GCN5L1 interacts with αTAT1 and RanBP2 to regulate hepatic α-tubulin acetylation and lysosome trafficking

GCN5L1 (also known as BLOC1S1) was identified as a protein with sequence homology to the nuclear acetyl transferase general control of amino acid synthesis 5 (Gcn5) (Driessen et al., 1997; Inoue et al., 1996). Subsequent studies showed that the genetic depletion of GCN5L1 reduced mitochondrial protein acetylation (Fukushima et al., 2016; Scott et al., 2012; Webster et al., 2013, 2014), and that GCN5L1 was enriched in mitochondria and interacted with canonical substrates of the established mitochondrial deacetylase (SIRT3) (Scott et al., 2012). Interestingly, short-term GCN5L1 depletion resulted in the induction of restricted mitochondrial autophagy (mitophagy) in parallel with enhanced mitochondrial protein turnover and a reduction in overall mitochondrial content and cellular oxidative metabolism (Webster et al., 2013). In contrast, primary GCN5L1-knockout MEF cells retained their mitochondrial content with a concurrent induction of mitophagy and mitochondrial biogenesis regulatory programs (Webster et al., 2014). Additional studies found that GCN5L1 modulated mitochondrial control of fat oxidation (Fukushima et al., 2016; Thapa et al., 2017) and mitochondrial-linked regulation of gluconeogenesis (Wang et al., 2017). Recently, the first validated lysine target of GCN5L1 was identified, namely, acetylation of the mitochondrial-associated kinesis Kif1Bα-binding protein (KBP), which required cooperation between GCN5L1 and the mitochondrial acetyl-CoA-generating enzyme L-threonine dehydrogenase (TDH) (Donato et al., 2017).

In parallel studies, GCN5L1 was found to interact with the multiprotein biogenesis of lysosome-related organelles complex-1 (BLOC-1) and was alternatively named as a BLOC-1 subunit (i.e. BLOS1 or BLOC1S1) (Starcevic and Dell'Angelica, 2004). Mutations in numerous BLOC-1 complex subunits, but not in GCN5L1 itself, result in the development of the Hermansky–Pudlak syndrome (Chiang et al., 2003; Li et al., 2003), a syndrome linked with abnormal vesicular trafficking to lysosome and related organelles such as melanosomes. Interestingly, Drosophila GCN5L1 deficiency produced eye pigment defects that implicated this protein in the control of melanosome trafficking (Cheli et al., 2010). Although the germline knockout of GCN5L1 in mice is embryonic lethal (Webster et al., 2014; Zhang et al., 2014), E14.5 embryos showed defects in eye pigmentation (Zhang et al., 2014) and the exploration of epidermal growth factor receptor (EGFR) endolysosomal trafficking demonstrated that GCN5L1 interacted with sorting nexin 2 (SNX2) and an endosomal sorting complex component TSG101, to modulate EGFR trafficking (Zhang et al., 2014). More recently, GCN5L1 was found to associate with BORC, a distinct multiprotein complex that regulates lysosome positioning (Pu et al., 2015). Although BORC has some overlapping proteins with the BLOC-1 complex, most subunits are unique. Nevertheless, the genetic knockdown of GCN5L1 phenocopied the effect of depletion of other BORC subunits and resulted in reduced dissemination of lysosomes to the periphery of cells (Pu et al., 2015).

Together these studies point to differential roles of GCN5L1 in distinct subcellular compartments that span from acetylation dependent effects on mitochondrial homeostasis and metabolism to impaired endosomal lysosome trafficking and function. Given that the acetylation of microtubules plays an important role in endosome-lysosome trafficking (Perdiz et al., 2011; Xu et al., 2017), we reasoned that a common link between the diverse effects/phenotypes of GCN5L1 may, in part, be due to the role of GCN5L1-mediated acetylation in modulating organelle positioning.

In this study, we confirm that the absence of GCN5L1 results in impaired centrifugal movement of lysosomes. Furthermore, we show that GCN5L1 modulates lysosomal positioning via the regulation of microtubule acetylation in concert with the canonical α-tubulin acetyltransferase αTAT1. Finally, we show that this regulation is dependent on the RAN-binding protein 2 (RanBP2), a component of nuclear pore complexes also known as Nup358, which facilitates the interaction of GCN5L1 with α-tubulin. This study identifies a novel mechanism underpinning the role of GCN5L1 in protein acetylation due to its interaction with αTAT1, and expands our understanding of its role in the regulatory control of lysosome positioning. Furthermore, this study uncovers a potential mechanism for the previously identified role of RanBP2 in the modulation of organelle positioning.

