SIRT4 interacts with OPA1 and regulates mitochondrial quality control and mitophagy

The stress-responsive mitochondrial sirtuin SIRT4 controls cellular energy metabolism in a NAD+-dependent manner and is implicated in cellular senescence and aging. Here we reveal a novel function of SIRT4 in mitochondrial morphology/quality control and regulation of mitophagy. We report that moderate overexpression of SIRT4, but not its enzymatically inactive mutant H161Y, sensitized cells to mitochondrial stress. CCCP-triggered dissipation of the mitochondrial membrane potential resulted in increased mitochondrial ROS levels and autophagic flux, but surprisingly led to increased mitochondrial mass and decreased Parkin-regulated mitophagy. The anti-respiratory effect of elevated SIRT4 was accompanied by increased levels of the inner-membrane bound long form of the GTPase OPA1 (L-OPA1) that promotes mitochondrial fusion and thereby counteracts fission and mitophagy. Consistent with this, upregulation of endogenous SIRT4 expression in fibroblast models of senescence either by transfection with miR-15b inhibitors or by ionizing radiation increased L-OPA1 levels and mitochondrial fusion in a SIRT4-dependent manner. We further demonstrate that SIRT4 interacts physically with OPA1 in co-immunoprecipitation experiments. Overall, we propose that the SIRT4-OPA1 axis is causally linked to mitochondrial dysfunction and altered mitochondrial dynamics that translates into aging-associated decreased mitophagy based on an unbalanced mitochondrial fusion/fission cycle.


INTRODUCTION
Aging is defined as a time-dependent loss of physiological integrity and organ function and is characterized by key hallmarks, among them genomic instability, stem cell exhaustion, mitochondrial dysfunction, and cellular senescence [1]. Senescence is either elicited endogenously, e.g. via telomere shortening in the case of replicative senescence [2][3][4], or triggered by extrinsic noxae such as UV radiation (stress-induced senescence) that is responsible for photo-aging of the skin [5][6][7]. Senescent cells occur and accumulate in-vivo, and clearance of senescent cells ameliorates aging-associated pathologies thereby extending healthy lifespan [8,9]. The hallmarks of aging are functionally interconnected. For instance, upregulation of the cyclin-dependent kinase (CDK) and cell cycle inhibitor p21 WAF , a p53 target gene and key effector of stress-induced cellular senescence [10,11], leads to a reduced capacity of stem cell dependent tissue regeneration [12]. Moreover, p21 WAF maintains a vicious circle in γ-irradiated fibroblasts by re-enforcing senescence signaling via the generation of more mitochondrial reactive oxygen species (mtROS) and thus more ROS induced DNA damage [13]. Overall, the pro-aging effects of senescence and the SASP (senescence associated secretory phenotype) seem to be at least in part dependent on the presence of mitochondria [14,15].
The morphology and dynamics of mitochondria underlie a stringent quality control via fusion-fission cycles which are primarily mediated by the key large GTPases OPA1 (optic atrophy 1), mitofusins (MFN1 and 2), and DRP1 (dynamin-related protein 1). Agingassociated changes in the expression levels of fusion proteins (OPA1, MFN1/2) or reduction of the fission factors DRP1 and FIS1 (a recruitment factor for DRP1 in the mitochondrial outer membrane) have been observed in senescent endothelial or mesenchymal stem cells, both resulting in increased mitochondrial elongation and fusion [16,17]. As part of mitochondrial quality control, dysfunctional/depolarized, fragmented mitochondria are continuously eliminated by mitophagy, a form of selective autophagy [18][19][20]. However, there is accumulating evidence for a decline of autophagy and mitophagy during aging [18,[21][22][23][24][25][26][27] thereby impairing mitochondrial quality and function. For instance, mitophagy was greatly reduced in aged muscle stem cells (satellite cells) resulting in accumulation of dysfunctional mitochondria, increased ROS generation, and a senescent phenotype [28,29].
Mitochondrial sirtuins (mtSIRT) comprise three members, SIRT3, SIRT4, and SIRT5 which are all involved in regulating energy metabolism and metabolic homeostasis [30][31][32][33] and whose expression is regulated by transcriptional and post-transcriptional mechanisms, including miRNAs [34,35]. Among the mtSIRTs, SIRT4 expression is upregulated during senescence triggered by different stimuli as well as in-vivo in photoaged human skin [35]. Upregulation of SIRT4 expression can per se induce senescence as shown in trophoblast stem cells [36]. Only limited information exists about the role of mtSIRTs in the regulation of mitochondrial morphology/dynamics and quality control mechanisms. SIRT3, the major deacetylase of metabolic targets in mitochondria, promotes mitochondrial networking and function by activating the fusion factor OPA1 [37]. In the case of SIRT5, an enzyme with demalonylase, deglutarylase, and desuccinylase activities [38,39] that regulates ammonia detoxification, mitochondrial size was increased and mitophagy decreased upon SIRT5 overexpression [40]. Lastly, overexpression of SIRT4 was linked to the regulation of mitochondrial dynamics via inhibition of ERK-mediated phosphorylation of the pro-fission factor DRP1, therefore inhibiting its activity and hence mitochondrial fission [41]. Similar to SIRT5, several enzymatic activities have been described for SIRT4, including ADP-ribosylation of glutamate dehydrogenase (GDH) [42], lipoamidase mediated targeting of the pyruvate dehydrogenase (PDH) complex [43], and lysine deacetylation in the control of leucine metabolism [44].
We demonstrated recently that up-regulation of endogenous SIRT4 expression inhibits the mitochondrial membrane potential (ΔΨm), increases mtROS level, and alters mitochondrial morphology towards aggregation [35]. Here, we further addressed the molecular basis for this phenotype in cells stably expressing wild-type and mutant forms of SIRT4 at low levels and characterized the role of SIRT4 in mitochondrial morphology/quality control and mitophagy under basal and mitochondrial stress conditions.

