ACSS2 S659 phosphorylation exposes the NLS of ACSS2 to bind to importin α5

To determine whether ACSS2 contains a NLS that is exposed for importin binding only after AMPKα-dependent phosphorylation of ACSS2, we mutated the Arg 664/665 in the putative NLS sequences (amino acids 656–668) close to the carboxy-terminus of ACSS2 into alanine (Figure 2A). Immunofluorescent (Figure 2B) and cell fractionation (Figure 2C) analyses demonstrated that Flag-ACSS2 R664/665A, unlike wild-type (WT) ACSS2, was unable to translocate into the nucleus upon glucose deprivation. This result indicated that the NLS containing R664/665 in ACSS2 is essential for glucose deprivation-induced nuclear translocation of ACSS2.

Open in a separate window

Figure 2

ACSS2 phosphorylation at S659 exposes the NLS of ACSS2 to bind to importin α5

(C–H) Immunoblotting analyses were performed with the indicated antibodies.

(A) Schematic of ACSS2 showing its potential NLS predicted by the NLStradamus tool.

(B and C) U87 cells expressing the indicated Flag-ACSS2 proteins were deprived of glucose for 1 h. Immunofluorescent analyses were performed with an anti-Flag antibody and the percentage of nuclear ACSS in 20 cells in each group were quantitated (right panel) using the ImageJ software program (B). Total cell lysates and cytosolic and nuclear fractions were prepared (C). A two tailed Student’s t test was used. represents P < 0.001. (D) U87 cells expressing the indicated SFB-tagged importin α proteins were deprived of glucose for 10 min. A pull-down assay with streptavidin agarose beads was performed.

(E) U87 cells were deprived of glucose for 10 min. Immunoprecipitation with an anti-importin α5 antibody was performed.

(F) U87 cells with or without importin α5 depletion were deprived of glucose for 1 h. Total cell lysates and cytosolic and nuclear fractions were prepared.

(G) Purified GST-importin α5 was mixed with the indicated purified His-ACSS2 proteins in the presence or absence of active AMPK. A GST pull-down assay was performed.

(H) U87 cells expressing the indicated Flag-ACSS2 proteins were deprived of glucose for 10 min. Immunoprecipitation with an anti-Flag antibody was performed.

(I) Parental and the indicated U87 cells with knock-in of ACSS2 S659A or R664/665A were deprived of glucose for 1 h. Immunofluorescent analyses were performed with an anti-ACSS2 antibody. The percentage of nuclear ACSS in 20 cells in each group were quantitated (right panel) using the ImageJ software program. A two tailed Student’s t test was used. represents P < 0.001.

See also Figure S2.

Importin α functions as an adaptor that links NLS-containing proteins with importin β, which then docks the ternary complex at the nuclear-pore complex to facilitate translocation of these proteins across the nuclear envelope. Six importin α family members (α1, α3, α4, α5, α6, and α7) have been identified in humans (Yang et al., 2012). Glucose deprivation induced the binding of ACSS2 to SFB-tagged importin α5 but not to importin α1, α3, α4, α6, or α7 (Figure 2D). This was validated by a co-immunoprecipitation assay demonstrating that endogenous ACSS2 bound to endogenous importin α5 upon glucose deprivation (Figure 2E). In addition, depletion of importin α5 using KPNA1 (coding for importin α5) short hairpin RNA largely blocked glucose deprivation-induced nuclear translocation of ACSS2 (Figure 2F), indicating an essential role for importin α in nuclear translocation of ACSS2.

To determine whether AMPK-mediated ACSS2 S659 phosphorylation plays a role in the binding of ACSS2 to importin α5, we mixed purified WT His-ACSS2, His-ACSS2 S659A, or His-ACSS2 R664/665A with GST-importin α5 in the presence or absence of purified His-AMPK and/or AMP. As shown in Figure 2G, WT His-ACSS2 alone did not bind to importin α5. However, inclusion of AMPK in the presence but not the absence of AMP enabled ACSS2 to interact with importin α5. In contrast, inclusion of His-ACSS2 S659A or His-ACSS2 R664/665A (Figure 2G) or presence of compound C (Figure S2A) inhibited the binding of ACSS2 to importin α5. These in vitro results were supported by a co-immunoprecipitation assay demonstrating that glucose deprivation induced the binding of endogenous importin α5 to WT Flag-ACSS2 but not to Flag-ACSS2 S659A or Flag-ACSS2 R664/665A (Figure 2H). These results strongly suggested that phosphorylation of ACSS2 S659 by AMPK leads to exposure of intramolecularly masked R664/665 in the ACSS2 NLS for binding to importin α5, which is required for nuclear translocation of ACSS2.

