SUMO peptidase ULP-4 regulates mitochondrial UPR-mediated innate immunity and lifespan extension

Animals respond to mitochondrial stress with the induction of mitochondrial unfolded protein response (UPRmt). A cascade of events occur upon UPRmt activation, ultimately triggering a transcriptional response governed by two transcription factors: DVE-1 and ATFS-1. Here we identify SUMO-specific peptidase ULP-4 as a positive regulator of C. elegans UPRmt to control SUMOylation status of DVE-1 and ATFS-1. SUMOylation affects these two axes in the transcriptional program of UPRmt with distinct mechanisms: change of DVE-1 subcellular localization vs. change of ATFS-1 stability and activity. Our findings reveal a post-translational modification that promotes immune response and lifespan extension during mitochondrial stress.


Introduction
The ability of an organism to cope with an ever-changing and challenging environment lies in its ability to activate stress responses. Failure to appropriately respond to different stresses and maintain cellular and organismal homeostasis could result in multiple diseases including metabolic and neurodegenerative disorders (Jovaisaite et al., 2014;Lee and Ozcan, 2014;Wang and Kaufman, 2012). Animals respond to mitochondrial stress with the induction of mitochondrial unfolded protein response (UPR mt ), a surveillance program that monitors mitochondrial function and initiates mitochondria-to-nucleus crosstalk to maintain mitochondrial protein-folding homeostasis (Benedetti et al., 2006;Yoneda et al., 2004) and coordinate the expression of electron transport chain (ETC) components in mitochondrial and nuclear genomes (Houtkooper et al., 2013). UPR mt also elicits global changes to reprogram metabolism (Nargund et al., 2015;Nargund et al., 2012), activate immune responses Melo and Ruvkun, 2012;Pellegrino et al., 2014) and extend lifespan (Durieux et al., 2011;Merkwirth et al., 2016;Tian et al., 2016).
UPR mt signaling ultimately activates a transcriptional response governed by two transcription factors: ATFS-1 and DVE-1. ATFS-1 contains an N-terminal mitochondrial targeting sequence and a C-terminal nuclear localization sequence. Under normal condition, ATFS-1 is imported into mitochondria, where it is degraded by mitochondrial protease LON. During mitochondrial stress, mitochondrial import efficiency is impaired, resulting in nuclear accumulation of ATFS-1 (Nargund et al., 2012). ATFS-1 controls approximately half of the mitochondrial stress response genes, including those encoding mitochondrial-specific chaperones, proteases and immune response genes (Nargund et al., 2012). ATFS-1 also regulates genes involved in metabolic reprogramming, such as those functioning in glycolysis (Nargund et al., 2015). Another axis of the UPR mt transcriptional program relies on DVE-1, a homeobox transcription factor homologous to human SATB1/SATB2. Upon mitochondrial perturbation, DVE-1 translocates from cytosol to nucleus, binds to the open-up chromatins devoid of H3K9me2, and initiates the transcription of mitochondrial stress response genes (Haynes et al., 2007;Tian et al., 2016). While several core components of UPR mt have been identified, the regulation, especially post-translational regulation of these components has not been reported.
Aberrant activity of SUMOylation drastically affects cellular homeostasis and has been linked with many diseases (Flotho and Melchior, 2013;Mo et al., 2005;Sarge and Park-Sarge, 2009;Seeler et al., 2007). It has been reported that SUMO could covalently modify Drp1, a protein essential for mitochondrial dynamics (Prudent et al., 2015). In addition, SUMOylation of a pathogenic fragment of Huntingtin, a PolyQ-repeats protein that specifically binds to the outer membrane of mitochondria and impairs mitochondrial function, has been reported to exacerbate neurodegeneration in a Drosophila Huntington's disease model (Costa and Scorrano, 2012;Panov et al., 2002;Steffan et al., 2004).
In the present study, we find that under mitochondrial stress, SUMO-specific peptidase ULP-4 is required to deSUMOylate DVE-1 and ATFS-1 to activate UPR mt in C. elegans. ULP-4 is also required eLife digest Most animal cells carry compartments called mitochondria. These tiny powerhouses produce the energy that fuels many life processes, but they also store important compounds and can even cause an infected or defective cell to kill itself. For a cell, keeping its mitochondria healthy is often a matter of life and death: failure to do so is linked with aging, cancer or diseases such as Alzheimer's.
The cell uses a surveillance program called the mitochondrial unfolded protein response to assess the health of its mitochondria. If something is amiss, the cell activates specific mechanisms to fix the problem, which involves turning on specific genes in its genome.
A protein named ULP-4, which is found in the worm Caenorhabditis elegans but also in humans, participates in this process. This enzyme cuts off chemical 'tags' known as SUMO from proteins. Adding and removing these labels changes the place and role of a protein in the cell. However, it was still unclear how ULP-4 played a role in the mitochondrial unfolded protein response.
Here, Gao et al. show that when mitochondria are in distress, ULP-4 removes SUMO from DVE-1 and ATFS-1, two proteins that control separate arms of the mitochondrial unfolded protein response. Without SUMO tags, DVE-1 can relocate to the area in the cell where it can turn on genes that protect and repair mitochondria; meanwhile SUMO-free ATFS-1 becomes more stable and can start acting on the genome. Finally, the experiments show that removing SUMO on DVE-1 and ATFS-1 is essential to keep the worms healthy and with a long lifespan under mitochondrial stress.
The experiments by Gao et al. show that the mitochondrial unfolded protein response relies, at least in part, on SUMO tags. This knowledge opens new avenues of research, and could help fight diseases that emerge when mitochondria fail.
to promote UPR mt -mediated innate immunity and lifespan extension. Our results reveal an essential and unexplored function of post-translational regulation in UPR mt signaling.

