p62/Sequestosome-1, Autophagy-related Gene 8, and Autophagy in Drosophila Are Regulated by Nuclear Factor Erythroid 2-related Factor 2 (NRF2), Independent of Transcription Factor TFEB*

Background: A possible role of Drosophila Nrf2/CncC in regulating p62/Ref(2)P and autophagy upon oxidative stress is unknown. Results: Both p62/Ref(2)P and DmAtg8a are positively regulated by the oxidative stress-induced transcription factor Nrf2/CncC. Conclusion: The oxidative stress response is directly linked to autophagy induction in several tissues in flies. Significance: Nrf2/CncC induces autophagy in flies and may regulate autophagic activity in other organisms. The selective autophagy receptor p62/sequestosome 1 (SQSTM1) interacts directly with LC3 and is involved in oxidative stress signaling in two ways in mammals. First, p62 is transcriptionally induced upon oxidative stress by the NF-E2-related factor 2 (NRF2) by direct binding to an antioxidant response element in the p62 promoter. Second, p62 accumulation, occurring when autophagy is impaired, leads to increased p62 binding to the NRF2 inhibitor KEAP1, resulting in reduced proteasomal turnover of NRF2. This gives chronic oxidative stress signaling through a feed forward loop. Here, we show that the Drosophila p62/SQSTM1 orthologue, Ref(2)P, interacts directly with DmAtg8a via an LC3-interacting region motif, supporting a role for Ref(2)P in selective autophagy. The ref(2)P promoter also contains a functional antioxidant response element that is directly bound by the NRF2 orthologue, CncC, which can induce ref(2)P expression along with the oxidative stress-associated gene gstD1. However, distinct from the situation in mammals, Ref(2)P does not interact directly with DmKeap1 via a KEAP1-interacting region motif; nor does ectopically expressed Ref(2)P or autophagy deficiency activate the oxidative stress response. Instead, DmAtg8a interacts directly with DmKeap1, and DmKeap1 is removed upon programmed autophagy in Drosophila gut cells. Strikingly, CncC induced increased Atg8a levels and autophagy independent of TFEB/MitF in fat body and larval gut tissues. Thus, these results extend the intimate relationship between oxidative stress-sensing NRF2/CncC transcription factors and autophagy and suggest that NRF2/CncC may regulate autophagic activity in other organisms too.

Cnc proteins are well conserved in worms, insects, fish, birds, and mammals but absent in plant and fungi (13). Cnc proteins were first identified in Drosophila melanogaster (14), where the Cnc locus gives rise to transcripts encoding three isoforms, CncA, CncB, and CncC (15). The CncB isoform is active in early embryogenesis in labral and mandibular segments that further develop into the fly's head, whereas the CncC isoform is active from late embryogenesis on and throughout life (14 -16). CncA lacks the transactivation domain entirely, and it is not known whether CncA acts as a repressor like the short forms of Nrf1 and Nrf2 (13).
CncC, the longest isoform of Cnc, contains a unique N-terminal domain with regions of similarity to the well conserved Keap1-binding motifs of mammalian Nrf2. A Keap1 orthologue in Drosophila acts as a negative regulator of CncC (17). Similar to NRF2, CncC is activated by different oxidants. This activation has a role in the fly's antioxidant defense (15,17,18). Loss of DmKeap1 resulted in high mRNA levels of gstD1, a CncC target gene, and this correlates with enhanced oxidative stress response and extended life span in Drosophila (17,19,20). Recently, overexpression of detoxifying enzymes has also been reported in two insecticide-resistant strains of Drosophila due to constitutive activation of CncC (21).
Autophagy is a catabolic process where an isolation membrane engulfs part of the cytoplasm to create a double-membrane vesicle called the autophagosome, which fuses with lysosomes and leads to degradation of their contents (22,23). Selective autophagy receptors bind to cargo and dock onto the forming phagophore through a direct interaction with ATG8 family proteins, enabling delivery and autophagic degradation of the cargo (24,25). Human p62/sequestosome 1 (hereafter named p62) interacts with LC3 and ubiquitin, is a selective autophagic substrate, and is the first identified cargo receptor for autophagic degradation of ubiquitinated targets (26,27). When autophagy is abolished in the liver of Atg7 conditional knock-out mice, p62 accumulates in aggregates, and antioxidant proteins and phase II detoxification enzymes are strongly induced (28). p62 is induced by various stressors both at the mRNA and protein levels, and this p62 induction is inhibited in cells from Nrf2 knock-out mice (29,30). Several groups, including ours, have reported that p62 competes with NRF2 for binding to KEAP1, resulting in stabilization of NRF2, whereas KEAP1 is sequestered into p62 bodies and subsequently degraded by autophagy (31)(32)(33)(34). It was also recently shown that phosphorylation of the KEAP1-interacting region (KIR) motif of p62 enhanced binding to KEAP1 (35,36). We reported earlier that NRF2 bound to an ARE site in the p62 promoter and induced p62 expression upon oxidative stress (31). Hence, we could conclude that p62 is involved in establishing a positive feedback loop inducing its own expression and prolonged NRF2 response under stress conditions (31).
D. melanogaster ref (2)P is the orthologue of mammalian p62 (25,(37)(38)(39)(40) and was first characterized as a modifier of virus multiplication (38,41,42). Ref (2)P has been reported to be a major component of protein aggregates in flies defective in autophagy or with impaired proteasomal function and in fly models of neurodegenerative diseases (39,43,44). However, it is not known if Ref (2)P binds directly to DmAtg8 via a functional LIR motif.
