were omitted from the analysis.

Statistics
Statistics were performed using GraphPad Prism 7. Normality of data distribution was determined using D'Agostino's and Pearson's normality test. For the genetic analysis in Figure 1 and the ERG measurements in Figure 1 Figure 3-figure supplement 3, significance was determined using one-way ANOVA, followed by Tukey's multiple comparisons tests. Statistical significance for non-parametric data, including the ERGs with light treatment quantifications in Figure 2, was determined by Kruskal-Wallis tests followed by multiple comparisons testing with Dunn's correction. For the image quantification data in Figure 4, significance was determined by two-way ANOVA followed by multiple comparisons with Benjamini-Krieger-Yekutieli's False Discovery Rate correction. All tests were two-sided with no experimental matching. RStudio (version 1.1.442, 2018, RStudio, Inc.) was used for Fisher's Exact Tests for the eye interaction screen, with Bonferroni's multiple comparison method to determine significance (Figure 2-figure supplement 1). Tests were two-sided. When possible, blinding of sample identities was performed for image acquisition and fluorescence intensity quantification. Sample sizes for ERG assays, EM experiments, fluorescence quantifications and fly genetic analysis were determined based from previous experience (Rahman et al., 2012;Stenesen et al., 2015;Nandi et al., 2017).