Disruption of SUMO-Specific Protease 2 Induces Mitochondria Mediated Neurodegeneration

Post-translational modification of proteins by small ubiquitin-related modifier (SUMO) is reversible and highly evolutionarily conserved from yeasts to humans. Unlike ubiquitination with a well-established role in protein degradation, sumoylation may alter protein function, activity, stability and subcellular localization. Members of SUMO-specific protease (SENP) family, capable of SUMO removal, are involved in the reversed conjugation process. Although SUMO-specific proteases are known to reverse sumoylation in many well-defined systems, their importance in mammalian development and pathogenesis remains largely elusive. In patients with neurodegenerative diseases, aberrant accumulation of SUMO-conjugated proteins has been widely described. Several aggregation-prone proteins modulated by SUMO have been implicated in neurodegeneration, but there is no evidence supporting a direct involvement of SUMO modification enzymes in human diseases. Here we show that mice with neural-specific disruption of SENP2 develop movement difficulties which ultimately results in paralysis. The disruption induces neurodegeneration where mitochondrial dynamics is dysregulated. SENP2 regulates Drp1 sumoylation and stability critical for mitochondrial morphogenesis in an isoform-specific manner. Although dispensable for development of neural cell types, this regulatory mechanism is necessary for their survival. Our findings provide a causal link of SUMO modification enzymes to apoptosis of neural cells, suggesting a new pathogenic mechanism for neurodegeneration. Exploring the protective effect of SENP2 on neuronal cell death may uncover important preventive and therapeutic strategies for neurodegenerative diseases.


Introduction
Emerging evidence suggests the importance of protein modification by Small Ubiquitin-related Modifier (SUMO) in neural development and function [1][2][3]. Abnormal SUMO modification has been found in several neurodegenerative diseases, characterized by progressive loss or dysfunction of neurons [4][5][6]. Unlike ubiquitin with a well-established role in protein degradation, SUMO is involved in protein trafficking, cell proliferation and survival, as well as ubiquitin-mediated proteolysis [7][8][9][10][11]. Covalent conjugation of SUMO to protein substrates, also known as sumoylation, is a reversible process catalyzed by SUMO ligases [12,13]. The removal of SUMO, also known as desumoylation, is mediated by SUMO proteases [14,15]. Although these proteins have been shown to reverse sumoylation in various physiological systems, their roles in mammalian development and disease remain largely unknown. SUMO-specific protease 2 (SENP2) is found in three alternatively spliced forms exhibiting differential subcellular localizations [16]. Genetic inactivation of Senp2 reveals its requirement in development of trophoblast stem cell niches and lineages during development of the placenta [17].
Although SENP2 mutants display embryonic defects including brain and heart abnormalities, they are likely associated with placental insufficiency which requires further investigation [17,18].
Enhanced sumoylation and accumulation of SUMO-conjugated proteins have been widely observed in patients with various neurodegenerative disorders [19][20][21][22]. Among the most notable ones are polyglutamine disorders, including Huntington's disease (HD) caused by a trinucleotide expansion, and neuronal intranuclear inclusion disease (NIID). The encoded CAG expansions result in production of toxic proteins carrying extended glutamine repeats. In HD, SUMO1 conjugation of the disease protein Huntingtin (Htt) contributes to the disease pathology possibly by stabilizing the toxic Htt [20]. SUMO-modified targets/substrates also accumulate in the nuclear aggregates of NIID, a multisystem neurodegenerative disease characterized by large intranuclear inclusions in neurons of the central and peripheral nervous systems [21]. In autosomal recessive juvenile parkinsonism, the SUMO pathway might affect protein degradation mediated by the disease protein Parkin, an E3-ubiquitin ligase [23]. Targeting the SUMO pathway may offer new strategies for disease prevention and therapy. However, there is no evidence indicating a direct involvement of SUMO modification regulators/enzymes in neurodegenerative disease. Information providing a causal link of SUMO dysregulation to neural cell survival is also very limited.