Here, we show that Ref(2)P interacts with DmAtg8a in vitro and in vivo through a LIR motif and that this is necessary for autophagic degradation of Ref (2)P. We also show that ref (2)P is a transcriptional target of CncC and contains a CncC-responsive ARE in its promoter. However, Ref(2)P does not bind directly to DmKeap1 via a KIR motif, as found for mammalian p62 and KEAP1. Consequently, ectopically expressed Ref(2)P does not induce the oxidative stress response in fly tissues. Very interestingly, we found CncC to induce atg8a and stimulate autophagy in the fat body and larval gut. Hence, the positive feedback loop between p62 and Nrf2 seen in mammals is not present in D. melanogaster. However, CncC can induce ref (2)P, atg8a, and autophagy.
Plasmid Constructs-Plasmids used in this study are listed in supplemental Table 1. They were made by conventional restriction enzyme-based cloning or by use of the Gateway recombination system (Invitrogen). Point mutants were made using the QuikChange site-directed mutagenesis kit (Stratagene, 200523). Oligonucleotides for mutagenesis, PCR, and DNA sequencing were from Invitrogen or Sigma. Plasmid constructs were verified by DNA sequencing with BigDye version 3.1 (Applied Biosystems, 4337455).
Total RNA Isolation, cDNA Construction, and Cloning-Total RNA was isolated from S2Rϩ cells using the RNeasy Plus minikit (Qiagen, 74134), and cDNA was constructed using Transcriptor Universal cDNA Master (Roche Applied Science, 05 893 151 001). PCR products were amplified from cDNA using Phusion high fidelity DNA polymerase with primers containing the recombination site for Gateway entry vector pDONR221 (Invitrogen, 12536-017) or restriction enzyme sites and cloned into pDONR221 or pENTR using Gateway recombination cloning.
Reporter Gene Assays-S2Rϩ cells were seeded at a density of 3.5 ϫ 10 5 cells/well in 24-well plates and transfected using TransIT-2020 or the NanoJuice transfection kit. Transfections were performed with 75 ng of the Cnc and 250 ng of other expression plasmids together with 150 ng of the luciferase reporter plasmids. Cells were harvested 48 or 72 h post-transfection, and luciferase activities were measured using the Dual Light luciferase kit (TROPIX) on a Luminoskan RT dual injection luminometer (Labsystems). All reporter gene assays were carried out in three parallel experiments and repeated at least four times.
GST and MBP Pull-down Experiments-GST-and MBP pull-down assays with in vitro translated 35 S-labeled proteins were done as described previously (26). GST and MBP fusion proteins were expressed in Escherichia coli BL21(DE3) or SoluBL21 (Amsbio).
Fly Cultivation and Stocks-Flies were cultivated at 25°C on our standard laboratory fly medium consisting of (per liter) 32.7 g of dried potato powder, 60 g of sucrose, 27.3 g of dry yeast, 7.3 g of agar, 4.55 ml of propionic acid, and 2 g of nipagin, giving a final concentration of 15.3 g/liter protein and 6 g/liter sugar. For starvation experiments, larvae were well fed overnight with additional yeast paste and transferred to filter paper soaked in PBS for the 4-h starvation period. The fly stocks For the generation of UAS-DmKeap1 and GFP-Dmkeap1 fly lines, the coding sequence of Dmkeap1 was retrieved from Dmkeap1 cDNA from pUASt-DmKeap1-HA plasmid (Berkeley Drosophila Genome Project) with the help of primers containing the recombination site for Gateway entry vector pENTR (Invitrogen, catalog no. 12536-017). PCR product was isolated and cloned into pENTR to generate Gateway entry clone using Gateway BP Clonase II enzyme mix (Invitrogen, catalog no. 11789-020). Entry clone was then used to generate expression clone (pPW-DmKeap1 and pPGW-DmKeap1) using Gateway LR Clonase II enzyme mix (Invitrogen, catalog no. 11791-100). For the generation of UAS-GFP-CncC, the coding sequence of CncC was retrieved from total RNA from S2Rϩ cells using the RNeasy Plus minikit (Qiagen, catalog no. 74134), and cDNA was constructed using Transcriptor Universal cDNA Master (Roche Applied Science, catalog no. 05 893 151 001). CncC was PCR-amplified from cDNA using Phusion high-fidelity DNA polymerase (New England Biolabs, catalog no. M0530S) and cloned into pENTR vector for the generation of pPGW-CncC expression vector. These expression vectors were sent for embryo injection and establishment of transgenes to Bestgene Inc. For overexpression or knockdown studies in the larval hindgut, fat body, or wing imaginal discs, we established gstD-GFP;en-Gal4,UAS-RFP/CyO and hsp70-Flp;pmChAtg8a/CyO;actϾCD2ϾGAL4,UAS-GFPnls stocks and combined them with overexpression or RNAi knockdown transgenic lines. Loss of function mitotic clones in the Drosophila midgut was obtained by 1-h 37°C heat shock during 0 -8 h of embryogenesis.
