SUMO promotes longevity and maintains mitochondrial homeostasis during ageing in Caenorhabditis elegans

The insulin/IGF signalling pathway impacts lifespan across distant taxa, by controlling the activity of nodal transcription factors. In the nematode Caenorhabditis elegans, the transcription regulators DAF-16/FOXO and SKN-1/Nrf function to promote longevity under conditions of low insulin/IGF signalling and stress. The activity and subcellular localization of both DAF-16 and SKN-1 is further modulated by specific posttranslational modifications, such as phosphorylation and ubiquitination. Here, we show that ageing elicits a marked increase of SUMO levels in C. elegans. In turn, SUMO fine-tunes DAF-16 and SKN-1 activity in specific C. elegans somatic tissues, to enhance stress resistance. SUMOylation of DAF-16 modulates mitochondrial homeostasis by interfering with mitochondrial dynamics and mitophagy. Our findings reveal that SUMO is an important determinant of lifespan, and provide novel insight, relevant to the complexity of the signalling mechanisms that influence gene expression to govern organismal survival in metazoans.

www.nature.com/scientificreports/ were divided to get the ratio. For confocal microscopy, animals were immobilized on a 5% agarose pad, in a 5 μl drop of Nanobeads (Nanobeads NIST Traceable Particle Size Standard 100 nm, Polysciences). Images were taken with an LSM710 Zeiss confocal microscope, Axio-observer Z1.
For the analysis of SUMO conjugated proteins upon smo-1(RNAi) and SMO-1 overexpression (FGP14 strain), animals were washed and collected from two plates in M9 buffer with 0.1% Triton-X 100. Animals were washed one time with M9 buffer with 0.1% Triton-X 100 and transferred to a new tube. 200 μl of 40% trichloroacetic acid (TCA) (Sigma-Aldrich) and 300 μl glass beads (Sigma-Aldrich) were added to each sample. Samples were lysed by using a Beadbeater (Biospec), for 3 min (1 min on-1 min off) in a cold room. Lysates were transferred to a new tube and the beads were washed with 5% TCA and mixed with the lysates. Samples then were centrifuged for 15 min at 13,000 rpm at 4 °C. The pellet was washed three times with 500 μl chilled acetone. Pellets were dried at room temperature and resuspended in 100 μl 1X Sample Reducing Agent (Thermo Fisher Scientific). Following a 5 min incubation at 70 °C, samples were sonicated 2× for 10 s (12% amplitude). Samples were incubated again at 70 °C and then centrifuged for 10 min at 11,000 rpm. Samples were transferred to a new tube and loaded onto a 4-12% Bis-Tris gel for separation, transferred to a nitrocellulose membrane (Amersham-GE Healthcare). The membrane was blocked in a blocking solution (Invitrogen) for 1 h and then incubated overnight with anti-SUMO antibody (sheep) 35 (1:1000) and with anti-α-tubulin antibody (DSHB Cat# AA4.3, RRID:AB_579793, 1:5000) in 3% BSA at 4 °C. The next day the membrane was washed with 1XPBS-T (0.1% Tween 20) and incubated with AlexaFluor 647 secondary antibody for 1 h at room temperature. After washes with 1XPBS-T, the membrane was analyzed with Amersham Typhoon Biomolecular Imager (GE Healthcare).

Mitochondria isolation.
Age matched animals were collected in M9 buffer, and incubated at 4 °C with rotation in the presence of 10 mM DTT. To remove the DTT from the sample, three washes were performed with M9. Worms were homogenized in incubation buffer [50 mM Tris-HCl pH:7.4, 210 mM mannitol, 70 mM sucrose, 0.1 mM EDTA, 2 mM PMSF, cOmplete mini protease inhibitors cocktail (ROCHE)] with 100 strokes in a 3 ml Potter-Elvehjem homogenizer with PTFE pestle and glass tube (Sigma-Aldrich). The lysate was centrifuged at 200g for 1 min. The pellet was subjected to another round of homogenization. The lysate was combined with the supernatant from the previous centrifugation step, and they were centrifuged together for 1 min at 200g. The supernatant was centrifuged again at 12,000g for 5 min. We kept the supernatant as the cytoplasmic fraction and resuspended the pellet/mitochondrial fraction in incubation buffer. Samples were analysed on 4-12% SDS-PAGE gradient gels with anti-SUMO (DSHB Cat# SUMO 6F2, RRID:AB_2618393), anti-MTCO1 (Abcam [1D6E1A8] (ab14705)) and anti-α-tubulin (DSHB Cat# AA4.3, RRID:AB_579793).

