The SUMO protease SENP3 regulates mitochondrial autophagy mediated by Fis1

Abstract Mitochondria are unavoidably subject to organellar stress resulting from exposure to a range of reactive molecular species. Consequently, cells operate a poorly understood quality control programme of mitophagy to facilitate elimination of dysfunctional mitochondria. Here, we used a model stressor, deferiprone (DFP), to investigate the molecular basis for stress‐induced mitophagy. We show that mitochondrial fission 1 protein (Fis1) is required for DFP‐induced mitophagy and that Fis1 is SUMOylated at K149, an amino acid residue critical for Fis1 mitochondrial localization. We find that DFP treatment leads to the stabilization of the SUMO protease SENP3, which is mediated by downregulation of the E3 ubiquitin (Ub) ligase CHIP. SENP3 is responsible for Fis1 deSUMOylation and depletion of SENP3 abolishes DFP‐induced mitophagy. Furthermore, preventing Fis1 SUMOylation by conservative K149R mutation enhances Fis1 mitochondrial localization. Critically, expressing a Fis1 K149R mutant restores DFP‐induced mitophagy in SENP3‐depleted cells. Thus, we propose a model in which SENP3‐mediated deSUMOylation facilitates Fis1 mitochondrial localization to underpin stress‐induced mitophagy.


