GCN5-like Protein 1 (GCN5L1) Controls Mitochondrial Content through Coordinated Regulation of Mitochondrial Biogenesis and Mitophagy*

Background: The balance between mitochondrial biogenesis and autophagy controls cellular mitochondrial content. Results: Mitochondrial deacetylation-induced mitophagy evokes concurrent and interdependent up-regulation of TFEB and PGC-1α to sustain cellular mitochondrial content. Conclusion: Mitochondrial content is coordinately regulated by the mitochondrial and lysosome biogenesis programs under GCN5L1 control. Significance: Counter-regulatory interdependent programs function to sustain mitochondrial content/homeostasis in a nutrient-sensing, acetylation-dependent manner. Cellular mitochondrial content is governed by the competing processes of organelle biogenesis and degradation. It is proposed that these programs are tightly regulated to ensure that the cell maintains sufficient organelles to meet its biosynthetic, energetic, and other homeostatic requirements. We recently reported that GCN5L1, a putative nutrient-sensing regulator, controls mitochondrial removal by autophagy. Here we show that genetic deletion of GCN5L1 has a direct positive effect on the expression and activity of Transcriptional Factor EB (TFEB), which acts as a master regulator of autophagy. Surprisingly, the induction of TFEB-mediated autophagy pathways does not diminish cellular mitochondrial content, as its activity is countered by induction of the mitochondrial biogenesis transcriptional co-activator PPARγ coactivator 1α (PGC-1α). Concurrent induction of the TFEB and PGC-1α pathways results in an increased mitochondrial turnover rate in GCN5L1−/− cells. Finally, we show that genetic knockdown of either TFEB or PGC-1α leads to a corresponding decrease in the expression of the other gene, indicating that these proteins act coordinately, and in opposition, to maintain cellular mitochondrial content in response to the modulation of nutrient-sensing signatures.

Autophagy, an essential recycling process, is central to a number of cellular homeostatic mechanisms. Perhaps best known for its recycling of cellular components during periods of nutrient deprivation, autophagy is also necessary for the orderly removal of proteins, aggregates and organelles that have reached the end of their useful life (1)(2)(3). In the case of mitochondria, removal by autophagy (also known as "mitophagy") is a key housekeeping process, which is used to safely remove damaged or redundant organelles from the cell (4). This mechanism is of particular importance to maintain the quality of the mitochondrial population, as aberrant mitophagy is implicated in numerous pathologies, such as the premature development of Parkinson Disease in young individuals (5).
The systems used to identify and remove particular mitochondrial organelles from the cell by mitophagy, and how these are regulated, is currently an area of intense focus. One of the first to be established was the Pink1-Parkin pathway, where the disruption of mitochondrial function through respiratory uncoupling, leads to the stabilization of Pink1 on the mitochondrial membrane (6). This acts as a trigger for the translocation of the ubiquitin ligase Parkin to the mitochondria, where it ubiquitylates mitochondrial outer membrane proteins and targets them for autophagic and proteosomal degradation (6,7). In particular, Parkin ubiquitylates the mitofusin proteins involved in mitochondrial membrane fusion, thereby preventing uncoupled organelles from fusing to, and potentially damaging, other organelles (8). A second mechanism, utilizing members of the proapoptotic Bcl-2 family BNIP3 and Nix, is prevalent during periods of hypoxia, or during the maturation of erythrocytes (9, 10), respectively. When triggered, BNIP3 or Nix dimerize on the outer membrane of targeted mitochondria and bind to LC3, an ubiquitin-like protein involved in autophagosome maturation. With the aid of the linker protein p62/SQSTM1, these organelles are then transferred to the autophagosome for eventual degradation (11,12). For an extensive review or these and other mechanisms involved in mitophagy, please see Refs. 4,13. To counter the effects of mitophagy and maintain sufficient organelles to meet their needs, the cell must stimulate the mechanisms that control mitochondrial biogenesis. Given the importance of mitochondria to the survival of the cell, it is expected that there is a high degree of regulation and coordination between these two opposing pathways. We therefore resolved to investigate these dual mechanisms in the context of General Control of Amino Acid Synthesis 5-like 1 (GCN5L1), 2 a mitochondrial protein that we recently reported as controlling organelle clearance by mitophagy (14). Here we show that genetic deletion of GCN5L1 leads to an up-regulation of Transcription Factor EB (TFEB), a transcription factor which acts as a master regulator of autophagy (15). Despite an induction of TFEB-mediated autophagy, there is no net change in mitochondrial content, as there is a coordinated increase in mitochondrial biogenesis through the transcriptional co-activator peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1␣). Finally, we show that TFEB and PGC-1␣ are dynamically regulated in GCN5L1 cells, such that an increase in the expression of one gene is countered by an increase in the other, and vice versa. We therefore suggest that TFEB and PGC-1␣ act in opposition to each other, under the control of GCN5L1, to maintain cellular mitochondrial content.

