Register for Mechanobiology 2016!

The SUMO protease SENP5 is required to maintain mitochondrial morphology and function
Rodolfo Zunino, Astrid Schauss, Peter Rippstein, Miguel Andrade-Navarro, Heidi M. McBride


Mitochondria are dynamic organelles that undergo regulated fission and fusion events that are essential to maintain metabolic stability. We previously demonstrated that the mitochondrial fission GTPase DRP1 is a substrate for SUMOylation. To further understand how SUMOylation impacts mitochondrial function, we searched for a SUMO protease that may affect mitochondrial dynamics. We demonstrate that the cytosolic pool of SENP5 catalyzes the cleavage of SUMO1 from a number of mitochondrial substrates. Overexpression of SENP5 rescues SUMO1-induced mitochondrial fragmentation that is partly due to the downregulation of DRP1. By contrast, silencing of SENP5 results in a fragmented and altered morphology. DRP1 was stably mono-SUMOylated in these cells, suggesting that SUMOylation leads to increased DRP1 mediated fission. In addition, the reduction of SENP5 levels resulted in a significant increase in the production of free radicals. Reformation of the mitochondrial tubules by expressing the dominant interfering DRP1 or by RNA silencing of endogenous DRP1 protein rescued both the morphological aberrations and the increased production of ROS induced by downregulation of SENP5. These data demonstrate the importance of SENP5 as a new regulator of SUMO1 proteolysis from mitochondrial targets, impacting mitochondrial morphology and metabolism.


Recent progress has been made towards the identification of many proteins that control mitochondrial dynamics in mammalian cells, including the GTPases such as mitofusin1 and mitofusin2 (Chen et al., 2003; Eura et al., 2003; Ishihara et al., 2004; Rojo et al., 2002; Santel and Fuller, 2001), DRP1 (Pitts et al., 1999; Smirnova et al., 2001), Opa1 (Cipolat et al., 2004; Frezza et al., 2006; Griparic et al., 2004; Misaka et al., 2002; Olichon et al., 2003; Olichon et al., 2002), the inner membrane rhomboid-like protease PARL (Cipolat et al., 2006; Sik et al., 2004) and membrane modifiers such as endophilin B1 (Karbowski et al., 2004b; Pierrat et al., 2001; Takahashi et al., 2005) among others. Although the functional implications of dynamic mitochondrial behavior have been best characterized in the context of mitochondria-dependent programmed cell death (Bossy-Wetzel et al., 2003; Karbowski and Youle, 2003), the steady-state requirements for these morphological phenomena are also beginning to be appreciated as an important contributor to metabolic stability (Chen et al., 2005; McBride et al., 2006). However, the question remains as to how the activity of the mitochondrial remodeling machinery is regulated in response to cellular signaling cascades. One obvious mechanism to regulate morphological changes is through the use of post-translational modifications such as phosphorylation, acetylation, ubiquitination or SUMOylation. From a yeast two-hybrid screen, we previously identified SUMO1 as an interacting partner for DRP1 and, through biochemical experimentation, demonstrated that DRP1 is a substrate for direct SUMOylation (Harder et al., 2004). These data showed that the overexpression of SUMO1 resulted in a stable DRP1 conjugation that protected it from degradation and ultimately led to a shift in mitochondrial morphology towards a fragmented phenotype (Harder et al., 2004). SUMOylation is an important modification known to regulate an increasing number of reactions within the nucleus, at the nuclear pores and within the cytosol (Bossis and Melchior, 2006b; Marx, 2005). It is biochemically and structurally very similar to ubiquitin, where the covalent conjugation of SUMO to its substrates occurs after a complex cascade of enzymatic reactions. Newly synthesized SUMO1 is first cleaved by the SUMO proteases [called ubiquitin-like proteases (ULP) in yeast] at the C-terminus to remove four amino acids and reveal the diglycine motif. Following this, the SUMO1 protein forms a thioester bond with the catalytic cysteine within the E1 (heterodimer activating) enzyme in an ATP-dependent process. This complex then passes the SUMO protein on to form a second thioester bond with the cysteine within the single E2 enzyme Ubc9. Finally, a SUMO1 E3 ligase helps to mediate the transfer of SUMO between Ubc9 and the ϵ-amino group of a lysine residue within the substrate (Bossis and Melchior, 2006b; Johnson, 2004).

The deSUMOylation of conjugated substrates occurs by the action of the SUMO cysteine proteases that show broad substrate specificity because they recognize only the SUMO protein within the conjugate (Reverter and Lima, 2004; Xu et al., 2006). The current understanding of the specificity of SUMO proteases is based on their restricted localizations within the cell where they have access only to a subset of conjugated substrates. Nevertheless, the SUMO proteases are highly active, which results in the transient and sub-stoichiometric nature of most SUMO conjugates, making them difficult to detect experimentally. There are only a few SUMO proteases within the human genome, each carrying the conserved protease domain together with divergent N-termini. SENP1 is localized to the nucleus and has been shown to deSUMOylate HDAC1 to activate transcription of multiple genes (Bailey and O'Hare, 2004; Cheng et al., 2004). SENP2 (also called axam in rat) is localized primarily to the nuclear envelope (Hang and Dasso, 2002; Itahana et al., 2006; Zhang et al., 2002) and has been shown to function in transcriptional regulation, and in the regulation of axin and/or β-catenin pathways (Kadoya et al., 2002; Nishida et al., 2001). A truncated form of SENP2 can alter the localization of promyelocytic leukemia (PML) body proteins in the nucleus (Best et al., 2002; Ross et al., 2002). SENP3 and SENP5 are both primarily nucleolar proteins that have been shown to preferentially deSUMOylate SUMO2 and SUMO3 from substrates, although they also cleave SUMO1 (Gong and Yeh, 2006; Nishida et al., 2000). SENP6 (also known as SUSP1) is enriched in reproductive cells where it is localized to the nucleus to negatively modulate RXRα transcription (Choi et al., 2006; Kim et al., 2000). SENP8 (more commonly called DEN1) has specific isopeptidase activity for NEDD8, another ubiquitin-like modifier similar to SUMO (Gan-Erdene et al., 2003; Reverter et al., 2005). In each case, the primary localization of the SUMO proteases is within the nucleus, making it difficult to understand how they may function in the regulation of cytosolic SUMO deconjugation events.

Fig. 1.