DAF-16 and SKN-1 purification.
We codon optimized the sequence of isoform "a" of DAF-16 (UniProt number: O16850) and isoform "c" of SKN-1 (UniProt number: P34707) and inserted into a vector suitable for bacterial expression, pHis-TEV-30a 36 . This vector also contains an N terminal 6xHis-MBP tag and a TEV cleavage site. The construct was transformed to BL21 Rosetta E. coli strain for protein expression. Bacterial cultures were grown at 37 °C until OD 600 = 0.8, then cooled down and induced with 1 mM IPTG at 20 °C for 4 h for DAF-16 expression and with 0.1 mM IPTG at 37 °C for 4 h for SKN-1 expression. The cells were pelleted by centrifugation (4500 rpm, 30 min, 4 °C) and resuspended in 50 ml lysis buffer (50 mM Tris-HCl, 0.5 M NaCl, 10 mM imidazole, pH 7.5) supplemented with 0.5 mM TCEP and cOmplete mini protease inhibitors cocktail tablets -ROCHE. Cells were lysed by sonication (Digital Sonifier, Branson) for 5X 20 s pulses at 50% amplitude with 20 s cooling period between pulses. Cell lysate was centrifuged (15,000 rpm, 45 min, 4 °C) to clear the sample from any insoluble material and supernatant was loaded onto a 2 ml Ni-NTA column (Qiagen) pre-equilibrated with lysis buffer. After sample binding, the column was washed with ten column volume of lysis buffer and with ten column volume of lysis buffer containing 30 mM imidazole. The protein was eluted from the column with elution buffer (50 mM Tris-HCl, 150 mM NaCl, 150 mM imidazole, 0.5 mM TCEP). DAF-16 was further purified by size exclusion chromatography with Superose 6 Increase 10/300 column (GE Healthcare). Purified and concentrated protein was aliquoted and stored at − 80 °C. in vitro SUMoylation assays. Conjugation assays contained 50 mM Tris-HCl, 0.5 mM TCEP, 5 mM MgCl 2 , 2 mM ATP, 5 μg SUMO, 0.5 μg SUMO-Alexa Fluor 680, 60 ng SAE1/SAE2 (SUMO E1), 5 or 20 ng UBC-9, 10, 50 or 100 ng GEI-17 and 5 μg DAF-16, SKN-1 or MBP. Reactions were incubated at 30 °C for 4 h and they were run parallel on 4-12% Bis-Tris (better visualization of free SUMO) and 3-8% Tris-Acetate gels (better www.nature.com/scientificreports/ visualization of SUMO modified proteins). Gels were analyzed by Coomassie staining and Amersham Typhoon Biomolecular Imager (GE Healthcare) 35 .