EXPERIMENTAL PROCEDURES
Cells, Culture Conditions, and Transfection-Mouse embryonic fibroblasts (MEFs) were harvested at E11.5 and cultured in DMEM/15% FBS/Antibiotic-Antimycotic (Invitrogen) at 37°C/5% CO 2 for a maximum of 20 passages. The genotype of individual MEF lines was determined by multiplex PCR using the following primers: WT-F TGTTCACTCATCTGTGTGG-GTGGG; WT-R GGGAAACAGCCTATAACAGTCGTGGG; KO-R CCAACAGCTTCCCCACAACGG. To knockdown TFEB and PGC-1␣, MEFs were transduced with two independent lines of lentiviral shRNA particles (Sigma) under selection by puromycin. To reintroduce GCN5L1 into knock-out cells, we expressed GCN5L1-FLAG (16) using the Amaxa MEF 1 kit according to the manufacturer's instructions.
Bioenergetic Measurements and FACS Analysis-Measurements of mitochondrial respiration and glycolysis rates were performed using the Seahorse XF24 bioanalyzer following the manufacturer's protocols. After completion, cells were harvested and the rates adjusted relative to protein content. Cellular mitochondrial content was measured by FACS analysis, using Mitotracker Green (Invitrogen) fluorescence intensity as a proxy for mitochondrial volume. Each experiment was carried out at least three times, and a representative result is shown.
Protein Turnover Measurements-For SILAC experiments, wildtype and knock-out GCN5L1 MEF cells were cultured for six passages in "light" DMEM media, then switched to "heavy" DMEM containing 13C-Lys and 13C/15N-Arg for a chase period of 72 h. Proteins were harvested in 1% CHAPS lysis buffer, separated by SDS-PAGE, and stained with Gelcode Blue (Pierce). Twenty matching bands per lane were excised, and peptides were identified by LC-MS following trypsin digestion. The heavy:light ratio for proteins that were identified by at least two unique peptides was used to measure protein turnover rate between the two samples. Proteins were analyzed for cellular localization function using the GO component search (www. uniprot.com). For in vitro measurements of p62 and LAMP1 turnover, cells were treated with either 5 g/ml cycloheximide or 250 nM Bafilomycin A1 for 4 or 20 h. Proteins were analyzed by Western blot, and the accumulation and degradation rates were calculated following densitometry measurements.
Electron Microscopy-WT and GCN5L1 KO MEFs were grown in Permanox dishes, fixed with 2.5% glutaraldehyde, 1% paraformaldehyde, 0.12 M sodium cacodylate buffer pH 7.3 at room temperature for 20 min, then on ice for 40 min. After buffer washes, cells were post-fixed for 1 h on ice in 1% OsO 4 , 0.12 M sodium cacodylate buffer pH 7.3. After water washes, samples were block stained in 1% uranyl acetate, dehydrated in graded ethanol solutions, transitioned with propylene oxide and embedded in EMbed-812 (Electron Microscopy Sciences). Thin sections were stained with uranyl acetate and lead citrate, then examined on a JEM 1400 electron microscope (JEOL USA) with an AMT XR-60 digital camera (Advanced Microscopy Techniques Corp.).
Animal Husbandry-Animals used in this study were handled in a manner consistent with the guidelines established by the NHLBI Animal Care and Use Committee. Heterozygous mice with a targeted insertion in GCN5L1 were obtained from the International Knock-out Mouse Consortium/EUCOMM.