To explore whether GCN5L1 levels effect microtubule acetylation, we used liver homogenates from wild-type (WT) and liver-specific GCN5L1 (LKO) knockout mice (Wang et al., 2017). LKO mice show a significant reduction in α-tubulin acetylation on its cognate functional K40 residue (LeDizet and Piperno, 1991) (Fig. 1A). This reduction in K40 acetylation of α-tubulin was even more readily apparent in LKO primary hepatocytes (Fig. 1B). Immunofluorescence images also showed reduced α-tubulin acetylation without evidence of changes in microtubule distribution in primary LKO hepatocytes (Fig. 1C). Furthermore, adenovirus-mediated overexpression of GCN5L1 rescued LKO hepatocyte α-tubulin acetylation as evident by immunoblotting and immunofluorescence experiments (Fig. 1D,E). To reduce the potential for a non-specific effect of GCN5L1 overexpression, the levels of α-tubulin acetylation was assayed in response to an incremental 4-fold reduction in the multiplicity of infection (MOI) of adenoviral infection. All doses of GCN5L1 induced α-tubulin acetylation (Fig. S1A), supporting the idea that this is a functional effect. Taken together, these findings support that GCN5L1 plays a regulatory role in the modulation of microtubule acetylation.

The concept that non-enzymatic acetylation of proteins is operational in mitochondria (Wagner and Hirschey, 2014) and in the cytosol (Moussaieff et al., 2015) is being actively investigated. Given this, and the known requirement of mitochondrial GNC5L1 to partner with an acetyl-CoA-generating enzyme to facilitate acetylation on the mitochondrial membrane (Donato et al., 2017), we then determined whether cellular levels of acetyl-CoA in hepatocytes could underpin the effect of GCN5L1 on α-tubulin acetylation. However, the levels of acetyl-CoA were not different when comparing WT and LKO primary hepatocytes as measured by using high-pressure liquid chromatography (HPLC) and liquid chromatography mass spectrometry (LC-MS) acetyl-CoA assays (Fig. 2A; Fig. S2A). Interestingly, this finding did not preclude acetyl-CoA level-dependent acetylation of α-tubulin, given that acetate and ethanol, which both increase hepatic intracellular acetyl-CoA levels (Cai et al., 2011; Cederbaum, 2012), increased the acetylation of α-tubulin irrespective of the presence or absence of GCN5L1 (Fig. 2B; Fig. S2B). Here, despite the low basal level of α-tubulin acetylation in the LKO hepatocytes, their parallel dose-dependent increase in acetylation supports the idea that the GCN5L1 effect is not-dependent on cellular acetyl-CoA levels within the physiological range. To validate that GCN5L1 did not modulate the kinetics of microtubule acetylation, but rather the absolute extent of α-tubulin acetylation, we employed nocodazole to depolymerize and deacetylate microtubules (Chabin-Brion et al., 2001; Perdiz et al., 2011). The kinetics of the restoration of α-tubulin acetylation was then observed following the washout of nocodazole. Again, we show that the kinetics of α-tubulin acetylation was similar in the presence or absence of GCN5L1, but that the absolute level of acetylation was lower in the GCN5L1-knockout hepatocytes (Fig. 2C). Taken together, these data suggest that the effect of GCN5L1 on microtubule acetylation within the cytosol requires a ‘catalytic’ rather than a non-enzymatic mass action effect.