Mitochondrial imaging. For TMRE (tetramethylrhodamine, ethyl ester) (Sigma) staining, age-matched animals were placed overnight on an RNAi plate containing 150 nM TMRE and the next morning placed in a 10 mM levamisole drop on a microscope slide, sealed with a cover slip. Images were taken with a Zeiss Axio-Imager Z2 epifluorescence microscope. For mitochondrial ROS staining, synchronized animals were placed on RNAi plates containing 150 nM MitoTracker Red CMXROS (Thermo Fisher Scientific) overnight and the next morning were mounted on microscope slides in a 10 mM levamisole drop, sealed with a cover slip to assess mitochondrial ROS production. Images were taken with a Zeiss AxioImager Z2 epifluorescence microscope. Images were quantified using the software Image J. For paraquat treated TMRE staining, age-matched animals were placed on RNAi plates containing 4 mM paraquat for 1 day, and the next day were transferred to a fresh RNAi plate containing 4 mM paraquat and 150 nM TMRE. The next morning the animals were placed on microscope slides in a 10 mM levamisole drop, sealed with a cover slip. Images were taken with a Zeiss AxioImager Z2 epifluorescence microscope. Images were quantified using the software Image J. For monitoring mitophagy we used the previously described mitochondrial targeted Rosella biosensor in body wall muscle cells of the animals 10 , with a minor modification. We created a new transgenic line with biolistic transformation and fed these animals with the test RNAi constructs. Animals were mounted on 5% agarose pads in a 10 mM levamisole drop and sealed with a cover slip. Images were taken with an Invitrogen EVOS FL Auto 2 Cell Imaging System. Images were quantified using the software Image J.
Messenger RNA quantification. Total RNA was extracted from worms, using Trizol (Sigma). cDNA was synthesized using the iScript kit (BioRad). Quantitative Real Time PCR was performed using a Bio-Rad CFX96 Real-Time PCR system, and was repeated three times. The following primer pairs were used to quan- Atp measurements. To quantify intracellular ATP levels, we followed the protocol described here 37 . In short, 100 age matched animals were collected in 50 µl of M9 buffer and frozen at − 80 °C. Frozen worms were boiled at 95 °C for 15 min. After a 10 min centrifugation step at 14,000 rpm at 4 °C, the supernatant was transferred to a fresh tube and diluted tenfold before measurement. ATP content was determined by using the Roche ATP bioluminescent assay kit HSII (Roche Applied Science) and a TD-20/20 luminometer (TurnerDesigns). ATP levels were normalized to total protein content.
oxygen consumption rate measurement. To determine oxygen consumption rates, we followed a previously described protocol 38 . In short, 4-day-old adult animals were collected in 1 ml M9, and transferred to the chamber of Oxygraph (Hansatech Instruments). Measurements were performed for 15 min at 20 °C. The oxygen consumption rate was obtained by the slope of the straight portion of the plot. The animals were recovered after measurement and were subjected to sonication and protein determination. The oxygen consumption rates were normalized to total protein content.
Mitochondrial DNA quantification. mtDNA was quantified using quantitative real time PCR as described previously 39 . 50 worms were collected per condition, lysed and diluted tenfold before performing the PCR reaction. The following primer set was used for mtDNA (mito-1): 5′-GTT TAT GCT GCT GTA GCG TG-3′ and 5′-CTG TTA AAG CAA GTG GAC GAG-3′. The results were normalized to genomic DNA using the following primers specific for ama-1: 5′-TGG AAC TCT GGA GTC ACA CC-3′ and 5′-CAT CCT CCT TCA TTG AAC GG-3′. Quantitative PCR was performed using a Bio-Rad CFX96 Real-Time PCR system, and was repeated three times.
Lifespan analysis. Experiments were carried out at 20 °C, unless noted otherwise. Animals were synchronized by placing 7-8 gravid adults on control RNAi plates for overnight egg laying and they were removed the next morning. L4 larvae were placed on experimental plates (containing 2 mM IPTG and seeded with HT115 (DE3) bacteria comprising the control vector (pL4440) or the test RNAi construct) and they were transferred to a fresh plate every 2-3 days. Animals were scored for survival with movement provoking touch every second day. Those who crawled out of the plate or died due to internal egg-hatching were considered censored and incorporated into the dataset as such. Each lifespan assay was repeated at least two times and figures represent typical assays. Lifespan assays on NAC (N-acetyl cysteine) plates (10 mM final concentration) were performed only one time. Statistical analysis was performed using the Prism software package (version 7; GraphPad Software; https ://www.graph pad.com), and the product-limit method of Kaplan and Meier.
