SIRT3 (sirtuin 3) modulates respiration via the deacetylation of lysine residues in electron transport chain proteins. Whether mitochondrial protein acetylation is controlled by a counter-regulatory program has remained elusive. In the present study we identify an essential component of this previously undefined mitochondrial acetyltransferase system. We show that GCN5L1 [GCN5 (general control of amino acid synthesis 5)-like 1; also known as Bloc1s1] counters the acetylation and respiratory effects of SIRT3. GCN5L1 is mitochondrial-enriched and displays significant homology with a prokaryotic acetyltransferase. Genetic knockdown of GCN5L1 blunts mitochondrial protein acetylation, and its reconstitution in intact mitochondria restores protein acetylation. GCN5L1 interacts with and promotes acetylation of SIRT3 respiratory chain targets and reverses global SIRT3 effects on mitochondrial protein acetylation, respiration and bioenergetics. The results of the present study identify GCN5L1 as a critical prokaryote-derived component of the mitochondrial acetyltransferase programme.
The mitochondrial sirtuin deacetylase SIRT3 plays an important role in regulating oxidative metabolism [1–3] and redox stress [4–6]. Consequently, the regulation of SIRT3 activity is emerging as an important factor in the mitochondrial contribution towards disease susceptibility (reviewed in ). To understand this system more fully, it is necessary to identify proteins that counteract SIRT3 activity. As such, the discovery of a mitochondrial lysine acetyltransferase has been actively pursued .
Given the evolutionary history of mitochondria, one avenue of research has focused on bacteria. In Salmonella enterica, the acetyltransferase Pat has been shown to counteract CobB, a sirtuin homologue, in the regulation of acetyl-CoA synthetase . However, eukaryotic orthologues to Pat have not been identified in either the mitochondrial or nuclear genome . An alternative scenario in eukaryotes could be that mitochondrial proteins are acetylated in the cytosol prior to mitochondrial import. However, as fasting and feeding result in a dynamic flux in these post-translational modifications [2,10], it would be most likely that these modifications occur within mitochondria. Additionally, ATP synthase Fo subunit 8, a mitochondrial-encoded protein, is acetylated under nutrient flux conditions, further supporting the existence of an in situ mechanism .
In the present study we identify GCN5L1 [GCN5 (general control of amino acid synthesis 5)-like 1; also known as Bloc1s1], a protein with significant homology with the nuclear acetyltransferase GCN5 , as a mammalian regulatory protein in the control of mitochondrial protein acetylation and respiration. We show that GCN5L1 includes prokaryote-conserved acetyltransferase substrate- and acetyl-CoA-binding regions, is localized within mitochondria, modulates mitochondrial ETC (electron transport chain) protein acetylation, alters mitochondrial oxygen consumption, and counters the SIRT3 effects on mitochondrial protein acetylation, respiration and ATP levels.
MATERIALS AND METHODS
Phylogenetic and structural analysis
BLAST searches for acetyltransferases related to GCN5L1 identified several prokaryotic proteins, the closest being the xenobiotic streptogramin acetyltransferase from Burkholderia thailandensis (ZP_02389502), which we designated BtXAT. The sequences for acetyltransferases (all human, except BtXAT) described were obtained from GenBank® and analysed using the phylogenetic tool PhyML (http://www.phylogeny.fr). The XAT-repeat region of GCN5L1 was identified by interrogation of the GCN5L1 protein sequence, looking for a hexapeptide repeat motif matching X-[STAV]-X-[LIV]-[GAED]-X (NCBI cd03349). Data from this region was also used to map the substrate- and acetyl-CoA-binding regions of BtXAT. Predictions of GCN5L1 protein biochemical properties were performed using a Kyte–Doolittle hydrophobicity plot.
Cell culture and transfection
HepG2 cells were grown in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% FBS (fetal bovine serum). Plasmids were transfected using FuGENE® HD (Roche). For in vivo acetylation studies, cells were starved for 4 h in HBSS (Hanks balanced salt solution) and then re-fed with DMEM for 1.5 h before harvesting. Electroporation was used to transfect non-targeting control, GCN5L1 and SIRT3 siRNA (small interfering RNA) (Dharmacon).
