Decoupling of Apoptosis from Activation of the ER Stress Response by the Drosophila Metallopeptidase superdeath

Abstract

Genetic diseases display a great deal of variability in patient outcomes, much of which is caused by differences in genetic background. The endoplasmic reticulum (ER) stress response commonly modifies degenerative disease...

Endoplasmic reticulum (ER) stress-induced apoptosis is a primary cause and modifier of degeneration in a number of genetic disorders. Understanding how genetic variation influences the ER stress response and subsequent activation of apoptosis could improve individualized therapies and predictions of outcomes for patients. In this study, we find that the uncharacterized, membrane-bound metallopeptidase CG14516 in Drosophila melanogaster, which we rename as SUPpressor of ER stress-induced DEATH (superdeath), plays a role in modifying ER stress-induced apoptosis. We demonstrate that loss of superdeath reduces apoptosis and degeneration in the Rh1^G69D model of ER stress through the JNK signaling cascade. This effect on apoptosis occurs without altering the activation of the unfolded protein response (IRE1 and PERK), suggesting that the beneficial prosurvival effects of this response are intact. Furthermore, we show that superdeath functions epistatically upstream of CDK5—a known JNK-activated proapoptotic factor in this model of ER stress. We demonstrate that superdeath is not only a modifier of this particular model, but affects the general tolerance to ER stress, including ER stress-induced apoptosis. Finally, we present evidence of Superdeath localization to the ER membrane. While similar in sequence to a number of human metallopeptidases found in the plasma membrane and ER membrane, its localization suggests that superdeath is orthologous to ERAP1/2 in humans. Together, this study provides evidence that superdeath is a link between stress in the ER and activation of cytosolic apoptotic pathways.

ER stress, apoptosis, metallopeptidase, modifier genes

ENDOPLASMIC reticulum (ER) stress-induced apoptosis is a primary or contributing cause of degeneration in a wide variety of diseases, such as type 2 diabetes and Alzheimer’s disease (Yilmaz 2017; Alicka and Marycz 2018; Kurtishi et al. 2018). Reducing stress-induced apoptosis could be the key to slowing the progression of these diseases, and indeed a major focus of therapeutic development is to identify compounds that can inhibit apoptosis. Therapeutics that target cell death without impacting the beneficial survival pathways activated by the ER stress response are essential in the treatment of degenerative diseases.

ER stress occurs when protein folding is disrupted, leading to the accumulation of misfolded proteins in the ER. ER stress activates the Unfolded Protein Response (UPR)—a massive transcriptional response that, if successful, will return the ER and cell to homeostasis (Schröder and Kaufman 2005). This response is regulated by three sensors that are located in the ER membrane: IRE1, PERK, and ATF6 (Sano and Reed 2013). Upon sensing misfolded proteins, the IRE1 endonuclease domain is activated, allowing it to noncanonically splice the mRNA for the transcription factor Xbp1. Once spliced, Xbp1 is translated and translocates to the nucleus, activating expression of UPR target genes (Cox and Walter 1996; Sidrauski and Walter 1997). IRE1 also degrades a number of ER-targeted transcripts under ER stress conditions (Regulated IRE1 Dependent mRNA Decay or RIDD) (Hollien and Weissman 2006; Hollien et al. 2009; Sano and Reed 2013). Upon ER stress, PERK phosphorylates the translation initiation factor eIF2α. This greatly reduces translation of mRNA transcripts with the canonical transcription initiation mechanism, while allowing increased translation of specific stress-regulated transcripts such as the transcription factor ATF4 (Sano and Reed 2013). Finally, under ER stress, ATF6 is trafficked to the Golgi, where it is processed, releasing the cytoplasmic domain to act as a transcription factor (Sano and Reed 2013). Under conditions of chronic or extreme stress, the UPR may eventually induce apoptosis (Sano and Reed 2013). In Drosophila, this is primarily through the activation of the Jun Kinase (JNK) signaling cascade downstream of the apoptotic regulator CDK5 (Kang et al. 2012; Sano and Reed 2013). The mechanism through which CDK5 is activated by ER stress and the UPR is unknown.

The ER stress response is strongly influenced by genetic variation (Dombroski et al. 2010; Chow et al. 2013, 2015, 2016). In a previous study, we modeled the impact of genetic variation on the ER stress response by overexpressing mutant rhodopsin (Rh1^G69D) in the developing Drosophila eye (Chow et al. 2016). We crossed this model into the ∼200 genetic backgrounds of the Drosophila Genetic Reference Panel (DGRP) (Mackay et al. 2012; Chow et al. 2016). We measured retinal degeneration and performed a genome-wide association analysis to identify modifier variation that is associated with differences in degeneration. We generated a list of 84 conserved candidate modifier genes, ∼50% of which have known roles in apoptotic pathways and/or the ER stress response (Chow et al. 2016). By characterizing these modifiers, we can learn more about the pathogenesis and progression of ER stress-related diseases.

Here, we report a novel function for one of these modifiers, the Drosophila metallopeptidase CG14516, which we rename SUPpressor of ER stress-induced DEATH (superdeath). We demonstrate that, in this Rh1^G69D model of ER stress, loss of superdeath results in partial rescue of degeneration. This reduced degeneration is accompanied by reduced apoptosis and JNK signaling. This is in the absence of any detectable changes in activation of the ER stress sensors IRE1 or PERK, suggesting that superdeath lies downstream of the UPR in the activation of apoptosis. Epistasis experiments indicate that superdeath lies genetically upstream of CDK5, and that the changes observed in degeneration are possibly due to reduced activation of CDK5. While superdeath is orthologous to a number of mammalian metallopeptidases, we show that Superdeath protein can localize to the ER, suggesting that it is functionally related to the ER-associated proteases 1 and 2 (ERAP1 and ERAP2) (Haroon and Inman 2010). Our results indicate that inhibition of Superdeath/ERAP1/ERAP2 would reduce apoptosis levels under conditions of ER stress, while retaining the beneficial effects of UPR activation, making it a valuable target for therapeutic development.