Introduction
Accurate and proper degradation of dysfunctional mitochondria by mitophagy is essential for maintaining control over mitochondrial quality and quantity. However, mitophagic mechanisms, and their control, remain unclear (Green, Galluzzi et al., 2011). Degradation of dysfunctional mitochondria occurs via a poorly understood quality control program involving mitochondrial fission of dysfunctional components (Twig, Elorza et al., 2008), their accumulation in autophagosomes (Jimenez-Orgaz, Kvainickas et al., 2018), and subsequent lysosomal fusion resulting in degradation (Sun, Yun et al., 2015).
SUMO proteases deconjugate SUMO from SUMOylated proteins in a process termed deSUMOylation (Hickey, Wilson et al., 2012. The extent of target protein SUMOylation in any given state arises from the balance of the opposing activities of SUMO conjugating and deconjugating enzymes. To date, three families of SUMO proteases have been identified: deSUMOylating isopeptidase 1 and 2 (DeSI1 and DeSI2), Ubiquitin-specific protease-like 1 (USPL1), and Sentrinspecific proteases (SENPs). The largest and best characterised family of SUMO proteases, SENPs consist of six cysteine proteases (SENP1-3 and 5-7), each having a distinct subcellular localisation and SUMO isoform substrate preference (Guo & Henley, 2014, Hickey et al., 2012. SENP3 is an essential deSUMOylating enzyme primarily responsible for deconjugating SUMO-2/3 from modified proteins, as evidenced by the findings that SENP3 depletion results in a significant increase in global SUMO-2/3-ylation in mammalian cells (Fanis, Gillemans et al., 2012, Guo, Hildick et al., 2013, Haindl, Harasim et al., 2008, Luo, Gurung et al., 2017. Although SENP3 is known to play key roles in many processes such as inflammation (Huang, Ghisletti et al., 2011), stem cell differentiation (Nayak, Viale-Bouroncle et al., 2014), ribosome biogenesis (Raman, Weir et al., 2016) and cell stress responses (Guo et al., 2013, Huang, Han et al., 2009, Yan, Sun et al., 2010, specific targets and physiological roles for this enzyme are largely unknown. Previously believed to be an exclusively nuclear protein, our recent work has revealed that, importantly, SENP3 also resides in the cytoplasm, and is required for mitochondrial fission (Guo et al., 2013, a process which ensures mitophagy-mediated removal of damaged or dysfunctional mitochondria (Twig et al., 2008). Moreover, emerging evidence has strongly implicated SENP3 in ageing and age-related degenerative diseases. For example, SENP3 gene expression is markedly reduced in brain samples either from healthy aged subjects (Lu, Pan et al., 2004) or from Alzheimer's disease patients (Weeraratna, Kalehua et al., 2007), and SENP3 levels are altered in brain samples from the patients of Down's syndrome (Binda, Heimann et al., 2017) which is associated with premature aging and is a risk factor for dementia development (Roth, Sun et al., 1996, Weksler, Szabo et al., 2013. However, the functional consequences and molecular mechanism(s) underlying the changes in SENP3 levels remain undetermined. Furthermore, accumulation of dysfunctional mitochondria is linked with aging and age-related degenerative diseases, and this is believed to be 5 due to defects in a mitochondrial quality control program enforced by mitophagy, suggesting the possible involvement of SENP3 in mitochondrial quality control.
Here, we have investigated the role of SENP3 in mitophagy-mediated mitochondrial disposal in model (HeLa) cells by investigating the effects of altered SENP3 levels on induced mitophagy and autophagosome formation. Cells were exposed to a model stressor, the iron chelator DFP, which specifically induces mitophagy, but not general autophagy (Allen, Toth et al., 2013, Sargsyan, Cai et al., 2015, Yamashita et al., 2016. DFP is used for treatment of iron overload in inherited and degenerative diseases such as Thalassemia major and Parkinson's disease. We show that levels of the SUMO-2/3-specific deSUMOylating enzyme SENP3 are greatly stabilised upon DFP treatment via a pathway that involves the reduction in levels of the E3 ubiquitin (Ub)-protein ligase CHIP. This stabilisation of SENP3 decreases Fis1 SUMO-2/3-ylation, enhances Fis1 mitochondrial localization and induces Fis1mediated autophagosome formation and mitochondrial disposal. Taken together, our results reveal a novel mechanism dependent on the presence of SENP3 in coordinated regulation of the SUMOylation status of Fis1 to ensure mitochondrial quality control.
Consistent with this, we detected induction of the autophagosomal marker LC3-II by DFP in HeLa cells that express little or no Parkin (Supplementary Figure 1A). To detect lysosomal-mediated acidification of engulfed mitochondria, we designed and characterised a novel probe termed MitoPHfluorin (Supplementary Figure 1B). This probe consists of mCherry-SEP (a pH-sensitive GFP-variant (Ashby, Ibaraki et al., 2004) tagged to the mitochondrial targeting sequence of the ActA protein of Listeria monocytogenes (Pistor, Chakraborty et al., 1994), resulting in a quenchable dual- Interestingly, DFP treatment also led to a significant decrease in global SUMO-2/3ylation ( Figure 1A) but not SUMO-1-lyation ( Figure 1B) in HeLa cells. Moreover, levels of the deSUMOylating enzyme SENP3 ( Figure 1C), but not SENP5 ( Figure   1D), were increased by DFP. Interestingly DFP treatment decreases the levels of CHIP ( Figure 1E), an E3 Ub ligase known to control SENP3 levels under basal conditions (Yan et al., 2010). Consistent with previous findings, knocking down CHIP in HeLa cells increased the levels of SENP3 but not SENP5 ( Figure 1F), and prevented the DFP-induced increase in SENP3 levels ( Figure 1G). Taken together, these findings suggest that the downregulation of CHIP upon DFP treatment underlies increased SENP3 levels, thereby modulating levels of cellular SUMO-2/3 conjugation.

SENP3 is required for mitophagy induced by DFP
Mitophagy-mediated disposal of dysfunctional mitochondria occurs via a multi-step process involving two intracellular events: mitochondrial autophagosome formation and subsequent autolysosome formation (Yamashita & Kanki, 2017). To explore if SENP3 is involved in DFP-induced mitochondrial autophagosome formation, we examined the effect of RNAi-mediated knockdown of SENP3 on the expression of 7 the autophagosomal marker LC3-II in HeLa cells treated with DFP. SENP3 knockdown significantly reduced DFP-induced LC3-II expression (Figure 2A), indicating an essential role for SENP3 in autophagosome formation. We then compared mitochondria-containing autolysosome formation in control and SENP3knockdown cells using MitoPHfluorin. SENP3 knockdown significantly reduced the levels of SEP fluorescence quenching induced by DFP, indicating a critical role for SENP3 in the disposal of dysfunctional mitochondria ( Figure 2B). Taken together, these results suggest that a protein deSUMOylation pathway mediated by SENP3 promotes DFP-induced mitophagy.