Contruction of plasmids
The cDNA for GCN5L1 (Open Biosystems) was cloned into p3XFLAG-CMV-14 (Sigma) to create GCN5L1–FLAG. The cDNAs for NDUFA9 (NADH dehydrogenase subunit A9) and ATP5a (ATP synthase subunit 5a) (Open Biosystems) were cloned into pCMV3TAG-4A (Stratagene).
Polyclonal antibody production
A synthetic peptide corresponding to 16 amino acids of human and mouse GCN5L1, along with a conjugating cysteine residue at the C-terminus (MLSRLLKEHQAKQNER-C), was produced and injected into NZW (New Zealand white) rabbits (Covance). Serum was affinity-purified against the immunizing peptide, and the antibody was validated for recognition of human and mouse GCN5L1.
GCN5L1 localization, immunoblot analysis and co-immunoprecipitation
Confocal microscopy was used to localize GCN5L1–FLAG and DsRed-mito (Clontech) in fixed hepatocytes, by indirect immunolabelling of FLAG using Alexa Fluor® 488 (Invitrogen). Immunogold labelling and electron microscopy was used to localize endogenous GCN5L1 and the ETC protein ATP5a. Sub-mitochondrial localization was performed by osmotic pressure subfractionation and proteinase K protection assays. Antibodies used for Western blot analysis were: (i) the monoclonal antibodies anti-Ac-K (acetyl-lysine), anti-OPA1 (optic atrophy 1), anti-GAPDH (glyceraldehyde-3-phosphate dehydrogenase) and anti-VDAC (voltage-dependent anion channel) (all from Cell Signaling Technology); and the polyclonal antibodies anti-Ac-K, anti-ATP5a, anti-NDUFA9, anti-GDH (glutamate dehydrogenase) (all from Abcam) and anti-FLAG (Sigma). In co-immunoprecipitation experiments between GCN5L1–FLAG and NDUFA9–Myc or ATP5a–Myc, cells were co-transfected with plasmids for 24 h. Lysates were harvested and incubated with FLAG- or Myc-conjugated beads (Sigma and Cell Signaling Technology respectively). Beads were washed and analysed by Western blotting. Endogenous co-immunoprecipitation experiments followed a similar protocol; however, lysates were incubated with the relevant antibodies overnight, and interacting proteins were captured using Protein A/G beads (Santa Cruz Biotechnology). Western blots were quantified using image analysis software, and those shown are representative of at least three independent experiments.
The use of mice in the present study was approved by the NHLBI (National Heart, Lung, and Blood Institute) Animal Care and Use committee, and animals were maintained according to their guidelines.
Oxygen consumption measurements were performed on the XF24 analyser (Seahorse Bioscience). Control- and GCN5L1-transfected siRNA cells were transferred to 24-well plates overnight and incubated in sucrose respiration medium for 1 h before analysis. To measure the response to the mitochondrial respiration substrates glutamate and malate, cells were permeabilized with digitonin (10 μg per 106 cells for 5 min) and incubated in sucrose respiration medium. Glutamate (10 mM)/malate (5 mM) were added after baseline measurements had been taken. ATP levels in intact cells were measured (n=5) using the EnzyLight assay kit (Bioassay Systems).