SUMO-specific peptidase ULP-4 is required for the activation of UPR mt
Previously, we have performed a genome-wide RNAi screen to identify genes that are required for the activation of mitochondrial unfolded protein response (UPR mt ) in C. elegans . ulp-4, a gene encoding ortholog of SUMO-specific peptidase in C. elegans, is one of the hits from our primary screen. RNAi of ulp-4 impaired the activation of UPR mt that is induced by mitochondrial inhibitor antimycin A, or RNAi of nuclear encoded mitochondrial gene spg-7 (mitochondrial metalloprotease) ( Figure 1A-B and Figure 1-figure supplement 1A). RNAi of cco-1 (nuclear-encoded cytochrome c oxidase-1 subunit) is also widely used to disrupt mitochondrial function and activate UPR mt (Durieux et al., 2011;Nargund et al., 2012;Pellegrino et al., 2014). Consistently, deficiency of ulp-4 also suppressed the induction of endogenous mitochondrial chaperone genes hsp-6 and hsp-60 under cco-1 RNAi ( Figure 1C and Figure 1-figure supplement 1B). Notably, transcript level of ulp-4 was also elevated during mitochondrial stress (Figure 1-figure supplement 1C). In contrast, ulp-4 RNAi did not affect the induction of endoplasmic reticulum (ER) stress reporter hsp-4p::gfp nor heat shock stress reporter hsp-16.2p::gfp ( Figure 1D-E). In addition, worms treated with ulp-4 RNAi were still able to induce the expression of endogenous hsp-4 or hsp-16.2 during ER or heat shock stress ( Figure 1F). RNAi of ulp-4 from L1 stage only delayed worm development a bit ( Figure 1-figure supplement 1D). To further exclude the possibility that the suppression of UPR mt by ulp-4 RNAi is due to developmental delay, we also treated worms with ulp-4 RNAi starting at L4 stage and observed the reduction of UPR mt as well (Figure 1-figure supplement 1E). Overexpression of ULP-4 in ulp-4 RNAi worms rescued UPR mt activation ( Figure 1G and Figure 1-figure supplement 1F). Lastly, we crossed an ulp-4(tm1597) mutant allele that lacks 404nt in the promoter region of ulp-4 with hsp-6p::gfp reporter, and showed that the induction of UPR mt was also impaired in ulp-4 mutants ( Figure 1H-I and Figure 1-figure supplement 1G).
To see if other SUMO peptidases have similar effects to regulate UPR mt , we treated C. elegans with ulp-1, 2, or 5 RNAi ( Figure 1J) and tested for their abilities to induce UPR mt . Deficiency of ulp-1, 2, or 5 failed to suppress antimycin-or spg-7 RNAi-induced UPR mt ( Figure 1K), suggesting a specific role of ulp-4 in mediating UPR mt . Sequence alignment (Katoh and Standley, 2013) of ulp-1, 2, 4 and 5 revealed that the only conserved region among them is the catalytic domain ( Figure 1-figure supplement 2). Thus, the specificity of ULP-4 in UPR mt signaling might be due to its ability to specifically interact with other protein components in UPR mt pathway.
Conversely, RNAi of the E1 SUMO activating enzyme aos-1 or the E2 SUMO conjugating enzyme ubc-9 in C. elegans induced UPR mt more potently ( . Taken together, these results suggest that ULP-4 plays a specific role to mediate mitochondrial stress response through its SUMO peptidase activity.