Survival assays. For oxidative stress assays we grew synchronized animals until day 6 of adulthood on control or smo-1(RNAi) plates. At day 6 animals were transferred to control or smo-1(RNAi) plates containing paraquat (methyl viologen dichloride hydrate, Sigma) in 2 mM final concentration. Additionally, bacteria were killed with UV before adding paraquat on the plates in order to prevent interference arising from bacterial metabolism. For acute heat stress survival the animals were grown until day 4 of adulthood on control or smo-1(RNAi) plates at 20 °C. At day 4 of adulthood, animals were subjected to 37 °C for 2 h and then placed back to 20 °C. In Scientific RepoRtS | (2020) 10:15513 | https://doi.org/10.1038/s41598-020-72637-9 www.nature.com/scientificreports/ both stress conditions animals were scored for survival with movement provoking touch every day. Those who crawled out of the plate or died due to internal egg-hatching were considered censored and incorporated into the dataset as such. Each stress survival assay was repeated at least two times and figures represent typical assays. Statistical analysis was performed using the Prism software package (version 7; GraphPad Software; https ://www. graph pad.com), and the product-limit method of Kaplan and Meier.

Statistical analysis.
Statistical analyses were carried out using the Prism software package (version 7; GraphPad Software; https ://www.graph pad.com) and the Microsoft Office 2010 Excel software package (Microsoft Corporation, Redmond, WA, USA). Mean values were compared using unpaired t-tests or one-way ANOVA.
Previous studies have established the prominent contribution of specific tissues in the control of the ageing process 43 . To examine whether SUMO promotes longevity in a tissue specific manner, we targeted smo-1 expression in different tissues by RNAi. We utilized transgenic strains carrying a mutation in rde-1 gene (RNAi defective), which encodes an Argonaute protein family member that is required for RNA interference 44 . A wild type version of rde-1 is then re-introduced to the animals in a tissue specific manner, restoring RNAi capacity in that tissue. Using this approach, we assessed the effects of SUMO downregulation in the intestine (VP303 strain), hypodermis (NR222 strain) and muscles (NR350 strain). Only intestine specific smo-1(RNAi) resulted in shorter lifespan, while knockdown of smo-1 in hypodermal and muscle tissues had no effect (Fig. 2C, Fig.S1E,F). We also investigated the consequences of neuron-specific SUMO deficiency. The nervous system of C. elegans is mostly resistant to RNAi, but specific neurons (amphids and phasmids) are susceptible 45 . To probe neuron specific SUMO depletion, we used a strain (TU3401), which carries a mutation in sid-1 (systemic RNA interference www.nature.com/scientificreports/ defective), a gene that encodes a transmembrane channel for dsRNA, required for systemic RNA interference 46 .
To render neurons susceptible to RNAi, a sid-1 transgene is introduced, driven by a pan-neuronal promoter, unc-119 47 . Similarly to the intestine, reduction of smo-1 expression, specifically in neurons, resulted in shorter lifespan (Fig. 2D). Importantly, overexpression of smo-1 specifically in the intestine (p vha-6 smo-1), led to an extended lifespan (Fig. 2E). On the contrary, neuron-specific smo-1 overexpression (p rab-3 smo-1) did not have any significant effect on the lifespan of the animals (Fig. 2F). This result suggests the importance of balanced SUMO levels in the nervous system. Taken together, these findings indicate that SUMO influences C. elegans lifespan through the intestine and the nervous system.

SUMO regulates the transcriptional activity of SKN-1 and DAF-16.