SIRT4 promotes stress induced mitochondrial ROS production
In the present study, we addressed the role of SIRT4, a miR-15b repressed and stress-inducible senescenceassociated mitochondrial sirtuin [35,45], in the regulation of mitochondrial dynamics and quality control. We generated HEK293 cell lines stably expressing C-terminal eGFP fusion proteins of SIRT4, including its catalytically inactive version (H161Y) or SIRT4(Δ28N) lacking the N-terminal mitochondrial targeting signal [46]. Expression and subcellular Figure 1. Catalytically active SIRT4-eGFP increases mitochondrial ROS production after CCCP-induced mitochondrial uncoupling. HEK293 cell lines stably expressing eGFP, SIRT4-eGFP, SIRT4(H161Y)-eGFP, or SIRT4(Δ28N)-eGFP were treated with CCCP (10 nM) for two hours either with or without pretreatment with MitoQ (100 nM; 16 hours). Thereafter cells were stained with the cell-permeant dyes MitoTracker® Deep Red and MitoSOX™ Red followed by flow cytometric analysis. Shown are mean ± s.d. values from nine experiments. To evaluate statistical significance (comparison of eGFP vs. wild-type or mutant SIRT4 or comparisons indicated by brackets) two-way ANOVA followedup by Tukey's test were performed (**p<0.01; ***p<0.001). The effect of MitoQ treatment was evaluated using a paired t-test ( # p<0.05).
localization of SIRT4-eGFP and its mutants were confirmed by confocal microscopy, immunoblotting, and flow cytometry analysis (Suppl. Fig. 1). We next subjected these cell lines to CCCP (Carbonylcyanide mchlorophenylhydrazone) treatment and hence to dissipation of the mitochondrial membrane potential (ΔΨm). As depicted in Fig. 1, cells expressing SIRT4-eGFP displayed a significant increase in stress-induced mitochondrial ROS (mtROS) levels which were twotimes higher as compared to eGFP-expressing control cells and which could be significantly inhibited by cotreatment with the mitochondria-specific anti-oxidant mitoQ [47,48]. In contrast, mtROS response was significantly weaker in cells expressing SIRT4(H161Y)-eGFP or SIRT4(Δ28N)-eGFP (Fig. 1). Here, co-treatment with mitoQ did not show an effect. Thus, consistent with the mtROS inducing and Ψmdecreasing role of SIRT4 in primary human dermal fibroblasts [35], SIRT4 expression promotes stressinduced mtROS production also in HEK293 cells via its enzymatic activity.