GCN5L1 Ϫ/Ϫ Mouse Embryonic Fibroblasts Display
Hallmarks of Up-regulated Autophagy-To further investigate the role of GCN5L1 in regulating mitochondrial clearance by autophagy, we attempted to produce a line of GCN5L1 Ϫ/Ϫ mice using a targeted insertion construct (Fig. 1A). Unfortunately, numerous attempts failed to produce viable mice, and further investigation established that GCN5L1 knock-out embryos fail to develop beyond ϳE12.5. However, we were successful in isolating numerous wild type, hemizygous, and homozygous knock-out GCN5L1 MEF lines from the ϳE11.5 stage (Fig. 1B). As previously shown (14), GCN5L1 knock-out MEFs display a vastly increased number of autophagsomes relative to wild type cells (Fig. 1C). This correlates with increased transcript production of the autophagosome-related gene microtubule-associated protein 1A/1B-light chain 3 (LC3; Fig. 1D), and increased levels of the active lipidated form of the protein product (LC3-II) in knock-out MEFs (Fig. 1E). To ensure that these changes are related to the loss of GCN5L1, we reconstituted the protein in knock-out MEFs using a GCN5L1-FLAG construct. Expression of this construct led to a reduction in both LC3 gene expression and cellular LC3-II protein levels ( Fig. 1, D and E).
Loss of GCN5L1 Induces the Expression of TFEB and Its Downstream Targets-We next investigated whether GCN5L1 Ϫ/Ϫ MEFs exhibited other alterations in their cellular autophagy programs. Proteins in the autophagy-regulated (ATG) pathway are involved in the conjugation and delivery of substrates to the autophagosome (17). On examination, we found that there was no difference in the cellular protein levels of Beclin-1, Atg7, Atg3, or Atg12-5 between wild type and knock-out cells, indicating that these pathways are intact in GCN5L1 Ϫ/Ϫ MEFs ( Fig.  2A). The mammalian target of rapamycin (mTOR) pathway acts a sensing mechanism to regulate autophagy induction in response to nutrient and growth stress (17). Again, we found that there was no significant difference in the expression or phosphorylation of mTOR, or its complex members Raptor, Rictor and G␤L, in wild type and knock-out cells ( Fig. 2A).
Additionally, there was equivalent phosphorylation of the mTOR substrate p70S6K ( Fig. 2A), indicating that this pathway was functioning normally. We next examined the expression of TFEB, a transcription factor that acts as a master regulator of autophagic and lysosomal pathways. Here we found that there was a significant increase in both TFEB transcript and protein levels in GCN5L1 knock-out cells (Fig. 2, B and C). Additionally, the transcript and protein levels of lysosomal membrane protein 1 (LAMP1), and the autophagy-substrate conjugation protein p62, both targets of TFEB-mediated transcription, were greatly elevated in GCN5L1 Ϫ/Ϫ MEFs (Fig. 2, B and C).
The presence of elevated p62 levels have variously been ascribed to either an increase in autophagic activity (14) or a block in autophagic flux (18). While the increased transcript levels (Fig. 2B) would argue for the former, we wished to establish whether the autophagic flux rate was impaired. Both wild type and knock-out cells were susceptible to similar changes in p62, LAMP1, and LC3 protein levels caused by the protein synthesis inhibitor cycloheximide, and the lysosomal inhibitor Bafilomycin A1 (Fig. 2D). While the higher initial protein levels in the GCN5L1 Ϫ/Ϫ MEFs meant that the absolute amount of p62 protein turnover was higher, there was no significant difference in relative percentage protein turnover between wild type and knock-out cells (Fig. 2E). As such, it appears that the autophagic flux is maintained in knock-out cells, and that the increase in autophagy-related proteins in GCN5L1 Ϫ/Ϫ cells is a consequence of up-regulated TFEB activity.
We then examined whether the accumulation of TFEB pathway proteins in knock-out MEFs was caused by their mislocalization, and subsequent reduced activity, within the cell. We separated wild type and knock-out MEF cells into fractions using differential centrifugation. While there was an increase in TFEB, p62, LAMP1, and LC3-II protein levels in knock-out cells, there was no disruption in the localization of these proteins to the autophagosome/lysosome membrane fraction (Fig. 2F).