SENP2 deficiency causes neurodegeneration
We previously created a mouse strain carrying a null allele of SENP2 [17]. The knockout of SENP2 led to severe developmental abnormalities in trophoblast stem cell niches and lineages during placentation [17]. Although brain and heart deformities were also detected in the SENP2-null embryos ( Figure S1, Maruyama et al.,unpublished,and [18]), we speculated these are secondary defects due to placental insufficiencies [17]. To analyze the involvement of SENP2 and the importance of SUMO modification in neural development and disease, we first examined its expression pattern. In situ hybridization detected the presence of SENP2 mRNA in the developing mouse brain at embryonic day 14.5 (E14.5) and postnatal day 0 (P0), P7 and P14 ( Figure 1A). SENP2 was expressed in subventricular neural progenitors and differentiated cells of the cerebral cortex ( Figure 1A). To definitively assess our speculation on the contribution of placental deficiencies to the embryonic deformities, we took a genetic approach by creating a mouse model deficient for SENP2 during neural development. A new mouse strain carrying a SENP2 DSUMO Fx allele, permitting removal of the protease core domain using the Cre-loxP system, was generated ( Figure S2). The presence of Cre caused an inframe deletion, resulting in production of a SENP2 mutant deficient for the SUMO protease activity. Using EIIa-Cre to remove the protease core domain, we generated a mouse strain carrying SENP2 DSUMO mutant allele expressing the truncated SENP2 ( Figure S3). The SENP2 DSUMO D/D embryos were significantly smaller or underdeveloped compared to their SENP2 DSUMO +/+ and SENP2 DSUMO +/D littermates at E10.5 ( Figure S3A-B). Development of all three trophoblast layers was severely impaired in the homozygous mutants ( Figure S3C-J). These extraembryonic and embryonic defects are highly reminiscent to the SENP2 nulls [17], suggesting that the protease core domain deletion results in a loss of function mutation. We also were able to obtain mice heterozygous for the deleted allele without any noticeable abnormality, further suggesting that there is no dominant phenotype associated with the mutation.
Next, we generated a SENP2 DSUMO-Nes model, in which SENP2 is ablated in the neural progenitor cells by Nestin-Cre ( Figure S4). At newborn, no obvious defects associated with the deletion could be detected, including neuronal differentiation (Fu and Hsu, unpublished), indicating that SENP2 is not essential for embryonic neural development. The embryonic deformities observed in the SENP2 nulls were attributed to placental insufficiency. However, the SENP2 DSUMO-Nes mice displayed movement difficulties at P10. They developed paralysis around P16 ( Figure 1B and Supplementary Video S1; 100% penetrance, n = 20 SENP2 DSUMO-Nes mutants collected from 10 litters), and died at the age of 3 weeks. The size of the mutant brains was slightly smaller comparable to the control at P0, but later on exhibited a gradual reduction ( Figure 1B). At P14, the mutant brain looked transparent, and was much smaller than the control ( Figure 1B; *, p,0.05; **, p,0.01, n = 3). Histology revealed no obvious defects at P0 but severe brain abnormalities at P7 and P14 associated with the SENP2 deficiency ( Figure 1C). The cerebral cortex of SENP2 DSUMO-Nes became significantly thinner and malformed. Other CNS regions, e.g. midbrain, cerebellum, hippocampus and spinal cord were also affected by the mutation although the phenotypes were less severe ( Figure S5). The results suggested an essential role of SENP2 in neural development at postnatal, but not prenatal, stages.