Materials and Methods

Fly stocks and maintenance

Flies were raised at room temperature on standard diet based on the Bloomington Stock Center standard medium with malt. The strain containing GMR-GAL4 and UAS-Rh1^G69D on the second chromosome (GMR > Rh1^G69D) has been described previously (Chow et al. 2016; Palu and Chow 2018). The following strains are from the Bloomington Stock Center: MS1096-GAL4 (8696), UAS-superdeath RNAi (42947), a second UAS-superdeath RNAi (35802), control attP40 (36304), control attP2 (36303), superdeath-GFP (64447). The UAS-CDK5 RNAi (104491), the third UAS-superdeath RNAi (108616), and the control attP (60100) lines are from the Vienna Drosophila Resource Center. The puc-LacZ enhancer trap is available from Kyoto (109029). The strains containing the UAS-Xbp1-EGFP transgenes were a gift from Don Ryoo (NYU).

Eye/wing imaging

For eye and wing images, adult females were collected under CO₂ anesthesia, aged to 2–7 days, then flash frozen on dry ice. Eyes were imaged at 3× magnification using a Leica EC3 camera. Wings were dissected away from the body, then imaged at 4.5× magnification using the same camera. Eye and wing area were measured in ImageJ as previously described (Chow et al. 2016).

Immunohistochemistry

Eye discs and salivary glands were dissected from wandering L3 larvae in cold 1× phosphate-buffered saline (PBS), then immediately transferred to cold 4% paraformaldehyde (PFA) on ice. S2 cells were treated while adhered to sterile plastic coverslips. Samples were fixed in 4% PFA for 15–20 min, then washed in 1× PAT (1× PBS with 0.1% TritonX100) prior to blocking with 5% normal donkey serum. Samples were stained with primary antibodies for rhodopsin (1:50; #4C5 Developmental Studies Hybridoma Bank), green fluorescent protein (GFP) (1:2000; #A6455 Thermo-Fisher, 1:100; #M048-3 MBL), LacZ (1:20-1:50; #40-1a DSHB), Calnexin 99A (1:50; #cnx99A 6-2-1 DSHB), Golgin-84 (1:50; #golgin-84 12-1 DSHB), Lamp1 (1:100; # ab30687 Abcam), and V5 (1:500; #13202S Cell Signaling, and 1:250; #R960-25 Thermo-Fisher). Apoptosi was monitored using the ApopTag Red In Situ Apoptosis Detection Kit (#S7165 Millipore). Samples were mounted in Slowfade Diamond Antifade Mountant (#S36967 Thermo-Fisher) and imaged with an Olympus FV1000 confocal microscope or a Nikon A1 confocal microscope.

Western blots

Protein was isolated from 10 wandering L3 larvae brain-imaginal disc complexes or from S2 cells, and homogenized in 1× Laemmli/radioimmunoprecipitation assay (RIPA) buffer containing 1× protease inhibitors (Roche cOmplete Mini EDTA-free protease inhibitor tablets) as well as the phosphatase inhibitors Calyculin A and okadaic acid. Equivalent amounts of protein were resolved by SDS-PAGE (10% acrylamide) and transferred to polyvinylidene fluoride (PVDF) membrane by semi-dry transfer. Membranes were then treated with either 5% bovine serum albumin (BSA) or 5% milk protein block in 1× Tris Buffered Saline with Tween (TBST) prior to immunoblotting. Blots were probed with antibodies for P-eif2α (1:1000; #32157 abcam), Pan-eif2α (1:500; #26197 abcam), and tubulin (1:2000; #12G10 Developmental Studies Hybridoma Bank). Blots shown are representative of at least three biological replicates, and quantification was performed using ImageJ.

Tunicamycin treatment

Crosses to generate the indicated genotypes were set up on egg caps containing yeast paste. L2 larvae were then treated with either 10 µg/ml Tunicamycin (diluted 1:1000 from a 10 mg/ml stock solution) or 1:1000 DMSO in Schneider’s media for 5 hr at room temperature. The larvae were then washed in 1× PBS twice and placed on standard media. Viability was determined by survival to pupation. Survival for each genotype was normalized to the DMSO-treated control condition. Each replicate represents 112–130 larvae per genotype.

S2 cells

DsRNA was generated using the MEGAscript T7 Transcription kit (#AM1334 ThermoFisher), with primers for EGFP (F: TTAATACGACTCACTATAGGGAGACCACAAGTTCAGCGTGTCC and R: TTAATACGACTCACTATAGGGAGAGGGGTGTTCTGCTGGTAGTG) and superdeath (F: TTAATACGACTCACTATAGGGAGATCCGGTGGTTAAGGTGTCAAGG and R: TTAATACGACTCACTATAGGGAGAGCCGGAGTTGACGAACATGG). S2 cells were treated with DsRNA against EGFP (as a control) or against superdeath at a density of ∼2 × 10⁶ cells/ml in a 24-well plate. Cells were incubated with DsRNA for 4–7 days before being split and treated with either 2 mM DTT or DMSO as a control. Cells were treated for 4 hr for Xbp1 splicing and P-eif2a measurements and for 7.5 hr for rpr and hid measurements. RNA was isolated from cells using the Direct-zol RNA Miniprep Kit and used to generate cDNA (Protoscript II, NEB). Protein was isolated from cells as described above.