ULP-4 deSUMOylates DVE-1 at K327 residue during mitochondrial stress
To understand the molecular mechanism of ULP-4 in mediating UPR mt , we sought to identify its protein targets. We first performed a cherry-picked yeast two-hybrid screen to test if ULP-4 could interact with known UPR mt pathway components. We found that DVE-1, a homeodomain-containing transcription factor in UPR mt (Haynes et al., 2007), interacted with ULP-4 ( Figure 2A). Notably, DVE-1 could specifically interact with ULP-4, but not ULP-2 or ULP-5 (Note: overexpression of ULP-1 in yeast is lethal) (Figure 2-figure supplement 1A). Consistently, smo-1 is one of the top hits from our yeast two-hybrid screen with DVE-1 as bait (Figure 2-figure supplement 1B).
To test if ULP-4 could directly deSUMOylate DVE-1, we employed a yeast three-hybrid experiment to induce the expression of ULP-4 or ULP-5 in yeasts. The expression of ULP-4 or ULP-5 was driven by a met17 promoter, which could be induced when growth media is deficient for methionine.Induction of ULP-4, but not ULP-5, prevented yeast growth caused by SMO-1-DVE-1 interaction ( Figure 2G), suggesting that ULP-4 may removes SUMO moiety from DVE-1. Overexpression of ULP-4 in worms decreased SUMOylation level of DVE-1 ( Figure 2H). In addition, expression of ULP-4 C48, the catalytic domain of ULP-4 (Letunic and Bork, 2018) in 293 T cells could deSUMOylate DVE-1 in mammalian cells ( Figure 2I). Lastly, if during mitochondrial stress, ULP-4 is indeed required to deSUMOylate DVE-1, mutation of K327 to arginine that prevents DVE-1 SUMOylation would bypass the requirement of ULP-4 for UPR mt induction. Indeed, we found that a CRISPR/Cas9 knockin strain with DVE-1 K327R mutation bypassed the requirement of ulp-4 and was capable to activate UPR mt under ulp-4 RNAi ( Figure 2J and

SUMOylation affects the subcellular localization of DVE-1
We next aimed to understand how SUMOylation affects DVE-1 function. During mitochondrial stress, DVE-1 translocates from cytosol to nucleus (Haynes et al., 2007;Tian et al., 2016) ( Figure 3A and B). Inactivation of ulp-4 by RNAi abolished the nuclear accumulation of DVE-1 in cco-1 RNAi-treated worms ( Figure 3A-C). Conversely, overexpression of ulp-4 increased the nuclear accumulation of DVE-1 ( Figure 3D). More importantly, we found that the induction of UPR mt correlated well with DVE-1 subcellular localization. When we fed worms with ulp-4 RNAi for one generation and treated their progeny with ulp-4 RNAi for twenty-four hours to allow efficient ulp-4 knockdown, and then fed animals with cco-1 RNAi, we observed cytosolic accumulation of DVE-1 and suppression of UPR mt ( Figure 3E). However, when we treated progeny with a mixture of ulp-4 and cco-1 RNAi, which induced mitochondrial stress before ulp-4 was efficiently knocked down, DVE-1 was still able to translocate to the nucleus and induce UPR mt ( Figure 3E).
We also tested the subcellular localization of DVE-1 K327R under ulp-4 RNAi. Different from wild type proteins, DVE-1 K327R constitutively localized in the nucleus of C. elegans, even if ulp-4 was knocked down by RNAi ( Figure 3F). Conversely, SUMO-mimetic DVE-1 constitutively localized in the cytosol ( Figure 3G). Taken together, during mitochondrial stress, ULP-4 deSUMOylates DVE-1 at K327 residue to allow its nuclear accumulation to initiate UPR mt .