SIRT4 expression promotes stress triggered autophagic flux, but decreases Parkin associated mitophagy considering the increased mitochondrial content
Mitochondria with sustained depolarization are removed by mitophagy, a selective degradation mechanism for damaged mitochondria, ensuring mitochondrial quality control [22,49]. To assess the role of SIRT4 in this process we determined the autophagic flux in HEK293 cell lines stably expressing SIRT4-eGFP or its mutants by analyzing the levels of the autophagy marker LC3B-II after treatment with rapamycin (mTORC1 inhibitor and general autophagy inducer) or CCCP (a mitochondrial uncoupler and selective mitophagy inducer). As depicted in Fig. 2, LC3B-II levels were approximately two-fold-increased in CCCP-treated HEK293 SIRT4-eGFP cells, but did not change significantly in cells expressing SIRT4(H161Y)-eGFP or SIRT4(Δ28N)-eGFP. We did not detect significant changes in the autophagic flux of any of these cell lines following rapamycin treatment. Consistent with the immunoblot analysis in Fig. 2, the number of LC3B dots per cell was also approximately two-fold higher in CCCP treated SIRT4-eGFP cells vs. eGFP controls as determined by confocal imaging analysis (data not shown). Given this potentially specific effect of SIRT4 on mitophagy, we next analyzed the translocation of the E3 ligase Parkin to damaged mitochondria, a process that promotes selective recruitment of mitochondria to the autophagic machinery [18]. HEK293 cells expressing SIRT4-eGFP (or its mutants) were transfected with a cDNA coding for mCherry-Parkin, subjected to treatment with CCCP and Bafilomycin A1 (BafA1), and analyzed by confocal microscopic imaging (Fig. 3A). Translocation of mCherry-Parkin to mitochondria was observed in all instances as expected after dissipation of the membrane potential. Yet, only for SIRT4-eGFP expressing cells the mitochondrial mass was increased when compared to the eGFP control (Fig. 3B). Taking this into account, we observed that HEK293 cells expressing SIRT4-eGFP, but not the catalytically inactive SIRT4(H161Y)-eGFP or the extra-mitochondrially localized SIRT4(Δ28N)-eGFP mutants, displayed a relative decrease in mCherry-Parkin/MTC02 double-positive dots by approximately 40% when standardized against the mitochondrial content (i.e., total MTC02 signal) (Fig. 3C). Taken together, these data indicate that under mitochondrial stress conditions SIRT4 despite promoting mtROS production leads to a net decrease in mitophagy and therefore subsequent decreased removal of dysfunctional mitochondria. (A) HEK293 cell lines stably expressing eGFP, SIRT4-eGFP, SIRT4(H161Y)-eGFP, or SIRT4(Δ28N)-eGFP were treated with DMSO (control), rapamycin (100 nM), or CCCP (10 µM) for two hours. In addition, during the second hour cells were either untreated (-) or co-treated (+) with BafA1 (Bafilomycin A; 100 nM) that stalls autophagic flux via inhibition of the fusion between lysosomes and autophagosomes. A representative experiment is depicted in which LC3B-I and LC3B-II levels were analyzed by immunoblotting. (B) LC3B-II signals (cotreatment with BafA1) were compared to the protein levels of β-Actin/ACTB as loading control using ImageJ based quantification. Data shown are mean ± s.d. values from four to seven experiments. To evaluate statistical significance (treatment vs. DMSO) two-way ANOVA followed by Tukey's test was performed (**p<0.01).