Immunohistochemistry, Microscopy, and Analysis-Larval fat body, gut, and discs were dissected from larvae; fixed in 4% formaldehyde, PBS (Polysciences, catalog no. 18814 ultrapure) for 30 min; washed three times for 10 min each in PBSBT con-taining 0.5% BSA and 0.3% Triton X-100; and labeled immediately afterward according to standard protocols. Polyclonal rabbit anti-MBP-Ref(2)P serum (1:5000) was prepared. 4 Secondary antibodies against rabbit IgG conjugated with Dy649 or Alexa 633 were from Jackson Immunoresearch Laboratories, Inc. (1:1000). Samples were counterstained with Hoechst (Sigma) to highlight nuclei and mounted in Vectashield H-1000 (Vectorlabs) for imaging on either Zeiss LSM510, -710, or -780 microscopes. HeLa cells were cultured in 8-well chambered coverslides (Nunc) and transiently transfected with 100 ng of pDest-mCherry-eGFP-Ref(2)P WT or W454A/I457A construct using TransIT-LT1. Cells were fixed in 4% paraformaldehyde 18 h after transfection. Images were obtained using a LSM510-META microscope. Images were processed using Canvas version 11 (ACD Systems). Relative GFP-DmKeap1 expression levels and cell size were analyzed by ImageJ using original lsm images obtained from a Zeiss LSM710 or LSM780 microscope. GFP-DmKeap1 expression levels and cell size area were evaluated between an atg13 mutant enterocyte (EC) cell and an immediate random neighboring cell at the mid-level of the nucleus. Cells were encircled with the region of interest tool, and pixel intensity levels measured by IntDen were normalized to total area of the cells. Normalized relative IntDen levels between atg13 mutant cells and neighboring cell pairs were compared. Pixel-based co-localization analysis is presented as Pearson coefficient values and was performed using the JACoP application of ImageJ. Results were presented using GraphPad Prism version 6.0.
Mass Spectrometry Analysis-Gel pieces were subjected to in-gel reduction, alkylation, and tryptic digestion using 6 ng/l trypsin (V511A, Promega, Madison, WI) (51). OMIX C18 tips (Varian, Inc., Palo Alto, CA) were used for sample clean up and concentration. Peptide mixtures containing 0.1% formic acid were loaded onto a Fisher EASY-nLC1000 system and EASY-Spray column (C18, 2 m, 100 Å, 50 m, 15 cm). Peptides were fractionated using a 2-100% acetonitrile gradient in 0.1% formic acid over 50 min at a flow rate of 250 nl/min. Separated peptides were analyzed using a Thermo Scientific Q-Exactive mass spectrometer. Data were collected in data-dependent mode using a Top10 method. Raw data were processed using Proteome Discoverer version 1.4 software. Fragmentation spectra were searched against the NCBInr_2013013 database using an in-house Mascot server (Matrix Science Ltd., London, UK). Peptide mass tolerances used in the search were 10 ppm, and fragment mass tolerance was 0.02 Da. Enzyme settings were set to trypsin with two allowed mismatches. Ubiquitination (GlyGly) of lysine and oxidation of methionine was set as a variable modification. Carbamidomethylation of cysteine was set as a fixed modification. Peptide ions were filtered using a false discovery rate of 1% for peptide identifications.
Statistical Analysis-Student's t tests for independent samples were performed by GraphPad Prism version 6.0 and presented as the S.E. Ϯ S.D. to compare group data. Differences were regarded as significant for two-tailed p values of Ͻ0.05. (2)P has a domain architecture similar to that of p62 with an N-terminal PB1 domain, a ZZ-type zinc finger domain, and a C-terminal UBA domain (Fig. 1A). It has not been tested whether Ref(2)P is able to bind directly to D. melanogaster Atg8, although Ref(2)P accumulates upon autophagy deficiency, suggesting that it may be a selective autophagy substrate (39,43,44). We first analyzed and showed binding of in vitro translated Ref(2)P to recombinant GST-DmAtg8a in a pull-down assay (Fig. 1B). By visual inspection based on knowledge of LIR motif sequences and reports about a putative LIRmotif in Ref (2)P (24, 25, 52, 53), we identified a potential LIRmotif ( 451 DPEWQLIDN 459 ) in Ref (2)P. This motif is also the top candidate predicted by the iLIR Web server (54). Deletion of this motif (⌬LIR) and point mutants of the aromatic and hydrophobic residues (W454A, I457A, and W454A/I457A) in the core LIR motif abolished the interaction between Ref(2)P and DmAtg8a (Fig. 1B). This clearly shows that this is an LIR-mediated interaction. This was also verified in vivo by coimmunoprecipitation of eGFP-DmAtg8a and 3xFLAG-Ref (2) (2)P is required for its accumulation in acidic vesicles and subsequent degradation. An important characteristic of p62 is its ability to polymerize via the N-terminal PB1 domain (56,57). Hence, we tested whether Ref(2)P could self-interact via its PB1 domain. In GST pull-down assays, full-length Ref(2)P interacted with itself, whereas a PB1 domain deletion mutant did not (Fig. 1F). Ref(2)P did not interact with p62, confirming the specificity of the PB1-mediated oligomerization of Ref(2)P (Fig. 1F).