To gain insight relevant to the role of SUMOylation in the regulation of ageing, we examined the effects of SUMO depletion in the context of well-characterized signalling pathways that modulate lifespan in C. elegans 5 . HSF-1, SKN-1 and DAF-16 are key stress response transcription factors, promoting longevity under conditions of heat shock, mild oxidative stress and low insulin/IGF1 signalling 8 . Considering, that HSF-1 has been shown to have the potential to be modified by SUMO, both in mammals 48 and in C. elegans 49 , we focused our studies on SKN-1 and DAF-16. We find that skn-1(RNAi);smo-1(RNAi) animals display shorter lifespan, similar to animals subjected only to skn-1 RNAi (Fig. 3A). Importantly, SKN-1 depletion annulled lifespan extension caused by SMO-1 overexpression (Fig. 3B). Thus, SKN-1 mediates the pro-longevity effects of SUMO. Given that the mammalian orthologue of SKN-1,  Table S1. www.nature.com/scientificreports/ NRF2 is a target for SUMOylation 50,51 , we examined whether SKN-1 is also modified by SUMO conjugation. Detection of SUMO modification in vivo poses multiple challenges (low expression of SKN-1 under normal conditions, low SUMOylation rates of total protein content), and we were unable to observe it in our Western blots (data not shown). To this end, we performed in vitro SUMOylation assays with bacterially expressed 6xHis-MBP tagged and purified SKN-1. We found that SKN-1 is not an optimal SUMO target in vitro (Fig. S2A).
To further investigate the role of SKN-1 in mediating the effects of SUMO on lifespan, we subjected animals to oxidative stress which activates SKN-1 52 . We exposed 6 day old wild type, FGP14 and smo-1(RNAi) treated animals to 2 mM paraquat, an inducer of oxidative stress. As expected, skn-1(RNAi) treated animals were sensitive to oxidative stress (Fig. 3C). Notably, smo-1 overexpressing animals were resistant to oxidative stress, while reduction in smo-1 expression did not alter survival on paraquat (Fig. 3C). We also assayed heat stress resistance upon perturbation of SUMO levels. Similarly to oxidative stress, smo-1 overexpression increased survival of day 4 animals to acute (2 h) heat shock at 37 °C, whereas, SUMO depletion did not affect heat stress resistance (Fig. 3D). Next, we monitored the expression of gst-4, a SKN-1 target gene encoding glutathione S-transferase, levels in day 2 animals on 2 mM paraquat. Using a p gst-4 GFP reporter fusion, we found that knockdown of smo-1 reduced the expression of gst-4, while, overexpression of smo-1 further increased gst-4 expression (Fig. 3E). Therefore, under stress conditions, increase of SUMO levels enhances stress resistance by potentiating expression of stress response genes, while depletion of SUMO compromises survival under stress by reducing the capacity to fully mount a response.
Interestingly, under normal conditions, smo-1 knockdown transiently increased gst-4 expression during adulthood in day 4 animals, while there was no observable difference by day 8 (Fig. S2B,C). By contrast, gst-4 expression was not altered upon smo-1 overexpression during adulthood (day 4 or day 8 animals; Fig. S2B,C). The mRNA levels of gst-4 mirrored the changes observed with the p gst-4 GFP reporter fusion (Fig. S2D). We find that SKN-1 becomes activated in the absence of the FOXO transcription factor DAF-16 that mediates insulin/ IGF1 signalling, in day 4 of adulthood, both in wild type and smo-1 overexpressing animals (Fig. S2B). SKN-1 activation is diminished by day 8 (Fig. S2C). Hence, although SKN-1 is likely not modified by SUMO in vivo, both are required for normal lifespan. Reduction of SUMO levels triggers activation of SKN-1, while SUMO abundance protects against stress and promotes longevity.
In addition to SKN-1, DAF-16 also promotes stress tolerance and longevity 6 . We asked whether SUMO exerts its effects on animal ageing by modulating DAF-16 activity. DAF-16 is activated in long-lived C. elegans mutants,  Table S1. www.nature.com/scientificreports/ carrying lesions in the insulin/IGF1 receptor DAF-2. We find that smo-1 knockdown shortened the lifespan of DAF-2 deficient animals (Fig. 4A,B). Furthermore, downregulation of daf-16 abolished lifespan extension caused by smo-1 overexpression (Fig. 4C). Notably, under lifespan-extending conditions, where DAF-16 is activated (in daf-2(e1370) mutants), or protein synthesis is reduced (in ife-2(ok306) mutants), downregulation of the SUMO protease ulp-1 extends lifespan (Fig. 4B,D). These findings indicate that DAF-16 is, in part, mediating the effects of SUMO on lifespan. Thus, we investigated whether DAF-16 is a target for SUMOylation. We conducted in vitro SUMOylation assays with purified, bacterially expressed and 6xHis-MBP tagged DAF-16. We detected SUMO conjugation to DAF-16 ( Fig. 4E top panel, Fig. S3A). To increase the sensitivity of our assay, we added fluorescently labelled SUMO (SUMO-AlexaFluor 680) 35 , which facilitated detection of the SUMO-modified form of DAF-16 (Fig. 4E, bottom panel). We also tested the capacity of the MBP tag to undergo SUMOylation. We found that MBP is not a SUMO target (Fig. S3B). Therefore, DAF-16 is a bona fide SUMOylation target.