Finally, we asked whether the increase in TFEB pathway protein expression was a direct effect of GCN5L1 ablation. As such, we reconstituted GCN5L1 expression in knock-out MEFs using transient transfection. As seen previously, loss of GCN5L1 led to an increase in the levels of TFEB and its downstream targets, p62 and LAMP1 (Fig. 2, G and H). However, expression of a GCN5L1-FLAG construct for 48 h negated this induction, bringing both the transcript and protein expression levels back down to those found in wild type cells (Fig. 2, G and H). In summary, we conclude that loss of GCN5L1 leads to a direct increase in the expression and activity of TFEB and its downstream targets.
Despite Reduced Respiration Capacity, GCN5L1 Ϫ/Ϫ MEFs Are Not Deficient in Mitochondrial Content-We previously demonstrated that HepG2 liver cells with shRNA-depleted GCN5L1 levels were deficient in mitochondrial content, and that this led to a reduction in their oxidative respiration capacity (14). We therefore examined whether MEF KO cells that are completely deficient in GCN5L1 exhibited a similar phenotype. While there was no significant difference in the baseline levels of oxidative phosphorylation between wild type and knock-out cells, GCN5L1 Ϫ/Ϫ MEFs were deficient in their maximal uncoupled respiration rate, reserve capacity and use of oxygen for ATP synthesis (Fig. 3, A and B). To counter this defect, GCN5L1 knock-out MEFs displayed an elevated level of glycolysis, both at baseline and when stimulated with glucose (Fig. 3,  C and D).
In concert, these data suggest that GCN5L1 knock-out MEF cells are depleted in mitochondrial bioenergetic output. Given that the TFEB autophagy pathway is elevated in these cells, we speculated that this may be due to an increased rate of mitophagy, which would result in a depleted mitochondrial cellular content. Surprisingly, we found that there was no significant difference between wild type and knock-out cells in the expression of numerous mitochondrial proteins (Fig. 3E), in mitochondrial DNA content (Fig. 3F), or in total mitochondrial Mitotracker Green fluorescence (Fig. 3G). Therefore, we conclude that despite the increased levels of autophagy proteins, the mitochondrial defects observed in the GCN5L1 Ϫ/Ϫ cells are not related to any reduction in mitochondrial content. GCN5L1 Ϫ/Ϫ MEFs Maintain Mitochondrial Content by Upregulation of PGC-1␣-mediated Biogenesis Pathways-As there was no difference observed in the mitochondrial content of wild type and knock-out GCN5L1 MEFs, we speculated that there may be a mechanism by which the cell counters the increased autophagic activity in GCN5L1 Ϫ/Ϫ cells. We therefore examined the expression levels of genes involved in mitochondrial biogenesis. Several genes (including NRF1, PGC-1␤, and PPAR␥) showed modest, yet significant increases in knock-out cells relative to wild type (Fig. 4A). However, more significantly, there was a vast increase in the expression levels of the mitochondrial biogenesis transcriptional co-activator, PGC-1␣ (Fig.  4A). To ensure that this increase was directly related to the loss of GCN5L1, we reconstituted expression of the protein in knock-out cells using the GCN5L1-FLAG construct. Transient transfection of this plasmid significantly reduced PGC-1␣ expression, suggesting that this increase was triggered by GCN5L1 loss (Fig. 4B).
We reasoned that this increase in PGC-1␣-mediated mitochondrial biogenesis in GCN5L1 Ϫ/Ϫ cells would reverse the loss of mitochondrial content promoted by the TFEB pathway induction. As this would be difficult to measure by looking at steady-state mitochondrial levels (as evidenced by the data shown in Fig. 3), we decided to look at the mitochondrial turnover in wild type and knock-out cells, using stable isotope labeling of amino acids in cell culture (SILAC)-based proteomics (Fig. 4C). Following a 72 h chase period with heavy SILAC media, we used LC-MS to identify over 4000 unique proteins, 1600 of which were identifiable by at least two unique peptides. Of these, 318 were identified as being localized to mitochondria using bioinformatic searches, which represents roughly onethird of the mitochondrial proteome (19). The heavy:light SILAC incorporation ratio was used as a proxy for protein turnover, and the values were compared between wild type and knock-out cells. Of the mitochondrial proteins identified, over 78% displayed higher turnover levels in GCN5L1 Ϫ/Ϫ cells, with an equal distribution of protein turnover ratios found in all compartments of the organelle (Fig. 4D, supplemental Table  S1). From these data, we conclude that GCN5L1 knock-out MEFs up-regulate their mitochondrial biogenesis mechanisms to counteract the increased TFEB-mediated mitophagy rate. This results in an increased rate of mitochondrial turnover in GCN5L1 Ϫ/Ϫ cells, without differences in the overall steadystate mitochondrial content between wild type and knock-out lines.