Knockdown of superdeath was confirmed using qPCR (primers: F: ATTCGCAGCAGTTTCCACCAC and R: TTCGTGGCGAACTTGAACAGC). Xbp1 splicing was evaluated from the cDNA using PCR (primers for Xbp1: F: TCAGCCAATCCAACGCCAG and R: TGTTGTATACCCTGCGGCAG). The spliced and unspliced bands were separated on a 12% acrylamide gel, and the proportion of these bands quantified using ImageJ software. rpr (F: TTGCGGGAGTCACAGTGGAG and R: AATCCTCATTGCGATGGCTTGC) and hid (F: TACCTACTACGCGGGCTACACG and R: TGGTACTCGCGCTCATCCTC) levels were analyzed by qPCR. Transcript levels were normalized to rpl19 (F: AGGTCGGACTGCTTAGTGACC and R: CGCAAGCTTATCAAGGATGG) and compared between matched DMSO or DTT-treated S2 cells.

Cloning

Superdeath was overexpressed in S2 cells using the TOPO TA Cloning Kit (Thermo-Fisher). The coding sequence for superdeath was expressed from the pMT-DEST48 inducible expression vector with a C-terminal V5 tag. S2 cells adhered to sterile glass cover slips were made competent using the Calcium Phosphate Transfection Kit (Thermo-Fisher), and expression of the construct was induced with 500 µM CuSO₄ for 66 hr. Cells were then stained for the V5 tag and other subcellular markers to determine Superdeath protein localization as described above.

Statistics

Statistics were calculated using R or Prism software. P-values were determined using either one-way or two-way ANOVA for eye size, fluorescence levels, and transcript levels in qPCR. A pairwise T-test was performed for larval tunicamycin treatment. A cutoff of P = 0.05 was used for significance.

Reagent and data availability

The authors state that all data necessary for confirming the conclusions presented in the article are represented fully within the article. Strains and stocks are available upon requests. Supplemental material is available at figShare (https://figshare.com/articles/SupplementalFiles_superdeath_pdf/11401827).

Results

A single nucleotide polymorphism in CG14516 is associated with variation in expression and degeneration in the Rh1^G69D model of ER stress

In a previous study, we examined the impact of genetic variation on ER stress-induced apoptosis using a model of ER stress in which we overexpressed mutant, misfolding rhodopsin in the developing eye imaginal disc using the GAL4/UAS system (GMR-GAL4 > UAS-Rh1^G69D) (Chow et al. 2016). We crossed this model of retinal degeneration and ER stress into the Drosophila Genetic Reference Panel (DGRP)—a collection of ∼200 wild-derived isogenic strains (Mackay et al. 2012). We performed a genome-wide associated analysis to identify candidate modifier genes of ER stress and degeneration (Chow et al. 2016). As half of the genome in the F1 flies came from the DGRP parent and the other from the model “donor” strain, all candidate modifiers were identified from dominant interactions. One of these candidate genes is CG14516—a previously uncharacterized membrane-bound metallopeptidase. A single nucleotide polymorphism (SNP) in the first intron of CG14516 is significantly associated with eye size in the Rh1^G69D model (P = 2.24 × 10⁻⁵, Figure 1A). Strains carrying the major “A” allele have larger, less degenerate eyes (21,922 ± 2279 pixels) compared to those carrying the minor “T” allele (19,863 ± 2873 pixels). We next asked if this SNP is associated with differences in CG14516 expression levels by mining previously published, publically available RNA-seq expression data from the DGRP (Huang et al. 2015). The minor “T” allele is associated with a small, but significant, increase in CG14516 expression in adult females [9.09 ± 0.24 log₂ (FPKM +1)] compared to the major “A” allele [8.84 ± 0.30 log₂ (FPKM +1), P = 4.92 × 10⁻³, Figure 1B] (Huang et al. 2015). Importantly, expression levels of CG14516 in adult females is inversely correlated with eye size in the presence of Rh1^G69D (r = −0.25, P = 0.0013, Figure 1C). These data show that reduced expression of CG14516 is associated with a decrease in Rh1^G69D-induced degeneration, suggesting that loss of CG14516 function should reduce ER stress-induced degeneration. We have therefore named this gene SUPpressor of ER stress-induced DEATH (superdeath).

Figure 1

A single nucleotide polymorphism (SNP) in CG14516 is associated with changes in expression and degeneration. Variation in sequence and expression of the Drosophila melanogaster gene CG14516 is associated with eye size in the Rh1^G69D model of ER stress. (A) The 3R:24966022 SNP in CG14516 (BDGP R5/dm3) is associated with Rh1^G69D-induced degeneration. Rh1^G69D DGRP eye size is plotted by allele identity. Strains carrying the minor “T” allele (19,863 ± 2873 pixels, N = 32) have significantly smaller eyes than those carrying the major “A” allele (21,922 ± 2279 pixels, N = 136). Representative strains are shown. (B) Expression of CG14516 in strains carrying either the “A” or the “T” allele was determined from previously published, publically available RNA sequencing data in adult females (Huang et al. 2015). CG14516 levels was significantly increased in strains carrying the minor “T” allele (9.09 ± 0.24 units, N = 33) as compared to those carrying the major “A” allele (8.84 ± 0.30 units, N = 147). (C) Eye size in the Rh1^G69D DGRP strains is inversely correlated with CG14516 expression levels in adult females (r = −0.25, N = 167, P = 0.0013). Raw data for (A) and (C) were taken from Chow et.al. 2016 and Huang et.al. 2015, respectively. Values are average ± SD **P < 0.005, ****P < 0.00005.

