MFN1 deacetylation activates adaptive mitochondrial fusion and protects metabolically challenged mitochondria

The ability to use a diverse set of fuel sources in response to changing nutrient environments is crucial for cellular and organismal survival. Under normal conditions, glucose metabolized through glycolysis is the preferred and main source of energy in animals. Upon food deprivation (fasting) and reduced glucose supply, cells can switch to use fatty acids, as well as amino acids, for ATP production. Unlike glycolysis, fatty acids and amino acids must be metabolized in mitochondria through the tricarboxylic acid cycle (TCA) and oxidative phosphorylation to generate ATP. This adaptive metabolic shift therefore increases the functional demands on mitochondria in order to meet the energy need. An increase in oxidative phosphorylation, however, also elevates reactive oxygen species (ROS) production due to leakages in the electron transport chain (Lesnefsky and Hoppel, 2003; Sanz et al., 2006). How the metabolically challenged mitochondria manage ROS production to prevent oxidative damage remains to be fully characterized.

Mitochondria undergo active fusion and fission in response to various stress conditions (Okamoto and Shaw, 2005; Detmer and Chan, 2007). The importance of these highly orchestrated morphological changes, in most cases, remains a mystery. In response to severe nutrient starvation conditions, formation of an inter-connected mitochondrial network has been proposed to prevent excessive loss of mitochondria by creating a physical barrier to autophagy (Gomes et al., 2011; Rambold et al., 2011). However, it is unlikely that the sole function of stress-induced mitochondrial fusion is to guard against autophagy. Additionally, it is not known whether the dynamic morphological remodeling plays an active regulatory role in mitochondrial function. The answers to these questions are crucial to delineate the physiological role of mitochondrial dynamics.

In mammals, the dynamin-related GTPases MFN1, MFN2 and OPA1 coordinately promote fusion of mitochondrial outer and inner membranes, respectively, whereas DRP1 (also known as DNM1L) stimulates fission (Detmer and Chan, 2007). Under extreme starvation, phosphorylation by protein kinase A (PKA) reduces DRP1 mitochondrial association and thereby increases mitochondrial connectivity (Gomes et al., 2011). The regulation of the pro-fusion proteins MFN1 and MFN2 is less well characterized. Lack of either MFN1 or MFN2 creates a fusion deficiency and results in mitochondria that are severely fragmented (Chen et al., 2003). In addition to this morphological aberration, accumulation of mitochondrial DNA (mtDNA) mutations has been observed in MFN1/MFN2 double-knockout cells (Chen et al., 2010). This finding indicates that mitochondrial fusion machinery is required for maintaining the integrity of the mitochondrial genome. Why fusion-deficient mitochondria are more prone to acquiring mtDNA mutations is not known.

The protein deacetylase, HDAC6, is a component of quality control autophagy. Unlike classically defined non-selective autophagy, quality control autophagy uses an ubiquitin-dependent mechanism to selectively dispose of protein aggregates and damaged organelles (Lee et al., 2010a; Lee et al., 2010b; Lee and Yao, 2010; Lee et al., 2013). In conjunction with the ubiquitin E3 ligase parkin, a causative gene in early-onset Parkinson's disease, HDAC6 facilitates the elimination of impaired mitochondria by mitophagy (Narendra et al., 2008; Lee et al., 2010a). The requirement of HDAC6 for efficient mitophagy suggests an important role for HDAC6 in mitochondrial quality control. However, HDAC6-knockout (KO) mice are grossly normal and without overt phenotypes commonly associated with mitochondrial defects (Gao et al., 2007; Zhang et al., 2008; Lee et al., 2010b). These findings raise the possibility that HDAC6 might be required for maintaining a healthy mitochondrial population only under stress conditions.

In this report, we provide evidence that glucose deprivation in cultured cells and fasting in mice induces mitochondrial fusion. This metabolic-stress-induced fusion failed in HDAC6-deficient cells and mice. We show that in response to glucose starvation, HDAC6 binds and deacetylates MFN1, leading to MFN1 activation and mitochondrial fusion. Importantly, the failure to undergo mitochondrial fusion did not lead to acute metabolic deficiency or loss of mitochondria by autophagy; rather it caused a buildup of oxidative stress leading to mitochondrial degeneration, a defect reversed by the expression of an acetylation-resistant MFN1 mutant. These findings indicate that mitochondrial fusion mediated by MFN1 and HDAC6 is a part of adaptive response that is crucial for protecting metabolically challenged mitochondria from excessive oxidative damage.