Ref(2)P Lacks a KIR Motif and Does Not Bind Directly to DmKeap1 but Can Bind to Ubiquitinated DmKeap1-Having
established that Ref(2)P shares important features with mammalian p62 with respect to PB1-mediated oligomerization and functional LIR-dependent interaction with DmAtg8a, we asked whether the interaction with Keap1 is conserved. DmKeap1 is a 744-amino acid-long protein with similar domain architecture as human KEAP1 with BTB, BACK, and KELCH repeats domains (Fig. 2G). Ref(2)P and DmKeap1 can be co-immunoprecipitated from Drosophila cells (58,59). However, Ref(2)P did not interact with in vitro translated DmKeap1 or with human KEAP1 in pull-down assays ( Fig. 2A). Nor did p62 interact with DmKeap1, whereas p62 and KEAP1 interacted strongly, as shown previously ( Fig. 2A). Tyr residues 525 and 572 in the KELCH domain of human KEAP1 important for the interaction with p62 are substituted with Phe residues in DmKeap1 (31). However, mutating both of these Phe residues to Tyr in DmKeap1 did not restore interaction, neither with Ref(2)P nor with p62 (data not shown). A sequence alignment of p62 orthologs from diverse metazoans shows that the KIR motif is first found in the Cephalochordates, the lancelets, represented here by Branchiostoma (Fig. 2B). The motif is not found in the most primitive chordates, the Urochordates, as represented by Ciona. All vertebrate p62 orthologs contain the KIR motif, whereas Ref (2) (Fig. 2D). Co-immunoprecipitation of ubiquitinated proteins, along with DmKeap1 itself, was also dependent on the presence of a functional UBA domain in Ref(2)P (Fig. 2E). A point mutation, M565V, in the MGF motif important for ubiquitin binding (60), prevented co-immunoprecipitation of Ref(2)P and DmKeap1 from transfected S2Rϩ cell extracts (Fig. 2F). Consistent with the UBA domain dependence for interaction, MS/MS analyses of immunoprecipitated DmKeap1 from Drosophila cells detected four ubiquitinated lysines in DmKeap1: one each in the BTB and the BACK domain and two in the KELCH repeats ( Fig. 2G and supplemental Fig. S1). This strongly suggests that DmKeap1 is co-immunoprecipitated with Ref (2)P along with other ubiquitinated proteins depen-dent on the UBA domain of Ref(2)P. Hence, the KIR-mediated p62-Keap1 interaction identified in mammals most likely evolved with the most primitive fish, the Cephalochordate lancelets.
DmKeap1 Interacts with DmAtg8a-The KELCH-BTB family proteins KBTBD6 and -7 (KELCH repeat and BTB domain-containing protein) were recently shown to interact with the human GABARAP subfamily of ATG8 family members in a LIR-dependent manner (61). Hence, we tested whether DmKeap1 could interact with DmAtg8a. Interestingly, DmKeap1 interacted with DmAtg8a in pull-down assays, and the interaction was not dependent on the KELCH repeats (Fig.  3A). No direct interaction between human KEAP1 and ATG8 family proteins was detected (Fig. 3B, top). DmKeap1 interacted with human GABARAPs, which are more similar to DmAtg8a than the LC3 subfamily proteins (Fig. 3B, middle). A direct interaction between KBTBD7 and ATG8 family proteins was also found (Fig. 3B, bottom).
A single putative LIR motif was identified within the interacting N-terminal part of DmKeap1 encompassing the BTB and BACK domains. However, a mutation of this putative LIR did not affect the interaction (data not shown). As mentioned above for Ref(2)P, ATG8 homologues bind to canonical LIR motifs through their LDS. Surprisingly, the LDS mutant of DmAtg8a bound equally well with DmKeap1 as the WT DmAtg8a (Fig. 3C), strongly suggesting that this binding has an interaction mode different from that of the canonical LIR-LDS interaction. To further investigate the mode of the DmKeap1-DmAtg8a interaction, we used a series of recombinant GST-DmAtg8a deletion constructs (residues 1-26, 1-71, 26 -71, and 26 -121). As expected from previous analyses of LIR interac-tions (26,62,63), Ref(2)P interacted with full-length DmAtg8a but failed to interact with any of the deletion constructs (Fig.  3D, top). However, DmKeap1 interacted with DmAtg8a(1-71) (Fig. 3D, bottom), confirming that DmKeap1 binds to DmAtg8a in an LIR-independent manner.