SENP5 regulates mitochondrial fission. (A) COS-7 cells transfected as indicated were fixed and examined for mitochondrial morphology by examining the OCT:CFP signal. Values represent the mean ± s.d. of three coverslips, examining 50-100 cells from each. Fragmented, mitochondria with a stokes radius of 1. (B) Representative images of the OCT-CFP matrix marker shown as an example of the rescue phenotype. (C) Lysates of COS-7 cells (transfected as indicated) examined by SDS-PAGE, and then probed with anti-His6, anti-FP, anti-DRP1, anti-Mfn2 and anti-Tom20 antibodies. (D) Quantification of DRP1 protein levels obtained from the blots in C was expressed as a percentage of the intensity found in cells transfected with SUMO1:His alone. The values were normalized to the levels of Tom20 in each lane to ensure equal loading (n=3). UTF, untreated cell fraction.

To understand the regulation of DRP1 SUMOylation, we searched for the SUMO/sentrin proteases that participate in the regulation of mitochondrial dynamics. We here demonstrate that the cytosolic pool of SENP5 plays a role in the regulation of mitochondrial morphology at least in part through its effects on DRP1 deSUMOylation. Our data reveal that SENP5 is required to maintain normal mitochondrial morphology and to control intracellular levels of reactive oxygen species (ROS). This is the first report to demonstrate a function for a SUMO protease in the regulation of mitochondrial dynamics and lends insights into potential post-translational modifications that control the action of DRP1.


To test the function of SUMO proteases in mitochondrial dynamics, we screened for those capable of rescuing the SUMO1-induced mitochondrial fragmentation. Both SENP2 and SENP5 showed significant probability to target the mitochondria using the Mitopredict algorithms (not shown), and SENP2 has been known to function outside of the nucleus on axin in the cytosol (Kadoya et al., 2002; Nishida et al., 2001). Therefore, we focused first on these two SUMO proteases. Quantification of the mitochondrial morphology in cells transfected with either SENP5:YFP or SENP2:YFP (Hang and Dasso, 2002; Itahana et al., 2006) indicated that only SENP5 expression affected the mitochondrial morphology by increasing the number of cells with elongated tubules (Fig. 1A). In addition, the data showed that SENP5 specifically rescued the SUMO1-induced fragmentation that we had reported previously (Fig. 1A,B) (Harder et al., 2004). SENP5 has been shown to preferentially deSUMOylate SUMO2 and SUMO3 proteins, with lower activity towards SUMO1 (Di Bacco et al., 2006; Gong and Yeh, 2006). Therefore, we confirmed that SENP5 can catalyze the cleavage of SUMO1-conjugated substrates in COS-7 cells co-transfected with SENP5:YFP and SUMO1:His6. By probing the total cell extracts with antihistidine (His) antibodies, we determined that overexpression of SENP5 significantly reduced the level of total SUMO1:His6 conjugates that was dependent upon the presence of the catalytic cysteine residue C713 [mutated in SENP5(C713L), Fig. 1C]. Our data indicate that the low enzymatic activity previously observed in vitro for SENP5 upon SUMO1 compared with SUMO2 and SUMO3 (Gong and Yeh, 2006) does not exclude SUMO1 substrates as relevant biological targets. Our previous investigations identified DRP1 as the first mitochondrial SUMO1 target (Harder et al., 2004). We showed that the stable SUMOylation of DRP1 prevented its degradation, explaining - at least partially - the increased mitochondrial fission in cells overexpressing SUMO1 (Harder et al., 2004). We therefore examined the levels of endogenous DRP1 in cells expressing SENP5:YFP and consistently observed a reduction in the levels of this protein in total solubilized extracts. Importantly, this effect was dependent on the presence of the catalytic cysteine residue (Fig. 1C,D).

Fig. 2.

SENP5 is broadly distributed and deSUMOylates multiple mitochondrial substrates. (A) COS-7 cells transfected with SENP5:YFP were fixed; YFP fluorescence indicates that SENP5 localizes mainly in nucleolus with clear cytosolic and nuclear staining (bar, 10 μm). (B) COS-7 cells were harvested and fractionated to obtain mitochondria and cytosol. In addition to the total extracts, 50 μg of each fraction were loaded in a gel to examine endogenous SENP5 distribution. Markers of different fractions (HSP70 for cytosol, HSP60 and Tom20 for mitochondria) are indicated. (C) Untransfected COS-7 cells or those expressing SENP5:YFP were harvested and mitochondria were isolated. Of total cell lysate (Tot) or mitochondriaenriched fractions (Mit) 100 μg were separated by SDS-PAGE to examine the endogenous mitochondrial SUMO1 conjugates. Shown are short and long exposures of the anti-SUMO1 western blots (top and middle panel), together with anti-DRP1 and anti-Tom20 detection (bottom panel). SUMO1 conjugates in the mitochondrial fraction are indicated by circles, arrows indicate DRP1 high molecular mass conjugates. UTF, untreated cell fraction.

As previously reported, we also observed that SENP5:YFP is primarily localized to the nucleoli (Fig. 2A) (Di Bacco et al., 2006; Gong and Yeh, 2006); however, we also observed a significant amount of cytosolic staining (Fig. 2A). To ensure that this was not due to overexpression of the tagged protein, we examined the subcellular distribution of endogenous SENP5 by separating cytosolic and mitochondrial fractions (Fig. 2B). As with the transfected YFP-tagged SENP5, we were seeing substantial levels of endogenous SENP5 in the cytosol and very low levels associated with the mitochondria. However, we next examined whether cytosolic SENP5 deSUMOylate other mitochondrial associated proteins (Fig. 2C). After brief exposure, the amount of endogenous SUMO1 conjugation in the total extracts is reduced upon overexpression of SENP5:YFP, consistent with the loss of His6:SUMO1 conjugates observed in Fig. 1C. In addition, after longer exposure, a number of SUMO1 conjugates can be seen in the mitochondrial fraction (Harder et al., 2004); these are significantly reduced in cells overexpressing SENP5 (Fig. 2C, circles). Although the total levels of DRP1 are reduced by ∼50% (Fig. 1D), overexposure of the blot also reveals a reduction in conjugates of higher molecular mass (arrows, Fig. 2C). These data indicate that SENP5 is a SUMO1 protease that shows broad substrate specificity for mitochondrial proteins, regulates the sensitivity of DRP1 for degradation and protects mitochondria from fragmentation.