In vitro acetylation analysis and in vitro immunoprecipitation
An in vitro acetylation assay was adapted from a previously published protocol . HepG2 cells were transfected with control or GCN5L1–FLAG plasmids for 24 h, or control and GCN5L1 siRNA for 72 h, at which time mitochondria were isolated. Samples were resuspended in reaction buffer [50 mM Tris/HCl, 50 mM NaCl, 4 mM MgCl2 and 5 mM nicotinamide (pH 7.4)], sonicated and then incubated for 1.5 h at 37°C, in the absence or presence of 2.5 mM acetyl-CoA. Mitochondrial proteins were used for global acetylation analysis (the reaction was stopped by boiling with SDS sample buffer, followed by SDS/PAGE and immunoblot analysis with a monoclonal anti-Ac-K antibody) or immunoprecipitation from non-denatured samples. For the reconstitution of GCN5L1 in mitochondria, HepG2 cells were depleted of GCN5L1 using siRNA as described above, after which intact mitochondria were isolated by sucrose gradient centrifugation. A total of 50 μg of pure mitochondria per sample were resuspended in acetyl-CoA assay buffer [100 mM Tris/HCl, 50 mM NaCl, 4 mM MgCl2 and 5 mM nicotinamide (pH 7.4)] on ice and sonicated. Samples were incubated at either 25°C or 95°C for 5 min, followed by the addition of 0.5 μg of GST (glutathione transferase)–GCN5L1 (Abnova) in PBS or PBS alone (control). After incubation for 1.5 h at 30°C, the reaction was stopped by boiling in SDS sample buffer. The histone assay followed this protocol, except for the addition of 2.5 μg of recombinant human histone H3 (New England Biolabs) to the mitochondrial samples where indicated. Following the 1.5 h incubation, acetylated histone H3 was recovered by immunoprecipitation from non-denatured samples with a polyclonal anti-Ac-K antibody. For in vitro immunoprecipitation, lysates from the in vitro acetylation analysis were incubated overnight with a polyclonal anti-Ac-K antibody. Immunoprecipitated proteins were purified using Protein G beads, washed and analysed by Western blotting.
Where required, data were tested for normality using the Kolmogorov–Smirnov test, followed by a one-tailed Student's t test or a Mann-Whitney U test using SigmaPlot 11 (Systat Software). A P value of <0.05 was regarded as statistically significant.
GCN5L1 is identified as a putative mitochondrial counter-regulator of SIRT3
To identify mitochondrial lysine acetyltransferases, we screened the human mitochondrial proteome database MitoCarta  for proteins harbouring acetyltransferase regions, and used yeast GCN5 as ‘bait’ in BLAST searches. Fluorescent-tagged candidate proteins were expressed to evaluate mitochondrial localization, and oxygen consumption was measured following siRNA knockdown to identify candidates that counter the known respiratory effects of SIRT3 depletion [1,14]. The strategy employed is illustrated in Supplementary Figure S1(A) (at http://www.BiochemJ.org/bj/443/bj4430655add.htm). Four of the candidate proteins identified included GCN5L1 (synonym BLOC1S1), NAT8b, NAT9 and NAT11 (NAT is N-terminal acetyltransferase). Expression of YFP (yellow fluorescent protein)tagged constructs and DsRed-mito showed that two proteins, namely GCN5L1 and NAT9, co-localized with the mitochondrial marker (Supplementary Figure S1B). As the knockdown of SIRT3 blunts mitochondrial respiration [1,15], we reasoned that the genetic depletion of a counter-regulatory component should show the inverse phenotype. siRNA was used to knockdown GCN5L1 and NAT9; however, only the loss of GCN5L1 resulted in increased oxygen consumption in HepG2 cells (Supplementary Figure S1C). Subsequent studies focused on the role of GCN5L1 in mitochondrial protein acetylation, and as a putative counter-regulatory protein to SIRT3.
GCN5L1 is mitochondrial-enriched and more highly expressed in oxidative tissues
GCN5L1 is a previously uncharacterized protein, although its sequence homology with GCN5, the nuclear acetyltransferase, has been described [11,16]. We first confirmed its mitochondrial localization using confocal microscopy on HepG2 cells transfected with FLAG-tagged GCN5L1 and DsRed-mito (Figure 1A), and by immunogold-labelled electron microscopy, which showed that the endogenous protein had a similar localization to ATP5a in mouse liver mitochondria (Figure 1B and Supplementary Figure S1D). Kyte–Doolittle hydropathy modelling of the GCN5L1 protein suggested that it is a non-transmembrane globular protein (Supplementary Figure S1E) , suggesting its location within mitochondria to be in either the intermembrane space or matrix-soluble fractions. Osmotic pressure subfractionation and proteinase K assays of isolated mouse mitochondria both support the hypothesis that GCN5L1 resides in the soluble matrix and intermembrane space fractions (Figures 1C and 1D). Its mitochondrial enrichment was further supported by the higher levels of GCN5L1 in slow- compared with fast-twitch skeletal muscle (Figure 1E).