SUMOylation affects the stability and transcriptional activity of ATFS-1
To see if SUMO also affects ATFS-1 localization, we used hsp-16.2 heat shock promoter to drive the expression of GFP-tagged full-length ATFS-1. Upon mitochondrial perturbation, ATFS-1 was still able to translocate to the nucleus when ULP-4 expression is diminished ( Figure 5A). It should be noted that it is very difficult to express full-length ATFS-1 in worms, probably due to toxicity (expression of full-length ATFS-1 in yeast is lethal). Therefore, we made truncations of ATFS-1, and found  Interestingly, we noticed that ulp-4 RNAi greatly reduced the protein level of ATFS-1 D 1-184 (Figure 5-figure supplement 1C), suggesting that ulp-4 RNAi may affect ATFS-1 expression, or stability. Seven hours after heat induction of ATFS-1 expression, levels of ATFS-1 D 1-184 were comparable in control or ulp-4 RNAi animals ( Figure 5B). However, after 24 hours, ATFS-1 D 1-184 level in ulp-4 RNAi treated animals significantly decreased ( Figure 5B). Treating worms with proteasome inhibitor MG132 partially rescued the reduction of ATFS-1 D 1-184 under ulp-4 RNAi ( Figure 5-figure supplement 1C), suggesting that ulp-4 RNAi affects ATFS-1 protein stability. Full-length ATFS-1 proteins could be detected when mitochondrial protease LON-1 was inhibited (Nargund et al., 2012). Treating worms with lon-1 RNAi, we showed that full-length ATFS-1 levels were also reduced under ulp-4 RNAi ( Figure 5C). Mutation (K326R) that abolished ATFS-1 SUMOylation partially restored its protein level under ulp-4 RNAi ( Figure 5D-E). In contrary, fusion of SMO-1 to mimic SUMOylated form of ATFS-1 significantly reduced its protein level ( Figure 5D-E). Taken together, these results suggest that SUMOylation reduces the stability of ATFS-1.
ulp-4 is essential for UPR mt -regulated innate immunity and lifespan extension During mitochondrial stress, UPR mt not only initiates mitochondrial protective responses, but also activates immune responses and extends worm lifespan Merkwirth et al., 2016;Pellegrino et al., 2014;Tian et al., 2016). The essential function of ulp-4 in signaling UPR mt makes it likely to be crucial for animal fitness during mitochondrial stress. Indeed, worms treated with cco-1 RNAi had a severe synthetic growth defect on ulp-4 RNAi ( Figure 6A). Consistently, ulp-4(tm1597) mutants revealed a more severe developmental delay when grown on spg-7 RNAi (Figure 6-figure  supplement 1A-B). Mutation of the SUMOylation sites of ATFS-1 and DVE-1 in C. elegans partially rescued the developmental delay of spg-7 mutants ( Figure 6-figure supplement 1C-E). The survival rate of worms exposed to high dosage of antimycin was also significantly reduced in ulp-4 RNAi ( Figure 6-figure supplement 1F-G).
A broad range of microbes isolated from natural habitats harboring wild C. elegans populations could perturb mitochondrial function and induce the expression of hsp-6p::gfp . ulp-4 RNAi also impaired UPR mt activation when we challenged worms with a Pseudomonas strain, a mitochondrial insult isolated from the natural habitat of C. elegans ) ( Figure 6B). Moreover, ulp-4 RNAi suppressed the activation of immune response and xenobiotic detoxification response ( Figure 6C). To further validate the requirement of ulp-4 in UPR mt -mediated immune response, we treated irg-1p::gfp transgenic worms, a reporter strain for pathogen-infected response (Estes et al., 2010), with control or ulp-4 RNAi and then challenged them with Pseudomonas. We showed that deficiency of ulp-4 greatly suppressed the induction of irg-1 ( Figure 6D). Deficiency of ulp-4 also impaired worm development and survival rate when they were infected with Pseudomonas ( Figure 6E-F) . The reduced survival rate of ulp-4-deficient worms could be rescued with atfs-1 K326R; dve-1 K327R mutations, further demonstrating that deSUMOylation of ATFS-1 and DVE-1 is the major function of ULP-4 during mitochondrial stress ( Figure 6G).
Finally, we analyzed the lifespans of control or ulp-4 RNAi worms. Under unstressed condition, ulp-4 RNAi did not affect worm lifespan. However, ulp-4 RNAi greatly suppressed the lifespan extension in cco-1 RNAi-treated worms, which could be rescued by mutations of SUMOylation site within ATFS-1 and DVE-1 ( Figure 6H). Overexpression of ulp-4 neither affects the basal level of UPR mt under unstressed condition ( Figure 1G), nor affects worm lifespans with or without mitochondrial stress ( Figure 6-figure supplement 1H). In contrary, smo-1 RNAi shortened worm lifespan, but greatly extended the lifespan of spg-7 mutants ( Figure 6-figure supplement 1I). Mutations of SUMOylation site within ATFS-1 and DVE-1 extended lifespan of spg-7 mutants as well ( Figure 6figure supplement 1J). Taken together, these results suggest that ulp-4 is required for mitochondrial stress-induced lifespan extension.
In summary, our studies indicate that mitochondrial stress signals through ULP-4, which deSU-MOylates DVE-1 and ATFS-1 to modulate their localization, stability and transcriptional activity. Consequences of these events are elevated innate immunity and prolonged lifespan (Figure 7).