To test the hypothesis that HDAC6 is required for mitochondrial quality control under stress conditions, HDAC6 KO mouse embryonic fibroblasts (MEFs) and mice were challenged by glucose starvation or fasting to increase mitochondria-dependent energy production (Ogawa et al., 2003; Romanello et al., 2010). When wild-type MEFs were cultured in glucose-free medium supplemented with dialyzed serum, we found that mitochondria formed a more-connected network, indicating a net increase in fusion (Fig. 1A). Quantification confirmed that the percentage of cells with a hyperfused mitochondrial network increased by ∼2-fold upon glucose starvation (Fig. 1B). In contrast, mitochondria in HDAC6 KO MEFs did not undergo fusion in glucose-free medium; instead, they became visibly fragmented (Fig. 1A,B). To confirm this phenotype, we examined the mitochondrial response to glucose starvation in HDAC6 KO MEFs stably reconstituted with wild-type (+HDAC6wt) or catalytic-dead mutant HDAC6 (+HDAC6cd) (Gao et al., 2007). As shown in Fig. 1C, expression of wild-type, but not the catalytic inactive mutant, HDAC6, effectively restored mitochondrial network morphology upon glucose starvation. To provide a quantitative estimate of mitochondrial connectivity, we used fluorescence recovery after photobleaching (FRAP) in MEFs expressing a mitochondria-localized YFP (mito-YFP) (Szabadkai et al., 2004). As shown in Fig. 1D, although all cell lines showed comparable recovery of mito-YFP after photobleaching (mobile fraction) in complete medium, upon glucose starvation, mitochondrial connectivity in HDAC6 KO and KO+HDAC6cd MEFs was significantly reduced compared to that in HDAC6 KO+HDAC6wt MEFs. These results show that HDAC6 and its catalytic activity are required for promoting mitochondrial fusion induced by glucose starvation.

Under severe nutrient starvation (Hank's solution), DRP1 becomes phosphorylated and inhibited by PKA, leading to mitochondrial fusion (Gomes et al., 2011; Rambold et al., 2011). However, we found that the inhibitory phosphorylation of DRP1 S637 was not induced in wild-type or HDAC6 KO MEFs subject to glucose starvation, although it was clearly elevated by treatment with Hank's solution (Fig. 2A). These results indicate that HDAC6 regulates mitochondrial fusion independently of DRP1. We therefore investigated whether HDAC6 functionally interacts with the pro-fusion factor MFN1. As shown in Fig. 2B, although no detectable MFN1–HDAC6 complex was observed by co-immunoprecipitation assays in cells in complete medium (Fig. 2B, lane 1, +glucose), endogenous MFN1 became markedly associated with HDAC6 upon glucose starvation (Fig. 2B, lane 3). Notably, the formation of the MFN1–HDAC6 complex was accompanied by a marked reduction in MFN1 acetylation (Fig. 2B, AcK Panel, lane 3; supplementary material Fig. S1B), which was not observed in HDAC6 KO MEFs (Fig. 2B, lane 4; supplementary material Fig. S1B). These results suggest that HDAC6 binds to and promotes MFN1 deacetylation under glucose starvation conditions. Supporting this proposition, reintroduction of wild-type (Fig. 2C, lane 2), but not the catalytically dead mutant, HDAC6 (Fig. 2C, lane 3), restored MFN1 deacetylation in HDAC6 KO MEFs subject to glucose deprivation. Notably, the MFN1-releated mitofusin 2 (MFN2) is also subjected to acetylation and interacts with HDAC6; however, MFN2 acetylation is not affected by glucose starvation or HDAC6 (Fig. 2D). Collectively, these findings indicate that HDAC6 binds and deacetylates only MFN1 in response to glucose starvation.