Since GCN5L1 does not contain a canonical acetyltransferase catalytic domain, and nor does it modulate α-tubulin acetylation in an acetyl-CoA dose-dependent manner or facilitate the α-tubulin acetylation in vitro (Fig. S3A), we explored potential GCN5L1 partners that would facilitate α-tubulin acetylation. Since αTAT1 is a major α-tubulin acetyltransferase, we tested whether it interacts with GCN5L1 (Akella et al., 2010; Kalebic et al., 2013). Bidirectional co-immunoprecipitation studies were performed following overexpression of GFP-tagged αTAT1 and Flag-tagged GCN5L1 in HEK293 cells. Immunoprecipitation with an anti-Flag antibody (Fig. 3A), or with an antibody directed against GFP (Fig. 3B) reveal a clear interaction between overexpressed GCN5L1 and αTAT1. To further validate this interaction, an antibody directed against endogenous GCN5L1 was used to immunoprecipitate Flag-tagged αTAT1 (Fig. 3C). To confirm this interaction, confocal microscopy was performed in HeLa cells following the overexpression of GFP–αTAT1 and Flag–GCN5L1 (Fig. 3D), or Flag–αTAT1 and GFP–GCN5L1 (Fig. 3E). Following cell fixation and immunostaining, the overlay of reciprocal GFP and Flag images show significant overlap between GCN5L1 and αTAT1 in the cytoplasm.

To evaluate whether this interaction between GCN5L1 and αTAT1 is functional, a Flag-tagged αTAT1 expression construct was transfected into WT and GCN5L1 KO hepatocytes. Confocal microscopy shows that the αTAT1-transfected cells have higher levels of acetylated α-tubulin in the WT and LKO cells (Fig. 4A). In contrast, in non-transfected cells (shown with an asterisk in the merged panels), the levels of acetylated α-tubulin are higher in the WT compared to the KO cells (Fig. 4A). To quantify these differences, the levels of acetylated α-tubulin was assessed by immunoblot analysis. As shown previously, non-transfected LKO cells had lower levels of acetylated α-tubulin compared to non-transfected WT cells, and the overexpression of αTAT1 increased acetylated tubulin levels in both WT and KO cells (Fig. 4B). We then explored whether the genetic knockdown of αTAT1 could abrogate the rescue of α-tubulin acetylation following reconstitution of GCN5L1 in LKO hepatocytes. This concept was validated by the observation that the α-tubulin acetylation in LKO cells was not restored upon GCN5L1 adenovirus infection when siRNA targeting αTAT1 was concurrently transfected (Fig. 4C). Similarly, WT cells transduced with GCN5L1 adenovirus had diminished levels of acetylated tubulin in the presence of αTAT1 siRNA (Fig. 4C).

Given that cytosolic GCN5L1 has been found to be operational in numerous multiprotein complexes (Pu et al., 2015; Starcevic and Dell'Angelica, 2004; Zhang et al., 2014) we employed the proximity-dependent biotin identification assay BioID (Roux et al., 2012), to identify potential novel interacting partners that could play additional regulatory roles in GCN5LI-dependent α-tubulin acetylation. To identify proteins that interacted with GCN5L1, a biotin-ligase (BirA)-tagged GCN5L1 (GCN5L1–BirA) was transiently expressed in 293T cells. In the presence of biotin supplementation, proteins biotinylated by GCN5L1–BirA were captured with streptavidin beads and verified on immunoblot analysis (Fig. 5A). MS experiments were then performed, to identify potential interacting proteins. In total, 324 proteins were identified with high confidence, including most subunits of BLOC1 (Falcón-Pérez et al., 2002), including BLOC1S2, BLOC1S4, BLOC1S6 and Snapin (Table S1). These proteins have previously been identified as GCN5L1-interacting proteins (Starcevic and Dell'Angelica, 2004). Interestingly, this proximity-dependent biotin identification assay identified all of the shared subunits with BLOC1, and LOH12CR1 of the BORC complex, as previously described (Pu et al., 2015). To approximate the relative abundance of proteins recovered in the MS analyses, and to correct for overall recovery between experiments, the peptide spectrum match (PSM) values for each protein were calculated (Pisitkun et al., 2012) and the top 123 most-enriched proteins (PSM>1) were analyzed by functional annotation clustering using DAVID (Fig. 5B; Table S2). Intriguingly, numerous proteins identified were enriched in pathways that were linked to microtubule-related functions, such as mitosis, cytoskeleton organization and microtubule-linked trafficking. These data further support that GCN5L1 plays a role in microtubule-related functions.