Since SUMOylation has been implicated in chromatin remodelling 55 , we considered whether smo-1 overexpression decreased transcription of DAF-16 target genes by inducing chromatin condensation. We found that knockdown of daf-2 increased sod-3 expression in smo-1 overexpressing animals, albeit to a lesser extent, compared to wild type controls (Fig. 3H, Fig. S3E). Taken together, these findings indicate that SUMO represses the transcriptional activity of DAF-16.

SUMo facilitates mitochondrial homeostasis. Mitochondrial metabolism is an important determi-
nant of ageing across diverse organisms. We found that SUMOylation levels increase not just in whole worm lysates (Fig. 1A), but also in the mitochondrial fraction (Fig. S4A). To examine whether SUMO impacts lifespan by altering mitochondrial function, we utilized the dye, TMRE (tetramethylrhodamine, ethyl ester), which stains mitochondria according to their membrane potential. Wild type animals display reduced mitochondrial membrane potential during ageing 1 . Nevertheless, knockdown of smo-1 alleviated age-associated mitochondrial membrane potential decline (Fig. 5A,B). Moreover, animals overexpressing smo-1 displayed increased TMRE staining early in adulthood (Fig. 5A,B).
Previous studies have shown that mitochondrial fission is required for, and precedes mitophagy 64 . Since SUMO impinges on mitochondrial dynamics, we tested whether mitophagy is also affected upon downregulation of smo-1 expression. To monitor mitophagy, we used a mitochondria targeted, ratiometric Rosella biosensor, expressed in body wall muscle cells 10 . Under mitophagy-inducing, low insulin/IGF1 signalling conditions, in daf-2 mutants, the ratio of GFP/DsRed fluorescence is reduced, signifying mitophagy induction (Fig. 6C, S6A). By contrast, knockdown of smo-1 increased the GFP/DsRed ratio, indicating that SUMO depletion inhibits www.nature.com/scientificreports/ mitophagy (Fig. 6C, S6A). This observation is consistent with the block of mitochondrial fission upon smo-1 downregulation (Fig. 6B). Additionally, we also assayed mitophagy in the nervous system 65 , in the RNAi sensitive, rrf-3(pk1426) mutant background. Surprisingly, reduced smo-1 expression resulted in the induction of mitophagy in neurons (Fig. 6D, S6B). Thus, SUMO is required to perform mitochondrial fission and mitophagy in a tissue specific manner.

Discussion
The regulatory role of SUMO in the ageing process remains elusive, up to date. Earlier studies in mice uncovered a progressive increase of protein SUMOylation in brain and plasma, during ageing 29 . Notably, we find a similar increase in the amount of SUMO, during adulthood in C. elegans, both in whole worm lysates and in the mitochondrial fraction (Fig. 1A, S4A). Interestingly, reducing the expression of SUMO proteases did not have any lifespan altering effect under normal conditions (Fig. S1B,C). Presumably, the function of these proteases is redundant. Furthermore, under various stress conditions mild heat stress (25 °C), reduced translational rates (ife-2(-) mutant background) and low insulin signalling (daf-2(-) mutant background) the loss of ulp-1 resulted in an extended lifespan (Figs. S1D, 4B,D). Therefore, specific SUMO proteases have the potential to regulate longevity upon stress conditions. To define the exact role of each SUMO protease under distinct stress insults is an interesting topic for future investigations. SMO-1 is expressed in all animal tissues and mainly localizes to the nucleus 66 . Accordingly, the most studied functions of SUMO are nucleus-associated 14,67,68 ; however, its role outside of the nucleus is emerging 69 , with a focus on the nervous system. Indeed, we find that tightly controlled expression level of smo-1 is a prerequisite for normal lifespan (Fig. 2). A recent study implicated SUMOylation of the germline RNA binding protein CAR-1 (cytokinesis, apoptosis, RNA-associated), in the regulation of lifespan by insulin signalling. Under low insulin signalling conditions (in daf-2 mutants), CAR-1 is less likely to be SUMO-modified, and can effectively inhibit glp-1 expression in the germline, allowing for lifespan extension 28 .