DmKeap1 Is Not Turned Over by Basal Autophagy but Is Removed upon Programmed Autophagy in Drosophila Gut
Cells-To follow DmKeap1 in vivo, we generated transgenic flies expressing untagged or GFP-tagged full-length DmKeap1. Expression of either transgene was detected by Western blot analysis when ubiquitously expressed in L3 larvae (Fig. 4A). Because DmKeap1 interacted with DmAtg8a, we addressed the question of whether DmKeap1 was turned over by autophagy by evaluating total DmKeap1 protein levels of whole L3 larval lysates between control larvae and larvae homozygous for null alleles of atg6, atg8a, atg13, or ref(2)P. No significant accumulation of DmKeap1 levels could be detected in larvae deficient for atg6, atg8a, atg13, or ref(2)P (Fig. 4A). Because whole animal

Nrf2/CncC Regulates ref(2)P, atg8a, and Autophagy
Western blot cannot reveal tissue-specific differences in DmKeap1 regulation, we generated atg13 mosaic animals ubiquitously expressing GFP-DmKeap1 and where homozygous atg13 mutant cells could be recognized by the lack of red fluorescent protein (RFP). In line with previous findings that lack of basal autophagy in epithelial tissues of imaginal discs leads to accumulation of Ref(2)P, atg13 mutant clones specifically accumulated Ref(2)P in wing imaginal discs (48). However, GFP-DmKeap1 levels did not accumulate in atg13 mutant clones relative to neighboring cells (Fig. 4B). To further investigate DmKeap1 under non-basal conditions, we chose to focus on the larval intestine because previous studies have established that DmKeap1 regulates CncC activity in Drosophila intestinal cells (17). The Drosophila midgut consists of chiefly two cell types, nutrient absorptive EC that undergo mitotic endocycles and grow in both nuclear and cytoplasmic size during larval life and adult midgut progenitor stem cells (AMPs) that multiply by mitotic cycles in clusters starting from one cell to reach 9 -12 cells by the end of the last larval stage (L3). Programmed autophagy commences by the late L3 stage (wandering L3) in EC cells and is responsible for the initial cell size reduction and removal of EC cells by cell death during the early pupal life, whereupon AMP stem cells repopulate and make the future adult gut during metamorphosis (64) (Fig. 4C). Immunolabeling of DmKeap1 and expression of GFP-DmKeap1 both revealed a similar dynamic behavior of DmKeap1 during midgut larval development. In early to mid-L3 larval stage midgut EC and AMPs, DmKeap1 localized to cytoplasmic punctate structures (Fig. 4, D and F). By the late larval stages, shortly before pupariation (wandering L3), DmKeap1 levels were strongly reduced and less punctate in EC cells, whereas high levels of expression were sustained in AMPs (Fig. 4, E and  G). Co-expression of GFP-DmKeap1 and mCherry-Ref(2)P revealed partial overlap of GFP-DmKeap1 or mCherry-Ref(2)P signal, suggesting that DmKeap1 and Ref(2)P are components of distinct but associated cytoplasmic structures in EC cells (Fig. 4H). As reported earlier, starvation induced a robust formation of mCherry-DmAtg8a structures, partially overlapping with Ref(2)P, but not GFP-DmKeap1, suggesting that DmKeap1 is not degraded by starvation-induced autophagy. Alternatively, GFP-DmKeap1 fluorescence is rapidly quenched in acidic autolysosomes (Fig. 4I). Programmed autophagy is initiated in EC cells of the larval midgut by the wL3 stage to facil-itate cytoplasmic cell size reduction during programmed cell death executed during pupal development (64). Because this transition coincided with the reduction of DmKeap1 levels in EC cells, we next investigated whether programmed autophagy could be responsible for this reduction. As expected, atg13 mutant EC cells failed to undergo cytoplasmic cell size reduction and accumulated Ref(2)P by the late wL3 stage, consistent with a block in autophagic activity. Strikingly, GFP-DmKeap1 levels remained high in atg13 mutant cells, relative to neighboring autophagy-proficient cells (Fig. 4, J-L). Taken together, this suggests that DmKeap1 is not turned over by basal autophagy but is removed upon programmed autophagy in Drosophila gut cells.
DmKeap1 Interacts with CncC-It is not known if the direct interaction between Keap1 and Nrf2 is conserved in Drosophila. The DLG and ETGE motifs responsible for the interaction in Nrf2 are found in the longest isoform of the Drosophila homologue, CncC (9). First, we tested all three isoforms of Cnc (A, B, and C) for binding to DmKeap1 in pull-down assays (Fig. 5A). Only CncC interacted with DmKeap1 (Fig. 5B). To further address the importance of the DLG and ETGE motifs, we made point mutations and deletion constructs of CncC affecting these motifs (Fig. 5A). A double point mutation of the DLG motif (D357A/L358A) did not affect the CncC-DmKeap1 interaction, but a double point mutation of the ETGE motif (E460A/ G462A) or both motifs (D357A/L358A/E460A/G462A) completely abolished the interaction (Fig. 5C). Consistently, the CncC(464 -1383) lacking both DLG and ETGE motifs did not bind (Fig. 5C). Similar to the finding in mammals (31), Ref(2)P and CncC did not interact in vitro (Fig. 5D).

Mapping of a CncC Binding Site in the ref(2)P Promoter-We have previously mapped an ARE site in the p62 promoter (31).