Loss of superdeath expression rescues Rh1^G69D-induced apoptosis and degeneration

To test the impact of superdeath expression on Rh1^G69D-induced degeneration, we expressed an RNAi construct targeting superdeath in the presence of the Rh1^G69D model of ER stress (Rh1^G69D/superdeathi). As in the original model, we used the GMR promoter to turn on expression of Rh1^G69D, RNAi, and other UAS-transgenes early in L3 development (FlyBase Curators 2017). Expression of Rh1^G69D begins to immediately activate the UPR. We performed all staining and Western blots on tissues collected at the late, wandering L3 stage, well after the UPR has been initiated. In previous studies using this model (Kang et al. 2012; Palu and Chow 2018), the late L3 stage has been successfully monitored for activation of the UPR and associated cell death, while adults exhibit stable, abnormal eye phenotypes upon eclosion. To confirm knockdown levels, ubiquitous expression of the superdeath RNAi construct results in a 75% reduction in superdeath expression (0.278 ± 0.072 relative to controls at 1.00 ± 0.16, P = 1.77 × 10⁻⁴, Supplemental Material, Figure S1A).

We found that, in the absence of superdeath, eye size is significantly increased (15,299 ± 1658 pixels) compared to a control that is only expressing Rh1^G69D (Rh1^G69D/control) (11,942 ± 473 pixels, P = 8.15 × 10⁻⁶, Figure 2A). This increase was recapitulated using two additional RNAi lines (Figure S1, B and C), indicating that the increase in eye size is indeed due to a loss of superdeath expression. We observe a slight increase in eye size when we reduce expression of superdeath in wild-type eyes compared to controls, but no qualitative difference (28,867 ± 1566 pixels vs. 25,968 ± 1026 pixels in controls, P = 1.16 × 10⁻⁴, Figure 2B). While this may indicate an impact on apoptosis under control conditions, the proportional difference is too small to detect dramatic differences. Indeed, wild-type eye imaginal discs display so little apoptosis at the stage we are observing that we would be unable to measure a reduction in cell death.

Figure 2

Loss of superdeath reduces ER stress-associated apoptosis. Reducing expression of superdeath reduces apoptosis and degeneration in models of ER stress. (A) Degeneration caused by overexpression of Rh1^G69D is partially rescued by RNAi-mediated knockdown of superdeath expression (15,299 ± 1658 pixels, N = 10 in Rh1^G69D/superdeathi flies as compared to 11,942 ± 473 pixels, N = 10 in Rh1^G69D/controls). (B). Eye size also showed a small increase when the superdeath RNAi construct was expressed in a wild-type background (28,867 ± 1566 pixels, N = 10) as compared to controls (25,968 ± 1026 pixels, N = 10), but no qualitative differences were observed. (C). Rh1^G69D/superdeathi eye imaginal discs display reduced apoptosis compared to Rh1^G69D/controls as measured by TUNEL staining. (D) S2 cells treated with DsRNA against EGFP showed increased expression of the apoptotic gene rpr after 7.5 hr of DTT exposure (2.23 ± 0.33, N = 6) as compared to DMSO-treated control cells (1.00 ± 0.09, N = 6). Activation of rpr expression was significantly reduced in S2 cells that were treated with DsRNA against superdeath (1.47 ± 0.07, N = 6 in DTT-treated cells compared to 1.00 ± 0.11, N = 6 with DMSO). (E) Activation of JNK signaling was reduced in Rh1^G69D/superdeathi eye imaginal discs compared to Rh1^G69D/controls as determined by expression of puc-LacZ. (F) When quantified, LacZ levels were significantly lower in Rh1^G69D/superdeathi eye discs (0.367 ± 0.082, N = 3) as compared to Rh1^G69D/controls (1.00 ± 0.39, N = 4). Values are average ± SD. Bar = 0.1 mm. *P < 0.05, **P < 0.005, ****P < 0.00005.

In our laboratory, we have validated 25 candidates from the original screen in this way. Loss of 7 of these genes enhances degeneration, loss of 8 genes enhances degeneration, and loss of 10 genes has no effect (Chow et al. 2016; Palu and Chow 2018; data not published). The varied impact of RNAi knockdown of candidate genes suggests that the phenotypes we observe are unlikely to be a result of either of GAL4 dilution or nonspecific activation of the RNAi system.

Because reduced degeneration in the Rh1^G69D model of ER stress is often accompanied by reduced apoptosis (Kang et al. 2012; Palu and Chow 2018), we measured cell death in the eye imaginal disc, a developmental structure that will eventually become the adult eye. This tissue is also the site of mutant rhodopsin overexpression, and where ER stress and apoptosis are being induced. TUNEL staining indicates that there is reduced apoptosis in the absence of superdeath (47 ± 27 cells) compared to Rh1^G69D/controls (104 ± 21 cells, P = 0.044, Figure 2, C and D). Our findings demonstrate that superdeath is required for high levels of ER stress-induced apoptosis and subsequent degeneration. The rescue effect observed upon loss of superdeath is likely due to this reduced apoptosis.

Because ubiquitous loss of superdeath is lethal during pupal stages (see below), we were unable to use null alleles to confirm our tissue-specific RNAi findings. To confirm these in an alternate system, we treated Drosophila S2 cells with DsRNA targeting either superdeath or EGFP as a control. The DsRNA targets a region of superdeath independent from that targeted by the RNAi, making it a good validation of the in vivo model. Treatment of S2 cells with DsRNA against superdeath resulted in 90% reduction in superdeath expression (0.102 ± 0.038 relative to controls at 1.00 ± 0.29, P = 1.9 × 10⁻⁵, Figure S1D).

S2 cells were then treated with dithiothreitol (DTT) to induce ER stress. DTT disrupts disulfide bond formation, and results in massive protein misfolding, ER stress, and the induction of the UPR (Jämsä et al. 1994). The use of a chemical induction method for ER stress allows us to test superdeath with an alternate, independent form of ER stress. Additionally, all cells in the culture will be undergoing ER stress. It is easier to detect stress responses in S2 cells than in the in vivo model, where the signal is often overwhelmed by the background signal from other, unaffected, cells in the tissue.