Fis1 is critical for DFP-induced mitophagy
Since SENP3 regulates mitophagy, we next wondered as to the identity of the specific substrate(s) of its deSUMOylating activity that may mediate this effect. We  Guo et al., 2013, Prudent, Zunino et al., 2015. Drp1 is recruited to the mitochondrial surface and acts by interacting with specific mitochondrial docking proteins such as Mff, MID49, MID51 and possibly Fis1 in mammalian cells (Loson, Song et al., 2013). However Drp1 does not appear to be essential for mitophagy (Yamashita et al., 2016, Yamashita & Kanki, 2017. Meanwhile, evidence has emerged for the critical role of Fis1 in the mitophagy-mediated disposal of defective mitochondria induced by stressors (Rojansky, Cha et al., 2016, Shen, Yamano et al., 2014, Yamano, Fogel et al., 2014. To investigate if Fis1 is involved in DFP-induced mitochondrial autophagosome formation, we first examined the effect of RNAi-  Figure 3B). Taken together, these results indicate that Fis1 is required for DFP-induced mitophagy.

Fis1 is a novel SUMO target and is deSUMOylated by SENP3
Since Fis1 had been identified as a potential target for SUMOylation in a proteomics screen for SUMO targets in vivo (Tirard, Hsiao et al., 2012), we next verified whether Fis1 was in fact a bona fide target for SUMOylation. Flag-Fis1 is efficiently conjugated by His-SUMO-2 in HEK293 cells ( Figure 4A). Subsequently under denaturing conditions, we detected conjugation of His-SUMO-2 to endogenous Fis1 in HEK293 cells ( Figure 4B). Furthermore, Fis1 antibody-reactive band(s) were detected in immunoprecipitations of endogenous SUMO-2/3 conjugates in HeLa cells ( Figure 4C), indicating the existence of SUMOylated endogenously Fis1. Importantly, SENP3 knockdown led to a significant increase in Fis1 SUMO-2-lyation, suggesting SENP3 directly deSUMOylates Fis1 ( Figure 4D). Taken together, these findings indicate that Fis1 is subject to SUMO-2/3-lyation under basal conditions, which can be reversed through the action of the deSUMOylating enzyme, SENP3.

Fis1 is primarily SUMOylated at lysine 149
To identify which lysine in Fis1 is SUMOylated, we performed site-directed mutagenesis. Fis1 contains one high-probability consensus lysine residue (K119) and three high-probability non-consensus lysine resides (K67, K149 and K151), predicted by SUMOsp (Xue, Zhou et al., 2006). While mutation of either the K119 or the K67 to a non-SUMOylatable arginine (R) did not reduce Fis1 SUMOylation in a

The SUMOylation status of Fis1 acts as a key switch to regulate its mitochondrial localization and mitophagy induction
Interestingly, K149 is known to be critical for localizing Fis1 to mitochondria. Nonconservative mutation of K149 to alanine (A) greatly reduces Fis1 mitochondrial localization (Delille & Schrader, 2008, Stojanovski, Koutsopoulos et al., 2004 whilst Fis1 with the conservative K149R mutation is localized to mitochondria (Alirol, James et al., 2006). However, whether the K149R mutation results in any changes in the  Figure 3), on LC3-II induction. Fis1 mutant-expressing cells show elevated levels of LC3-II, in the absence of DFP ( Figure 5A). Taken together, these findings suggest that the SUMOylation status of Fis1 serves as a key molecular switch to regulate its mitochondrial localization and mitophagic autophagosome formation.

Discussion
Healthy mitochondria are vital to generate energy for eukaryotic cells and accumulation of dysfunctional mitochondria has been linked to ageing and agerelated diseases. This age-related pathological feature is believed to be due to defective mitophagy, which mediates mitochondrial quality control via lysosomal degradation. However, the mechanisms that regulate mitophagy and the key pathways involved in mitochondrial quality control remain largely unknown. Here we report a previously uncharacterized pathway that regulates mitophagy via modulation of the SUMOylation status of the mitochondrial protein Fis1 by the SUMO-2/3specific protease SENP3 (Figure 6). To the best of our knowledge, this is the first example showing functional crosstalk between protein SUMOylation and regulation of mitophagy.