SIRT4 expression inhibits mitochondrial respiration already under basal conditions
The findings above raised the question whether ectopic SIRT4 expression predisposes cells to mitochondrial stress by impacting already on basal mitochondrial functions such as oxidative phosphorylation. Conflicting results have been described regarding the impact of increased SIRT4 levels on mitochondrial oxygen consumption [50,51]. We therefore measured the oxygen flow in HEK293 cells expressing SIRT4-eGFP (or its mutants) under basal conditions and upon inhibition of specific respiratory complexes of the electron transport chain (ETC) [52]. As depicted in Fig.   4A and B, HEK293 SIRT4-eGFP cells displayed a significantly reduced oxygen flow both under basal conditions and at maximum respiration, the latter mirroring a reduced electron transport system capacity (ETS) upon complete mitochondrial uncoupling of the electron transport system using CCCP. Surprisingly, cells expressing SIRT4(Δ28N)-eGFP, which cannot localize to mitochondria, showed a significantly increased O 2 flow both at basal and maximum respiration (Fig. 4C). This finding suggests a dominantnegative function of SIRT4(Δ28N) that could be caused by inhibition and/or trapping of endogenous SIRT4 in the cytosol. This interpretation is consistent with an increased oxygen flux and improved mitochondrial function upon SIRT4 depletion [51]. To evaluate statistical significance two-way ANOVA followed by Tukey's test was performed (*p<0.05).

SIRT4 expression leads to a higher degree of elongated/fused mitochondria associated with higher levels of the mitochondrial fusion regulator L-OPA1
Mitochondria, dependent on their (dys)functional status/depolarization, undergo either increased mitochondrial fusion or fission, the latter of which leads to mitophagic clearance upon excessive damage [20,27]. To further analyze the impact of SIRT4 on mitochondrial quality control we analyzed the mitochondrial network, i.e. mass and structure, in HEK293 cells expressing SIRT4-eGFP (or its mutants). We employed confocal microscopy using staining with the anti-mitochondria antibody MTC02 and subsequent ImageJ based analysis. As depicted in Fig. 5A, cells expressing SIRT4-eGFP underwent an increased mitochondrial aggregation already under basal (untreated) conditions. This aggregation was clearly less pronounced in GFP control cells or cells expressing the SIRT4 mutants. These findings were corroborated by flow cytometry where HEK293 SIRT4-eGFP cells, but not cells expressing SIRT4(H161Y)-eGFP or SIRT4(Δ28N)-eGFP, depicted significantly increased MitoTracker® staining intensity and hence an increased mitochondrial mass (Suppl. Fig. 2A and 2B). Further quantitative analysis revealed that the mean length of fused mitochondrial tubes was significantly increased only in SIRT4-eGFP expressing cells ( Fig. 5B and Suppl. Fig. 3; Suppl. Movies 1 to 4). To corroborate this finding at the molecular level we analyzed the proteolytic processing of the large GTPase and key mitochondrial fusion/fission regulator OPA1 [53,54].  [52,64]. Oxygen consumption was measured under basal and stressed conditions (leak: treatment with oligomycin A, 2 µg/ml; ETS: mitochondrial uncoupling of the electron transport system using CCCP, 400-500 nM; residual oxygen consumption, ROX: treatment with 500 nM rotenone and 2.5 µM antimycin A) as compared to control cells expressing eGFP. To evaluate statistical significance twoway ANOVA followed by Tukey's test was performed [**p<0.01, ***p<0.001; SIRT4-eGFP: n=5; SIRT4(H161Y)-eGFP, n=4; SIRT4(Δ28N)-eGFP: n=4].
Indeed, in HEK293 SIRT4-eGFP cells, the ratio of the proteolytically processed forms of OPA1 was significantly shifted from S-OPA1 (associated with mitochondrial fission and mitophagy) towards L-OPA1 (associated with mitochondrial fusion) ( Fig. 6A and  6B). Taken together, these data suggest that SIRT4-eGFP expression inhibits mitochondrial respiration that in terms of mitochondrial quality control translates into increased mitochondrial fusion, the latter possibly resulting in reduced basal mitophagy.