SENP2 is essential for protection of mitochondriadependent apoptosis
The neurodegenerative phenotype of SENP2 DSUMO-Nes prompted us to examine programmed cell death affected by the mutation. Immunostaining of active Caspase 3 and TUNEL analysis revealed that abnormal apoptosis is, not detectable at P0, but increased at P4 (Caspase 3: 0.4660.12% in control vs. 1.5260.33% in mutant) and highly enhanced at P7 (Caspase 3: 0.8260.08% in control vs. 9.460.59% in mutant; TUNEL: 0.9160.17% in control vs. 12.0960.87% in mutant) (Figure 2A, p,0.01, ,700 cells were counted in each of 3 independent experiments, mean 6 SEM). The apoptotic abnormality, albeit less severe at this stage, was also observed in other CNS regions ( Figure S6). To further elucidate the mechanism underlying the neural cell death of SENP2 DSUMO-Nes , we examined expression of the activated form of Bak, a proapoptotic effector which promotes programmed cell death through modulation of mitochondrial morphogenesis [24,25]. In the SENP2 DSUMO-Nes cerebral cortices, Bak activation is stimulated at P0 (1.3960.41% in control vs. 3.0360.17% in mutant) and P7 (3.460.36% in control vs. 8.2160.59% in mutant), suggesting an association of mitochondrial dysfunction with the SENP2 mutation ( Figure 2B, p,0.01, ,700 cells were counted in each of 3 independent experiments, mean 6 SEM). Neurons derived from the cerebral cortices of mouse embryonic brains were then cultured in vitro for examination of mitochondrial dynamics. Fluorescent labeling of the mitochondria revealed a more than 2.5-fold increase of neurons containing fragmented, but not tubular/rod-like, mitochondria in the cell body and neurite caused by the mutation (20.864.4% in control vs. 55.367.8% in mutant) ( Figure 2C, p, 0.002, ,200 neurons were counted in each of 3 independent experiments, mean 6 SEM). Electron microscopy analysis of the P7 brain sections further identified fragmentation of the mitochondria in the cerebral cortical neurons of SENP2 DSUMO-Nes ( Figure 2D). The mitochondrial cisternae are generally intact although few of them show alterations on the inner membrane. The results thus suggested a protective effect of SENP2 on neuronal cell survival. SENP2 plays an essential role in the

Author Summary
Protein modification by SUMO is a reversible and evolutionarily conserved process. Members of the SUMOspecific protease (SENP) family are known to reverse SUMO-conjugation in many defined systems, but their importance in mammalian development and pathogenesis remains largely elusive. Although SUMO-conjugated proteins have been shown to aberrantly accumulate in patients with neurodegeneration, there is no evidence supporting a direct involvement of SUMO modification enzymes in human diseases. This study reveals that disruption of SENP2 causes neurodegeneration through modulation of mitochondrial morphogenesis. Our findings provide a causal link of SUMO modification enzymes to cell survival, suggesting a new pathogenic mechanism for neurodegeneration. Exploring the protective effect of SENP2 on neuronal cell death may uncover important preventive and therapeutic strategies for neurodegenerative diseases.
regulation of mitochondrial dynamics during postnatal development of CNS.
We then examined whether the SENP2 deficiency causes imbalances of sumoylation, resulting in accumulations of SUMOconjugated proteins. Immunostaining of SUMO1 showed increased levels of the sumoylated proteins (26.161.5% in control vs. 39.464.5% in mutant), indicating that SENP2 deficiency induces hyper-sumoylation ( Figure 3A, p,0.01, ,700 cells were counted in each of 3 independent experiments, mean 6 SEM). Although SENP2 was shown to regulate the Mdm2-p53 pathway [16,17], the expression and the activity of p53 and Mdm2 were not altered in these mutants (Fu and Hsu, unpublished). The neural defects caused by the SENP2 deletion most likely were not associated with p53-induced apoptosis, which is a mitochondrial independent event. Examination of protein extracts isolated from the P7 cerebral cortices revealed an elevation of SUMO1 association in the mutants ( Figure 3B). The loss of SENP2 activated Bak ( Figure 2B), which has been shown to promote sumoylation of Dynamin regulated protein 1 (Drp1) and its association with mitochondria during programmed cell death [24,25]. Therefore, we tested if Drp1 is affected in the SENP2 DSUMO-Nes mutants. Not only the stability (1.9-fold), but also SUMO1 association with Drp1 (2.7-fold), was enhanced by the mutation while RanGAP1, a known substrate of SENP2, did not appear to be affected ( Figure 3B). We then examined the mitochondrial association of Drp1 in primary neurons derived from the cerebral cortices of mouse embryonic brains. The mutation apparently promoted Drp1 association with the mitochondria (Fig. S7). The results implied that dysregulation of Drp1 may cause mitochondrial defects, leading to the development of neurodegeneration in the SENP2 DSUMO-Nes mutants.