An interesting candidate interacting protein was the nuclear pore complex protein RanBP2. Although this protein is most highly enriched on the cytosolic surface of the nuclear pore and regulates nuclear protein transport (Dickmanns et al., 2015), it also functions as an allosteric activator of kinesin-1 (Cho et al., 2009) and modulates microtubule-linked processes, including mitosis, α-tubulin acetylation and mitochondrial localization and function (Cho et al., 2007; Ganley et al., 2011; Joseph and Dasso, 2008). Given these putative overlapping functional pathways, we validated the interaction between GCN5L1 with RanBP2 by immunoprecipitation analysis. Anti-Flag immunoprecipitation of cell lysis from 293T cells overexpressing Flag–GCN5L1 and HA–RanBP2 expression vectors with subsequent immunoblot analysis showed a modest affinity between GCN5L1 and the full-length RANBP2 (Fig. 6A). The inverse interaction using anti-HA immunoprecipitation similarly showed the interaction of RanBP2–HA with GCN5L1–Flag (Fig. 6B). As an additional control to exclude a non-specific interaction between HA-RanBP2 and endogenous GCN5L1, immunoprecipitation was assessed in response to using either an anti-IgG antibody or an antibody directed at GCN5L1. Here too, an interaction between overexpressed RanBP2 and endogenous GCN5L1 is found (Fig. 6C). RanBP2 is a large protein (3224 amino acids) with multiple functional domains (Kassube et al., 2012; Lin et al., 2013; Werner et al., 2012). To pinpoint the interacting region, multiple length C-terminal deletion mutants of RanBP2 (Wälde et al., 2012) were employed in immunoprecipitation studies. All the C-terminal deletion mutants of RanBP2 bound the Flag-tagged GCN5L1 (Fig. S4A), suggesting that GCN5L1 interacted with the N-terminal region (amino acids 1–1133) of RanBP2. The interaction of GCN5L1 with RanBP2 was most robust when the 1133 amino acid N-terminal region of RanBP2 was co-expressed. This affinity was reduced when the initial 805 amino acids was excluded from the 1133 N-terminal fragment of RanBP2 and co-expressed with GCN5L1 (Fig. 6D). Interestingly the N-terminal region of RanBP2 (amino acids 1–900) had previously been shown to be required for the association of RanBP2 with microtubules (Joseph and Dasso, 2008). In this study, we find that RanBP2 directly interacted with α-tubulin. This was performed in 293T cells following the overexpression of HA-tagged RanBP2 (amino acids 1–1133) and α-tubulin. Subsequent immunoprecipitation with an antibody directed at the HA tag showed the interaction between RanBP2 and α-tubulin (Fig. S4B).

Given the interaction of RanBP2 with GCN5L1, we then explored whether RanBP2 also binds to αTAT1. Here, bidirectional co-immunoprecipitation studies show that αTAT1 binds to the N-terminal of RanBP2 (Fig. 7A,B), and confocal microscopy showed that this domain of RanBP2 is distributed along the microtubular network in concert with αTAT1 (Fig. 7C). To confirm the role of RanBP2 in this complex, we show that the overexpression of RanBP2 (amino acids 1–900) mediated an increase in α-tubulin acetylation and that the concurrent knockdown of αTAT1 abolished this acetylation effect (Fig. 7D; Fig. S5A). The non-transfected cells in the same image panels, as seen with DAPI staining, serve as a negative control. Collectively, these results indicate that RanBP2 recruits the GCN5L1–αTAT1 complex to promote α-tubulin acetylation. We then confirmed that the N-terminal region of RanBP2 (amino acids 1–900) was sufficient to bind to microtubules (Fig. 7D; Fig. S5B), and to increase the acetylation of α-tubulin (Fig. 7D; Fig. S5C). The requirement of this domain was confirmed by the finding that a truncated fragment RanBP2 (amino acids 1–600) disrupted microtubule binding and α-tubulin acetylation (Fig. S5B,C).

Given the finding that GCN5L1 played a role in modeling α-tubulin acetylation, we then assessed whether GCN5L1 itself interacts with α-tubulin. However, neither conventional nor pre-emptive crosslinking co-immunoprecipitation studies could verify a direct interaction (Fig. 7E). We then explored whether an indirect association exists between GCN5L1 and α-tubulin by employing RanBP2 as a structural intermediary. Here, following co-expression of Flag–GCN5L1 and the HA–RanBP2 fragment (amino acids 1–1810), subsequent immunoprecipitation with an antibody recognizing Flag (GCN5L1) pulled down both RanBP2 and α-tubulin (Fig. 7E). Only the co-expression of GCN5L1 and RanBP2 had the ability to immunoprecipitate α-tubulin (Fig. 7E). Collectively, the results suggest that GCN5L1 is recruited to microtubules by RanBP2, which then coordinately facilitate interaction with, and acetylation of, α-tubulin by αTAT1.