SIRT4 interacts with the mitochondrial fusion regulator L-OPA1
To get further insight into the molecular mechanism of how SIRT4 regulates basal mitochondrial quality control we performed co-immunoprecipitation and mass spectrometry to identify interaction partners of SIRT4 in HEK293 cells. The mitochondrial fusion regulator OPA1 was identified to interact specifically with SIRT4-eGFP. Next, we verified this interaction by Western blot analysis after co-immunoprecipitation both in untreated and in CCCP-stressed cells using single-domain-anti-GFP antibodies (Fig. 7A). Here, both forms of OPA1, L-OPA1 and S-OPA1, coimmunoprecipitated with SIRT4-eGFP as well as with its mutants. However, we observed that the amount of L-OPA1 co-immunoprecipitating with SIRT4(H161Y)-eGFP was significantly lower as compared to wild-type SIRT4-eGFP when standardized to the total protein input (Fig. 7B). The (direct or indirect) interaction between SIRT4-eGFP and OPA1 could be confirmed via reverse co-immunoprecipitation using a rabbit anti-OPA1 antibody [55] (Suppl. Fig. 4). Taken together, SIRT4 may interact in an enzyme activity-dependent manner with L-OPA1 via a currently unknown mechanism and thereby regulate mitochondrial fusion.

Upregulation of endogenous SIRT4 expression in fibroblast models of cellular senescence increases L-OPA1 levels and mitochondrial fusion in a SIRT4dependent manner
Altered mitochondrial dynamics with decreased autophagy or mitophagy are hallmarks to cellular aging [22-24, 27, 56]. We therefore expanded our analyses to primary human dermal fibroblasts to confirm the role of the SIRT4 -L-OPA1 axis in mitochondrial morphology and dynamics in a cellular model more relevant to aging. To upregulate the expression of SIRT4 fibroblasts were either transfected with miR-15b inhibitors or subjected to γ-irradiation (γIR; single dose of 20 Gy) [35]. Under both conditions fibroblasts showed an anti-proliferative response [13,35], as e.g. measured by decreased BrdU incorporation (Suppl.   SIRT4-eGFP, SIRT4(H161Y)-eGFP, or SIRT4(Δ28N)-eGFP were either untreated or treated with CCCP (10 µM, 2h) and thereafter subjected to OPA1 coimmunoprecipitation (IP) analysis using sepharose beads coupled anti-GFP singledomain-antibodies (nanobodies). Total cell lysates were loaded as input control (5%). CCCP treatment caused a complete proteolytic processing of L-OPA1 to S-OPA1. TACC3 was detected using specific antibodies and served as a representative negative co-immunoprecipitation control. (B) SIRT4 enzymatic activity is required for efficient interaction of SIRT4 with L-OPA1. The amount of L-OPA1 co-immunoprecipitated with SIRT4-eGFP, SIRT4(H161Y)-eGFP, or SIRT4(Δ28N)-eGFP was determined in relation to the protein input and subjected to ImageJ-based densitometric analysis. To evaluate statistical significance, twoway ANOVA followed by Tukey's tests was performed (*p<0.05; n=4). Figure 8. Aging-associated SIRT4 upregulation leads to a shifted L-OPA1 to S-OPA1 ratio in two different fibroblast senescence models. Primary human dermal fibroblasts were either transfected with miR-15b inhibitors (or control oligonucleotides) or subjected to γ−irradiation (γIR; 20 Gy) (both in the presence or absence of siRNA duplexes against SIRT4) [35] followed by analysis of OPA1-L and OPA1-S expression by immunoblotting after four days (A) As a control for complete proteolytic processing of L-OPA1 to S-OPA1, fibroblasts were treated with CCCP (10 µM) for two hours. The ratio between the expression levels of L-OPA1 and S-OPA1 was determined by ImageJ-based densitometric analysis in miR-15b inhibitor transfected fibroblasts (B) and cells subjected to γIR (C). To evaluate statistical significance (compared to control), two-way ANOVA followed by Tukey's test was performed (*p<0.05; n=4). mitochondrial profile/morphology were analyzed in the presence or absence of siRNA duplexes against SIRT4. Consistent with the findings in SIRT4-eGFP expressing HEK293 cells, in both fibroblast senescence models the ratio of L-OPA1 to S-OPA1 was significantly upregulated, both in a SIRT4-dependent manner (Fig. 8A and 8B) and consistent with corresponding changes in the mitochondrial profiles as visualized by confocal microscopy and quantitative ImageJ based analysis. Increased mitochondrial mass and significantly elevated mitochondrial fusion (indicated by an increased mean length of fused mitochondrial tubes) were observed upon SIRT4 upregulation through miR-15b inhibition (Fig. 9, Suppl. Fig. 6, Suppl. Movies 5 to 7) or γirradiation (Fig. 10, Suppl. Fig. 7, Suppl. Movies 8 to 10).