TFEB and PGC-1␣ Expression Levels Are Dynamically Regulated to Maintain Mitochondrial Content in GCN5L1 Ϫ/Ϫ MEFs-
Contrasting previous studies have shown that PGC-1␣ acts as a transcriptional co-activator of TFEB expression (20), and that TFEB acts as a transcription factor for PGC-1␣ expression (21). We therefore sought to determine whether PGC-1␣ or TFEB act in a linear pathway as suggested by these studies, or whether they act in opposition to each other at the same level, under the control of GCN5L1, to regulate mitochondrial content. We first used two independent lentivrial shRNA constructs to knockdown the levels of PGC-1␣ in GCN5L1 Ϫ/Ϫ MEFs. Expression of these constructs reduced the transcription of PGC-1␣ in knock-out MEFs to levels similar as those found in wild type cells (Fig. 5A). We then examined how these changes affected the expression of TFEB and its downstream targets, p62 and LAMP1. In GCN5L1 Ϫ/Ϫ cells, PGC-1␣ shRNA knockdown led to a corresponding decrease in TFEB expression, with a subsequent reduction in p62 and LAMP1 transcript levels (Fig. 5A).
We next examined how knockdown of TFEB would affect the expression of PGC-1␣. As above, expression of two independent lentiviral shRNA vectors reduced the transcription levels of TFEB to wild type levels, with a corresponding effect on the transcription of p62 and LAMP1 (Fig. 5B). This was mirrored by a similar reduction in the expression levels of PGC-1␣ in GCN5L1 knock-out MEFs treated with TFEB shRNA (Fig. 5B). Additionally, knockdown of both PGC-1␣ and TFEB reduced the transcript levels of LC3, indicating that both have an effect on autophagy pathways (Fig. 5C). At the mitochondrial level, the net result of these coordinated changes is maintenance in the size of the total mitochondrial pool in both TFEB and PGC-1␣ knockdown cells (Fig. 5D). We therefore conclude that PGC-1␣ and TFEB expression levels are dynamically-and coordinately-regulated via a negative feedback loop, and that this occurs downstream of GCN5L1 (see model, Fig. 6). In this manner, the cell can closely regulate the opposing mechanisms of mitochondrial biogenesis (via PGC-1␣) and lysosomal biogenesis (via TFEB), to maintain a functional population of healthy mitochondria.

DISCUSSION
The first identification of a role for GCN5L1 in the regulation of mitochondrial function came through a search for proteins involved in mitochondrial lysine residue acetylation. Here it was shown that the expression of GCN5L1 led to an increase in lysine acetylation in the mitochondria, and that this worked in opposition to the mitochondrial deacetylase SIRT3 (16). We then found that a decrease in acetylation caused by GCN5L1 shRNA knockdown in HepG2 cells acted as a molecular marker of mitochondrial nutrient depletion, and that this resulted in an increase in mitophagy (14). These findings were in agreement with previous studies that suggested GCN5L1 (also known as BLOC1S1) may be involved in the formation of lysosome-related organelles (22), and that its loss was associated with an increase in mitophagy in a unbiased genetic siRNA library screen (23). We now show that genetic deletion of GCN5L1 stimulates the cellular pathways that target mitochondria for degradation via autophagy, and that this is countered by an increase in the systems that bolster mitochondrial volume within the cell. This indicates that one of the functions performed by GCN5L1 is to act as a negative regulator of both mitochondrial biogenesis and degradation systems.
The up-regulated expression of both mitophagy-and biogenesis-related genes in GCN5L1 Ϫ/Ϫ MEFs led to an increase in the mitochondrial turnover rate in these cells. However, the coordination of these two pathways led to no net difference being observed in mitochondrial content between wild type and knock-out cells. It is interesting to note that despite the parity in organelle content and baseline respiration, there is a deficit in GCN5L1 Ϫ/Ϫ mitochondrial output that leads to an increased reliance on glycolysis (Fig. 3). It is tempting to speculate that the increased turnover rate in knock-out MEF cells results in a preponderance of immature or poorly formed mitochondria that may be limited in their bioenergetic capabilities. It will be of interest in future studies to determine whether the increased turnover rate seen in GCN5L1 Ϫ/Ϫ MEFs results in some dysregulation of the oxidative phosphorylation architecture of these organelles.