Acetylation of a conserved lysine residue (K222) in the GTPase domain of MFN1 has previously been identified by mass spectrometry (Choudhary et al., 2009). To investigate whether K222 acetylation affects MFN1 activity, we assessed the ability of acetylation-mimicking (K222Q), and acetylation-resistant (K222R) MFN1 mutants to restore mitochondrial fusion in MFN1 KO MEFs. Of note, glutamine has been shown to functionally substitute for lysine acetylation (Ren and Gorovsky, 2001; Wang et al., 2014) and the MFN1-K222R mutant showed reduced acetylation (supplementary material Fig. S1C). As expected, expression of the wild-type MFN1 reestablished a network of mitochondria in MFN1 KO MEFs (Fig. 2E, top panels). The acetylation-resistant MFN1-K222R mutant was highly active in this complementation assay as well (Fig. 2E, middle panels). In contrast, the acetylation-mimicking MFN1 K222Q mutant failed to induce efficient mitochondrial fusion (Fig. 2E, bottom panels), suggesting that acetylated MFN1 is less active. Further supporting this conclusion, in HDAC6 KO MEFs, where MFN1 was more acetylated (Fig. 2B), the acetylation-resistant K222R mutant MFN1 was significantly more active than wild-type MFN1 in promoting mitochondrial fusion, whereas the K222Q mutant again showed reduced activity (Fig. 2F,G). Collectively, these results indicate that HDAC6-mediated deacetylation increases MFN1 activity.

Glucose starvation forces cells to utilize alternative fuel sources, including fatty acids, to produce ATP through oxidative phosphorylation in mitochondria. We next investigated whether the fusion status affects mitochondrial activity. HDAC6 KO MEFs showed lower basal oxygen consumption in normal medium, consistent with a recent report (Kamemura et al., 2012). However, in response to glucose starvation, oxygen consumption in HDAC6 KO MEFs eventually elevated to a level comparable to that in wild-type MEFs (Fig. 3A). Similar to wild-type MEFs, glucose-starved HDAC6 KO MEFs were also able to increase β-oxidation and utilize palmitic acid (Fig. 3B). Consistent with comparable oxygen consumption and β-oxidation, both cell types also generated similar ATP levels in glucose-free medium (Fig. 3C,D). These results indicate that mitochondria in HDAC6 KO cells are capable of meeting acute energetic demands in response to glucose starvation. Similarly, the fusion-deficient MFN1 KO MEFs were able to produce ATP normally under glucose starvation conditions (Fig. 3E,F). These findings indicate that mitochondrial fusion is not required for the elevation of mitochondrial activity in response to glucose starvation.

Increase in oxidative phosphorylation could potentially lead to ROS production due to leakages in the electron transport chain (Lesnefsky and Hoppel, 2003; Sanz et al., 2006). We therefore determined whether mitochondrial fusion affects mitochondrial ROS production. To this end, mitochondrial ROS levels were determined in wild-type and HDAC6 KO MEFs cultured in normal or glucose-free medium by MitoSox staining. As shown in Fig. 4A, wild-type MEFs produced similar levels of mitochondrial ROS under basal conditions and glucose starvation. In contrast, glucose starvation resulted in an increase in mitochondrial ROS in HDAC6 KO MEFs (Fig. 4A). Interestingly, MFN1 KO MEFs, which have severely fragmented mitochondria, showed increased mitochondrial ROS levels under basal conditions, which were further elevated upon glucose starvation (Fig. 4A). OPA1 KO MEFs, which are defective in mitochondrial inner membrane fusion, also displayed significant mitochondrial ROS accumulation after glucose starvation (Fig. 4A).

Importantly, the aberrant mitochondrial ROS production observed in HDAC6 KO MEFs could be significantly suppressed by the acetylation-resistant (K222R), but not the acetylation-mimicking (K222Q), MFN1 mutant, supporting an important role of MFN1 deacetylation in limiting ROS production in metabolically challenged mitochondria (Fig. 4B). These results indicate that mitochondrial fusion deficiency can lead to aberrant buildup of ROS upon glucose starvation. Supporting this conclusion, we observed a significant accumulation of oxidized proteins in the mitochondria (Fig. 4C, compare lanes 6 and 8), but not the cytosol (lanes 2 and 4), of HDAC6 KO MEFs upon glucose starvation (Fig. 4D). An oxidized protein population was not detected in mitochondria from wild-type MEFs subjected to glucose starvation (Fig. 4C, lane 7, Fig. 4D). Taken together, these findings indicate that HDAC6- and mitochondria fusion-deficient cells generate higher mitochondrial ROS stress upon glucose starvation.