We monitored expression of the apoptosis-associated transcripts reaper (rpr) and hid to determine if there is a difference in the induction of apoptosis when superdeath expression is reduced in S2 cells; reaper and hid are inhibitor proteins that target the Inhibitor of Apoptosis Proteins (IAPs), enabling caspase activation and the initiation of apoptosis (Hay et al. 1995; Kuranaga et al. 2002). Their expression is upregulated under most apoptosis-inducing conditions (Shlevkov and Morata 2012; Mollereau and Ma 2014). Additionally, overexpression of both rpr and hid is sufficient to induce apoptosis (Hay et al. 1995; Goyal et al. 2000). Their expression is a good read-out of the activation of a variety of apoptotic pathways (Bilak and Su 2009; Shlevkov and Morata 2012; Zhai et al. 2012). As expected, treatment of control S2 cells with DTT results in a ∼2.25-fold increase in rpr expression (2.23 ± 0.33) compared to cells treated with DMSO (1.00 ± 0.09, P < 10⁻⁷), indicating that a larger percentage of the DTT-treated cells are undergoing cell death (Figure 2E). In contrast, S2 cells lacking expression of superdeath that are treated with DTT display an ∼1.5-fold increase in rpr expression (1.47 ± 0.07) as compared to DMSO-treated cells (1.00 ± 0.11, P = 1.2 × 10⁻³, Figure 2E). This response is significantly weaker than that seen in control cells (P = 3.5 × 10⁻⁶). A similar trend was observed for hid expression levels (Figure S2). Treatment of control S2 cells with DTT results in a threefold increase in hid expression (2.93 ± 1.98) compared to cells treated with DMSO (1.00 ± 0.284, P = 0.020). In contrast, S2 cells lacking expression of superdeath that are treated with DTT do not display a significant increase in hid expression (1.74 ± 0.75) compared to cells treated with DMSO (1.00 ± 0.40, P = 0.250). These results support a role for superdeath in apoptosis activation.

superdeath regulates JNK signaling independently of UPR activation

In Drosophila models of ER stress, and specifically in this model, apoptosis is initiated through activation of the JNK signaling cascade (Kang et al. 2012; Zhang et al. 2016). To determine if JNK signaling is disrupted upon loss of Superdeath activity, we monitored the expression of a known Jun target gene, puckered (puc). We used an allele of puc wherein the coding sequence has been replaced by the coding sequence for LacZ, such that LacZ expression is driven by the promotor and regulatory sequences that normally govern puc expression (Kanda and Miura 2004). LacZ levels serve as a direct readout for binding of the puc promotor by the Jun transcription factor. As expected, we detected high expression of LacZ in the Rh1^G69D/control eye imaginal discs (1.00 ± 0.39, Figure 2F). This expression is significantly reduced in the absence of superdeath (0.367 ± 0.082 relative to controls, P = 0.043, Figure 2, F and G). Our findings support a model of reduced signaling through the JNK cascade, which ultimately results in reduced apoptosis.

Loss of superdeath does not impact the activation of UPR signaling pathways

We hypothesized that the reduction in apoptosis and JNK signaling observed in the absence of superdeath might be caused by reduced activation of the UPR. We therefore monitored the activation of two of the UPR sensors: IRE1 and PERK. We have chosen to focus on these two pathways as there are no validated targets of ATF6 in Drosophila that are independent from other ER stress pathways. IRE1 is the most conserved of the ER stress sensors. When activated by the accumulation of misfolded proteins, the RNAse domain of IRE1 is responsible for the noncanonical splicing of the mRNA for the transcription factor Xbp1 (Sidrauski and Walter 1997; Sano and Reed 2013). The spliced isoform of Xbp1 is then translated and travels to the nucleus, where it activates expression of UPR target genes. We monitored IRE1 activity in the eye imaginal disc using an Xbp1 transgene where the 3′ end of the transcript has been replaced with the coding sequence for EGFP, such that EGFP is expressed only under conditions that induce IRE1 activity and Xbp1 splicing (Ryoo et al. 2007, 2013; Sone et al. 2013; Huang et al. 2017). IRE1 activity was measured by staining for EGFP in eye imaginal discs dissected from Rh1^G69D/control and Rh1^G69D/superdeathi flies. We also monitored Rh1^G69D levels using an antibody against rhodopsin to determine if there are differences in the amount of misfolded protein being expressed. We detected no significant differences in either EGFP (0.905 ± 0.057 relative to 1.00 ± 0.25 in controls) or rhodopsin levels (0.899 ± 0.071 relative to 1.00 ± 0.10 in controls, Figure 3, A–C). Similarly, there was no significant difference in Xbp1 splicing after exposure to DTT in S2 cells treated with DsRNA targeting EGFP or superdeath (Figure 3D). These results collectively demonstrate that loss of superdeath does not influence IRE1 activation in response to ER stress.

Figure 3

Loss of superdeath does not alter IRE1 or PERK activation. Activation of the UPR is not altered by loss of superdeath in models of ER stress. (A) Rh1^G69D/superdeathi eye discs do not display altered expression of Xbp1-EGFP or rhodopsin as compared to Rh1^G69D/controls. Eye discs were dissected from wandering L3 larvae expressing Rh1^G69D and UAS-Xbp1-EGFP, stained for rhodopsin and GFP and counterstained with 4′,6-diamidino-2-pheneylindole (DAPI). (B) Loss of superdeath does not significantly alter Xbp1-EGFP expression (0.905 ± 0.057, N = 4) compared to Rh1^G69D/controls (1.00 ± 0.25, N = 4). (C) Rhodopsin levels were also not significantly altered (0.899 ± 0.071, N = 4 relative to 1.00 ± 0.10, N = 4 in controls). (D) DTT treatment increased Xbp1 splicing in S2 cells compared to control cells treated with DMSO. This increase was similar in S2 cells treated with DsRNA against either EGFP or superdeath. (E) Rh1^G69D/superdeathi eye discs had similar levels of P-eif2α as compared to Rh1^G69D/controls. (F) Loss of superdeath does not significantly alter the ratio of P-eif2α/Pan-eif2α compared to Rh1^G69D controls (1.37 ± 0.81, N = 3 relative to 1.06 ± 0.54, N = 3). (G) DTT treatment in S2 cells increased levels of P-eif2α compared to the control DMSO treatment. This increase was similar to cells treated with DsRNA against either EGFP or superdeath. H. Loss of superdeath does not significantly alter the increase in relative P-eIF2α levels seen upon treatment with DTT as compared with control cells (3.42 ± 2.61, N = 3 relative to 5.28 ± 5.85, N = 3). Relative ratio of P-eIF2α/Pan-eIF2α upon DMSO treatment is indicated by the dotted line (N = 3/group). Values are average ± SD. Bar = 0.1 mm.