As a subunit of BORC, knockdown of GCN5L1 had previously been shown to impair the centrifugal movement of lysosomes in HeLa cells by decreasing the recruitment of kinesin-1 molecules to lysosomes (Pu et al., 2015). To validate this in primary hepatocytes, we examined the intracellular distribution of lysosomes in WT and LKO primary hepatocytes using an antibody directed against lysosome-associated membrane protein-1 (LAMP-1). GCN5L1-KO hepatocytes showed increased levels of LAMP-1 compared to WT (Fig. S6A). Concurrently, we found the lysosomes were distributed throughout the cytoplasm in WT cells, but were tightly clustered in the perinuclear region in LKO hepatocytes (Fig. 8A). This abnormal distribution of lysosomes in LKO cells was restored by adenovirus-driven overexpression of GCN5L1 (Fig. S6B). As the acetylation of α-tubulin promotes kinesin-1 binding to microtubules with increased cargo delivery to the peripheral region of cells (Reed et al., 2006), we employed tubacin, a selective inhibitor of the α-tubulin deacetylase HDAC6 (Haggarty et al., 2003), to assess its effect on lysosomal distribution. Here, we find that tubacin increases microtubule acetylation in both WT and LKO hepatocytes (Fig. 8B). In parallel, tubacin promoted the peripheral distribution of lysosomes in both genotypes, although the response in the KO cells was only partial (Fig. 8B). This partial response may reflect the excess of accumulated perinuclear lysosomes in the LKO cells. Together, these experiments support the idea that GCN5L1-dependent α-tubulin acetylation contributes to the centrifugal movement of lysosomes.

In this present study, we show that a novel multiprotein complex comprising GCN5L1, RanBP2 and αTAT1 functions in concert to acetylate α-tubulin, and that this acetylation modification plays an important role in lysosome positioning in the cell. Moreover, we identify that GCN5L1 and RanBP2-mediated microtubule acetylation is dependent on αTAT1 and that RanBP2 appears to function as the intermediary linking GCN5L1 to α-tubulin.

Prior to this study, the only functionally characterized target of GCN5L1 acetylation was the kinesin motor-binding protein KBP (Donato et al., 2017). In that study, performed in murine embryonic stem cells (mESCs), the genetic depletion of GCN5L1 promoted mitochondrial biogenesis and mESC differentiation (Donato et al., 2017). Interestingly, GCN5L1-mediated acetylation of KBP required the presence of the mitochondrial acetyl-CoA generating enzyme L-threonine dehydrogenase (TDH) (Donato et al., 2017). This finding was intriguing, given that a GCN5L1 includes a prokaryote-conserved acetyltransferase substrate- and acetyl-CoA-binding regions (Scott et al., 2018), but not a canonical catalytic domain (Scott et al., 2012). In our study, we show that GCN5L1 facilitated acetylation of α-tubulin is dependent on the αTAT1 acetyltransferase, and that in the absence of GCN5L1 α-tubulin acetylation is diminished. Interestingly, although the nuclear acetyltransferase GCN5 does contain an acetyltransferase catalytic domain, it requires additional partners to function as a canonical enzyme (Grant et al., 1999). Putting this together, GCN5L1 may function as a molecular component of an acetyltransferase machinery, and, at least in the cytosol, can partner with αTAT1 and RanBP2 to confer this activity. At the same time, and in contrast to the requirement of TDH to generate the acetyl-CoA for GCN5L1 mediated acetylation of KBP (Donato et al., 2017), the high abundance of acetyl-CoA in the liver may preclude the requirement for its generation for GCN5L1 functioning in this tissue.