To examine the functional importance of SENP5 on DRP1 and mitochondrial function, we used short hairpin RNAs (shRNAs) designed to knock-down the expression of SENP5 and selected for stable transfectants. Although we tested a number of shRNA specifically targeting SENP5, we achieved only ∼60±15% reduction of the protein (Fig. 3A,B, n=8). Silencing of SENP5 resulted in ∼70±29% increase of total DRP1 protein levels, providing further evidence that SUMOylation protects DRP1 from degradation (Fig. 3A,B, n=8). Although very high exposures reveal a 150 kDa DRP1 conjugate band, we did not see a significant increase in this product upon the reduction of SENP5 (not shown). However, immunoprecipitation of DRP1 consistently revealed a SUMO1 immunoreactive band at ∼90 kDa (Fig. 3C), which was significantly stabilized upon the loss of the SENP5 SUMO protease. Western blot analysis of DRP1 shows either doublet or triplet bands from different tissue sources, but the origin of these different DRP1 species has not been determined. The presence of a 90 kDa SUMO1-positive band suggested that the loss of SENP5 resulted in the stabilization of a mono-SUMOylated DRP1 product. To further examine this possibility, we transfected the Flag:SUMO1 construct into cell lines that had been stably transfected with either vector alone or with or shRNA targeting SENP5 and repeated the immunoprecipitation with anti-DRP1 antibodies. As seen with the endogenous protein, Flag:SUMO1 was immunoprecipitated with DRP1 and the band precisely overlapped with the highest molecular mass DRP1 product (90 kDa; arrows Fig. 3D). In addition, probing with anti-SUMO1 antibody again detected the 90 kDa SUMO1 conjugate in the vector-transfected control cells, and was enriched within SENP5-silenced cells (Fig. 3D). The use of Flag-tagged monoclonal antibodies, monoclonal anti-SUMO1 and polyclonal anti-SUMO1 antibodies all confirmed the same result: the loss of SENP5 stabilizes a mono-SUMOylated conjugate of DRP1. Although the migration of this doublet appears to reflect a shift of approximately 3 kDa to 6 kDa, it remains possible that this does reflect a ∼10 kDa SUMO1 conjugate.

Fig. 3.

Silencing of SENP5 results in increased DRP1 SUMOylation. (A) SENP5 silencing increased DRP1 conjugates. COS-7 cells transfected with either shRNA vector alone (vector) or with SENP5-specific shRNA (SENP5 RNAi1), were selected in the presence of 5ug/ml puromycin for 12 days before collection; 100 μg of the total lysates were separated by SDS-PAGE and western blots with the indicated antibodies are shown. (B) Quantification of the SENP5 and DRP1 protein levels in control (Vector) or SENP5-silenced [shRNA (SENP5)] cells from eight different experiments. In each case, 100 μg of total cell extract was loaded and signals from the western blots were quantified using the Alpha Innotech HD Gel documentation system. (C) Immunoprecipitation of DRP1 from control (Vector) or SENP5-silenced (RNAi1) cells reveals an increase in mono-SUMOylated DRP1. COS-7 lysates from cells stably expressing vector alone or SENP5 shRNA were immunoprecipitated with either anti-DRP1 antibody or mouse IgG. Immunoprecipitates were run in duplicates on a 4-20% gradient gel; one blot was probed for endogenous DRP1 and the other for SUMO1. (D) Immunoprecipitation of DRP1 from SUMO1:FLAG-transfected COS-7 cells stably expressing vector alone or SENP5 shRNA. COS-7 cells stably expressing vector alone or SENP5 shRNA were transfected with SUMO:Flag for 24 hours prior to immunoprecipitation with either anti-DRP1 antibody or mouse IgG. Immunoprecipitates were run on duplicate gels; one blot was probed with a monoclonal anti-Flag antibody, the second with the anti-SUMO antibody. Following this, both gels were probed with anti-DRP1 antibody. The increased total level of DRP1 precipitated within the SENP5-silenced cells appears as a doublet. Only the top band of this doublet overlaps directly with the anti-Flag and anti-SUMO1 antibody signal (arrow).

Having established that the silencing of SENP5 leads to increased DRP1 SUMO1 conjugation, we next examined the consequences on mitochondrial morphology. Cells stably expressing vector or SENP5 shRNA were fixed, stained with anti-Tom20 antibodies and the mitochondrial morphology was examined (Fig. 4A). Consistent with a positive function of DRP1 SUMOylation in the process of mitochondrial fission, the cells with reduced levels of SENP5 contained many fragmented mitochondria. However, the overall phenotype was mixed, containing many fragments in addition to a number of round and flattened mitochondria (Fig. 4A,B). These mitochondria remained motile and time-lapse imaging revealed flexible amoeboid-like movements, including multiple fission events (supplementary material Movie 1). The unusual mitochondrial shape in the SENP5-silenced cells suggested alterations in the cristae architecture of the organelle. Upon examination of the ultrastructure of the mitochondria by electron microscopy, we observed small fragments of mitochondria in the SENP5-silenced cells (Fig. 4C, arrows); however, this was observed to be a partial effect, similar to that seen in the fluorescent images. The larger mitochondria appeared much less tubular, most appearing highly spherical, consistent with the round, flat shapes seen by confocal microscopy. The cristae within these cells were generally intact, although few organelles showed some alterations of their inner membranes. We consider it likely that these differences reflect the increased probability of sectioning laterally through the widened and flat lamellar cristae (Fig. 4C, asterisks). The phenotype is distinct from apoptotic cristae remodeling - where the opened cristae result in very spherical mitochondrial fragments (Germain et al., 2005; Scorrano et al., 2002) - because the confocal reconstructions of mitochondria from SENP5 silenced cells show a flat, sickle-cell-like morphology rather than the swollen phenotype (supplementary material Movie 2). In cells infected with adenoviral expressed DRP1 conjugated to YFP (DRP1:YFP), we observed an increased recruitment of DRP1 around the peripheral edges of the enlarged, flattened organelles that often appeared as ridges of fluorescence (Fig. 4D, arrows). This is a localization distinct from the more punctate distribution of DRP:YFP observed along the tubular organelles in control cells (Fig. 4D, top panels). SUMO1:YFP was observed to be recruited to mitochondria in a punctate pattern both in control and SENP5-silenced cells. There were increased numbers of SUMO1:YFP puncta in cells treated with SENP5 siRNA. Although some of these puncta were found at the sites of mitochondrial fission (Fig. 4E and supplementary material Movie 3), many remained stably associated with the fragmented mitochondria. The different patterns of SUMO1 and DRP1 recruitment indicates that mitochondrial DRP1:YFP is SUMOylated primarily at sites of fission. This is consistent with the sub-stoichiometric nature of the conjugate observed biochemically (Fig. 3C,D), even in cells treated with SENP5 siRNA. We next wanted to examine the requirement of functional DRP1 in generating the phenotype observed in the SENP5-silenced cells. We first grouped the phenotypes into four categories, normal (including thin tubules and rods), fragmented (uniformly fragmented mitochondria), networked (interconnected reticulum) and enlarged/fragmented (a mixture of fragmented and enlarged organelles that appear flattened). The two latter phenotypes are shown in Fig. 4B; enlarged are indicated with an asterisk and fragmented organelles by an arrow. The quantifications of these phenotypes are shown in Fig. 5B, where the SENP5-silenced cells exhibit a significant increase in these aberrant (round, flat) mitochondria. To confirm a functional role for DRP1 in this transition, we infected the stable cell lines with a dominant-interfering DRP1 mutant, the adenovirally expressed vector DRP1(K38E):CFP (Fig. 5Aii,v). We also knocked down the DRP1 mRNA in both control cells and cells stably transfected with SENP5 shRNA (Fig. 5Aiii,vi, and supplementary material Fig. S1). In both groups of cells fragmentation and formation of aberrant morphologies observed in cells containing SENP5 shRNA was inhibited (Fig. 5B), and the mitochondria appeared as a fused network (Fig. 5A). This further supports DRP1 as a functional target of SENP5.