To study whether Nrf2-mediated p62 induction is conserved in flies, we made a reporter construct encompassing nucleotides Ϫ1425 to ϩ173 of the D. melanogaster ref(2)P upstream region fused to luciferase (named pGL3-Pref(2)P(Ϫ1425/ϩ173)-Luc). As shown in Fig. 6, A and E, co-transfection of the pGL3-Pref(2)P(Ϫ1425/ϩ173)-Luc reporter plasmid and the 3xFLAG-CncC expression vector in S2Rϩ cells resulted in up to 40-fold induction of reporter gene activity, depending on the duration of transfection, compared with the control 3xFLAG expression vector. This indicates that ref (2)P is a transcriptional target for CncC.   JUNE 12, 2015 • VOLUME 290 • NUMBER 24

JOURNAL OF BIOLOGICAL CHEMISTRY 14953
To investigate the transactivation ability of the different isoforms of D. melanogaster Cnc, we co-expressed CncA, -B, and -C with pGL3-Pref(2)P(Ϫ1425/ϩ173)-Luc in S2Rϩ cells. CncC strongly induced the ref(2)P promoter, CncB gave no induction, and CncA had a negative effect (Fig. 6A). Co-transfection of CncA or CncB with CncC along with pGL3-Pref(2)P(Ϫ1425/ ϩ173)-Luc showed that CncA and CncB inhibited CncC-mediated activation of the ref(2)P promoter (Fig. 6A). CncA and -B may act as inhibitors to modulate or reduce the CncC mediated induction. Overexpression of the 3xFLAG-Cnc isoforms in S2Rϩ cells showed that CncA and CncB were very well detectable under normal conditions, whereas we did not see any CncC (Fig. 6B). This may be due to rapid proteasomal degradation of CncC mediated by DmKeap1. This was confirmed by analysis of the expression level of 3xFLAG-CncC in S2Rϩ cells treated with the proteasomal inhibitor MG132 (Fig. 6C). Consistent with a previous study (65), we observed accumulation of CncC upon proteasomal inhibition. This did not occur after BafA1 treatment (Fig. 6C). BafA1 inhibits autophagosome-lysosome fusion and degradation of macromolecules in the lysosomes due to neutralization of the lysosomal pH. Hence, CncC is rapidly turned over by the proteasome in S2Rϩ cells under our experimental conditions. Analysis of the ref(2)P promoter sequence revealed three ARE-like consensus sequences positioned at Ϫ1070, Ϫ525, and Ϫ25. Luciferase reporter gene assays using constructs with mutated or deleted putative ARE sites revealed that CncC mainly mediates ref (2)P promoter induction via the Ϫ25 site (GTGGTGAGTTCGATA). Deletion of the two other putative sites did not give any significant reduction (Fig. 6E). The three single nucleotide point mutations of the Ϫ25 ARE at Ϫ31, Ϫ29, and Ϫ27 positions (TGAGT to AGGGA) in the longer (Ϫ1425/ ϩ173) and shorter (Ϫ426/ϩ173) variants of the ref(2)P promoter significantly reduced the CncC-mediated induction of the ref(2)P promoter (Fig. 6F). This clearly indicates that the sequence 5Ј-GTGGTGAGTTCGATA-3Ј between nucleotides Ϫ36 and Ϫ22 is a minimal functional ARE in the ref(2)P promoter. The Ϫ25 ARE site identified in the ref(2)P gene promoter is conserved (Fig. 6D). To investigate the association of CncC with the ref(2)P promoter in vivo, we performed ChIP analyses using S2Rϩ cells. However, due to lack of a CncC antibody, we co-transfected pGL3-Pref(2)P(Ϫ1425/ϩ173)-Luc and 3xFLAG-CncC in S2Rϩ cells and used an anti-FLAG antibody for the ChIP experiments. The Pref(2)P fragment encompassing the Ϫ25 ARE co-precipitated with anti-FLAG antibodies, strongly suggesting that CncC is associated with the upstream regulatory region of ref(2)P in S2Rϩ cells (Fig. 6G).
In mammalian cells, p62 can activate its own promoter by sequestering Keap1. Because Ref (2) (Fig. 6H). Interestingly, co-overexpression of these along with CncC clearly showed that DmKeap1 and DmMafS repressed the ref(2)P promoter, whereas Ref(2)P co-expression had no effect (Fig. 6H). (2)P levels in D. melanogaster, using a well characterized oxidative stress reporter with the Drosophila glutathione S-transferase D1 promoter fused to a GFP-expressing cassette (gstD-GFP) as a positive control (17,66). CncC overexpression in Drosophila larvae was performed using the engrailed-Gal4 driver, which is expressed in the dorsal domain in the hindgut, posterior domain in wing discs, and scattered cells in the fat body. Induction of gstD-GFP in response to ectopic expression of CncC was observed in hindgut, wing discs, and fat bodies (Fig. 7, A-C) (data not shown). Importantly, overexpression of CncC induced strong accumulation of Ref(2)P protein in hindgut, wing imaginal disc epithelium, and fat body cells (Figs. 7C and 8) (data not shown) compared with neighboring control cells with no ectopically expressed CncC (Fig. 7B) and ref (2)P knockdown using ref (2)P-IR (Fig. 7H). This strongly indicates that ref (2)P is a transcriptional target of CncC in D. melanogaster. Furthermore, the ref(2)P induction upon oxidative stress is mediated by CncC, and this regulatory relationship is not restricted to one organ or cell type. We have shown that p62 enhances NRF2 expression via KEAP1. Because the KIR motif-mediated DmKeap1-Ref(2)P interaction is not conserved, we asked whether the positive feedback loop in mammals was present in Drosophila. Overexpression of Ref(2)P led to a strong increase in Ref(2)P protein levels but did not give any effect on gstD-GFP expression compared with control with no overexpressed Ref(2)P and ref (2)P-IR conditions (Fig. 7, B, G, and H). These results confirm

Nrf2/CncC Regulates ref(2)P, atg8a, and Autophagy
that there is no positive feedback regulation of Ref(2)P and CncC in Drosophila as seen in mammals. This loop probably evolved later in evolution or may have been lost in insects.