The role of the acetylation of α-tubulin in enhancing microtubular strength (Xu et al., 2017) and its function in intracellular organelle trafficking is well established (Geeraert et al., 2010; Reed et al., 2006). Additionally, the major enzymes regulating α-tubulin acetylation and deacetylation, respectively, have been identified as αTAT1 and HDAC6 (Akella et al., 2010; Hubbert et al., 2002; Shida et al., 2010). The genetic depletion of GCN5L1 has been found to disrupt endolysosomal trafficking (Zhang et al., 2014), lysosome positioning (Pu et al., 2015) and lysosome function (Gaidt et al., 2017), and here we find that GCN5L1 directly interacts with αTAT1 and that the absence of GCN5L1 inhibits α-tubulin acetylation. Taken together, these data support the role of GCN5L1 as a novel regulatory protein for the function of αTAT1 in modulating α-tubulin acetylation and subsequent lysosome positioning/distribution.

Traditionally RanBP2, which is located on the cytoplasmic filaments of the nuclear pore complex, contributes to the transport of proteins into or out of the nucleus (Dickmanns et al., 2015). However, additional studies strongly suggested that RanBP2 may play a role in intracellular trafficking as evident by a requirement of RanBP2 in the regulation of interphase microtubules (Pichler et al., 2002; Joseph and Dasso, 2008), in microtubule–kinetochore interactions (Joseph et al., 2004) and in the appropriate intracellular distribution of mitochondria (Cho et al., 2007). In this study, we find that GCN5L1 and RanBP2 interact with each other and that RanBP2 facilitates recruitment of GCN5L1 to α-tubulin for subsequent acetylation.

Advancing proteomics studies have identified a greater diversity in protein subcellular location than had been previously appreciated (Thul et al., 2017), and this emerging recognition is pertinent to both our findings with respect to GCN5L1 and RanBP2. Proteomic analysis of GCN5L1-binding partners and subsequent DAVID analysis revealed cellular processes operating in distinct subcellular locations. However, whether the mechanism of action of GCN5L1 and its binding partners are the same or specialized in different subcellular locations needs to be explored. GCN5L1 has been found to modulate the acetylation of proteins in the mitochondria, on the outer mitochondrial membrane and, as shown here, in the cytosol (Donato et al., 2017; Scott et al., 2012). Additionally, its functional link with αTAT1 in the cytoplasm and the requirement for an acetyl-CoA-generating enzyme for its function on the outer mitochondrial membrane further support its role in protein acetylation. In mitochondria, the abundance of acetyl-CoA, and the presence of a GCN5L1 acetyl-CoA-binding region suggests that GCN5L1 could function as an adaptor protein to spatially align acetyl-CoA with interacting lysine residues. This prospect will be supportive of the observation that non-enzymatic acetylation occurs within mitochondria. However, these concepts will need to be characterized further.

Historically the function of lysosomes was thought to be restricted to the degradation and recycling of cellular waste. However, in the past few decades, we have begun to appreciate that these organelles play an integral role in a diffuse array of cellular processes, including roles in autophagosome and endosome biology, cellular secretion, cell membrane repair, nutrient sensing, signal transduction and energy metabolism (Korolchuk et al., 2011; Settembre et al., 2013). To orchestrate these functions, the regulation of lysosomal positioning within the cell is fundamental, as is the role of microtubules in the modulation of bound molecular motors, which support the spatial-dependent role for distinct lysosomal functions (Matteoni and Kreis, 1987; Mrakovic et al., 2012). Concordant with these observations, a case for a potential multi-organellar role for GCN5L1 can be made, as we found profound lysosomal distribution defects in knockout cells compared to WT cells. In parallel, microtubules are subject to numerous post-translational modifications, and a change in tubulin acetylation is generally proposed to modify tubulin stability (Sadoul and Khochbin, 2016; Wloga et al., 2017). Additionally, the acetylation of tubulin may have distinct functions in different tissues, and proteomics analysis now suggests that tubulin acetylation is not restricted to the K40 residue (Sadoul and Khochbin, 2016). Furthermore, how changes in acetylation modify the function of bound molecular motor proteins additionally requires further validation. Despite these uncertainties, the identification of the more intricate machinery regulating αTAT1 through GCN5L1 and RanBP2 sheds additional insight into the complexity of this regulation. Furthermore, given that GCN5L1 has been shown to modulate the stability of a kinesin motor-binding protein (Donato et al., 2017), the role of GCN5L1 in microtubular function and its effects on organelle biology will be an intriguing arena for ongoing studies.