Fig. 4.

Silencing of SENP5 alters mitochondrial morphology, and increases DRP1 and SUMO1 recruitment. (A) COS-7 cells treated or not with shRNA targeting SENP5 were infected with DRP1:YFP, fixed and stained with anti-Tom20 antibody. Images were taken with the Poloychrome IV monochrometer on the Olympus IX70 microscope; Tom20 staining (white), DRP:YFP (green). Bars, 15 μm. (B) Amplified images of mitochondria from different cells treated with shRNA targeting SENP5 stained with anti-Tom20 antibody. Confocal images show enlarged mitochondria (*) and mitochondrial fragments (arrows), typical of cells that express shRNA targeting SENP5. Bars, 5 μm. (C) Cristae are not significantly altered in SENP5-silenced cells. COS-7 cells stably expressing vector (left panels) or SENP5 shRNA (right panels) were processed for EM. Two representative images are shown for each condition. Many smaller mitochondria were observed in the SENP5-silenced cells (arrows), larger mitochondria lost their thin, tubular morphology and appeared more spherical. Asterisks represent cristae sectioned longitudinally through the lamella, which were more readily observed within those cells expressing SENP5 shRNA. Bars, 500 nm. (D) Enlarged boxed areas shown in A showing recruitment of DRP1:YFP to mitochondria in control cells (top panels) and SENP5-silenced (bottom panels). Bars, 2 μm. Note the relatively even distribution of DRP1 puncta along mitochondria within control cells. Mitochondria that lack SENP5 show many enrichments of DRP1:YFP along the ridges of the enlarged mitochondria (arrows). (E) Images shown represent enlarged regions of confocal sections taken from live cells expressing either control shRNA (top panels) or SENP5 shRNA (bottom panels) that were transfected with SUMO1:YFP (green) and the matrix marker pOCT:CFP (white). In control cells, SUMO1:YFP is occasionally observed at active sites of fission as shown previously (Harder et al., 2004). In SENP5-silenced cells, similar SUMO1:YFP puncta are also observed at sites of fission; however, there are an increased number of SUMO1:YFP puncta on the mitochondria that did not appear to be at active fission sites (at least during the time of imaging). These still images are taken from a live confocal video which shows SUMO1:YFP puncta with respect to the dynamic fission activity of mitochondria (supplementary material Movie 3A,B). Bars, 2 μm.

Another important question in the regulation of mitochondrial morphology is the relationship between DRP1-mediated fission and the process of mitochondrial fusion. We therefore wanted to test whether the SENP5-silenced cells containing the aberrant, fragmented mitochondria were altered in their rates of fusion. To test this we targeted photoactivatable GFP (PAGFP) to the matrix of mitochondria by conjugating it to the 32-amino-acid-long matrix-targeting signal of ornithine carbamyl transferase (OCT), resulting in OCT:PAGFP (Karbowski et al., 2004a; Patterson and Lippincott-Schwartz, 2002). Stably transfected cells expressing either vector shRNA or SENP5 shRNA were then transfected with pre(p)OCT:PAGFP and the signal was activated in approximately one-quarter of each cell using a laser at 405 nm (Fig. 6A). The spread of the signal throughout the mitochondrial reticulum was then captured by obtaining between ten and twelve 1-μm stacks through the cell, every 10 minutes for a total of 40 minutes. Following this, all of the PAGFP within the mitochondria was activated with at 405 nm. By creating a mask that scored one binary signal per voxel, we calculated the percent of the voxels containing a PAGFP signal at each time point within the cell relative to the total signal upon complete illumination of the mitochondria (see Materials and Methods). Using this method, we plotted the increase of mitochondria (as a percentage) that shared the PAGFP at the end of the 40-minute period for at least ten cells in each condition (Fig. 6B). The data showed that mitochondria of control cells shared their signal with 10-50% of unactivated mitochondria in the 40-minute period. Cells silenced for SENP5 did not fuse as extensively as control cells, with a mean increase in fusion of ∼4% (Fig. 6). This indicates that the phenotype observed in the SENP5-silenced cells is partly due to the effects on mitochondrial fusion in addition to (or as a consequence of) the increased stability of DRP1 and fragmentation.

Fig. 5.

Interference with DRP1 rescues the SENP5-silenced phenotype. (A) COS-7 cells stably expressing vector only (panels i,ii,iii) or SENP5-specific shRNA (panels iv,v,vi) were infected with the dominant-negative DRP1-K38E-CFP-HA (infection controls are shown in insets of ii and v), or treated with dsRNA to knock-down expression of DRP1 (panels iii,vi; knock-down controls are shown in supplementary material Fig. S1) were fixed in 4% PFA for staining with monoclonal antibody against cytochrome c. Bars, 10 μm (i,ii,iv,v) and 5 μm (iii,vi). (B) Quantification of the mitochondrial phenotype observed in A. Each bar represents the mean ± s.d. of cells from three coverslips, with 50 cells examined per coverslip.