We next used the CRISPR/Cas9 genome editing knock-in technology to replace endogenous ACSS2 with ACSS2 S659A, ACSS2 R664/665A, or catalytically inactive ACSS2 T363K in U87 and U251 glioblastoma cells (Figures S2B and S2C) (Ingram-Smith et al., 2006). Expression of ACSS2 S659A and ACSS2 R664/665A, which had catalytic activity comparable with that of WT ACSS2 (Figure S2D), blocked glucose deprivation-induced nuclear translocation of ACSS2 as demonstrated by immunofluorescent (Figure 2I) and immunoblot (Figure S2D) analyses. Of note, expression of ACSS2 R664/665A was still phosphorylated at S659 (Figure S2E). In contrast, the catalytically inactive ACSS2 T363K mutant (Ingram-Smith et al., 2006) (Figure S2C) was still able to translocate into the nucleus of U87 and U251 cells (Figures S2F and S2G), indicating that the observed subcellular redistribution of ACSS2 occurred independently of its catalytic activity. Taken together, these results indicated that AMPK phosphorylates ACSS2 at S659, which promotes the binding of ACSS2 to importin α5 and subsequent nuclear translocation of ACSS2.

TFEB-associated ACSS2 binds to the promoter regions of lysosomal and autophagy genes and promotes expression of these genes

To determine the nuclear function of ACSS2, we immunoprecipitated it, analyzed its interacting proteins (Table S1), and observed that an approximately 70-kDa protein interacted with ACSS2 in U87 cells only under glucose-deprivation condition (Figure 3A). We identified this protein as TFEB via mass spectrometry (Figure S3A). TFEB functions as a master gene in lysosomal biogenesis by coordinating expression of lysosomal hydrolases and membrane proteins and in expression of genes involved in autophagy (Sardiello et al., 2009). Both lysosomal biogenesis and autophagy are specifically required for maintenance of intracellular amino acid pools (Perera et al., 2015). It was shown that AMPK regulates TFEB-dependent regulation of lysosomes (Young et al., 2016). It was also reported that mammalian target of rapamycin (mTOR)-dependent TFEB phosphorylation at S211, which triggered the binding of 14-3-3 proteins to TFEB, caused TFEB to be retained in the cytosol (Roczniak-Ferguson et al., 2012; Settembre et al., 2012). Co-immunoprecipitation analyses demonstrated that TFEB bound to ACSS2 in the nucleus but not in the cytosol in U87 cells (Figure 3B). As expected, glucose deprivation, which inhibited mTOR activity (Mo et al., 2015; Shackelford and Shaw, 2009), resulted in the nuclear translocation and co-localization of TFEB with ACSS2 (Figure S3C). Of note, depletion of TFEB and reconstituted expression of an RNA interference-resistant (r) TFEB R245/248A nuclear translocation-deficient mutant (Roczniak-Ferguson et al., 2012) blocked glucose deprivation-induced nuclear translocation of TFEB but not of ACSS2 (Figure S3B), indicating that TFEB and ACSS2 separately translocated into the nucleus, where they formed a complex. This finding was further supported by mTOR inhibition induced by amino acid withdrawal or treatment with the mTOR inhibitor Torin 1, which did not affect AMPK activity, resulted in nuclear translocation of TFEB but not of ACSS2 (Figure S3C).

Open in a separate window

Figure 3

TFEB-associated ACSS2 binds to the promoter regions of lysosomal and autophagy genes and promotes expression of these genes

(B, G and I) Immunoblotting analyses with the indicated antibodies were performed.