To assess the physiological relevance of HDAC6-dependent mitochondrial fusion, we subjected wild-type and HDAC6 KO mice to fasting, forcing skeletal muscle to generate more energy from mitochondrial oxidative phosphorylation. We first investigated whether fasting, similar to glucose starvation, caused MFN1 deacetylation. Although more variable, a general trend of lower MFN1 acetylation in skeletal muscle was indeed observed in wild-type, but not HDAC6 KO, mice after fasting (supplementary material Fig. S1D–F). We next analyzed mitochondrial morphology in the tibialis anterior muscle by electron microscopy. As shown in Fig. 5A, under the fed state, mitochondria in both wild-type and HDAC6 KO tibialis anterior muscles resided stereotypically in pairs on either side of the Z-disc (Fig. 5A, fed, top panels). Upon fasting, mitochondria in wild-type tibialis anterior muscles underwent dramatic realignment and fused into large elongated mitochondria that often spanned multiple sarcomeres (Fig. 5A, fasted, black arrowheads). Remarkably, in fasted HDAC6 KO mice, individual mitochondria were aggregated but failed to fuse (Fig. 5A, white arrowheads). Thus, HDAC6 is required for fasting-induced mitochondrial fusion in mice.

Interestingly, unfused mitochondria in fasted HDAC6 KO mice were frequently swollen and contained less densely packed cristae, indicative of mitochondrial damage (Fig. 5A, white arrowheads). To confirm mitochondrial damage, the activities of mitochondrial electron transport chain complexes in tibialis anterior muscle were evaluated by histological staining that assessed cytochrome c oxidase (COX, complex IV, brown) and succinate dehydrogenase (SDH, complex II, blue) activity. As shown in Fig. 5B,C, under the fed condition, wild-type and HDAC6 KO mice tibialis anterior muscle showed similar brown and light-blue checkerboard appearances, which reflects mitochondrial activities in the different muscle fiber types. However, upon fasting, a marked decrease in COX complex IV activity (Fig. 5B, brown), which is encoded by the mitochondrial genome, was observed in HDAC6 KO but not in wild-type tibialis anterior muscle. Conversely, an increase in SDH activity (Fig. 5C, blue), which is encoded by the nuclear genome, was specifically detected in fasted HDAC6 KO muscle. A decrease in COX complex IV activity with a compensatory increase in SDH activity is a hallmark of mitochondrial dysfunction (Lee et al., 1998; Chen et al., 2010). These results show that HDAC6 is required to promote mitochondrial fusion and prevent mitochondrial damage in skeletal muscle challenged by fasting.

Fasting and glucose shortage activate metabolic reprogramming that simultaneously elevates energy production from mitochondria and the risk of mitochondrial oxidative damage. In this report, we have presented evidence that metabolically challenged mitochondria undergo active fusion to limit oxidative stress. The highly orchestrated adaptive mitochondrial fusion requires the protein deacetylase HDAC6, which binds, deacetylates and activates MFN1. The loss of HDAC6 prevents glucose-starvation- and fasting- induced MFN1 deacetylation and mitochondrial fusion, resulting in excessive mitochondrial oxidative stress and damage. Our findings identify active mitochondrial fusion as an integral part of the stress response that protects metabolically challenged mitochondria.