Given the evidence that mitochondrial morphology is a reflection of metabolic stability (Chen et al., 2005; Pich et al., 2005; Yu et al., 2006), we were interested to examine whether this aberrant and fragmented mitochondrial phenotype may reflect alterations in metabolism (ΔΨ) or increases in ROS production. Increased free radicals have been shown to lead to mitochondrial swelling and general alterations in morphology (Zorov et al., 2005). Therefore, we incubated cells stably transfected with the vector or SENP5 shRNA with both MitoFluorRed 633 and dihydroethidium (hET), a free radical sensor dye. Upon oxidation hET becomes cleaved, resulting in a shift in fluorescence from excitation at 355 nm and emission 420 nm to the ethidium fluorescence at excitation 518 nm and emission 605 nm. The resulting ethidium intercalates with mtDNA and, upon high rates of conversion, the dye also fills the cytosol and nucleus as it binds chromosomal DNA. Cells were incubated with the dyes at 37°C and images were taken using at 543 nm (to visualize oxidized hET) and 633 nm to visualize the MitoFluorRed 633 (Fig. 7A,B). Images were collected from many fields from live cells using identical laser intensities and image resolution to compare fluorescence intensities. The total fluorescence for the two dyes in each image was then divided by the cell number to obtain an average intensity per cell. Data from 244 SENP5-silenced cells and 270 cells containing the vector alone demonstrated no difference in the electrochemical potential (uptake of photometric dye) between the two cell types (Fig. 7A,B). This was consistent with the generally unaltered appearance of the cristae morphology (Fig. 4C), whose disruption is often correlated with disruptions in metabolic state (Mannella et al., 2001). However, there was a significant increase in the production of ethidium in cells containing SENP5 shRNA (Fig. 7A,B). The standard deviation between those cells (SENP5 shRNAi) was larger than the control cells, which is partly explained by the fact that the knock-down was not 100% efficient. In addition, the fluorescence intensity of the oxidized hET dye increased (relative to control cells) during the 90-minute imaging period. We therefore plotted the average fluorescence intensity per cell relative to the time in which the image was taken (Fig. 7C). This analysis indicates that cells lacking SENP5 produce radicals at a faster rate than the control cells. In addition, the images of cells incubated with hET indicate that the SENP5-silenced cells show a highly diffuse pattern of oxidized fluorescence with high enrichment in the nucleoli, whereas even at 60 minutes of incubation, the control cells still show clear mitochondrial localization of the dye (Fig. 7A). We further confirmed the quantification of individual cells by collecting total cells, and measuring the oxidized hET fluorescence in a fluorimeter after 40 minutes of incubation. These data demonstrate a two-fold increase in the production of free radicals in the SENP5-silenced cells (Fig. 7D). To determine whether this metabolic phenotype could be rescued by reassembling the tubular morphology, we infected the vector-containing and SENP5-silenced cells with adenovirally expressed DRP1(K38E):CFP, as well as silencing DRP1 with siRNA. The data reveal a partial but significant rescue of the increased ROS production in silenced cells upon inhibition of DRP1 activity by silencing or through the dominant-negative DRP1(K38E) mutant (Fig. 7E and supplementary material Fig. S1). The partial rescue suggests that, although SENP5 functions in the regulation of mitochondrial morphology, the metabolic effects on the production of ROS may be due to the modification of additional cytosolic or mitochondrial SENP5 SUMOylated substrates. Nevertheless, these data provide further evidence to support the idea that mitochondrial morphology and metabolic state are molecularly linked.

Fig. 6.

Mitochondrial fusion is inhibited upon silencing of SENP5. (A) COS-7 cells stably transfected with shRNA vector alone (top panels) or to shRNA targeting SENP5 (bottom panels) were transfected with pOCT:PAGFP and the regions of interest indicated were activated at 405 nm. Stacks were obtained by imaging cells at 488 nm every 10 minutes for 40 minutes, after which all of the mitochondria containing PAGFP were activated. Masks were obtained where each voxel having a fluorescent signal above the threshold was scored as white and single slices taken from the stacks are shown at 1 minute post activation, at 40 minutes and after complete activation. Bars, 10 μm. (B) Quantification of the data obtained from control cells (n=10) and from SENP5-silenced cells (n=13). The percent increase in the spread of the activated signal is shown for each cell. The circled data points represent the cell whose corresponding images are shown in A. The data show that there is an average of 21% of mitochondrial fusion within 40 minutes in COS-7 cells, whereas SENP5-silenced cells have significantly reduced rates of fusion (P<0.001).

Fig. 7.

Downregulation of SENP5 leads to increased reactive oxygen species (ROS) production. (A) COS-7 cells stably expressing either vector (top panels) or silenced for SENP5 shRNA (middle and bottom panels) were incubated at 37°C with 10 μM hET; oxidized hET (red) and MitofluorRed (green) were observed at the times indicated. Note the increased intensity of red fluorescence at 60 minutes in cells that contain shRNA targeting SENP5. The bottom panel shows a higher magnification of the aberrant mitochondria within these cells. (B) Average fluorescence intensity per cell of oxidized hET and MitofluorRed from 244 SENP5-silenced cells and 270 cells containing the vector. Data were collected (using identical laser intensities and scanning parameters) from live cells over a 90-minute period, imaging 10-20 cells per field. (C) Average fluorescence of oxidized hET per cell was plotted for each field as a function of imaging time. Each point represents the average hET fluorescence of 10-20 cells. Control cells did not significantly increase their ROS production over time, whereas the rate of ROS production in SENP5-silenced cells is increased. (D) COS-7 cells stably expressing either vector or shRNA targeting SENP5 were harvested and duplicate samples of 105 cells were incubated at 37°C with 10 μM hET for 40 minutes in culture medium. Then, fluorescence emission scans were collected to examine the reduced (emission peak, 420 nm; excitation, 355 nm) and oxidized hET (emission peak, 620 nm; excitation, 518 nm). Each value represents the mean ± s.d. of two independent aliquots (*P<0.01). (E) Same as D, but cells had been previously infected or not with DRP1-K38E-CFP-HA. Each value represents the mean ± s.d. of two independent aliquots (*P<0.01, **P<0.01).