(A) U87 cells with or without expressing Flag-ACSS2 were deprived of glucose for 1 h. Immunoprecipitates of Flag-ACSS2 with an anti-Flag antibody were separated using sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and stained with Coomassie Brilliant Blue. The indicated protein band was excised for mass spectrometric analysis and identified as TFEB (peptide hits, 92).

(B) U87 cells were deprived of glucose for 1 h. Cytosolic and nuclear fractions were prepared. Immunoprecipitation with an anti-TFEB antibody was performed.

(C) Parental U87 cells and the indicated clones of cells with ACSS2 S659A, R664/665A, or T363K knock-in were infected with luciferase-reporter lentiviruses carrying four tandem copies of intact CLEAR elements. These cells were treated with or without glucose starvation for 10 h. Transcription was measured according to firefly luciferase activity. Luciferase activities were normalized to the cell numbers obtained from the duplicated samples. The data are presented as the mean ± standard deviation (SD) for triplicate samples.

(D) Parental and the indicated U87 cells with knock-in of ACSS2 S659A, R664/665A, or T363K mutant were deprived of glucose for 1 h. ChIP analyses using an anti-TFEB antibody were performed. The histogram shows the amount of immunoprecipitated DNA expressed as a percentage of the total input DNA. The data are presented as the mean ± SD for triplicate samples.

(E) Endogenous TFEB-depleted U87 cells with reconstituted expression of WT rTFEB or rTFEB R245/248A were treated with or without glucose starvation for 1 h. ChIP analyses were performed using an anti-ACSS2 antibody. The histogram shows the amount of immunoprecipitated DNA expressed as a percentage of the total input DNA. The data are presented as the mean ± SD for triplicate samples.

(F and G) Parental and the indicated U87 and U251 cells with knock-in of ACSS2 S659A, R664/665A, or T363K mutant were deprived of glucose for 10 h. Quantitative polymerase chain reaction (PCR) analysis (F) of the mRNAs of the indicated genes in U87 cells (F) and immunoblot analyses of U87 and U251 cells (G) with the indicated antibodies were performed. Data were normalized for β-actin mRNA levels and presented as fold changes induced by glucose deprivation (F).

(H) Parental and the indicated U87 cells with knock-in of ACSS2 S659A, R664/665A, or T363K mutant were deprived of glucose for 10 h. Quantitative PCR analysis of the mRNAs of the indicated genes was performed. Data were normalized for β-actin mRNA levels and presented as fold changes induced by glucose deprivation. The data are presented as the mean ± SD for triplicate samples.

(I) Parental and the indicated U87 or U251 cells with knock-in of ACSS2 S659A, R664/665A, or T363K mutant were deprived of glucose for 10 h. Immunoblot analyses with the indicated antibodies were performed.

See also Figures S3, S4, S5 and Tables S1, S2, S3.

To determine the role of ACSS2 S659 phosphorylation in the interaction between ACSS2 and TFEB, we depleted endogenous ACSS2 in U87 cells and expressed NLS-ACSS2 S659A fusion protein (Figure S3D), which contained the NLS of SV40 large T-antigen and ACSS2 S659A and was localized in the nucleus without glucose deprivation (Figure S3E). Glucose deprivation induced the binding of NLS-ACSS2 S659A to TFEB (Figure S3F). These results indicated that ACSS2 S659 phosphorylation, which is required for its nuclear translocation, is not directly involved in the interaction with TFEB.

We next expressed the TFEB R245/248A in ACSS2 R664/665A knock-in U87 cells (Figure S3G) and revealed their interaction upon glucose deprivation (Figure S3H). These results suggested that cytosolic and importin binding-deficient TFEB R245/248A released from 14-3-3 by AMPK-dependent mTOR inhibition is able to bind to cytosolic and importin-binding deficient ACSS2 R664/665A. However, WT TFEB did not bind to ACSS2 R664/665A, suggesting that importin has greater ability to bind to TFEB than does ACSS2 R664/665A. This finding was supported by immunblotting analysis showing that depletion of importin 8 (IPO 8), which is specific for TFEB-nuclear translocation (Perera et al., 2015), enabled WT TFEB to bind to ACSS2 R664/665A in response to glucose deprivation (Figure S3I). These results strongly suggested that glucose deprivation resulted in accessibility of AMPK-phosphorylated WT ACSS2 and dephosphorylated WT TFEB to importin for separate nuclear translocation.