SENP2 in SUMO-mediated regulation of Drp1
Drp1 has been implicated in neural degenerative diseases with disruption of mitochondrial dynamics [26,27]. To test if Drp1 plays a role in this pathogenic process, we investigated its regulation by SENP2. Our previous report showed that three gene products of SENP2 (SENP2, SENP2M and SENP2S), generated by alternative splicing, leading to the use of distinct translation initiation sites, exhibit distinct subcellular localizations and functions [16]. The SENP2, SENP2M and SENP2S isoforms are predominately located to the nucleus, cytoplasmic vesicles and perinuclear region, and cytoplasm, respectively [16]. First, we examined which of these isoforms might be involved in the regulation of Drp1 using a parental cell line and its stably transformed variants, which express high levels of different isoforms [16]. Whole cells or mitochondria only prepared from these cell lines were used to isolate extracts, followed by protein analysis. Overexpression of a HA tagged SUMO1 led to hypersumoylation of total as well as the mitochondrial proteins in the parental cells which occurs less effectively in all SENP2 variants ( Figure 4A). SUMO1 also promotes total cell, cytoplasmic and mitochondrial accumulation of Drp1, suggesting that its stability is modulated by sumoylation. However, this regulatory process, not affected by SUMO2 and SUMO3, is apparently a SUMO1specific regulation ( Figure S8A, B). Moreover, high levels of SENP2S, but not SENP2 and SENP2M, prevented the SUMO1induced accumulation of Drp1 to the mitochondria ( Figure 4A). SENP2S also decreased the SUMO1-induced accumulation of Drp1 in the cytoplasm. Thus suggests that the Drp1 reduction mediated by SENP2S is caused by protein degradation rather than decreased targeting to the mitochondria ( Figure 4A). Immunoprecipitation-immunoblot analysis further showed that the SU-MO1-association of endogenous Drp1 is eliminated by SENP2S, but not other isoforms ( Figure 4B). Although certain levels of reduction were detected in the SENP2 and SENP2M analyses,  these might be attributed to the disruption of cellular compartmentalization in vitro. To further examine the ability of SENP2 to remove SUMO1 from Drp1, we used in vitro reconstitution analysis ( Figure 4C). Recombinant enzymes, including Ubc9 and SAE1/2, were first utilized to perform the SUMO1 conjugation of Drp1. The addition of purified SENP2 efficiently was able to reverse this sumoylation process (a ,3.8-fold decrease), suggesting Drp1 as a direct substrate of SENP2 ( Figure 4C). Because of differential subcellular distributions of the SENP2 isoforms (SENP2 in nucleus; SENP2M in Golgi; SENP2S in cytoplasm) [16], their co-localizations with Drp1 were then investigated. Double labeling analysis indicated an extensive co-localization between Drp1 and SENP2S ( Figure 4D). Using a proximity ligation assay examining protein-protein association within the cells, we found that SENP2S exhibited an isoform-specific interaction with Drp1 ( Figure 4E). The interaction apparently took place in the mitochondria and cytoplasm ( Figure 4F). Furthermore, using siRNA specifically knocking down SENP2 to an expression level at ,17% ( Figure 4H), we found that its reduction promotes Drp1 association with the mitochondria (Figure 4G), resulting in a 2.2-fold increase compared to the control ( Figure 4H). A mitochondrial protein with higher molecular mass, which is probably the SUMO1-associated Drp1, was also increased in the SENP2 siRNA treated cells. Consistent with our analysis in the primary neuron ( Figure 2C), the knockdown of SENP2 also enhanced mitochondrial fragmentation in the cell line ( Figure 4I, p,0.01, ,200 cells were counted in each of 3 independent experiments, mean 6 SEM). These results imply an isoform-specific effect of SENP2 on Drp1 stabilization and mitochondrial accumulation through modulation of SUMO1specific conjugation.