SUMOylation is a reversible, sub-stoichiometric regulatory modification that plays an essential role in modulating dynamic protein complex transitions (Johnson, 2004). Most of the known targets of SUMOylation reside in the nucleus or nuclear envelope, with functional effects on transcription, chromatin remodeling, protein transport and cell cycle transition. The mechanistic consequence of SUMOylation is dependent on the substrate, where it may help to stabilize a protein complex from degradation (Ulrich, 2005a), affect the activity of enzymes and channels (Rajan et al., 2005; Ulrich, 2005b), trigger import or exclusion from the nucleus (Marx, 2005) or, in the case of septin GTPases, to disassemble oligomeric complexes (Johnson and Blobel, 1999; Johnson and Gupta, 2001). We have previously reported the first mitochondrial SUMO1 target to be DRP1 and demonstrated that overexpression of SUMO1 led to increased mitochondrial fragmentation and the stabilization of DRP1 protein. We have here extended these studies by examining the SUMO protease SENP5 and shown that its overexpression reduces the total of SUMO1 conjugates and DRP1 levels. Conversely, silencing of SENP5 leads to the stabilization of DRP1, to a fragmented and widened mitochondrial morphology and is coincident with the increased production of ROS. SUMO proteases have broad substrate specificity (Mossessova and Lima, 2000; Xu et al., 2006) because they primarily recognize the SUMO protein and not the SUMOylated substrate. This point is highlighted by the fact that there are only two known SUMO/sentrin proteases in yeast (Fig. 1A) and they are responsible for deSUMOylating all of the cellular conjugates. The yeast SUMO proteome has identified more than 150 substrates (Hannich et al., 2005; Panse et al., 2004; Wykoff and O'Shea, 2005; Zhou et al., 2005), which clearly indicates that the two SUMO proteases recognize many conjugated substrates. However, evidence is emerging that the mammalian SUMO proteases show some specificity; for example, SENP5 was shown to remove SUMO1 from only one of three conjugated residues within PML (Gong and Yeh, 2006). In addition, SENP2, a protease that has known cytosolic targets (Kadoya et al., 2002; Nishida et al., 2001), did not affect mitochondrial SUMO conjugates or morphology, which further indicates that these proteases may exhibit more substrate specificity than previously anticipated. We have shown that DRP1 is one of the SENP5 mitochondrial substrates whose mono-SUMOylation is stabilized when the protease levels are reduced. Interestingly, the previously reported SUMO1 conjugate of DRP1 at 150 kDa (Harder et al., 2004) was also observed in these experiments; however, this product was present in very low stoichiometric amounts and was not significantly stabilized upon the loss of SENP5. Instead, the immunoprecipitated DRP1 shows that the top, 90-kDa band of the DRP1 doublet is also SUMO1 immunoreactive and significantly stabilized upon the loss of SENP5. This suggests that the SUMOylation-deSUMOylation cycle of DRP1 is complex and may involve the activity of more than one SUMO1 protease. The functional examination of each specific SUMOylated conjugate of DRP1 is the subject of ongoing work.

The two mitochondrial phenotypes induced upon silencing of SENP5, the morphology and the increased ROS production, are rescued by interference with DRP1. This provides further evidence that the SUMOylation of DRP1 positively regulates its function with respect to mitochondrial fission and membrane remodeling, and that the morphology of the mitochondria reflects the metabolic status of the organelle. The loss of SENP5 also led to a decrease in mitochondrial fusion, which may suggest some additional targets within the fusion machinery. However, it is important to note that the loss of DRP1 resulted in highly fused and interconnected mitochondria, even in SENP5-silenced cells. This would suggest that the inhibition of fusion is not because of a direct effect of altered SUMOylation within the core components of the fusion machinery. Instead, the effects on fusion may have been secondary to increased ROS production. Alternatively, actions of DRP1 and the fusion machinery might be mutually exclusive, where the activation of DRP1 by stable SUMOylation might inhibit the activity of the fusion machinery. In either case, these data indicate an important new relationship between SUMOylation and the regulation of mitochondrial dynamics.

Finally, our observation of a relationship between ROS production and SUMOylation lends insights into the functional requirement for SUMOylation in mitochondrial biology. It has been shown that ROS production results in the direct inhibition of the SUMOylation machinery by inducing the formation of a disulfide bridge between the catalytic cysteines of Uba2 of the E1 activation enzyme and Ubc9 (Bossis and Melchior, 2006a). Those data demonstrated that SUMOylation was inhibited upon incubation with H2O2, upon oxidative stress and during the respiratory burst in macrophages, making it a key sensor in these pathways (Bossis and Melchior, 2006a). Our data support the idea that some of the targets of the ROSsensitive SUMOylation machinery may directly impact mitochondrial function and morphology. Future work will focus on the characterization of other mitochondrial SUMO targets whose modification functions to regulate the metabolic and structural stability of the organelle.

Materials and Methods


COS-7 cells were purchased from American Type Culture Collection, Manassas, VA. cDNA encoding human SENP5 (accession number NM_152699) was purchased from the Riken NIH Gene Collection (Invitrogen, Carlsbad, CA). Polyclonal anti-Tom20 antibody was a generous gift from Gordon Shore, McGill University, Montreal. Other antibodies: monoclonal anti-cytochrome c (BD Biosciences, Mississauga, ON), anti-DRP1 (Transduction Laboratories, BD Biosciences, Mississauga, ON), anti-FP (BD Biosciences, Mississauga, ON), anti-HSP60 (Sigma, Oakville, ON), anti-HSP70 (Stressgen, Victoria, BC), anti-Tubulin (Molecular Probes, Oregon), anti-SUMO1 (ZYMED, San Francisco, CA), monoclonal anti-His6 (Cell Signaling, Beverly, MA). Alexa-Fluor-647-conjugated secondary antibodies were from Molecular Probes (Invitrogen, Carlsbad, CA). hET (dihydro-ethidium) and MitofluorRed 633 were obtained from Molecular Probes. The DRP1-K38E-HA adenoviral vector was constructed in Ruth Slack's laboratory, Ottawa, Canada (Germain et al., 2005). Protease inhibitors, N-ethyl maleimide (NEM) and puromycin were obtained from Sigma (Oakville, ON). Lipofectamine 2000 was obtained from Invitrogen (Carlsbad, CA).