GCN5L1 localizes to mitochondria
GCN5L1 possesses prokaryotic features and its knockdown disrupts mitochondrial protein acetylation
Phylogenetic mapping places GCN5L1 in a clade containing a histone acetyltransferase (HAT1), a bacterial xenobiotic acetyltransferase (BtXAT), and the nuclear lysine acetyltransferase GCN5 (Figure 2A). As such, we compared the GCN5L1 amino acid sequence with BtXAT and yeast GCN5. This comparison revealed a 53% similarity to the acetyl-CoA- and substrate-binding motifs of BtXAT. In contrast, homology with yeast GCN5 was limited to the non-catalytic N-terminal region of the protein (Figure 2B). The deduced GCN5L1 protein sequence, and especially the prokaryotic-like XAT repeat region, is highly conserved in metazoans (results not shown), suggesting a functional role for this region.
GCN5L1 levels modulate the degree of mitochondrial protein acetylation
To begin exploring the functional role of GCN5L1, we determined whether siRNA knockdown of this protein would modulate protein acetylation. Although GCN5L1 is predominantly mitochondrial, cytosolic expression was also observed. GCN5L1 levels were depleted in both subcellular compartments by siRNA; however, only the mitochondrial fraction showed a significant reduction in protein acetylation in HepG2 cells, supporting its functional role in mitochondria (Figure 2C). The acetylation function of GCN5L1 was further supported by a ≈60% attenuation of mitochondrial protein acetylation in an in vitro acetylation assay using GCN5L1-depleted cells (Figure 2D). Furthermore, the overexpression of GCN5L1 in HepG2 cells increased mitochondrial protein acetylation to a similar degree in the presence of acetyl-CoA (Figure 2E).
Histone H3 is acetylated by GCN5L1-enriched mitochondrial extracts
To further characterize the effect described above, in vitro assays were performed to assess whether recombinant GCN5L1 can acetylate purified histones, a classic acetylation substrate. In contrast with the in vivo studies, GCN5L1 promoted only weak histone acetylation when co-incubated in vitro (results not shown). To investigate whether this discrepancy may be due to the requirement of additional mitochondrial factors in the functioning of GCN5L1, we assessed mitochondrial protein acetylation in response to the addition of recombinant GCN5L1 to mitochondrial extracts purified from GCN5L1-depleted cells. Here, the reconstitution of GCN5L1 restored protein acetylation, although this was effect was limited to native, and not denatured, mitochondrial extract samples (Figure 3A). It is interesting to note in Figure 3(A) that the denaturing of mitochondria appears to facilitate ‘auto-acetylation’, as the boiled samples show enhanced protein acetylation in the presence and absence of GCN5L1. To explore further the requirement of additional mitochondrial cofactors in the acetylation function of GCN5L1, recombinant histone H3 was added to native GCN5L1-depleted mitochondrial extracts in parallel with the reintroduction, or not, of GCN5L1. The subsequent immunoprecipitation of acetylated lysine proteins, followed by immunoblot analysis with an antibody recognizing histone H3, showed increased H3 acetylation in the presence of GCN5L1 (Figure 3B). These results show that GCN5L1 positively regulates mitochondrial protein acetylation. However, its activity appears to require additional factors that reside in mitochondria.