To determine the interaction regions of TFEB and ACSS2, we performed an in vitro binding assay by mixing purified WT ACSS2, WT TFEB, and different truncation mutants of both ACSS2 and TFEB. Figure S3J shows that non-phosphorylated WT ACSS2 binds to non-phosphorylated WT TFEB, and their interaction requires the N-termini of both proteins. In addition, co-immunoprecipitation analyses did not detect complex formation of ACSS2 with the other TFE family members (Perera et al., 2015), TFE3, TFEC or MITF (Figure S3K), suggesting that ACSS2 specifically binds to TFEB.

To determine whether the binding of ACSS2 to TFEB regulates the transcriptional activity of TFEB, we infected U87 cells with luciferase-reporter lentiviruses containing the Coordinated Lysosomal Expression and Regulation (CLEAR) element, a motif recognized by TFEB. A luciferase assay demonstrated that glucose deprivation induced TFEB activation in U87 cells with knock-in of WT ACSS2. This induction was largely inhibited in U87 cells with knock-in of ACSS2 S659A, ACSS2 R664/665A, or inactive ACSS2 T363K (Figure 3C). These results indicated that TFEB activation is dependent on nuclear translocation of ACSS2 and its catalytic activity.

TFEB activation upregulates expression of lysosomal genes, including the lysosomal enzymes cathepsin A (encoded by CTSA), β-glucuronidase (encoded by GBA), and β-glucosidase (encoded by GUSB), as well as the lysosomal membrane protein LAMP1 (encoded by LAMP1) (Sardiello et al., 2009). A chromatin immunoprecipitation (ChIP) assay with antibodies against TFEB or ACSS2 demonstrated that glucose deprivation results in the binding of TFEB (Figure 3D) and ACSS2 (Figure 3E) to the promoter regions of CTSA, GBA, GUSB, and LAMP1; in addition, the binding TFEB and ACSS2 was blocked by knock-in of ACSS2 mutants (S659A, R664/665A, and T363K) and reconstituted expression of the nuclear translocation-deficient rTFEB R245/248A mutant, respectively. These results indicated that TFEB and ACSS2 are mutually required for their binding to the promoter regions of lysosomal genes. In line with these findings, glucose deprivation induced mRNA (Figure 3F) and protein (Figure 3G) expression for these lysosomal genes, which was largely abrogated by knock-in of ACSS2 mutants (S659A, R664/665A, and T363K). The effect of expression of these mutants was rescued by overexpression WT Flag-ACSS2 (Figure S4A). In contrast, ACSS2 R664/665A expression did not alter the mRNA levels of glucose deprivation-upregulated and non-TFEB target genes p21, cpt1c, and cyclin G (Figure S4B) (Bungard et al., 2010). These findings indicated that the binding of ACSS2 to the promoter regions of lysosomal genes and the catalytic activity of ACSS2 are required for expression of these genes.

It is known that TFEB activation also upregulates expression of genes involved in autophagy formation (Perera et al., 2015), and glucose deprivation increases autophagy (Russell et al., 2014). As expected, we found that glucose deprivation induced the binding of TFEB (Figure S4C) and ACSS2 (Figure S4D) to the promoter regions of MAP1LC3B, ATG3, and WIPI-1 as well as mRNA (Figure 3H) and protein (Figure 3I) expression of these genes; this induction was blocked by knock-in of the ACSS2 S659A, R664/665A, and T363K mutants (Figures 3H and ​and3I)3I) whereas the lysosomal inhibitor, chloroquine, had no effect on the inhibited LC3B expression mediated by ACSS2 R664/665A (Figure S4E). These results suggested that glucose deprivation enhanced-autophagosomal gene expression, which is regulated by nuclear ACSS2/TFEB-dependent gene transcription, is not induced by potential lysosomal defect. In addition, compound C treatment blocked glucose deprivation-induced expression of both autophagosomal and lysosomal genes in GBM cells (Figure S4F) while reconstituited expression of ACSS2 R664/665A in H1299 non-small cell lung cancer cells and PANC-1 pancreatic carcinoma cells (Figure S4G) also blocked these gene expression (Figures S4H and S4I) and lysosomal and autophagic expansion (Figures S4J and S4K), suggesting the observed effects are not cell line- or tumor type-dependent. Of note, reconstituted expression of phosphomimic ACSS2 S659D mutant (Figure S4L), which translocated into the nucleus in the absence of glucose deprivation (Figure S4M), failed to induce TFEB-dependent gene expression (Figure S4N), suggesting that nuclear translocation of both ACSS2 and TFEB are required for TFEB-dependent gene expression.