pOCT:CFP, SUMO1:YFP, DRP1:YFP, DRP1mycHisB and SUMO1mycHisB constructs have been described previously (Germain et al., 2005; Harder et al., 2004). The SENP5 coding sequence was inserted into the pEYFP-N1 vector (Clontech) by PCR amplification from the EST-containing vector (Riken clone 4824344) using primers 5′-GGGGTACCGATGAAAAAACAGAGGAAAATT-3′ (forward) and 5′-CGGGATCCCGGTCCATGAGCCGGCACTCACA-3′ (reverse), which introduced unique KpnI (5′) and BamHI (3′) restriction sites for subcloning into pEYFP-N1 (Clontech, BD Biosciences, Mississauga, ON). To produce SENP5(C713L), a quick-change mutation was performed in pEYFP-N1:SENP5. The catalytic cysteine at position 713 was replaced for leucine using primers 5′-CAGAAAAACGACAGTGACCTAGGATTTGTGCTCCAG-3′ and 5′-CTGGAGCACAAAGACTCCTAGGTCACTGTCGTTTTTCTG-3′, introducing a new AvrI site for screening. To obtain pQE31:SENP5 (for recombinant protein expression and antibody production) pEYFP-N1:SENP5 was digested with SalI and BamHI. The fragment was ligated to pCDNA3.1MycHisB (Invitrogen, Carlsbad, CA), cut with XhoI and BamHI. This new pCDNA3.1mycHis:SENP5 was in turn digested with PmeI and HindIII to ligate the SENP5 coding region into pQE31 (Qiagen, Mississauga, ON) digested with SmaI and HindIII. Sumo:FLAG was obtained by digestion of pcDNA3-SUMO1mycHisB with BamHI and XhoI to excise the full SUMO1 sequence, and ligation into pCMV-Tag2A, already cut with BamHI and XhoI. To label the mitochondrial matrix with PAGFP, the vector pOCT:PAGFP was constructed. For this, the coding sequence of PAGFP was amplified by PCR (using the primers 5′-CGGGATCCGTGAGCAAGGGCGAGGAGCTGTTC-3′ and 5′-CGGATCTAGATTACTTGTACAGCTCGTC-3′) with pPA-GFP-N1 (generous gift from J. Presley, McGill University) as a template. The PCR fragment was digested with BamHI and XbaI, and inserted in-frame behind the OCT targeting sequence using the previously described pOCT:CFP vector (Harder et al., 2004) exchanging the CFP sequence with PAGFP. All constructs were confirmed by sequencing.

Recombinant SENP5 purification and antibody production

To obtain recombinant SENP5-His6, E. coli bacteria (BL21 strain) were transformed with pQE31:SENP5 and grown to log phase in 6 liters of LB medium containing 50 μg/ml ampicillin. After induction with 0.3 mM IPTG for 3 hours at 37°C, bacteria were pelleted at 3000 g, and resuspended in buffer A (50 mM Tris pH 8.0, 100 mM NaCl). Following addition of protease inhibitors (in μg/ml: APMSF, 10; aprotinin, 5; antipain, 5; leupeptin, 1; pepstatin A, 1; chymostatin, 10) and 5 mM β-mercaptoethanol, cells were broken using a French cell press at 2500 psi. The lysate was centrifuged at 35,000 rpm in Ti70 rotor (Beckman Coulter) for 30 minutes, and the supernatant incubated with pre-equilibrated nickel-NTA agarose beads (Qiagen, Mississauga, ON) for 2 hours at 4°C. After three washes in buffer A containing 1 mM β-mercaptoethanol and 20 mM imidazole, the protein was eluted with 100 mM imidazole, and dialyzed overnight against buffer A. For antibody production, rabbits were immunized with 200 μg of recombinant SENP5 per boost using standard protocols.

Cell culture

COS-7 cells were cultured in DMEM (Invitrogen, Carlsbad, CA) containing 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, penicillin and streptomycin. Transfections with OCT:CFP, SENP5:YFP, SENP5(C713L):YFP, SUMO:His, DRP1:His, were performed overnight with COS-7 cells at 80-90% confluency, using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). Transfections with SUMO-FLAG were performed with COS-7 cells stably expressing either vector alone or vector containing shRNAi, in the same conditions, but in the absence of selecting agent (puromycin). Infections with DRP1-K38E-CFP-HA adenovirus were performed overnight on cells in culture dishes at 8×107 PFU/ml of complete medium (Germain et al., 2005).

shRNA silencing

The pShag Magic2 shRNA vector alone and vector containing sequences (5′-TGCTGTTGACAGTGAGCGACCAGTTACTTGGAATAGACAGTAGTGAAGCCACAGATGTA-3′, and 5′-TGCTGTTGACAGTGAGCGCGCAGATGGTTTGTTACTTGAATAGTGAAGCCACAGATGTA-3′) to silence SENP5 were obtained from Open Biosystems. The plasmids were transfected into COS-7 cells using Arrest-In (Open Biosystems, Huntsville AL), and stably transfected cells were selected with 5 μg/ml puromycin for 2 weeks before use in imaging and biochemical experiments. To silence DRP1, either non-targeted or four different double-stranded RNA oligonucleotides (antisense sequences: 5′-PUUAAUAUCUAGCUGGCUCCUU-3′, 5′-PUGGUAAACAAUCUCUGAUGUU-3′, 5′-PUAACAGGCAACCUUUUACGUU-3′ and 5′-PUUAUCAUCCACGGGUUCACUU-3′, Dharmacon RNA Technologies, Lafayette CO) targeted against different regions of DRP1 mRNA, were transiently transfected to either vector control or SENP5 shRNA stable cell lines, using Dharmafect3 (Dharmacon, Lafayette CO) as lipid carrier, for 4 days, after which cells were processed for imaging, biochemical and spectrofluorometric experiments.

Mitochondrial fractionation

Transfected or untransfected COS-7 cells were washed in PBS, harvested by scraping and resuspended in 1 ml homogenization buffer (220 mM mannitol, 68 mM sucrose, 20 mM HEPES pH 7.4, 80 mM KCl, 0.5 mM EGTA, 2 mM MgOAc, protease inhibitors). Cells were broken in a cell cracker with diameter 8.02 mm for ten passages using an 8.002-mm ball and the nuclear fraction was removed by centrifuging at 800 g for 10 minutes at 4°C. The post-nuclear supernatant was centrifuged at 9300 g for 20 minutes at 4°C to pellet mitochondria. Mitochondria were resuspended in homogenization buffer and repelleted to remove contaminants. The post-mitochondrial supernatant was centrifuged at 200,000 g to clear the cytosolic fraction of light membranes. Total extracts, mitochondrial and cytosolic fractions were normalized for protein content, and 50-200 μg loaded for electrophoresis and western blotting.