Intact mitochondrial contents are required for GCN5L1-mediated protein acetylation
GCN5L1 promotes ETC protein acetylation
As the predominant mitochondrial protein lysine deacetylase SIRT3 activates ETC proteins [1,18], we investigated whether known SIRT3 ETC targets interact with, and are modulated by, GCN5L1. Co-expression of GCN5L1 with Myc-tagged complex I protein NDUFA9 or complex V protein ATP5a, followed by reciprocal immunoprecipitation, showed that there was a physical interaction between both NDUFA9 and ATP5a with GCN5L1 (Supplementary Figure S2 at http://www.BiochemJ.org/bj/443/bj4430655add.htm). To confirm that these interactions occur between the endogenous proteins in vivo, untransfected HepG2 cell extracts were used for immunoprecipitation studies. In these studies the antibody directed against GCN5L1 was used for immunoprecipitation, and antibodies directed against NDUFA9 and ATP5a were used in immunoblot analysis (Figure 4A). These results confirm the protein–protein interaction between GCN5L1 and the ETC proteins. To functionally characterize whether these ETC proteins are targets of GCN5L1-mediated acetylation, we then undertook in vitro acetylation procedures using total mitochondrial protein in cells displaying either knockdown or overexpression of GCN5L1. Using mitochondria from control and GCN5L1-depleted cells, followed by immunoprecipitation with an anti-Ac-K antibody and immunoblot analysis, it was shown that acetylation of endogenous NDUFA9 and ATP5a was diminished following GCN5L1 knockdown (Figure 4B). In parallel, immunoprecipitation with an antibody against acetylated lysine demonstrated that endogenous NDUFA9 and ATP5a exhibited increased acetylation in cells overexpressing GCN5L1 relative to control samples (Figures 4C and 4D).
GCN5L1 modulates the acetylation of mitochondrial ETC proteins
Mitochondrial respiration is attenuated following GCN5L1 knockdown
As GCN5L1 targets ETC proteins, we then evaluated the effect of GCN5L1 knockdown on mitochondrial respiration. Following knockdown of GCN5L1 compared with scrambled control siRNA in HepG2 cells (Figure 5A), oxygen consumption was assessed using the Seahorse apparatus. Basal cellular oxygen consumption was increased by GCN5L1 knockdown, as was the maximal oxygen consumption following mitochondrial uncoupling (Figure 5B). The depletion of GCN5L1 in HepG2 cells resulted in a 36% increase in oxygen consumption (Figure 5C). To confirm that this oxygen was consumed in the mitochondria, the HepG2 cells were permeabilized with digitonin and oxygen consumption was determined in the presence of glutamate and malate as specific mitochondrial respiratory substrates. In these studies, oxygen consumption was increased by 32% in GCN5L1-knockdown cells (Figure 5D). In parallel with these changes in oxygen consumption, the depletion of GCN5L1 resulted in significantly higher cellular ATP levels (Figure 5E).
GCN5L1 knockdown increases mitochondrial respiration
GCN5L1 functionally opposes the effects of SIRT3
The results described above show that the acetylation and bioenergetic phenotype of GCN5L1 depletion is in direct contrast with that observed following the loss of SIRT3. As these findings suggest that GCN5L1 may counteract the function of SIRT3, we explored the response to the combined genetic manipulation of these proteins. We employed an in vivo model where SIRT3 was knocked down in HepG2 cells, with or without the concurrent knockdown of GCN5L1. Knockdown of SIRT3 alone increased mitochondrial protein acetylation, which was significantly reversed by the concurrent knockdown of GCN5L1 (Figure 6A). Accordingly, the knockdown of SIRT3 diminished mitochondrial oxygen consumption and cellular ATP levels [1,14], and the concurrent knockdown of GCN5L1 reversed these metabolic phenotypes (Figures 6B and 6C). The ability of GCN5L1 knockdown to reverse SIRT3 acetylation and respiratory phenotypes was confirmed in SIRT3−/− MEFs (mouse embryonic fibroblasts) (results not shown).
GCN5L1 counteracts the effects of SIRT3
Identification of the mitochondrial acetyltransferase machinery has proven elusive. This is in contrast with the nuclear compartment, where the acetyltransferases GCN5, p300 and TIP60 [Tat (transactivator of transcription)-interactive protein 60 kDa] have been found to function as counter-regulatory enzymes to SIRT1 . The identification GCN5L1 in the present study reveals a novel mediator of mitochondrial protein acetylation to counter sirtuin deacetylase function.