The isoform-specific regulation of Drp1 by SENP2S suggests its potential involvement in modulating mitochondrial dynamics. Using DsRed2-Mito labeling, mitochondrial morphology was examined in HCT116 and HCT116-SENP2S cells. Similar to previous findings [28], overexpression of Drp1 and SUMO1 caused fragmentation of the mitochondria in these cells (Figure 5A-B, A9-B9, I). However, the SUMO1-induced mitochondrial fission was prohibited by high levels of SENP2 (Figure 5C, C9, G, G9, I, p,0.01, ,100 cells were counted in each of 3 independent experiments, mean 6 SEM). This might be attributed to the regulatory effects of SENP2 on Drp1 sumoylation and stability. Therefore, we examined if Drp1 is involved in the SENP2-mediated protection of mitochondrial fragmentation. High levels of Drp1 were able to overcome the protective effect of SENP2 on the SUMO1-induced mitochondrial fission (Figure 5D, D9, H, H9, I, p,0.01, ,300 cells counted, n = 3, mean 6 SEM). In contrast, high levels of SENP2S did not seem to affect the Drp1-induced mitochondrial fission, suggesting that Drp1 acts downstream of SENP2S in the regulatory pathway. These results not only indicated a role of SENP2 in controlling mitochondrial dynamics but also suggested that SENP2 exerts its effects through modulation of Drp1.

Discussion
This study demonstrates that SENP2 controls the SUMO1mediated modification of Drp1 essential for the regulation of mitochondrial dynamics. Targeted disruption of SENP2 induces neurodegeneration through promotion of Drp1 sumoylation and mitochondrial fragmentation. Impaired desumoylation results in neural cell death suggesting a new pathogenic mechanism for neurodegenerative diseases. Dysregulation of several aggregationprone proteins which are sumoylation substrates have been implicated in neurodegeneration [19,20,22,29,30]. However, there is no evidence showing a direct involvement of SUMO modification enzymes in human diseases. Our findings suggest that enhanced sumoylation may also be attributed to mutations in the SUMO regulators in addition to the substrates. A balanced sumoylation is pivotal for neuronal cell survival. Hyper-sumoylation resulting from stimulation of SUMO ligases or disruption of SUMO proteases can lead to neural cell death. Our findings imply that targeting the SUMO protease may correct an imbalance of sumoylation and desumoylation. The SENP2 DSUMO-Nes mice might have potential in modeling human diseases associated with the SUMO pathway.
An association of the SUMO pathway with the regulation of mitochondrial dynamics has also been demonstrated in this study. Mitochondrial dysfunction has a strong association with neurodegenerative diseases [31][32][33]. Mitochondria possess a highly dynamic nature, undergoing frequent fusion and fission [34]. Due to large energy demands and long extended processes of the neurons, they are particularly sensitive and vulnerable to mitochondrial abnormalities. Enhanced mitochondrial fission induces apoptosis during neurodegeneration [31][32][33]. Mitochondrial dynamics is regulated by the GTPase dynamin-related protein Drp1, whose function is modulated by SUMO modification. In cells, overexpression of SUMO1 prevents Drp1 degradation, resulting in its stabilization and activation [35]. The SUMO1-induced Drp1 promotes mitochondrial fission which can be altered by manipulating the SENP activity [36,37]. Data presented in this study strongly suggest that SENP2 is the physiological enzyme essential for this regulation. SENP2 controls mitochondrial dynamics through modulation of Drp1  in neural development and disease. Furthermore, Drp1 regulation by the SUMO pathway is causally linked to neural degeneration.