The second major sensor of the UPR is the kinase PERK, which, upon activation by the accumulation of misfolded proteins, phosphorylates the translation initiation factor eif2a. This modification reduces the efficiency of canonical translation initiation while allowing for increased translation of select UPR regulators such as the transcription factor ATF4 (Sano and Reed 2013). To assess PERK activity, we monitored eif2α phosphorylation by Western blot of samples isolated from brain-imaginal disc complexes of Rh1^G69D/superdeathi and Rh1^G69D/control larvae. We detected no significant differences in P-eif2α accumulation in these samples relative to Pan-eif2α (1.34 ± 0.81 compared to 1.06 ± 0.54 in controls, P = 0.651), suggesting that PERK activity is also unaffected by reduced expression of superdeath (Figure 3, E and F). Phosphorylation of eif2α after DTT treatment is also similar between cells treated with DsRNA against either EGFP or superdeath (P = 0.641) (Figure 3, G and H).

Collectively, these results indicate that loss of superdeath activity does not impact the UPR, and the reduced apoptosis and JNK signaling observed in the absence of superdeath is independent of UPR activation or the accumulation of misfolded proteins.

superdeath functions upstream of CDK5 in ER stress-induced apoptosis

superdeath could be regulating general apoptosis signaling or could act more specifically on pathways activated by the UPR. To test if superdeath is generally involved in the regulation of cell death, we expressed RNAi targeting superdeath in the developing eye imaginal discs of flies overexpressing the cell death initiators p53 and rpr. p53 is primarily activated by the DNA damage response and can initiate apoptosis by transcriptionally activating the IAP inhibitors rpr, grim, and hid (Mollereau and Ma 2014). rpr is activated transcriptionally by p53 and the JNK signaling cascade (Kanda and Miura 2004; Shlevkov and Morata 2012; Mollereau and Ma 2014). Overexpression of either of these factors in the eye imaginal disc is sufficient to induce extensive apoptosis and a retinal degenerative phenotype in adult flies in the absence of any additional stressor (Hay et al. 1995; Jin et al. 2000). We can test the impact of superdeath expression on general apoptotic pathways by expressing the RNAi construct targeting superdeath in models of p53 or rpr overexpression and evaluating changes in eye degeneration.

We first tested loss of superdeath in a model of p53 overexpression (p53/superdeathi) to determine whether general cell death pathways are impacted. We found no difference in eye size between p53/control (14,852 ± 1126 pixels) and p53/superdeathi flies (15,315 ± 1000 pixels, Figure 4A). We next tested superdeath function in a model of rpr overexpression (rpr/superdeathi) to see if the function of this gene lies upstream or downstream of the transcriptional program that commonly activates apoptosis. As with p53, there was no difference in eye size between rpr/control (18,953 ± 834 pixels) and rpr/superdeathi flies (19,288 ± 664 pixels, Figure 4B). Our findings suggest that superdeath functions upstream of the transcriptional program that initiates apoptosis and in a pathway that is specific to ER stress. These data also indicate that superdeath does not ubiquitously modify general apoptotic pathways.

Figure 4

superdeath is upstream of CDK5 in the initiation of ER stress-induced apoptosis. superdeath acts specifically in ER stress-associated cell death, upstream of the transcription of apoptotic activators. (A) Degeneration caused by overexpression of p53 is not altered by RNAi-mediated knockdown of superdeath expression (15,315 ± 1000 pixels, N = 18 in p53/superdeathi flies compared to 14,852 ± 1126 pixels, N = 18 in p53/controls). (B) Degeneration caused by overexpression of rpr is also unaffected by loss of superdeath expression (19,288 ± 664 pixels, N = 13 in rpr/superdeathi flies compared to 18,953 ± 834 pixels, N = 15 in rpr/controls). (C) Degeneration caused by overexpression of Rh1^G69D in the absence of both CDK5 and superdeath (16,560 ± 1320 pixels, N = 12) does not significantly differ from degeneration in the absence of CDK5 alone (16,552 ± 1179 pixels, N = 15). Degeneration is qualitatively improved by RNAi-mediated knockdown of CDK5 expression although eye size is not significantly increased compared to Rh1^G69D/controls (15,307 ± 1482 pixels, N = 15). This is in contrast to the significant increase in eye size when superdeath expression is reduced (20,346 ± 1292 pixels, N = 15). Values are average ± SD. ****P < 0.00005.