The transcriptional regulation of the autophagosome-lysosome degradation system has recently come under a great deal of scrutiny. The main player in this field is TFEB, which has been shown to induce the transcription of over 400 genes, many of which are involved in autophagosome and lysosome biogenesis, including p62/SQSTM1, LAMP1, and the LC3-related protein GABARAP (24,25). TFEB recognizes a CLEAR box sequence (5Ј-GTCACGTGAC-3Ј) in the promoter region of lysosomal genes, and its activity leads to the coordinated upregulation of degradation pathways. TFEB normally resides in the cytosol, where it is sequestered by 14-3-3 proteins following phosphorylation of either mTOR or ERK1/2 (25). Upon activation, TFEB is dephosphorylated, which induces it release and translocation to the nucleus to effect transcription of genes in the "CLEAR network" (25). Interestingly, one of the downstream targets of TFEB -p62, has also been shown to have roles  in the transcriptional response to autophagy. While generally identified as an adaptor protein for autophagic substrates, accumulation of p62 can also lead to the up-regulation of antioxidant and detoxifying enzymes, via an interaction with the complex regulating the Nrf2 transcription factor (26). As such, p62 appears to have multiple roles in the regulation of autophagy, and its increased abundance in GCN5L1 Ϫ/Ϫ MEFs may have additional cytoprotective functions outside the trafficking of autophagic cargo.
The first link between the expression of TFEB and PGC-1␣, the transcriptional co-activator responsible for the induction of numerous mitochondrial biogenesis systems (27), was discovered in a model of Huntington Disease (HD). HD causes neurodegeneration by the accumulation of polyglutamine-rich proteins that are resistant to degradation (20). Overexpression of PGC-1␣, whose signaling is abrogated in HD, led to the clearance of protein aggregates in a mouse HD model. Further examination showed that PGC-1␣ induced the expression of TFEB, whose expression alone could clear polyglutamine aggregates through lysosomal degradation, which placed PGC-1␣ upstream of TFEB in this pathway. Conversely, it was recently shown that in mouse and Caenorhabditis elegans models of nutrient deprivation, TFEB could up-regulate lysosomemediated lipid hydrolysis via the activation of PGC-1␣ and its downstream target PPAR␣. Here the authors discovered the presence of multiple CLEAR motifs in the PGC-1␣ promoter region, and deletion of these abrogated the response of this TFEBϾPGC-1␣ pathway in response to reduced food availability (21).
In this study, we demonstrate that the loss of GCN5L1 leads to an increase in the expression of both TFEB and PGC-1␣ (Figs. 2 and 4), and that the depletion of either transcriptional activator leads to a coordinated decrease in the other in GCN5L1 Ϫ/Ϫ cells (Fig. 5). We therefore postulate that there is a tightly regulated coordination of TFEB and PGC-1␣ expression, and that this mechanism has evolved in order for the cell to maintain its mitochondrial content. For example, in response to an increase in mitophagy (enacted through increased TFEB expression), the cell may up-regulate the activity of PGC-1␣ to replenish their mitochondrial content. Given that both proteins have been shown to induce the expression of the other, it is conceivable that this may be a direct consequence of their transcription factor activity. As such, we suggest that instead of a linear pathway relationship, both TFEB and PGC-1␣ (under the control of GCN5L1) function in a feedback loop to control cellular mitochondrial content (Fig. 6).
This study demonstrates that GCN5L1 occupies a pinnacle role in the regulation of mitochondrial turnover, acting as a brake on the opposing processes of mitophagy and biogenesis. Future work will be required to determine the molecular mechanisms behind this regulation: for example, whether TFEB/ PGC-1␣ interacts directly with some portion of the GCN5L1 pool, or whether there is some retrograde signaling pathway which controls these factors emanating from mitochondria. It will also be necessary to determine how the dysregulation of this pathway, caused by GCN5L1 loss, will affect physiological processes that rely on the correct removal of mitochondria at the end of their usable life.