Active fusion has recently been proposed to prevent mitochondria from being degraded by autophagy under more extreme nutrient starvation (e.g. in Hank's solution) (Gomes et al., 2011; Rambold et al., 2011). Our analysis of glucose-starved cells or fasted HDAC6-deficient mice, however, did not reveal a significant loss of mitochondria (supplementary material Fig. S2A–D) or mitochondrial respiratory complexes (supplementary material Fig. S2E) despite a prominent defect in mitochondrial fusion. Instead, we found that a failure to undergo mitochondrial fusion upon metabolic challenge is accompanied by oxidative stress and mitochondrial damage (Figs 4 and 5). These findings suggest that mitochondrial fusion elicited by glucose deprivation or fasting and extreme starvation represents a distinct physiological adaptation: the former protects metabolically active mitochondria from oxidative stress whereas the latter shields mitochondria from excessive mitophagy. Consistent with this proposal, mitochondrial fusion under these two stress conditions is activated by different mechanisms: HDAC6-dependent MFN1 deacetylation in response to glucose starvation or fasting (this study), and inhibitory DRP1 phosphorylation upon extreme starvation (Gomes et al., 2011; Rambold et al., 2011). Supporting this view, HDAC6 KO cells can form mitochondrial networks upon treatment with Hank's solution (supplementary material Fig. S3A,B), similar to wild-type MEFs under glucose starvation (Fig. 1 and supplementary material Fig. S3C,D), indicating that HDAC6 is not required for all forms of stress-induced mitochondrial connectivity and its deficiency does not non-specifically prevent mitochondrial fusion.

We found that mitochondrial fusion induced by glucose starvation in cultured cells and fasting in mice was accompanied by a reduction in MFN1 acetylation (Fig. 2B; supplementary material Fig. S1D–F). Both MFN1 deacetylation and mitochondrial fusion were impaired in HDAC6-deficient cells and mice. These findings indicate that HDAC6 promotes mitochondrial fusion by binding, deacetylating and activating MFN1. Of note, this interaction does not require HDAC6 catalytic activity (supplementary material Fig. S1A). The location of the acetylatable K222 within the GTPase domain suggests that acetylation might inhibit the MFN1 GTPase activity important for mitochondrial fusion (Santel et al., 2003). Indeed, the acetylation-mimicking K222Q mutant MFN1 is severely defective in promoting mitochondrial fusion whereas the acetylation-resistant K222R mutant MFN1 is more active (Fig. 2E–G). Thus, reversible acetylation regulated by HDAC6 might allow for fine-tuning of MFN1 activity and mitochondrial dynamics in response to metabolic challenges or stresses. It is also worth noting that HDAC inhibitors have been shown to increase mitochondrial fusion under non-stressed conditions. However, this activity is independent of HDAC6 and involves down-modulation of DRP1 (Lee et al., 2012; Tailor et al., 2013).

Our analysis indicates that adaptive mitochondrial fusion is not required for acute energetic adaptation in MEFs (Fig. 3). Aberrant buildup of mitochondrial ROS, however, was detected in HDAC6-, MFN1- and OPA1-KO MEFs subject to glucose deprivation (Fig. 4A). Notably, expression of MFN1 K222R, but not K222Q, in HDAC6 KO MEFs could reverse mitochondrial ROS accumulation upon glucose starvation, supporting a crucial role of MFN1 deacetylation in limiting ROS production (Fig. 4B). These results indicate that active mitochondria fusion suppresses ROS accumulation, which could arise from an increase in mitochondrial energy production from fatty acid or amino acid oxidation upon glucose deprivation. Consistent with the proposal that networked mitochondria generate less oxidative stress, mitochondrial fission has been shown to contribute to excessive ROS production in other models (Yu et al., 2006; Nakamura et al., 2011). How mitochondrial connectivity affects ROS accumulation is a crucial issue that awaits further investigation. However, we noticed that HDAC6-, MFN1- and OPA1-KO MEFs all showed some degrees of elevated mitochondrial membrane potential (supplementary material Fig. S3E), which has been proposed to increase the electron backflow to complex I and subsequent ROS production (Anderson et al., 2009; Lanza and Nair, 2009; Suski et al., 2012). Thus, we favor a model where increased mitochondrial connectivity under stress conditions might lower trans-membrane potential, and thereby limit ROS accumulation (Fig. 4A; supplementary material Fig. S3E). Regardless of the specific mechanism, our findings suggest a model whereby the formation of a mitochondrial network would ensure that mitochondria meet increased energetic demands without incurring excessive oxidative damage. Given that food insecurity or starvation is a universal stress condition in nature, HDAC6-dependent adaptive mitochondrial fusion, although dispensable under standard laboratory conditions, likely plays a crucial role in maintaining individual fitness and survival in the wild. Furthermore, mitochondria also appear to be more connected in tumor cells when oxidative phosphorylation is elevated by replacing glucose with galactose in the medium (Rossignol et al., 2004). Thus, active mitochondrial remodeling could also be important for the metabolic fitness of tumors as well.