Our findings indicate that DAF-16 and SKN-1 mediate the effects of SUMO on ageing, in the soma. SUMOylation of transcription factors is mostly coupled with transcriptional repression 70 . Indeed, we found that DAF-16 is a target for SUMOylation, upon which, its transcriptional activity is quenched (Fig. 4). Under basal conditions this modification could serve to inhibit the unnecessary activation of stress response genes by nuclear localized DAF-16. Indeed, SUMO overexpression leads to the nuclear accumulation of DAF-16 (Fig. 4F); however, without the activation of DAF-16 target genes (Fig. 4G-J). Consequently, the long lifespan of SUMO overexpressing animals is only marginally shortened upon daf-16 knockdown (Fig. 4C). On the other hand, reduced skn-1 expression in the SUMO overexpressing background completely abolished lifespan extension in SUMO overexpressing animals (Fig. 3B). These results indicate that the interplay between DAF-16 and SKN-1 is an important determinant of the ageing mechanism. Notably, numerous pathways converge on organismal lifespan 71 ; further studies could determine the DAF-16 independent, lifespan extending mechanisms upon SMO-1 overexpression in C. elegans. Removal of SUMO results in the upregulation of stress responsive genes, controlled by DAF-16 and SKN-1 (Figs. 4G-J, S2B,D). These findings are consistent with the reported functions of SUMOylation in the regulation of stress responses (e.g. DNA damage response, ER stress, heat shock), which are indispensable for cellular survival 72 . Furthermore, a recent study indicates that activation of SKN-1 negatively regulates DAF-16, which is in agreement with our data showing the increased expression of DAF-16 target genes upon skn-1(RNAi) 73 (Fig. 4G,H). It would be interesting to analyse the role of SUMO in this interaction. Our results hint towards the possibility of the requirement of SUMO modification of DAF-16 for the successful SKN-1 repression. Thus, SUMO depletion compromises cellular homeostasis, and may, additionally, generate a state of cellular stress that contributes to early organismal death (Fig. 7).
SUMOylation has also been linked to mRNA translation regulation 74 , and chromatin organization 55 . Notably, hyposumoylation of chromatin has been associated with pluripotency reprogramming and enhanced lineage trans-differentiation 75 . We considered whether the changes in transgene expression upon SUMO depletion, or overexpression, in C. elegans are an indirect consequence of alterations in chromatin structure, and/or mRNA translation. We observed that downregulation of daf-2 is sufficient to restore expression of sod-3, a DAF-16 target gene, which is significantly reduced upon SUMO overexpression (Fig. 4H), indicating that the chromatin is still accessible under this condition. Furthermore, expression of gst-4, a SKN-1 target gene, is not affected by SUMO overexpression (Fig. S2B-D) and increases following paraquat treatment (Fig. 3E). By contrast, SUMO deficiency results in the upregulation of both sod-3, and gst-4 expression (Figs. 4H, S2B). Therefore, loss of SUMO engages stress response pathways, involving both DAF-16 and SKN-1 transcription factors (Fig. 7).