Discussion
We have identified a SUMO-specific peptidase ULP-4 that participates in C. elegans UPR mt . ULP-4 regulates the entire transcriptional program of UPR mt , underscoring the importance of ULP-4-mediated deSUMOylation in UPR mt signaling. However, how mitochondrial stress signals to ULP-4 warrants future analysis.
SUMOylation affects protein function through several mechanisms, including changes of protein conformation, protein-protein interaction, protein stability and subcellular localization (Chymkowitch et al., 2015). Interestingly, we find that SUMOylation affects DVE-1 and ATFS-1 through two distinct mechanisms: change of DVE-1 subcellular localization vs. changes of ATFS-1 stability and transcriptional activity. DVE-1 and ATFS-1 constitute the two axes in the transcriptional program of UPR mt , each might regulate a different subset of downstream genes. For instance, ATFS-1 has been shown to be the primary factor that controls the expression of genes involved in mitochondrial protein folding, glycolysis, xenobiotic detoxification and immune response (Nargund et al., 2012;Pellegrino et al., 2014). A detailed analysis of DVE-1 and ATFS-1 substrate selection may facilitate the understanding of why cells employ such intricate regulation of transcriptional response during mitochondrial stress.
Several mitochondrial quality control processes have evolved to maintain and restore proper mitochondrial function, including mitochondrial unfolded protein response (UPR mt ), mitochondrial dynamics, and mitophagy (Andreux et al., 2013). Cells selectively activate each quality control pathway, depending on the stress level of mitochondria (Andreux et al., 2013;Pellegrino et al., 2013). Mild mitochondrial inhibition is often associated with the activation of UPR mt to maintain and restore proteostasis. As stress exceeds the protective capacity of UPR mt , cells may employ mitochondrial fusion to dilute damaged materials, and activate mitochondrial fission to isolate severely damaged mitochondria for removal through mitophagy. Drp1, the central protein that controls mitochondrial fission, could be SUMOylated (Prudent et al., 2015). A RING-finger containing protein MAPL functions as the E3 ligase to promote Drp1 SUMOylation on the mitochondria. SUMOylated Drp1 . Model for ULP-4-mediated UPR mt signaling. (A) Under normal condition, ATFS-1 is imported into mitochondria and degraded by mitochondrial protease. DVE-1 is mainly localized in the cytosol. No UPR mt is induced. The animal is normal lived and with normal innate immunity. (B) Upon mitochondrial stress, DVE-1 is deSUMOylated by ULP-4 and translocated to the nucleus. Mitochondrial importer efficiency is compromised, leading to ATFS-1 nuclear accumulation. ULP-4 deSUMOylates ATFS-1, leading to increased stability and transcriptional activity (active form). UPR mt is induced. Animals are long-lived, with increased innate immunity. (C) Reduction of ULP-4 activity during mitochondrial stress results in SUMOylation of DVE-1 and ATFS-1. SUMOylated of DVE-1 localizes in the cytosol; whereas SUMOylated ATFS-1, which has lower transcriptional activity, is prone to degradation. Induction of UPR mt is impaired. UPR mt -regulated innate immunity and lifespan extension is partially suppressed. DOI: https://doi.org/10.7554/eLife.41792.015 facilitates cristae remodeling, calcium flux and release of cytochrome c, and stabilizes ER/mitochondrial contact sites. Whether SUMOylation affects other proteins in the mitochondrial quality control processes, such as those govern mitophagy, are worth to explore. The discovery of SUMOylation in modulating UPR mt opens up a new research direction to study post-translational regulation of UPR mt , and UPR mt -mediated immunity and longevity. For generation of hsp-16.2p::dve-1::smo-1::gfp worms, hsp-16.2 promoter, dve-1::smo-1 and gfp sequences were sub-cloned into pDD49.26 vector. Three glycine residues were placed as linker between smo-1 and dve-1.