The SENP2 deficiency causes cell survival issues through increases in mitochondrial fission, leading to the development of neurodegeneration. As Drp1 appears to be a direct substrate of SENP2, dysregulation of mitochondrial dynamics is likely the primary cause of defects induced by the SENP2 disruption. Further study of mice with aberrant expression of Drp1 in the neural cells promises new insight into this regulatory mechanism. It remains possible that the aberrant mitochondrial phenotype is one of the main causes, which acts parallel with another cellular abnormality or is a consequence of other cellular abnormalities, e.g. failure of neural connection. Therefore, it is interesting to test if prevention of mitochondrial apoptosis can alleviate the defects caused by SENP2 deficiency. Further examination on the role of mitochondria dynamics promises new insight into the SENP2mediated neuronal cell death.
The involvement of SENP2 in neural development and degeneration opens new opportunities to develop therapeutic targets in the SUMO pathway. As sumoylation has been shown to counter against ubiquitination, manipulation of the SUMO pathway may also alter the ubiquitination-mediated degradation for the prevention and treatment of neurological disorders. Although SENP2 may have a general effect on the neurons, it remains possible that a specific subtype is more sensitive to the loss of SENP2. In the SENP2 mutants, we identify different degrees of neurodegeneration in the cerebral cortex, hippocampus, cerebellum and spinal cord with the cerebral cortex being most severely affected. A disruption of SENP2 in a specific neuronal subtype may further divulge its role in neurodegenerative diseases. Testing the protective role of SENP2 in neural cell survival in disease conditions is also likely to gain a knowledge base of neurodegenerative diseases, leading to new therapeutic strategies.

Mouse strains
The SENP2 DSUMO Fx ES cell lines were generated by electroporation of a targeting vector, containing the insertion of an orphan loxP site in intron 15 and another loxP site and a pgk-neo cassette flanked by two FRT sites in intron 16, into CSL3 ES cells [17,40,41]. Twenty mouse ES cell clones heterozygous for the targeted allele were obtained by homologous recombination (targeting efficiency: 23/112). Two independent clones were injected into blastocysts to generate chimeras which were bred to obtain mice carrying the targeted allele. These mice were then crossed with the R26Flp mice to remove the pgk-neo cassette to obtain the SENP2 DSUMO Fx mouse strain. Mice were genotyped by PCR analysis using primers (59-TCTCACTTGAAACCG-TAGGGACC-39 and 59-GAAGGAAGGACTGGAGGAGAGA-AG-39) to identify the 59 loxP locus, primers (59-TTGTCA-GAAGCAGTGTCCTGCG-39 and 59-GACTGGGAAGATAT-GAACTCGGC-39) to identify the 39 loxP locus. The deleted allele was identified using primers (59-TCTCACTTGAAACCGTAGG-GACC-39 and 59-GACTGGGAAGATATGAACTCGGC-39). The PCR was performed by denaturation at 95uC for 5 min and 35 cycles of amplification (95uC for 30 s, 67uC for 30 s, and 72uC for 60 s), followed by a 7-min extention at 72uC. The SENP2 lacZ and Nestin-Cre mouse strains and genotyping methods were reported previously [17,40]. To generate the SENP2 DSUMO mouse strain expressing a deficient protein, mice carrying the SENP2 DSUMO Fx allele were crossed with EIIa-Cre transgenic mice to delete the protease core domain in the germ cells [41]. To examine the production of SENP2 transcript, the reverse transcription products were subject to PCR amplifications using

Histology, in situ hybridization and TUNEL analysis
Samples were fixed, paraffin embedded, sectioned and stained with hematoxylin/eosin for histological evaluation [17,42]. The in situ hybridization was performed as described [17,38,43,44]. In brief, sections were incubated with the digoxygenin labeled RNA probes generated by in vitro transcription [17,43], followed by recognition with an alkaline phosphatase conjugated anti-digoxygenin antibody and visualization with BM-purple [38,44]. TUNEL staining was performed with ApopTag (Millipore) as described [45,46]. For electron microscopy, mice were fixed by perfusion with fixative (2% paraformaldehyde, 2.5% Glutaraldehyde, 0.1M sodium cacodylate, 6.8% sucrose). The dissected tissues were then fixed with 1% osmium tetroxide, embedded in EPON/Araldite resin and cut in seventy nm sections, followed by staining with aqueous uranyl acetate and lead citrate and examined using Hitachi 7650 transmission electron microscope.