DISCUSSION
This study provides novel insight into the role of the mitochondrially localized and aging associated sirtuin SIRT4 in basal and stress-induced regulation of mitochondrial morphology and turnover mechanisms. Our data show that SIRT4 promotes mtROS generation triggered by dissipation of the membrane potential across the inner membrane. Although this led to an increase in autophagic flux, overall Parkin-linked mitophagy was apparently decreased due to an increased mitochondrial content. Mechanistically, we propose that SIRT4 functions via interacting with L-OPA1 representing a novel regulatory axis in mitochondrial quality control, i.e. promotion of   mitochondrial fusion and therefore likely reduction of mitophagy (summarized in Fig. 11).
Mitochondria represent a major source of ROS [57] that dependent on its concentration negatively impacts on mitochondrial function eventually resulting in autophagic clearance of damaged mitochondria. MtROS are typically increased in cells overexpressing enzymatically active SIRT4 [35] (Fig. 1) together with a reduced oxygen consumption under basal and mitochondrial stress conditions (Fig. 4) [51]. In contrast, depletion of SIRT4 in HEK293 cells results in increased oxygen flux and improved mitochondrial function [51]. Consistent with this, we observed an increased O 2 flow in HEK293 SIRT4(Δ28N)-eGFP cells both at basal and maximum respiration (Fig. 4), possibly reflecting a dominant-negative function of SIRT4(Δ28N). The latter may bind to wild-type SIRT4 and prevent it from mitochondrial translocation (that is dependent on the first 28 a.a.), given that recombinant SIRT4 as well as SIRT4 expressed in HEK293 cells forms homo-oligomeric to multimeric complexes in gel filtration analysis (own unpublished results). The molecular basis for the inhibitory and promoting effects of SIRT4 on mitochondrial respiration and mtROS generation, respectively, is likely connected to the functional inhibition of one or more SIRT4 targets in the inner mitochondrial membrane and matrix. For instance, SIRT4 negatively regulates key steps in anaplerosis of the citric acid cycle via different NAD +dependent enzymatic reactions, i.e. ADP-ribosylation and inhibition of glutamate dehydrogenase (GDH) [42] or lipoamidase-mediated targeting and inhibition of the pyruvate dehydrogenase (PDH) complex [43]. Moreover, mass spectrometric characterization of the protein interaction network of SIRT4 identified components of mitochondrial complexes involved in oxidative phosphorylation, namely complex I, IV, and V [58], with complex I representing a major source of mtROS [59].
Besides their functions in energy metabolism and mitochondrial respiratory chain complexes [58], all three mtSIRTs participate in the regulation of mitochondrial morphology/dynamics and hence mitochondrial quality control through interaction with key fusion/fission regulators. SIRT3 promotes mitochondrial networking and functions by deacetylation and therefore activation of the fusion factor OPA1 thereby increasing its GTPase activity [37]. In the case of SIRT4, we now observed increased mitochondrial fusion (Fig. 5, 9, and 10) concomitant with an elevated mitochondrial content in