Cell line
HEK293T cell was obtained from ATCC, which was authenticated by ATCC. Cells were validated to be free of mycoplasma contamination. No commonly misidentified cell lines were used.
C. elegans, yeast and cell culture C. elegans were cultured at 20˚C and fed with E.coli OP50 on Nematode Growth Media unless otherwise indicated. Yeast strain AH109 for yeast two-hybrid and three-hybrid assays was cultured with YPDA media at 30˚C unless otherwise indicated. 293 T cells were cultured with DMEM (10% fetal bovine serum) at 37˚C.

Induction of UPR mt
For RNAi-induced UPR mt , RNAi bacteria were grown in LB containing 50 mg/ml carbenicillin at 37˚C overnight. 200 ml of RNAi bacteria was seeded onto 6 cm worm plates with 5 mM IPTG. Dried plates were kept at room temperature overnight to allow IPTG induction of dsRNA expression. Synchronized L1 worms were raised on the RNAi plates at 20˚C. After 24 hr, 200 ml 10X concentrated RNAi (atp-2, cco-1 or spg-7) bacteria were provided. GFP expressions were imaged after 48 hr.
For antimycin A induced UPR mt , synchronized L1 worms were raised on 6 cm worm plates for 48 hr. 200 ml of 20 mg/ml antimycin were then provided. Fluorescent images were taken 24 hr after the addition of antimycin.
For P. aeruginosa induced UPR mt , synchronized L1 worms were raised on 6 cm worm plates for 24 hr before exposure to P. aeruginosa. Worms were imaged at adulthood day 1.

Microscopy
Worms with each indicated fluorescent reporter were dropped in 100 mM NaN3 droplet on 2% agarose pads and imaged with a Zeiss Imager M2 microscopy.

RNA isolation and Q-PCR
Worms were fed with each indicated RNAi or exposed to pathogen. Adulthood day one worms were washed off plates and washed several times with M9 buffer until supernatant was clear. Worm pellets were re-suspended with TRIzol reagent (cwbiol, CAT#cw0580A). Samples were frozen and thawed six times to crack worms. Total RNA was isolated by chloroform extraction, followed by ethanol precipitation and DNase treatment. cDNA was then synthesized by reverse transcription (transgene biotech, CAT#AT311-03). Quantitative real-time PCR was carried out using SYBR GREEN PCR Master Mix (Bio-Rad, CAT#1725121). Quantification of transcripts was normalized to act-3.

Yeast two-hybrid
Each indicated gene was cloned into pGADT7 or pGBDT7 plasmid (Clontech). Plasmids were then transformed into AH109 yeast strain. We first seeded yeast on -Trp and -Leu double dropout solid culture medium. After yeast colonies were formed, we picked individual yeast colony into sterile water, adjusted it to the same OD and dropped onto -Trp, -Leu, -His and -Ade four dropout solid culture medium. Images of yeast were taken after culturing at 30˚C for 2-4 days.
Yeast two-hybrid screen Full-length dve-1 was cloned into pGBDT7, and transformed into AH109 yeast strain to test for autoactivation and protein expression. We then transformed C. elegans AD library plasmids into yeast that already contains dve-1 BD plasmid. Yeasts were seeded onto four dropout solid culture medium and grown at 30˚C for a week. Each colony was picked, followed by mini-prep to extract AD plasmid for PCR and sequencing.