CncC Induces Autophagy Independent of TFEB/MitF-Loss of keap1 in the Drosophila larval brain was previously reported to activate CncC-driven GstD1 reporter gene expression, showing that this tissue harbors latent CncC activity (17). To investigate whether autophagy regulation is a part of this response, we investigated labeled keap1 loss of function clones carrying an autophagy reporter transgene, pmCherry-Atg8a. This reporter is expressed under the control of an endogenous 2-kb upstream gene regulatory sequence of atg8a, thus serving as a dual gene expression and autophagy activity reporter (67). Increased mCherry-Atg8a fluorescence was seen in keap1-deficient (GFP-positive) clones in brain lobes, suggesting that endogenous CncC may regulate atg8a transcription or possibly post-translational stability and autophagy activity (Fig. 8A). To test this hypothesis, we therefore expressed CncC and GFP-CncC in random flip out clones of the gut and fat body, two tissues often used to assess autophagy response due to their large cell size. Expression of CncC led to increased Lysotracker activity in fat body and midgut cells of early L3 larvae, suggesting increased autophagy and lyososomal activity (Fig. 8D). Expression of both CncC and GFP-CncC transgenes cell-autonomously increased mCherry-Atg8a levels and puncta formation of pmCherry-Atg8a flies (Fig. 8, B and C). Because mammalian Atg8 family proteins have been reported to incorporate into protein aggregates not representing bona fide autolysosomes and CncC overexpression induced high levels of Ref(2)P protein and aggregation, we sought to test whether the structures represent protein aggregates or autophagic structures. CncC-induced mCherry-Atg8a structures co-localized with Lysotracker, suggesting that they indeed represent autolysosomes (Fig. 8, D and E). Moreover, knockdown of atg9 in cells overexpressing CncC selectively inhibited mCherry-Atg8a puncta formation, but Ref(2)P up-regulation and aggregation perdured, suggesting that mCherry-Atg8a structures indeed represent autophagic activity (Fig. 8, G and H). To partially elucidate how CncC may induce autophagy, we addressed the intersection with the TORC1-regulated autophagy. In mammals, the transcription factor TFEB acts downstream of TORC1 and is required for autophagy upon starvation. Expression of a dominant negative version of the potential Drosophila TFEB orthologue, MitF, blocked autophagy induction upon hydrogen peroxide-mediated oxidative stress, suggesting that MitF indeed is the Drosophila TFEB orthologue and is required for autophagy (Fig. 8F) (50). In contrast to atg9 knockdown, co-expression of MitF-DN with GFP-CncC failed to reverse Ref(2)P and mCherry-Atg8a accumulation and puncta formation (Fig. 8I), suggesting that CncC-induced autophagy is independent of TORC1 regulation. To address whether up-regulation of mCherry-Atg8a levels was due to transcriptional or post-translational activities, we repeated the experiments using an independent autophagy reporter where mCherry-Atg8a is expressed under a synthetic fat body-expressed promoter element, r4Cherry-Atg8a. As expected, this reporter again showed robust cell-autonomous accumulation of mCherry-Atg8a structures that could be reversed by atg9 but not MitF inhibition (Fig. 8, J-L). In all experiments, Ref(2)P levels were strongly up-regulated and formed protein aggregates within the cytoplasm. Importantly, using this reporter line, mCherry-Atg8a levels were not up-regulated in CncC-expressing cells compared with its neighbors, suggesting that accumulation of mCherry-Atg8a in pmCherry-Atg8a flies is not due to post translational regulation but rather regulated at the transcriptional level and that this regulation is separable from autophagy induction. Taken together, this suggests that CncC activity upon loss of keap1 increased atg8a transcription. Moreover, CncC can induce autophagy in gut and fat body cells independently of the TORC1 effector MitF.

Discussion
Ref (2)P, the single p62 orthologue in D. melanogaster, is an established signaling adapter (37,38,40,68). Similar to p62, Ref(2)P accumulates with ubiquitin-containing protein aggregates in the brain of autophagy-deficient and neurodegenerative mutants of Drosophila (39,44). Ref (2)P is involved in maintenance of the viable mitochondria pool by acting downstream of Pink1 and Parkin, where Ref(2)P recycles excessive unfolded proteins via autophagy (69). This indicates a role for Ref(2)P as an autophagy receptor, similar to mammalian p62. Moreover, a recent report suggests that Ref(2)P is a selective autophagy substrate (70), but direct binding of Ref(2)P to DmAtg8a has not been shown. We found that Ref(2)P interacted with DmAtg8a in a LIR-dependent manner. The Ref(2)P LIR also fulfills the requirement for a canonical LIR motif (24,25,54). The func-   Only the longest Cnc isoform, CncC, contains the Keap1interacting DLG and ETGE motifs. Homology with the Neh2 domain of NRF2 suggests CncC to be the direct homologue of NRF2 (9). Similar to NRF2, CncC is thought to interact with DmKeap1 (9), and the activity of CncC is inhibited by DmKeap1 (71). In this study, we tested all three isoforms of Cnc (A, B, and C) for binding to DmKeap1, and only CncC interacted. The binding is mediated by the conserved ETGE motif. It has been reported both in humans and rodents that p62 binds directly to Keap1 using an ETGE-like motif, and this interaction positively regulates Nrf2 by blocking the interaction between Keap1 and Nrf2 (31)(32)(33)72). A simple protein-protein interaction map involving p62/Ref(2)P, KEAP1/DmKeap1, NRF2/ CncC, and ATG8/DmAtg8a shown in Fig. 9 highlights the major differences and similarities between humans and D. melanogaster. Surprisingly, we did not find any direct interaction between Ref(2)P and DmKeap1 in Drosophila. This was supported by reporter gene assays where Ref(2)P did not activate its own promoter, as p62 does in mammals. Recently,

Nrf2/CncC Regulates ref(2)P, atg8a, and Autophagy
Ref (2)P and DmKeap1 were reported to be co-immunoprecipitated from Drosophila cells (58,59). Here, we found this to be mediated by the UBA domain of Ref(2)P, which probably recognized ubiquitinated DmKeap1. Hence, the direct KIR-mediated interaction between p62 and Keap1 evolved with the vertebrates, consistent with the lack of a KIR motif in p62 orthologues in non-chordate metazoans.