cells overexpressing wild-type SIRT4, but not the catalytically inactive mutant H161Y. These findings were further corroborated by the novel physical link between SIRT4 and OPA1 that was evident in mass spectrometry analysis of SIRT4-eGFP binding partners (Lang et al., manuscript in preparation) and GFP-singledomain antibody (Fig. 7) as well as anti-OPA1 antibody (Suppl. Fig. 4) based co-immunoprecipitation experiments. The molecular mechanism leading to SIRT4 triggered increase in the L-OPA1/S-OPA1 ratio as well the interaction of SIRT4 with L-OPA1 ( Fig. 7A and Suppl. Fig. 4) were dependent on the enzymatic activity of SIRT4 (Fig. 7B), and could involve the stabilization of L-OPA1 via (in)direct protein-protein interaction or protection from stress-induced and OMA1 mediated processing [60]. The elevated L-OPA1 levels in all cell models analyzed were associated with an increased mitochondrial fusion (Fig. 5, 9, and 10). Moreover, in line with our findings, a recent report linked SIRT4 to the regulation of mitochondrial dynamics via inhibition of ERK-mediated phosphorylation of the pro-fission factor DRP1, therefore inhibiting its activity and hence mitochondrial fission [41]. However, similar to the SIRT4-OPA1 interaction, the molecular mechanism by which SIRT4 impacts on the ERK-DRP1 axis is currently unclear and may require one or more of the known enzymatic activities of SIRT4 (ADP-ribosyltransferase, lipo-amidase, or lysine deacylase; [42][43][44]). Lastly, in the case of SIRT5, mitochondrial size was increased and mitophagy decreased upon SIRT5 overexpression, whereas the opposite effect was observed in SIRT5 silenced cells or upon treatment with the SIRT5 inhibitor MC3482 [40]. SIRT5 overexpression led to increased levels of the fusion factors OPA1 and MFN2, whereas at the same time the positive mitophagy regulators PINK2 and PARK1 were downregulated [40]. Consistent with this, SIRT5 has a protective role by preventing mitochondrial fragmentation during starvation, given that SIRT5 deficiency increases mitochondrial DRP1 levels and mitophagy [61]. Taken together, all three mtSIRTs seem to promote mitochondrial fusion and/or inhibit fission, and thus might attenuate mitophagic clearance of dysfunctional mitochondria. Of note, Ho et al. observed that higher SIRT4 levels trigger increased retrograde AMPK/PGC1α signaling from mitochondria to the nucleus [51]. However, this mechanism may rather mediate an increase in mitochondrial biogenesis to compensate for the accumulation of dysfunctional mitochondria which eventually did not undergo proper mitophagic clearance.
In summary, our findings are consistent with an inverse correlation between increased levels of the metabolic tumor suppressor SIRT4 in cellular senescence and aging [35,45] and cumulative evidence for declining autophagy with advanced age [18,22,[24][25][26][27]56]. An increased content of mitochondria, which accumulate in a dysfunctional status, is characteristic for senescent cells [1,13]. Thus, we propose that stress/senescence induced SIRT4 functions via L-OPA1 as a novel determinant in the down-regulation of mitophagy during aging by shifting the mitochondrial fusion/fission cycle towards fusion.