SUMOylation has been linked to mitochondrial biogenesis and network remodelling through PGC-1α and Drp1 in mammals 63,76,77 . Moreover, SUMO modification of ATFS-1 and DVE-1 play a regulatory role in the process of mitochondrial unfolded protein response ( mt UPR) 78 . Here, we examined the involvement of SUMO in mitochondrial homeostasis. We observed an increase in mitochondrial content, but not in mitochondrial DNA copy number upon depletion of SUMO in C. elegans (Fig. S5C-F). Additionally, ATP and ROS production increases in aged animals, indicating elevated mitochondrial activity (Fig. 5C-F). Moreover, we found that SUMO promotes mitochondrial fragmentation during ageing (Fig. 6A,B) and it is also required for mitochondrial turnover via mitophagy in muscle cells (Fig. 6C). Intriguingly, SUMO blocks the process of mitophagy in the nervous system (Fig. 6D). Admittedly, increased mitophagy has been shown to have detrimental consequences on the homeostasis of the cell 79,80 . The tissue specific effects of SUMO on mitophagy merits further research. Notably, while in long-lived mutants, mitochondria display elongated morphology, coupled with reduced ROS production 58 , we find that short-lived, SUMO-depleted animals also feature elongated mitochondrial network, but elevated generation of ROS. This discrepancy indicates that mitochondrial turnover is differentially affected in long-lived, and short-lived, SUMO deficient mutants. In long-lived animals a healthy interconnected mitochondrial network is maintained by increased mitophagy, which moderates ROS production 10  www.nature.com/scientificreports/ knockdown interferes with mitochondrial fission and fusion, which in turn impairs mitophagy, resulting in accumulation of mitochondrial damage, higher ROS levels, and shorter lifespan. Our observations suggest that SUMO decreases mitochondrial function and promotes mitochondrial fission during ageing in C. elegans (Fig. 7). These findings are consistent with recent studies suggesting that impairment of mitochondrial dynamics contributes to the decline of mitochondrial function during ageing and the onset of age-related diseases 81 . Collectively, our findings indicate that SUMO influences the ageing process by modulating the transcriptional activity of the stress response transcription factors, DAF-16 and SKN-1 (Fig. 7). Admittedly, these are not the only transcription factors that could be affected by the depletion of SUMO. Further research is required to fully map the molecular changes in response to altered SUMO levels. Notably, recent studies have implicated SUMO in senescent decline. SUMOylation of p53 causes cellular senescence 82 , while deSUMOylation of Bmi1, a polycomb repressive complex member, is likewise, required for senescence 83 . Moreover, the SUMO E2 enzyme, Ubc9, regulates senescence by relocation of SUMOylated proteins to PML nuclear bodies 84 . Thus balanced protein SUMOylation is critical for stress resistance and survival. The ageing process disrupts SUMOylation balance while manipulations that fine-tune protein SUMOylation promote longevity.
Here, we demonstrate that the abundance of SUMO regulates lifespan. Reduced SUMO levels shorten lifespan, while increased smo-1 expression results in extended lifespan (Figs. 2, 7). SUMO influences ageing mainly through DAF-16 and SKN-1 (Figs. 3A,B, 4A,B), in a tissue specific manner, through the intestine and the nervous system (Fig. 2). We further show that DAF-16 is a target for SUMOylation, and that SUMO attachment represses the transcriptional activity of DAF-16 (Figs. 4, 7). We propose that under normal conditions, modification of DAF-16 by SUMO could prevent uncontrolled activation of nuclearly localized DAF-16 (Fig. 7). Accordingly, overabundance of SUMO leads to strong DAF-16 inhibition (Fig. 4H-J). Nonetheless, SUMO overexpressing animals exhibit long lifespan, which is dependent on SKN-1 (Fig. 3B) and in a lesser extent on DAF-16 (Fig. 4C). In addition, SUMO plays a critical role in the maintenance of mitochondrial homeostasis. Altering SUMO levels affects mitochondrial ATP and ROS production (Fig. 5), as well as, mitochondrial dynamics and clearance (Figs. 6, 7). Combined, our findings indicate that balanced protein SUMOylation is a prerequisite for healthy animal ageing.