Activation of JNK-induced apoptosis in the Rh1^G69D model of ER stress is regulated by the ser-thr kinase CDK5. Loss of CDK5 leads to reduced activation of apoptosis without altering the activation of the UPR (Kang et al. 2012). These molecular changes are accompanied by qualitative improvements in eye size and pigmentation (Kang et al. 2012; Chow et al. 2016). We hypothesized that superdeath might function in this pathway. To test this, we expressed RNAi constructs targeting CDK5 and superdeath, individually and concurrently, in the developing eye imaginal discs expressing the misfolded Rh1^G69D protein. We monitored degeneration using eye size in adult flies. As expected, loss of superdeath (Rh1^G69D/superdeathi) results in a substantial and significant increase in eye size as compared to the Rh1^G69D/controls (20,345 ± 1292 pixels vs. 15,307 ± 1482 pixels in controls, P < 0.00005, Figure 4C). In line with previous reports (Kang et al. 2012; Chow et al. 2016), expressing the misfolded Rh1^G69D protein and RNAi against CDK5 (Rh1^G69D/CDK5i) results in a qualitative eye rescue and improvement, but no change in eye size as compared to controls (16,552 ± 1179 pixels, P = 0.060, Figure 4C). Because the phenotype associated with loss of CDK5 is distinguishable from the phenotype associated with loss of superdeath, we can perform an epistasis experiment to determine which of these genes lies downstream of the other. We found that flies simultaneously expressing both RNAi against CDK5 and superdeath (Rh1^G69D/CDK5i-superdeathi) display similar phenotypes to Rh1^G69D/CDK5i flies, with no quantitative improvement in eye size as compared to controls (16,560 ± 1320 pixels, P = 0.081, Figure 4C). We concluded that superdeath must operate upstream of CDK5 to regulate the activation of JNK signaling and apoptosis.

superdeath regulates ER stress-induced apoptosis in multiple tissues

Because CDK5 and JNK signaling occur across different ER stress conditions, we tested whether superdeath can act as a modifier across different tissues and methods of ER stress initiation. The wing imaginal disc is a developmental structure that will eventually mature and become the adult wing. Wing imaginal disc expression of a misfolded protein such as Rh1^G69D using the MS1096-GAL4 driver (wing-Rh1^G69D/conrol) induces ER stress and apoptosis, resulting in a small, degenerate wing that fails to unfold upon eclosion (Figure 5A). Concurrent expression of Rh1^G69D and RNAi against superdeath (wing-Rh1^G69D/superdeathi) results in partial rescue of the degenerate wing phenotype (Figure 5A), similar to what was observed for the eye. This is also reflected in wing area, which increases from wing-Rh1^G69D/superdeathi (94,627 ± 26,363 pixels) flies compared to wing-Rh1^G69D/control (73,931 ± 19,988 pixels, P = 0.044, Figure 5A) flies.

Figure 5

superdeath is a general regulator of ER stress-induced cell death. The ER stress response and subsequent apoptosis is subject to regulation by superdeath across ER stress conditions. (A) Loss of superdeath partially rescues the vestigial wing phenotype caused by expression of Rh1^G69D in the wing disc (94,627 ± 26,363 pixels, N = 16 in wing-Rh1^G69D/superdeathi  vs. 73,931 ± 19,988 pixels, N = 10 in wing-Rh1^G69D/controls). (B) Larvae with ubiquitous knockdown of superdeath are significantly more resistant to tunicamycin-induced ER stress than control larvae. Four paired experimental replicates are shown, representing a combined total of N = 113 DMSO-treated and N = 130 TUN-treated Ubi-control larvae, and N = 112 DMSO-treated and N = 127 TUN-treated Ubi-superdeathi larvae. Values are average ± SD. *P < 0.05, **P < 0.005.

To determine whether superdeath also responds to diverse mechanisms of initiating ER stress, we ubiquitously expressed RNAi targeting superdeath using the Tub-GAL4 driver (Ubi-superdeathi). We then exposed Ubi-superdeathi larvae along with controls expressing only Tub-GAL4 (Ubi-control) to tunicamycin or DMSO for 4 hr. Tunicamycin inhibits N-linked glycosylation in the ER, inducing a massive ER stress response through the UPR (Samali et al. 2010). This 5-hr treatment is sufficient to significantly activate the UPR in the larvae, leading to developmental delay and lethality (Huang et al. 2015; Kang et al. 2015). Because ubiquitous loss of superdeath leads to almost complete lethality during pupal stages, we used survival to the pupal stage as our readout for sensitivity to tunicamycin. We found that Ubi-superdeathi larvae are significantly more resistant to tunicamycin-induced lethality as compared to Ubi-control larvae (Figure 5B and Table S1). We concluded from these results, and the S2 cell data above, that superdeath is generally important for the downstream consequences of ER stress.

Superdeath is localized to the ER

Based on sequence analysis using TMHMM and Scan Prosite tools, Superdeath is hypothesized to contain a single transmembrane domain near the N-terminal of the protein, with the active zinc metalloprotease site localized to the cytoplasm of the cell between amino acids 404 and 413 (Krogh et al. 2001; de Castro et al. 2006). This active site and the position of the transmembrane domain, are highly conserved when compared to human orthologs. Based on our results described above, we hypothesized that Superdeath could be localized to the ER membrane. Here, among other possibilities, it may be able to sense the activation of ER stress pathways, and activate the cytosolic CDK5-JNK signaling cascade.

To test this, we performed immunofluorescence staining for Superdeath in S2 cells and looked for colocalization with known subcellular markers: Calnexin 99A (ER), Golgin-84 (Golgi), and Lamp1 (lysosome). We induced expression of a transgenic Superdeath tagged with the V5 epitope in S2 cells, and stained for V5 and each subcellular marker. We found that, whereas Superdeath appears to colocalize with Calnexin 99A (Figure 6A), it does not colocalize with Golgin-84 (Figure 6B). Whereas the vast majority of the Superdeath and Lamp1 signals are distinct and nonoverlapping, we do detect a small minority of overlapping signals (Figure 6C). Importantly, in all cases, Superdeath does not appear to localize to the plasma membrane, the primary location of several potential human orthologs such as ANPEP, ENPEP, and LVRN. The staining for V5 is not detectable in cells not expressing the superdeath transgene (Figure S3). We concluded that Superdeath localizes and functions primarily at the ER membrane.