Yeast three-hybrid
Yeast three-hybrid for DVE-1 deSUMOylation were carried out using pBRIDGE and pGADT7 vectors. smo-1 was cloned into pGADT7. dve-1 alone, or together with ulp-4/5, were cloned into pBRIDGE. pGADT7 and pBRIDGE plasmids were transformed into AH109 strain and cultured on -Leu and -Trp double dropout solid culture medium for yeast growth, on -Leu, -Trp and -His dropout medium for SMO-1 and DVE-1 interaction assay, on -Met, -Leu, -Trp and -His dropout medium to induce the expression of ULP-4 or ULP-5 and test for deSUMOylation activity.

MG132 treatment of worms
Worms were grown on control or ulp-4 RNAi plates for 48 hr before heat shock at 37˚C for 1 hr. The worms were then transferred in M9 containing indicated RNAi bacteria and 100uM MG132. Images were taken after 24 hr.

Developmental delay and survival assay
To assay developmental delay induced by P. aeruginosa, worms were pre-treated with control or ulp-4 RNAi at 20˚C for 16 hr. P. aeruginosa was then dropped onto worm plates. Images were taken two days later.
To assay survival rate of worms under antimycin treatment, worms were grown at 20˚C on 6-well RNAi plates seeded with control or ulp-4 RNAi. After 48 hr, 20 mg antimycin A were provided.

Lifespan analysis
Lifespan analyses were conducted on RNAi plates at 20˚C. More than 100 synchronized L1 were seeded onto 6 cm worm plates with control or ulp-4 RNAi. Animals that did not move when gently touched were scored as dead. Worms were transferred every 2 days to new plates during the first 10 days and were transferred every 3-5 days afterwards. Lifespan experiments were performed twice.

PA14 survival assay
PA14 survival assay was carried out as described previously . PA14 was freshly streaked from frozen stock and cultured 37˚C for 16 hr. 10 mL PA14 were then spread onto 3.5 cm slow killing agar plates (3.5 mg/mL peptone, 3 mg/mL NaCl, 17 mg/mL Agar, 1 mM MgSO4, 25 mM 1MKH2PO4, pH 6, 1 mM CaCla, 5 mg/mL cholesterol). The slow killing plates seeded with PA14 were cultured at 37˚C for 24 hr. FUDR was spread on the plates and L4 worms were picked onto plates and cultured at 25˚C. Worms were scored at least three times per day after16 hours, until all worms were dead. PA14 survival assay was performed 2 times with three replications each time.

CRISPR/Cas9 knock-in
The pDD162 CRISPR/Cas9 expression plasmid was obtained from Addgene (#47549). C. elegans Cas9 target prediction tool (https://crispr.cos.uni-heidelberg.de) was used to design target sequences. Templates for recombination were cloned into pDD49.26 vector with~500 bp overhang upstream and downstream of the target. Cas9 pam sequences were mutated in the templates. The plasmid was injected into worms, with Cas9-rol-6 as co-injection marker. Rolling worms were singled and validated by PCR and sequencing.

Quantification and statistical analysis
All experiments in this paper, if not specifically indicated, have been repeated for at least three times. Statistical analysis was performed with GraphPad. DVE-1 subcellular localization, transcriptional activity assay, QPCR and worm length were analyzed by Student's t-test. PA14 survival and lifespan assays were analyzed using Log-Rank method. *p<0.05, **p<0.002 and ***p<0.0002. Worm lengths, relative gfp expression and immunoblot quantification were analyzed by Image J. DVE-1 subcellular localization was counted in more than 40 worms per plate and three independent replicates were analyzed for each condition. Transcriptional activity assay wasanalyzed with three independent replicates for each condition. PA14 survival assay was performed with more than 35 worms per plate and at least two independent replicates were analyzed for each condition. The experiment was repeated for at least two times. Lifespan assay was performed with two biological replicates.