Modulation of SENP3 stability by mitophagy induction
In mammals the Ub-proteasome system (UPS) controls SENP3 levels through proteasome-mediated degradation (Kuo, den Besten et al., 2008, Yan et al., 2010. It has been shown previously that the E3 Ub ligase CHIP mediates UPS-dependent degradation of SENP3 (Yan et al., 2010). In this present study, we found that upon DFP-mediated iron chelation SENP3 is stabilised whilst CHIP levels are reduced, suggesting that CHIP reduction may lead to SENP3 stabilisation through a reduction in SENP3 degradation.
A key finding of this study is that SENP3 is pivotal for mitophagy induction by iron chelation. SENP3 stabilisation is temporally correlated with decreased levels of global SUMO-2/3 conjugation. Consistent with this, knockdown of SENP3 abolishes DFP-induced formation of both mitophagic autophagosomes and autolysosomes, further suggesting that sufficient levels of SENP3 are required for the occurrence of mitophagy upon iron chelation. We further identify Fis1 as a key target for SENP3mediated deSUMOylation in mitophagy regulation since expression of a SUMOylation-deficient mutant of Fis1 abolishes the effect of SENP3 knockdown on DFP-induced mitophagy.
In conclusion, we define a previously unidentified SUMO-dependent pathway in which UPS-dependent stabilisation of SENP3 upon iron chelation enhances mitochondrial targeting of Fis1 to facilitate mitophagy induction. This pathway represents a novel example of a target-specific function of SENP3 in a mitochondrial quality control programme and reveals a potential therapeutic target for restoring impaired or defective mitophagy in health and disease.

DFP treatment for mitophagy induction
Confluent HeLa cells grown in 6-or 12-well plates were treated with deferiprone (DFP; 379409, Sigma-Aldrich; working concentration, 1mM) for the time durations as indicated.

DNA and siRNA transfections
For DNA, siRNA, or DNA-siRNA co-transfection into HeLa or HEK293 cells, we used jetPRIME reagents (

Subcellular fractionation
Cytosol and mitochondria fractions were prepared from HeLa cells using a Cell Fractionation Kit (Abcam), according to the manufacturer's instructions. endogenous SUMO-2/3 conjugates from the lysate samples were immunoprecipated as described previously .

Immunoblotting
Lysate, IP or His-PD Samples were resolved by SDS-PAGE (10-15% gels) and transferred to Immobilon-P membranes (Millipore Inc.), which were then Each immunoblot presented is representative of at least three experiments carried out using different cell populations.

Fluorescence and microscopy imaging
HeLa cells were plated on 35mm glass bottom dishes and transfected after 24 hr.
After 48 hr, cells were treated with DFP (1 mM) or PBS for a further 24 hr. Cells were fixed for 12 min at RT in 4% paraformaldehyde/PBS. Following three 5 min-washes with PBS to remove residual PFA, cells were permeabilised in 0.3% Triton X-

15
For comparison of conjugates of SUMO-2/3 or SUMO-1, and levels of SENP3, SENP5, CHIP, or LC3-II in lysate or His-PD samples between time-matched two groups, paired Student's test with two-tail p-value was performed. For analysis of mitophagic autolysosomes, unpaired Student's t-tests with two-tail p-value was performed comparing mean of red punctum numbers. Following quantification using ImageJ software, levels of proteins of interest were normalized to β-actin or GAPDH control levels. Each value is presented as mean ± SEM and is expressed as percentage of control value. To avoid bias all experiments were performed blind.

Acknowledgments
Start-up fund (BMS314826 to CG), Royal Society Research Grant (157589-11-1 to CG) and a PhD studentship (BMS 315637 to EW) from the Department of Biomedical Science at the University of Sheffield supported this work. We are also grateful to BBSRC (BB/R00787X) and Parkinson's UK for financial support (to KAW and REC, respectively). We thank C Smythe for critical reading of the manuscript and great logistical support, and EH Hettema for generous technical help with initial imaging experiments. We thank JM Henley for supplying plasmids and RT Hay for supplying the HEK293 N3S cells.

Author contribution
CG and EW conceived the project with valuable input from KAW. EW and CG designed and performed all biochemical and molecular biological analysis, EW performed all cell imaging assays in cell culture, and KAW and REC conducted some cell culture and sample preparation work. CG provided project management and wrote the manuscript with hypothesis development, experimental design and data interpretation contributed by all authors. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflict of Interest
The authors declare no conflict of interest.