Proximity ligation assay, immunostaining, immunoblot and immunoprecipitation
Proximity ligation assay (PLA) was performed using Duolink In Situ reagents (Duolink Bioscience). Briefly, cells were fixed and incubated with rabbit anti-Drp1 and mouse anti-myc tag antibodies. Two oligonucleotide-labeled anti-rabbit and anti-mouse PLA probes, which bind to each other when they are in close proximity, were then used to generate fluorescent signals. Immunostaining of cells [47] and tissue sections [48][49][50] were performed by incubation with primary antibodies, followed by detection with fluorescenceconjugated or horseradish peroxidase-conjugated secondary antibodies. Images were taken using Zeiss Axio Observer microscope equipped with deconvolution analysis. To determine the mitochondrial morphology, cells were either stained by MitoTracker or transient expression of DsRed2-Mito. For statistical analysis, cells containing MitoTracker or DsRed2-Mito positive mitochondria were counted and scored for tubular/rod-like or fragmented mitochondria. Immunoblot was performed by isolation of protein extracts from mitochondria, cells or tissues using M-PER (Pierce) in the presence of protease inhibitor cocktail, followed by electrophoresis as described [16,17,38,48]. Isolation of mitochondria was performed using a Mitochondria Isolation Kit (Thermo Fisher) according to the manufacture's description. Immunoprecipitation was performed using Pierce Classic IP Kit. Briefly, cells were lysed in buffer containing 0.025M Tris, 0.15M NaCl, 0.001M EDTA, 1% NP-40, 5% glycerol. Approximately 500 mg of protein lysates were mixed with 2 mg of antibodies overnight, followed by incubation with Protein A/G agarose for 1 hour at 4uC. The antibody-bound complex was then incubated with elution buffer for 5 min at 100uC, and collected by centrifugation for SDS-PAGE analysis. Mouse monoclonal antibodies, Actin (Thermo Fisher), HA (Cell Signaling), and myc tag (Santa Cruz); rabbit polyclonal antibodies Bak (Novus Biologicals), caspase-3 (BD Biosciences), Cox IV (Cell Signaling), Drp1 (Novus Biologicals), and SUMO1 (Cell Signaling), were used in these analyses.

Sumoylation and desumoylation
Recombinant GST-SAE1/SAE2, GST-Ubc9, GST-Drp1, HA-SUMO1 and myc tagged (MT)-SENP2 proteins expressed in Escherichia coli were affinity purified. The 20-ml reaction buffer containing 50 mM Tris-HCl (pH 7.6), 5 mM magnesium chloride, 10 mM ATP, 1 mg of GST-ASE1/2, 2 mg of GST-Ubc9, 10 mg of GST-HA-SUMO1 and 200 ng of GST-Drp1 with the presence of protease inhibitor cocktail was incubated for 3 h at 37uC. The desumoylation reaction was then carried out in 10 ml of the above sumoylation mixture with the addition of purified MT-SENP2 for overnight at 37uC. The samples were then analyzed by SDS-PAGE and immunoblot analysis of Drp1 and SUMO1. Video S1 A clip shows movement difficulties of the P10 SENP2 DSUMO-Nes mice which further developed paralysis around P16. (MOV)