CncC has a central role in regulation of xenobiotic response, cellular stress, and electrophilic stress (3,13,73). We found that CncC significantly induced the ref(2)P promoter, CncB gave no induction, and CncA had a significant negative effect. This is most probably due to the lack of the N-terminal transactivation domain found in CncC. Overexpressed CncC is a proteasome substrate, not detectable under normal conditions, but is stabilized by inhibition or depletion of proteasome subunit S5a (65). Consistently, we could not easily detect overexpressed CncC by Western blot in cultured cells unless proteasomal degradation was inhibited. However, CncA and CncB were very well expressed under similar conditions. This is interesting, because only CncC has the DLG and ETGE motifs for Keap1 binding. This indicates that DmKeap1 may work as an E3 ligase adaptor protein to recruit CncC for its proteasomal degradation as found for mammalian Keap1. Our data suggest that CncA and -B may compete with CncC for DNA binding to the ARE on the ref(2)P promoter. This indicates a role for the CncA and CncB isoforms as competitive repressors of CncC under non-stress conditions, to maintain homeostasis in the D. melanogaster antioxidant defense system. This is in analogy to p65 (truncated isoform of Nrf1) and a caspase-cleaved form of Nrf2, which are reported to act as transcriptional repressors in vertebrates (74,75).
Consistent with our results from cell culture experiments, CncC overexpression induced ref (2)P and the target gene gstD in hindgut, wing discs, and fat bodies of D. melanogaster. These in vivo results strongly support our finding that ref (2)P is a target gene for CncC. Interestingly, overexpression of Ref(2)P did not induce an oxidative stress response, at least not as measured by gstD-GFP expression in hindgut and wing discs. This correlates with the lack of a ETGE-like motif in Ref (2) Surprisingly, we found a direct interaction between DmKeap1 and DmAtg8a. The canonical LIR-LDS interaction is dependent on both the N-terminal part (residues 1-28) and the C-terminal part (residues 30 -125) of LC3B (26,62,63). However, the mode of the DmKeap1-DmAtg8a interaction seems different because DmKeap1 could bind to the N-terminal 71 amino acids of DmAtg8a(1-71). We have not been able to map any motif mediating this non-canonical interaction. However, the interaction of DmKeap1 with DmAtg8a is interesting, both because DmKeap1 appears to recognize a different binding surface in DmAtg8a than Ref(2)P and other LIR-containing proteins and because the interaction might have an important role for DmAtg8a-mediated autophagic degradation of DmKeap1 during Drosophila development. Recent studies show mammalian Keap1 to be degraded by autophagy under nutritional starvation and oxidative stress in a p62-dependent manner (34,76), whereas degradation of mammalian Keap1 by basal autophagy has not been clearly demonstrated. Consistent with this, we observed no degradation of DmKeap1 by basal autophagy in Drosophila, but DmKeap1 was degraded under programmed autophagy during Drosophila development. The significance of the DmAtg8a-DmKeap1 interaction for the degradation of DmKeap1 by autophagy remains to be tested. We favor the idea that autophagic degradation of DmKeap1 depends on a co-recruitment of autophagy receptors like Ref(2)P, and this is strongly supported by our finding that Ref(2)P interacts with ubiquitinated DmKeap1 in cell culture. Possibly, a combined binding of Ref(2)P and DmKeap1 to DmAtg8 may help to increase the local concentration of DmAtg8 to secure an efficient encapsulation of the aggregate. However, further work is needed to reveal the underlying mechanism and significance of this interaction.
Our finding that there is no positive feedback loop between CncC and Ref(2)P in flies was quite unexpected. However, the introduction of a KIR motif in p62 homologs during evolution of the most primitive fish, the amphioxus, suggests that the interdependent role of p62 and NRF2 in oxidative stress regulation developed early in vertebrate evolution. A related type of positive feedback loop has been reported for mammalian p62 and NFB (77), predicting a putative cross-talk between NRF2 and NFB pathways. Both gain of function mutations in NRF2 and loss of function mutations in Keap1 have been identified in human cancers and are believed to contribute to cancer cell survival and stress resistance upon cancer treatment (78).
Here we find, for the first time in vivo, that loss of Keap1 or CncC gain of function induces Atg8a up-regulation and autophagy. These results are thought provoking, because autophagy, like NRF2 gain of function, has been shown to prevent initial tumor development. Once established, however, autophagy promotes cancer cell survival during stress conditions and cancer treatment (79). Under which physiological set- tings CncC-mediated control of autophagy may function remains an open question. The most obvious possibility is that CncC controls autophagy in response to reactive oxygen species. Previous studies have established that Drosophila utilizes a TRAF6/Atg9/Jun N-terminal kinase (JNK) stress pathway to activate autophagy upon oxidative stress provoked by hydrogen peroxide feeding (80,81). In concordance with those studies, we found no evidence that CncC activity is required for autophagy induced upon hydrogen peroxide feeding (results not shown). It remains possible that more subtle and physiological conditions of reactive oxygen species formation during aging, mitochondrial dysfunction, or oncogene-induced stress may enlist CncC in stress coping mechanisms involving autophagy. It will be interesting to pursue potential roles of CncC in in vivo cancer models.