To determine the global genomic regulation by ACSS2 and TFEB, we performed chromatin immunoprecipitation combined with high-throughput sequencing (ChIP–seq) experiments using an anti-ACSS2 or anti-TFEB antibody. Notably, we found 81% of the ACSS2 target genes were also TFEB target genes, including genes related autophagy and lysosomal pathways (Figures S5A and S5B, Tables S2 and S3). The peak scores corresponding to the union of binding sites between ACSS2 and TFEB ChIP showed significant correlation (Pearson correlation coefficient = 0.53, P < 0.001) (Figure S5C). In addition, signals from both ACSS2 and TFEB ChIP were enriched at transcription start site (TSS) regions (−1k to +1k bp) upon glucose deprivation and the overall binding profile of ACSS2 ChIP is highly similar to TFEB ChIP (Figure S5D). Transcription factor motif analysis indicated a single dominant de novo motif for both ACSS2 and TFEB ChIP with a high degree of concordance (Figure S5E). The overlapping region of both motifs corresponded to the E-box sequence (5′-CANNTG-3′) that TFEB recognizes. In addition, the combination of the left flanking GT of ACSS2 motif, the overlapping region and the right flanking AC of TFEB motif exactly corresponded to the CLEAR-box sequence (5′-GTCACGTGAC-3′) which is specifically recognized by TFEB and presents in the regulatory regions of many lysosomal and autophagosomal genes. Indeed, we found the genes related to lysosomal and autophagosomal pathways were significantly enriched in the ACSS2 ChIP peaks (P < 0.001 for both pathways) (Figure S5F). Together, these results support the theory of TFEB, in complex with nuclear ACSS2, binding to the promoter regions of lysosomal and autophagosomal gene in order to promote the expression of these genes.

Nuclear ACSS2 mediates local generation of acetyl-CoA and subsequent histone H3 acetylation in the promoter regions of lysosomal and autophagosomal genes

ACSS2 produces acetyl-CoA from acetate (Watkins et al., 2007). However, glucose deprivation and limited amount of acetate in regular DMEM restrict the production of acetyl-CoA from citrate by ACL, from pyruvate by PDC, and from medium-uptake acetate by cytosolic ACSS2. As expected, glucose deprivation and regular DMEM culture resulted in a large reduction of nuclear acetyl-CoA levels in U87 cells (Figure 4A). Intriguingly, knock-in of ACSS2 S659A, R664/665A, and T363K caused a further reduction of nuclear acetyl-CoA levels (Figure 4A), indicating that nuclear ACSS2 is involved in production of nuclear acetyl-CoA. In line with this finding and the results revealing the binding of ACSS2 to the promoter regions of the lysosomal (Figure 3E) and autophagosomal genes (Figure S4D), a ChIP assay with an anti-total acetylated histone H3 antibody demonstrated that glucose deprivation induced histone H3 acetylation in the promoter regions of lysosomal (Figure 4B) and autophagosomal (Figure 4C) genes in U87 cells. Similar results were also observed using an anti-H3K9 acetylation and an anti-total acetylated histone H2B antibodies (Figures S5G and S5H). Importantly, knock-in of ACSS2 S659A, R664/665A, and T363K mutants (Figures 4B and ​and4C)4C) or compound C treatment (Figure S5I) blocked histone H3 acetylation in these promoter regions, whereas depletion of ACL or PDHA1 had little effect on this acetylation (Figure S5J). Of note, ACSS2 R664/665A expression also blocked glucose deprivation-induced histone H3 Arg17 dimethylation in the autophagosomal gene promoters (Figure S5K), and H3 Arg17 dimethylation was known be regulated by co-activator-associated arginine methyltransferase 1 (CARM1) in complex with TFEB (Shin et al., 2016). These results indicated that nuclear ACSS2 plays an instrumental role in nuclear acetyl-CoA production and subsequent histone H3 acetylation and methylation at the promoter regions of lysosomal and autophagosomal genes upon glucose deprivation.