All animal protocols were in accordance with Institutional Guidelines and approved by the Institutional Animal Care and Use Committee of the National Heart, Lung and Blood Institute of the National Institutes of Health, USA. The mice were maintained on a 12-h-light–12-h-dark cycle and housed at 3–5 mice per cage with free access to water and normal chow diet. All mice were generated in the C56BL/6 background and backcrossed for >10 generations (Wang et al., 2017). GCN5L1 liver knockout mice (LKO) were generated by crossing GCN5L1 flox/flox mice with albumin-Cre flox mice and the littermate GCN5L1 flox/flox mice were used as the controls (WT). For immunoblot analysis, 8-week-old male mice were anesthetized, then livers were removed and homogenized in RIPA buffer, 40 μg total protein was used for each sample.

All cell lines were obtained from the American Type Culture Collection (Manassas, Virginia) and were tested routinely tested for contamination. Primary hepatocytes were isolated from normal chow fed mice at the age of 8–12 weeks as previously described (Liu et al., 2011). Briefly, mice were anesthetized with isoflurane, and the liver was perfused with Krebs–Ringer buffer with glucose and EGTA (100 mM) for 3 min, followed by continuous perfusion with the same buffer containing CaCl₂ (1.4 mM) and collagenase (7000 IU/mouse, type I, Worthington) for 8 min. The livers were extracted, and dissociated cells were filtrated through a 100-μm cell strainer and purified by Percoll (Sigma-Aldrich). Hepatocytes were plated (6×10⁵ cells per well in a six-well plate) onto collagen (Sigma-Aldrich)-coated plates and cultured in DMEM containing 10% FBS and 1% penicillin/streptomycin (P/S). HeLa cells and 293T cells were cultured in DMEM containing 10% fetal bovine serum, 1% P/S and 2 mM glutamine at 37°C and 5% CO₂.

For overexpression studies, primary hepatocytes and HeLa cells were transfected with Lipofectamine 3000 (Life Technologies) following the manufacturer's instructions. 293T cells were transfected with PolyJet In Vitro DNA Transfection Reagent (SignaGen Laboratories). Cells were analyzed 48 h after transfection. For knockdown experiments, cells were transfected with indicated 25 nM siRNAs (ON-TARGETplus SMART-pool or ON-TARGETplus Non-Targeting SMARTpool; GE Dharmacon) with Lipofectamine RNAiMAX Reagent and analyzed 48 h later.

GCN5L1 overexpression adenoviruses were produced using the Adeasy Adenoviral System (Agilent) as described previously (Wang et al., 2017). For primary hepatocyte infection, adenovirus overexpressing either empty vector (control) or GCN5L1 were added to cells at a dose of 20 plaque-forming units/cell and analyzed 48 h later.

Acetyl-CoA levels were measured by HPLC or LC-MS as previously described with slight modification (Shurubor et al., 2017; Yang et al., 2017). Briefly, 10⁹ primary hepatocytes were lysed in buffer A (100 mM sodium phosphate monobasic, 75 mM sodium acetate, pH 4.6). After adding 10% 5-sulfosalicylic acid to reach the final concentration of 5%, the lysates were centrifuged (17,000 g) for 10 min at 4°C, then the supernatants were analyzed by HPLC or LC-MS.

GCN5L1–BirA-transfected 293T cells were grown in 15 cm tissue culture plates for 24 h, then incubated with or without 50 µM biotin (Sigma-Aldrich) for 15–17 h. The cells were then harvested, and total protein levels equalized by protein estimation using a BCA protein assay kit (Pierce). The cell pellets were lysed in 1 ml RIPA lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 1 mM EGTA, 0.1% SDS), supplemented with protease cocktail inhibitor (Roche). The lysates were centrifuged for 30 min at 20,000 g at 4°C. Next, the biotinylated proteins were purified from the supernatant by incubating with Sera-Mag streptavidin-coated magnetic particles (GE healthcare life sciences) for 4 h at 4°C. The beads were separated using a magnet and washed twice with RIPA lysis buffer followed by two washes with TAP lysis buffer (50 mM HEPES-KOH pH 8.0, 100 mM KCl, 10% glycerol, 2 mM EDTA, 0.1% NP-40) and three washes with 50 mM ammonium bicarbonate pH 8.0. Proteins on beads were reduced with 10 mM Tris(2-carboxyethyl)phosphine hydrochloride, and alkylated with 20 mM chloroacetamide, then digested using 5 μg/ml trypsin in 2 M urea at 37°C. The supernatant was harvested, and beads were washed twice with 50 mM ammonium bicarbonate pH 8.0, and washes were pooled together with the supernatant. The pooled tryptic digest was lyophilized and resuspended in 8 M urea, 50 mM ammonium bicarbonate and acidified with 1% (v/v) trifluoroacetic acid (TFA) prior to desalting on SDB-XC (Empore) StageTips as previously described.