The maintenance of sequence homology, from prokaryotes to nuclear-encoded eukaryotic mitochondrial proteins, is consistent with the symbiotic hypothesis of mitochondrial evolution . This phylogenetic fidelity is also evident, for example, in eukaryotic mitochondrial kinases , and in the iron/sulfur assembly regulatory protein frataxin . As with these examples, the conservation of the BtXAT protein acetyl-CoA- and substrate-binding regions in GCN5L1 supports the strong evolutionary pressure for retention of prokaryotic features in mitochondrial regulatory proteins.
The characterization of GCN5L1 in the present study shows that its expression, in isolation, is insufficient to significantly augment protein acetylation. However, the loss of acetylation following GCN5L1 depletion, and the re-establishment of this activity after its reconstitution, both demonstrate the requirement for GCN5L1 in mitochondrial protein acetylation. The findings of the present study suggest that GCN5L1 is one critical component of the mitochondrial acetyltransferase machinery. This concept is compatible with the function of yeast GCN5, where integration of this protein into a multi-subunit complex is required for maximal acetyltransferase activity . Taken together, these results have compelled us to begin to investigate whether additional mitochondrial proteins form multimeric complexes with GCN5L1 in order to orchestrate acetyltransferase activity.
The early characterization of the bioenergetic role of the mitochondrial deacetylase SIRT3 was performed in genetic knockout or knockdown conditions in cell and murine models. The initial studies showed that the absence of SIRT3 resulted in a marked reduction in mitochondrial oxygen consumption and a decrease in cellular ATP levels [1,14]. Subsequent studies in an in vivo context showed that the bioenergetic effects of SIRT3 were most pronounced under fasted conditions, and that the activity of numerous enzymes in multiple mitochondrial metabolic pathways are modified by SIRT3-mediated protein deacetylation. In the present study we used a similar approach, and identified that the absence of GCN5L1 inversely perturbs mitochondrial oxygen consumption and cellular ATP levels. However, more comprehensive biochemical analysis of the effects of mitochondrial protein acetylation will need to be performed as additional substrates of GCN5L1 are delineated, particularly following the identification of the putative functional partners of GCN5L1.
In summary, the present study shows that GCN5L1 functions as an essential component of the mitochondrial lysine acetyltransferase machinery, has phylogenetic linkage to prokaryotic acetyltransferases, and modulates mitochondrial respiration via acetylation of ETC proteins. GCN5L1 also counters the mitochondrial deacetylase function of SIRT3. Further investigation of GCN5L1, its putative interacting proteins and substrates, and the role of acetyl-CoA, should enhance our understanding of how acetylation modulates mitochondrial function.
ATP synthase subunit 5a
xenobiotic streptogramin acetyltransferase from Burkholderia thailandensis
Dulbecco's modified Eagle's medium
electron transport chain
general control of amino acid synthesis 5
NADH dehydrogenase subunit A9
small interfering RNA
Iain Scott, Bradley Webster and Michael Sack designed the studies; Iain Scott, Bradley Webster and Jian Li performed the studies; and Iain Scott, Bradley Webster and Michael Sack prepared the paper.
We thank Mary McKee (Partners Healthcare, Boston, MA, U.S.A.) for her help with electron microscopy studies, Michael Lazarou [NINDS (National Institute of Neurological Disorders and Stroke), Bethesda, MD, U.S.A.] for advice on proteinase K assays, and Toren Finkel [NHLBI (National Heart, Lung and Blood Institute), Bethesda, MD, U.S.A.] and Richard Youle [NINDS (National Institute of Neurological Disorders and Stroke), Bethesda, MD, U.S.A.] for their critical review of the paper prior to submission.
This work was supported by the Division of Intramural Research of the NHLBI (National Heart, Lung and Blood Institute), NIH (National Institutes of Health).