Measurement of mitochondrial reactive oxygen species (mtROS)
Cells were stained with cell-permeable dyes MitoTracker® Deep Red FM (100 µM; M22426) and MitoSOX™ Red (5 µM; M36008), both obtained from Thermo Fisher Scientific. The mitochondria-specific antioxidant MitoQ (10-(6'-ubiquinonyl) decyl triphenylphosphonium bromide) [47], kindly provided by Dr. Mike Murphy (Cambridge), was used to treat cells at a concentration of 100 nM for two days. Treatment of cells with the mitochondrial uncoupler CCCP (Carbonylcyanide m-chlorophenylhydrazone; 10 µM) occurred for two hours followed by harvesting and staining living cells in 1x PBS for 20 min. Analyses were performed with a BD FACSCanto™ II system (BD Biosciences) employing Flowing Software 2.5.1 (University of Turku, Finland).

Immunofluorescence and confocal laser scanning microscopy to analyze mitochondrial morphology and mitochondrial fusion/tube formation
To obtain high resolution pictures for analysis of mitochondrial morphology HEK293 cells and primary human dermal fibroblasts were co-stained with antibodies against MTC02 (Abcam, ab3298; 1:500) and α-Tubulin (Abcam, ab52866, 1:500) overnight at 4°C. To increase detection of SIRT4-eGFP fusion proteins (in the case of stably transfected HEK293 cells) primary antibodies against GFP (Nacalai Tesque, Inc., GF090R, 1:1000) were employed. Secondary antibodies were Alexa Fluor 488-conjugated goat anti-rat, Alexa Fluor 546-conjugated goat anti-mouse IgG, and Alexa Fluor 633-conjugated goat anti-rabbit IgG. Acquisitions were performed with the UltraVIEW spinning disk confocal microscope (Perkin Elmer) with emission wavelengths of 468 nm, 488 nm, 543 nm, and 633 nm and the Volocity 6.3 software (Perkin Elmer). Pictures were further analyzed using ImageJ software v1.49k employing specific macros for the analysis of mitochondrial mass and mitochondrial tube formation/length analysis (Suppl. Material & Methods).

Transfection of oligonucleotides
Primary human dermal fibroblasts were transfected with miR-15b inhibitors (or control oligonucleotides) with or without SIRT4 siRNA duplexes to modulate SIRT4 expression essentially as described [35]. Transfection efficiency was controlled by co-transfection of siGLO Red transfection indicators (GE Dharmacon).

Ionizing γ-irradiation (γIR)
Primary human dermal fibroblasts were transfected with siRNA duplexes against SIRT4 essentially as described [35]. Following 24h, cells were exposed to γIR (one dose of 20 Gy; 175 kV and 15 mA) using a Gulmay RS225 X-ray system from X-Strahl (Camberley, UK) and were cultured for another three days before analysis.

Statistical analysis
Data are presented as mean ± s.d. Multiple comparisons were analyzed by one-way or two-way analysis of variance (ANOVA) employing the GraphPad Prism software.

Measurement of BrdU incorporation
Primary human dermal fibroblasts were transfected with miR-15b-inhibitors with co-transfection of siRNA duplexes against SIRT4 or control siRNA duplexes.
Following two days, cells were incubated for another two days in culture media containing 10 µM BrdU. Staining of fixed cells and further FACS analysis were performed as described in the BrdU staining kit (556029, BD Biosciences). DNA was stained for 10 min with 7-AAD (10 µl/10 6 cells; 00-6993-50, eBioscience™).

Quantification of mitochondrial mass by MitoTracker® FACS
HEK293 cells were stained with the cell-permeable dye MitoTracker® Deep Red FM (100 µM; Thermo Fisher Scientific) for 20 min followed by flow cytometry using a BD FACSCanto™ II system (BD Biosciences) and Flowing Software 2.5.1 (University of Turku, Finland) or FlowJo V10 (FlowJo, LLC).