Open in a separate window

Figure 4

Nuclear ACSS2 mediates local generation of acetyl-CoA and subsequent histone H3 acetylation in the promoter regions of lysosomal and autophagosomal genes

(A–C) Parental and the indicated U87 cells with knock-in of ACSS2 S659A, R664/665A, or T363K mutant were deprived of glucose for 1 h. The levels of nuclear acetyl-CoA were measured (A). ChIP analyses using an anti-acetylated H3 antibody were performed. PCR with the primers against the promoter regions of the indicated lysosomal (B) and autophagosomal (C) genes was performed. The histogram shows the amount of immunoprecipitated DNA expressed as a percentage of the total input DNA. The data are presented as the mean ± SD for triplicate samples.

(D and E) U87 and U251 cells were deprived of glucose for 1 h in the absence or presence of TSA (500 nM). Acid-extracted histones visualized using Coomassie Blue staining were subjected to immunoblot analyses of acetylated histone levels with the indicated antibodies (D). Nuclear fractions of the cells were prepared, and the levels of nuclear acetyl-CoA were measured (E).

(F) Parental and the indicated U87 cells with knock-in of ACSS2 S659A, R664/665A, or T363K mutant were deprived of glucose for 1 h. The cells were permeabilized with digitonin (30 μg/ml) for 5 min and treated with the indicated concentrations of acetyl-CoA for 30 min. ChIP analyses using an anti-acetylated H3 antibody were performed. The histogram shows the amount of immunoprecipitated DNA expressed as a percentage of the total input DNA. The data are presented as the mean ± SD for triplicate samples.

See also Figure S5.

Given the limited amount of exogenous acetate in DMEM, we hypothesized that ACSS2 uses acetate released from deacetylated proteins to produce acetyl-CoA for epigenetic regulation (Lyssiotis and Cantley, 2014). Because the half-life of histone acetylation is on the order of minutes, a considerable fraction of acetate can be produced in vivo by the turnover of histone acetylation (Comerford et al., 2014). To confirm this, we examined the total acetylation levels in U87 and U251 cells for histones H3, H4, and H2B, which decreased markedly upon glucose deprivation (Figure 4D), although we detected increased histone H3 acetylation in the promoter regions of lysosomal and autophagosomal genes. Of note, pretreatment of the cells with the histone deacetylase inhibitor TSA, which blocked glucose-deprivation-induced reductions in the total acetylation levels of the three histones (Figure 4D), decreased nuclear acetyl-CoA levels (Figure 4E) and largely reduced glucose deprivation-induced histone H3 acetylation in the promoter regions of lysosomal genes (Figure S5L) and these gene expression (Figure S5M). In agreement with these results, addition of acetyl-CoA to regular DMEM-cultured and digitonin-permeabilized cells rescued inhibited histone H3 acetylation in the promoter regions of lysosomal (Figure 4F) and autophagosomal (data not shown) genes in U87 cells with knock-in of ACSS2 S659A, R664/665A, and T363K mutants in a dose-dependent manner (Figure 4F). Taken together, these results strongly suggested that lysosomal and autophagosomal gene promoter-associated ACSS2 utilizes the acetate generated by turnover of histone acetylation to produce acetyl-CoA, which is required for local histone H3 acetylation in the promoter regions of lysosomal and autophagosomal genes.