Immunoprecipitation and western blotting

Six 10-cm dishes of COS-7 cells stably expressing pShag Magic2 vector or vector containing shRNAi, were washed and harvested by scraping at 4°C with 1 ml per 10-cm dish of the following buffer: 10 mM HEPES pH 7.4, 50 mM NaCl, 0.5 mM EDTA, 2 mM MgCl2, protease inhibitors, 20 mM NEM. After a brief sonication to break the cells, 1% Triton X-100 was added and lysates were incubated at 4°C for 20 minutes, followed by centrifugation at 70,000 rpm in a TLA100.4 rotor, for 45 minutes at 4°C. Supernatants were separated, assayed for total protein concentration, and 1 mg per IP point (1 mg/ml conc.) was incubated overnight with either 2.5 μg G-protein agarose-coupled mouse monoclonal anti-DRP1 or 5 μg G-protein agarose-coupled mouse IgG (Sigma, Oakville, ON), rotating at 4°C. After washing three times, the beads were mixed with loading sample buffer, separated on 5-20% acrylamide gels, and blotted with the indicated antibodies. For the immunoprecipitation with monoclonal anti-DRP1 on lysates from COS-7 stably expressing vector or vector containing shRNAi and transiently oxerexpressing SUMO-FLAG, conditions were the same as before.

Fluorescence microscopy and spectrophotometry

Live or fixed cells transfected with fluorescent constructs were imaged on Olympus IX70 microscope with a 100× objective U Plan Apochromat, NA 1.35-0.50 objective, exited at 514 nm (YFP), 434 nm (CFP), with the Polychrome IV monochrometer (TillPhotonics, Grafelfing, Germany). The emitted light was filtered through a Till double CFP/YFP pass filter. Images acquired were saved as .tif files and overlaid in Adobe Photoshop for image assembly as described (Harder et al., 2004). Confocal images and videos were obtained with a 100× objective NA1.4 on an Olympus IX81 inverted microscope with appropriate lasers (515 nm argon for Alexa Fluor-515, 548 nm argon for oxidized hET, and 633 nm for mitofluorRed 633), using Olympus FV1000 confocal scanning microscope (Olympus Canada, Markham, ON). The images shown in Fig. 4E were rendered through a Gaussian filter (binned 4 pixels to 1) in Adobe Photoshop to smooth the cytosolic SUMOylation signal within the cells. Otherwise, no further processing was used in images shown in this study. For hET quantification in fluorimeter, 10 μM hET was added to harvested and suspended cells in culture medium, and after a 40-minute incubation at 37°C, 50 μl aliquots were placed in a quartz cuvette and scanned for emission fluorescence of reduced (420 nm, excitation at 355 nm) and oxidated hET emission at 605 nm (excitation at 518 nm), in a Quanta Master 6000SE (Photon Technology International, London, ON). The background fluorescence of hET was subtracted, and each value was also normalized for total cell mass (determined by protein concentration using the DC protein assay, Bio-Rad).

Electron microscopy studies

Cells were fixed in 1.6% glutaraldehyde and processed as previously described (Germain et al., 2005). Images were obtained using a JEOL1230 TEM at 60kV using a Hammamatsu Digital 2k × 2k camera (JEOL, Montreal PQ), saved as .tif files and assembled using Adobe Photoshop.

Quantification of mitochondrial fusion

For the mitochondrial fusion assay, cells transfected with matrix-targeted PAGFP were put in a chamber with 20 mM HEPES in DMEM media at 37 degrees to obtain an optimal environment. Confocal images were taken on an Olympus IX81 inverted microscope with a 100× objective NA 1.4 (Olympus Canada, Mississauga ON). Cells were pre-screened using a standard mercury lamp and a double CFP/YFP pass filter. For image acquisition, a 405 nm solid UV laser and the 488 nm laser line of an argon-krypton laser was used. Before activation, a stack of the cell was taken at 488 nm excitation and detection at 515-560 nm. This image was used for background correction of already activated PAGFP fluorophores. Then, a region of interest including about one-quarter of the cell was set and illuminated with the 405 nm laser. Directly afterwards and every 10 minutes (for 40 minutes), the whole stack of the cell was imaged using the 488 nm laser line to follow the distribution of the photoactivated fluorophore. After 40 minutes, the stack was completely activated with the 405 nm laser and imaged activated signal collected with the 488 nm laser. This stack was used as the reference for the total amount of mitochondria within the cell. For all image recordings, the following microscope settings were used: excitation beam splitter DD 405/488, pinhole diameter 1 Airy unit, 512×512 pixels, zoom 1, pixel size 0.248×0.248×1.0 μm.

For data analysis, the 3D-image stacks were cropped manually to center a single cell in the image. The colocalization tool within the Olympus Fluoview Software was used to create a digital mask around the mitochondria containing the activated fluorophore. Therefore, the threshold was set manually in each stack. Pixels that were above this threshold result in a white signal, all other pixels in a black signal. The sum of white pixels result in the mask surrounding the part of mitochondria with activated PAGFP. After fusion, the amount of mitochondria containing the activated fluorophore and thereby the size of the mask increases. The number of pixels within the mask divided by the number of pixels of all mitochondria after the complete activation results in the percentage of fused mitochondria at different time points.


This work was supported by the Canadian Institutes of Heath Research operating grant no. 68833 and a CIHR New Investigators Award to H.M.M. We are grateful to John Presley (McGill University) for sharing the PAGFP plasmid with us, to Ruth Slack (Ottawa Health Research Institute) for providing the adenovirus expressing DRP1(K38E)CFP and to Joerg Bewersdorf (Jackson Laboratories, Bar Harbor, MI) for help with the quantification of the PAGFP fusion assay. We thank Gordon Shore (McGill University), and members of the laboratory for their insightful comments on the manuscript. Finally, a special thanks to Ian and Jan Craig (Ottawa, ON) for their significant contribution towards the purchase of the confocal microscope used in this study.


  • Accepted January 23, 2007.


View Abstract