The resulting peptides from BioID were separated on a nanoLC system and simultaneously detected in data-dependent analysis mode on an LTQ Orbitrap Fusion machine (Thermo Fisher Scientific, San Jose, CA). By searching the resulting LC-MS/MS raw data against the SwissProt Mouse protein database, the peptide and protein IDs were assigned using Mascot V2.5 (Matrix Science Inc., Boston, MA) on the Proteome Discoverer 2.1 platform (Thermo Fisher Scientific). All samples were analyzed in triplicates from three independent experiments. Protein inclusion criterion required a protein be present in at least two of the three experiments. After proteins were compiled, keratins, histones and endogenously biotinylated carboxylases were discarded before calculating the total peptide spectrum match (PSM) for each individual experiment and the normalized PSM for each protein. The average normalized PSM/observable peptide number (OPN) (av n-PSM/OPN) was then calculated as previously described (Pisitkun et al., 2012).

Primary hepatocytes or HeLa cells were plated on 12 mm diameter, #1 thickness coverslips coated with rat tail type I collagen (neuvitro) and grown to 80% confluency before transfection. At 24 h after transfection, coverslips were washed in PBS and fixed with 4% paraformaldehyde (PFA) in PBS for 10 min. Coverslips were washed twice for 5 min in PBS and permeabilized for 15 min in 0.2% Triton X-100. Cells were blocked for 60 min in 1% BSA, 10% goat normal serum and 0.3 M glycine in PBS and stained for 60 min with primary antibodies (Table S3) in 10% goat normal serum in PBS at room temperature. The secondary antibody was an Alexa-Fluor-488- or 594-conjugated antibody used at a 1:1000 dilution in PBS for 1 h. Coverslips were again washed twice with PBS and mounted on slides using Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific). Images were obtained on an inverted confocal laser-scanning microscope (LSM880; CarlZeiss) fitted with a 63×, 1.4 numerical aperture (NA) objective.

For immunoprecipitation, transfected 293T cells were lysed in 50 mM Tris-HCl pH 7.4, 137 mM NaCl, 10 mM MgCl₂, 10% glycerol and 1% Triton X-100 containing protease inhibitors and centrifuged at 17,000 g for 10 min at 4°C. Supernatants (1–2 mg of total protein in 1 ml) were incubated with 2 µg primary antibody (Table S3) overnight at 4°C. 20 µl Sera-Mag Streptavidin-Coated Magnetic Particles (GE healthcare life sciences) were added to each sample and incubated at room temperature for 1 h. The beads were separated using a magnet and washed four times in TBS containing 0.1% Triton X-100, then the bound proteins were eluted in SDS sample buffer and analyzed by SDS-PAGE and immunoblotting.

In vitro tubulin acetylation assays were performed as previously described (Conacci-Sorrell et al., 2010). Briefly, cytosolic extract from HeLa cells (30 µg) were incubated with in vitro translated GCN5L1 (IVT-GCN5L1) or IVT vector for 1 h in the presence of acetylation buffer (50 mM Tris-HCl pH 8.0, 50 mM EDTA, 10% glycerol, 20 mM acetyl CoA, 10 mM Na butyrate, 10 mMTSA, 1 mMGTP) at 37°C, then the reactants were assayed by immunoblotting.

The recombinant DNA constructs and antibodies used in this work are described in Table S3.

Results are displayed as the mean±s.e.m. A comparison of groups was performed using two-tailed unpaired Student's t-test. A value of P<0.05 was considered to indicate statistical significance.

We thank the NHLBI Light Microscopy, Biophysics and Biochemistry Cores for the use of equipment and technical expertise, and Mary C. Dasso (National Institute of Child Health and Human Development, NIH, Bethesda, MD) for the GFP-tagged RanBP2 plasmids.