The electron microscopy analysis of skeletal mitochondria has revealed a striking morphological remodeling in response to fasting (Fig. 5A). The appearance of giant mitochondria indicates that mitochondria in fasted muscles undergo extensive spatial re-organization and fusion, a process clearly disrupted in HDAC6 KO mice. We also observed smaller and fragmented mitochondria in HDAC6 KO brain under fasting conditions, suggesting a similar regulation of mitochondria in other tissues (supplementary material Fig. S4). Interestingly, although mitochondria in HDAC6 KO fasted tibialis anterior muscles failed to fuse properly, they were clustered and congregated (Fig. 5A). These findings indicate that fasting-induced mitochondrial fusion is likely preceded by active transport and concentration of mitochondria. In this context, HDAC6 is required for fusion, but dispensable for mitochondrial movement. The dramatic mitochondrial morphological remodeling suggests that it is a part of an elaborate program that enables mitochondria to adapt to changing metabolic demands. Characterization of the components and signaling events that constitute this adaptive mitochondrial response would be of great interest in the future.

The unfused mitochondria in fasted HDAC6 KO muscles showed mitochondrial defects characterized by degenerative morphology, decreased mitochondrial COX complex IV activity and increased SDH activity (Fig. 5). Although we could not exclude the involvement of other mechanisms, such as HDAC6-dependent mitochondrial transport and mitophagy (Chen et al., 2010; Lee et al., 2010a), it is important to point out that analogous mitochondrial defects have been previously reported in skeletal-muscle-specific MFN1/MFN2 double-knockout mice (Chen et al., 2010). These mitochondrial phenotypes support the hypothesis that the MFN1 fusion machinery is a relevant target of HDAC6. It has been previously proposed that mitochondrial fusion can maintain mitochondrial function by complementing recessive pathogenic mtDNA with their healthy counterparts (Nakada et al., 2009; Chen et al., 2010). Our data suggest that, beyond a passive dilution mechanism, mitochondrial fusion could actively suppress mitochondrial damage by preventing the buildup of ROS.

Wild-type and HDAC6 KO MEFs reconstituted with various HDAC6 constructs were prepared as described previously (Gao et al., 2007; Lee et al., 2010b). Plasmids expressing Mito-YFP and CFP–MFN1 were a generous gift from Richard Youle (National Institutes of Health, Bethesda, MD). Lipofectamine LTX (Invitrogen) was used for transfection according to manufacturer's protocol.

Anti-mouse HDAC6 antibody was generated against amino acids 991 to 1149 as described previously (Gao et al., 2007). Antibodies against the following proteins were used: MFN1 (Santa Cruz Biotechnology, sc100561), acetyl-lysine (Cell Signaling, 9441 and Thermo Scientific, MA1-2021), Tom20 (Santa Cruz Biotechnology, sc11415), cytochrome c (BD, 556432), GAPDH (Cell Signaling, 2118), phosphorylated DRP1 (Cell Signaling, 4867 and 3455) and DRP1 (BD, 611113).

MEFs were maintained in Dulbecco's modified Eagle's medium (DMEM; Invitrogen, product number 11995) with 10% fetal calf serum (FCS). For glucose starvation experiments, cells at 70% confluence were washed with phosphate-buffered saline (PBS) three times and incubated with glucose-positive DMEM (Invitrogen, product number 11995) or -negative DMEM (Invitrogen, product number 11966) with 10% dialyzed serum (Sigma F0392) for 5 h.

Immunostaining was performed as described previously (Hubbert et al., 2002; Lee et al., 2004). Cells were cultured on glass coverslips, followed by the incubation with glucose-positive or -negative medium for 5 h. Cells were washed with PBS and then processed for immunostaining. Images were acquired by a Leica SP5 confocal microscope (Leica DMI6000C). For mitochondrial morphology analysis, cells are categorized according to following criteria: ‘1’, hyperfused, majority of mitochondria are interconnected; ‘2’, normal, mixed population of interconnected and non-connected mitochondria; or ‘3’, hyperfragmented, majority of mitochondria are not connected. The number of cells in each category was counted from three independent experiments (more than 100 cells were analyzed) and presented as mean of cell number (±s.d.).