Nuclear ACSS2 promotes lysosomal biogenesis and autophagy

To determine the functional consequence of nuclear translocation of ACSS2, we performed immunofluorescent analyses of U87 cells, which demonstrated increased staining intensity for the lysosome membrane protein LAMP1 under glucose deprivation (Figure 5A); this increased staining was blocked by knock-in of the ACSS2 S659A, R664/665A, and T363K mutants. Similarly, glucose deprivation-induced autophagy, as evidenced by increased numbers of endogenous LC3B (Figure 5B) and GFP-LC3 puncta (Figure S6A) in the cells and increased conversion of LC3B-I to LC3B-II (a phosphatidylethanolamine derivative of LC3-I) (Figure 3I), was inhibited by knock-in of the ACSS2 S659A, R664/665A, and T363K mutants. Consistently, electron microscopic analyses of U87 cells demonstrated that glucose deprivation induced marked increases in lysosome and autophagosome numbers per cell, and this increase was blocked by knock-in of the ACSS2 R664/665A and T363K mutants (Figure 5C). As expected, glucose deprivation for 24 h decreased the survival of U87 cells, and the observed cell death rate was largely enhanced by knock-in of ACSS2 S659A, R664/665A or T363K mutant (Figure 5D), indicating that glucose deprivation-induced autophagy and lysosomal biogenesis are required for cell survival in response to this metabolic stress.

Open in a separate window

Figure 5

Nuclear ACSS2 promotes lysosomal biogenesis and autophagy

(A and B) Parental and the indicated U87 cells with ACSS2 S659A, R664/665A, or T363K knock-in were deprived of glucose for 10 h. Immunofluorescent analyses were performed with anti-LAMP1 (A) and anti-LC3B (B) antibodies. The immunofluorescence intensity in 100 cells (A) or the numbers of LC3B puncta in 20 cells (B) was quantitated using the ImageJ software program.

(C) Parental and the indicated U87 cells with knock-in of ACSS2 R664/665A or T363K mutant were deprived of glucose for 10 h. Electron microscopic analyses were performed. The blue and red arrows point to the lysosome and autophagosome, respectively (left panel). The lysosomes and autophagosomes in 10 cells in each group were quantitated (right panel).

(D) Parental and the indicated U87 cells with knock-in of ACSS2 S659A, ACSS2 R664/665A, or T363K mutant were deprived of glucose for 24 h and stained with trypan blue (0.5%). The viable cells were counted.

See also Figure S6.

Experimental Model and Subject Details

Histologic evaluation and immunohistochemical staining

Mouse tumor samples were fixed and prepared for staining. The samples were stained with Mayer’s hematoxylin and eosin (BioGenex). Slides were then mounted using Universal mount (Research Genetics).

The immunohistochemical staining was performed using the VECTASTAIN ABC kit (VECTOR LABORATORIES, CA) according to manufacturer’s instructions. Sections of paraffin-embedded xenograft tissue were stained with antibodies against ACSS2, ACSS2 pS659, LAMP1, LC3B or nonspecific IgG as a negative control. The tissue sections from paraffin-embedded human GBM specimens were stained with antibodies against phospho-ACC S79, phospho-ACSS2 S659, or nonspecific IgG as a negative control. We quantitatively scored the tissue sections according to the percentage of positive cells and staining intensity, as previously defined (Li et al., 2016). We assigned the following proportion scores: 0 if 0% of the tumor cells showed positive staining, 1 if 0% to 1%, 2 if 2% to 10%, 3 if 11% to 30%, 4 if 31% to 70%, and 5 if 71% to 100%. We also rated the intensity of staining on a scale of 0 to 3: 0, negative; 1, weak; 2, moderate; and 3, strong. We then combined the proportion and intensity scores to obtain a total score (range: 0, 2–8). All patients had received standard adjuvant radiotherapy after surgery, which was followed by treatment with an alkylating agent (temozolomide in most cases). The use of human brain tumor specimens and the database was approved by the Institutional Review Board at Capital Medical University and The University of Texas MD Anderson Cancer Center.

Cell lines and cell culture conditions

The human glioblastoma cell lines U87 and U251 authenticated using short tandem repeat profiling in The University of Texas MD Anderson Cancer Center and AMPKα1/2 double-knockout mouse embryonic fibroblasts (a gift from B. Viollet, Institut Cochin, Paris, France), H1299 human lung carcinoma cells, and PANC-1 human pancreas ductal adenocarcinoma cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% dialyzed bovine calf serum (HyClone).