Figure 6

Superdeath is localized to the ER. Superdeath predominantly localizes to the ER membrane. (A) Superdeath localizes to the ER. S2 cells expressing Superdeath-V5 were stained for V5 (green) and Calnexin 99A (red) and counterstained with DAPI. A’ and A” represent the highlighted panels from (A). White arrows highlight select sites of V5 and Calnexin 99A overlap. (B). Superdeath does not localize to the Golgi. S2 cells expressing Superdeath-V5 were stained for V5 (green) and Golgin-84 (red) and counterstained with DAPI. B’ and B” represent the highlighted panels from (B). White arrows indicate select sites of independent V5 staining or Golgin-84 staining. (C). Superdeath does not primarily localize to the lysosome. S2 cells expressing Superdeath-V5 were stained for V5 (green) and Lamp1 (red) and counterstained with DAPI. C’ and C” represent the highlighted panels from (C). White arrows indicate select sites of independent V5 staining or Lamp1 staining. Bar = 0.01 mm.

To confirm this localization, we also performed immunofluorescence staining for Superdeath in the salivary glands of L3 larvae. Salivary gland cells are large and have an extensive ER network due to their secretory functions. They are an ideal tissue in which to identify the subcellular localization of Superdeath. As in S2 cells, we stained for Calnexin 99A, Golgin-84, and Lamp1 in the salivary glands of larvae carrying a GFP-tagged allele of Superdeath (Superdeath-GFP; #64447 BDSC ). Superdeath colocalizes with Calnexin 99A, but not with Golgin-84 or Lamp1 (Figure S4), similar to the colocalization observed in S2 cells. This is detectable by the extensive overlap as indicated by orange signals when costained with Calnexin 99A (Figure S4A). The vast majority of signals from Golgin-84 (Figure S4B) and Lamp1 (Figure S4C) lie adjacent to Superdeath-GFP, producing distinct red and green signals. Again, we do detect some potential overlap, though this time between Superdeath and Golgin-84. This might be due to close associations between the Golgi and ER membranes, or it may reflect some leakage of Superdeath protein to the Golgi, where it is then recycled back to the ER. We conclude that Superdeath localization is primarily to the ER membrane, and is similar in different cell types.

Discussion

Activation of the ER stress response and the subsequent cell death is a major contributor to the pathogenesis of a number of human diseases. Degenerative diseases are commonly caused or complicated by the accumulation of misfolded proteins in the ER and by stress-induced cell death (Hartong et al. 2006; Hetz and Saxena 2017; Zhu et al. 2017; Kurtishi et al. 2018; Yamanaka and Nukina 2018). In order to treat these degenerative diseases, it is essential to specifically target stress-associated cell death without inhibiting the beneficial stress-induced pathways that restore homeostasis.

In this study, we examined the metallopeptidase superdeath, a modifier of Rh1^G69D-induced degeneration (Chow et al. 2016). Loss of superdeath activity reduces apoptosis and degeneration without impacting the activation of the ER stress sensors IRE1 and PERK. Both of these sensors, while capable of initiating apoptosis upon chronic activation, have important cell survival functions that are essential for returning the stressed cell to homeostasis (Sano and Reed 2013). Loss of superdeath leaves these beneficial functions largely intact, and, instead, reduces the activation of JNK signaling through CDK5 activation. In this manner, superdeath orthologs could serve as important therapeutic targets to tip the balance in favor of cell survival in degenerative diseases.

Additionally, superdeath fills an important gap in ER stress-associated cell death biology. While previous studies have shown that CDK5 is responsible for activating the JNK signaling cascade under ER stress (Kang et al. 2012), the mechanism of how the stress signal is communicated to the kinase is still unknown. Our data suggest that the ER-associated Superdeath, which is likely orthologous to ERAP1 or ERAP2 in humans, may serve as an important bridge between ER stress and CDK5. Importantly, the active site of Superdeath is predicted to be present on the cytoplasmic side of the ER, suggesting that changes in ER membrane conformation and ER luminal environment could alter Superdeath activity, directly or indirectly activating CDK5.

There is already evidence that ERAP2 could be playing an important role in autoimmune disorders such as Crohn’s disease and inflammatory arthritis, as well as in the response to viral infection (Franke et al. 2010; Popa et al. 2016; Ye et al. 2018). The types of stress responses induced in these diseases, including ER, oxidative, and mechanical stresses, frequently activate cell death through the JNK signaling cascade. This raises the possibility that Superdeath and ERAP2 are regulating apoptosis in response to many cellular stresses, and that these roles are not limited to the induction of ER stress; this is an exciting avenue of future study.

Our findings suggest that the Drosophila gene superdeath regulates stress-induced apoptosis. We have demonstrated a role for this gene in known apoptotic pathways and have shown that its role lies downstream of the beneficial activation of the UPR. The position of superdeath in the cellular stress response makes its orthologs attractive candidates for therapeutic targeting for a variety of diseases associated with ER stress-induced degeneration. Understanding how modifiers of stress-induced apoptosis are functioning in the cell also increases our understanding of degenerative diseases and provides new avenues for personalized therapies.

Acknowledgments

We thank Drs. Julie Hollien, Aylin Rodan, Anthea Letsou, and Carl Thummel for use of equipment and sharing reagents. We thank Dr. D. Allan Drummond for tremendous insight on Twitter and the name superdeath. This research was supported by an National Institutes of Health/National Institute of General Medical Sciences (NIH/NIGMS) R35 award (1R35GM124780) and a Glenn Award from the Glenn Foundation for Medical Research to C.Y.C. C.Y.C. was the Mario R. Capecchi Endowed Chair in Genetics. R.A.S.P. is supported by a National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) T32 fellowship (5T32DK110966).

Footnotes

Supplemental material available at figshare: https://doi.org/10.6084/m9.figshare.11401827

Communicating editor: H. Bellen

Author notes

1

Current address: Biology Department, Purdue University-Fort Wayne, Fort Wayne, IN 46805.