FRAP was performed as described previously (Szabadkai et al., 2004). Briefly, cells were transiently transfected with the mito-YFP-expressing plasmid using Lipofectamine LTX (Invitrogen) according to manufacturer's protocol. Transfected cells were imaged with a Leica SP5 confocal microscope (Leica DMI6000C). For FRAP analysis, Leica LAS AF2.0 software was used. Circular regions of interest (ROIs) (2.5 µm in diameter) were imaged, using an 100× objective, before and after photobleaching with two iterations of the 514-nm laser. Images were taken in 1-s intervals and the fluorescence intensity in imaged ROIs was quantified with Leica LAS AF2.0 software. Mobile fractions were calculated according to published methods (Reits and Neefjes, 2001).

ROS generation in response to glucose starvation was determined using MitoSOX red mitochondrial superoxide indicator (Molecular Probes). The cells were washed with PBS, incubated with glucose-positive or -negative medium for 5 h, loaded with MitoSOX (5 µM) for 30 minutes, and subjected to fluorescence-activated cell sorting (FACS) analysis (Mukhopadhyay et al., 2007).

β-Oxidation of palmitic acid was measured by modifying previously described assays (DeBerardinis et al., 2006). Cells were counted and plated in duplicate 6-cm dishes using either glucose-positive or glucose-negative medium with dialyzed FBS serum and penicillin-streptomycin. After 10 h of incubation, the medium was exchanged for fresh medium. The mixture was prepared by adding 40 µCi of [9,10-3H] palmitate to 1 ml of a 20 µM unlabeled palmitate stock in a solution of 10% essentially fatty-acid-free BSA. The mixture was vortexed for at least 1 min and then 20 µl of the solution was added to each dish. Etomoxir (200 µM), which inhibits mitochondrial β-oxidation (Sigma, E1905), was added to half the plates to determine non-specific counts. The samples were incubated overnight (12–13 h) and 400 µl of medium was collected from each dish. The samples were spun, conjugated to Dowex Resin (1× 80–200 ion-exchange resin, Sigma, 217425), and eluted with water. The eluent was subject to scintillation counts for quantification.

Intracellular ATP was quantified by the luciferase-driven bioluminescence method (ATP Bioluminescence assay kit HS II, Roche, Germany) in freshly prepared cellular lysates according to the manufacturer's protocol.

Cytosolic and mitochondrial fractions were prepared, using the Mitochondrial isolation kit (Pierce, 89874) according to the manufacturer's protocol. 2 µg of each fraction were subjected to the oxidized protein analysis. Oxidized proteins in both cytosol and mitochondrial fractions were detected using Oxyblot protein oxidation detection kit (Chemicon, S7510) according to the manufacturer's protocol.

Tibialis anterior muscle was used for electron microscopy. After 48 h fasting, tibialis anterior muscles were collected from wild-type and HDAC6 KO mice. Tibialis anterior muscles were immediately fixed in 4% glutaraldehyde solution and processed by EM facility at Duke University for imaging. All animal experiments were performed according to approved guidelines. All mice were housed at the Duke University mouse facilities in accordance with the IACUC. Mice were fasted for 48 h with water supply and subjected to further analysis.

Tibialis anterior muscle was dissected and embedded in optimal cutting temperature (OCT) compound (Tissue-Tek) and frozen in liquid nitrogen. Slides were prepared and stained for COX activity, washed with distilled water, stained for SDH activity, washed, and mounted in Fluoromount G (SouthernBiotech). Detailed protocol for COX and SDH staining were obtained from the Washington University Neuromuscular Disease Center website (http://neuromuscular.wustl.edu/pathol/histol/sdh.htm and http://neuromuscular.wustl.edu/pathol/histol/cox.htm).

We thank David Chan (California Institute of Technology, CA) for MFN1 and OPA1 KO MEFs. We thank Allie McClure (Duke University, NC) and Yanhua Rao (Duke University, NC) for critically reading the manuscript.