Method Details

In vitro kinase assay

Kinase reactions were performed as described previously (Ji et al., 2009). In brief, purified recombinant AMPK or CAMKKβ (10 ng) was incubated with ACSS2 or AMPKα (100 ng) in 25 μl of kinase buffer (50 mM Tris-HCl, pH 7.5, 100 mM KCl, 5 mM MgCl₂, 1 mM Na₃VO₄, 50 μM DTT, 5% glycerol, 50 μM ATP, 10 μCi of [γ³²P]-ATP) at 25 °C for 1 h. The reactions were terminated with the addition of SDS-PAGE loading buffer and heated to 100 °C. The reaction mixtures were then subjected to SDS-PAGE.

Purification of recombinant proteins

Expression of His-ACSS2, His-ACSS2 truncations, His-TFEB, Flag-ACSS2, His-TFEB truncations, Flag-TFEB, and glutathione S-transferase (GST)-importin α5 was induced in bacteria, and these proteins were purified as described previously (Xia et al., 2007). Briefly, BL21(DE3) cells expressing (His)₆-, GST-tagged or Flag-tagged ACSS2, TFEB, and importin α5 were cultured in 250 ml of LB medium and treated with isopropyl β-D-1-thiogalactopyranoside (IPTG) for 16 h at 30 °C before lysis via sonication. Flag-tagged WT AMPKα2 and AMPKα2 T172A recombinant proteins were expressed in 293T cells. Briefly, a pcDNA3.1 plasmid (10 μg) expressing Flag-tagged WT AMPKα2 or AMPKα2 T172A recombinant proteins was transfected into 293T cells seeded in a 15cm-diameter dish with confluence of 70%. Forty eight hours after transfection, the cells were incubated with fresh DMEM supplemented with 10% bovine calf serum for 1 h and then harvested with a non-denaturing lysis buffer (20 mM Tris HCl pH 8.0, 137 mM NaCl, 1% NP-40). For (His)₆-tagged proteins, cell lysates were loaded onto a Ni-NTA column (GE Healthcare Life Sciences) followed by washing with five column volumes of 20 mM imidazole and subsequent elution with 250 mM imidazole. For GST-tagged proteins, cell lysates were loaded onto a GSTrap HP column (GE Healthcare Life Sciences) followed by washing with five column volumes of phosphate-buffered saline (PBS) and subsequent elution with 10 mM reduced glutathione. For Flag-tagged proteins, cell lysates were loaded onto a column containing ANTI-FLAG M2 agarose affinity gel followed by washing with five column volumes of phosphate-buffered saline (PBS) and subsequent elution with 100 μg/ml 3 × Flag peptide. The proteins in 10-kDa spin columns were desalted via washing twice with ice-cold PBS. The proteins were then loaded onto a HiPrep 16/60 Sephacryl S-200 HR gel filtration column (GE Healthcare Life Sciences) to remove contaminated proteins. The purification efficiency was examined using SDS-PAGE and colloidal Coomassie Brilliant Blue (G-250) staining, with superior sensitivity (as low as 1 ng of protein per band) (Dyballa and Metzger, 2009).

Immunoprecipitation and immunoblot analysis

Extraction of proteins from cultured cells was performed with a lysis buffer (50 mM Tris-HCl [pH 7.5], 0.1% SDS, 1% Triton X-100, 150 mM NaCl, 1 mM dithiothreitol, 0.5 mM EDTA, 100 M PMSF, 100 M leupeptin, 1 M aprotinin, 100 M sodium orthovanadate, 100 M sodium pyrophosphate, and 1 mM sodium fluoride). Cell extracts were clarified by centrifugation at 13,400 g, and the supernatants (2 mg protein/ml) were subjected to immunoprecipitation with the indicated antibodies. After overnight incubation at 4^°C, protein A or G agarose beads were added and left for an additional 3 h. Immunocomplexes were washed with lysis buffer 3 times and then subjected to immunoblot analyses with corresponding antibodies as described previously (Lu et al., 1998). The band intensity was quantified using the Image Lab software program (Bio-Rad).

Transfection

Cells were plated at a density of 4 × 10⁵ per 60-mm diameter dish 18 h prior to transfection. Transfection was performed using HyFect transfection reagent (Denville Scientific, NJ) according to the manufacturer’s instructions (Xia et al., 2007).

Reverse transcriptase-PCR

Total RNA was extracted from cultured tumor cells using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Total RNA (1 μg) was used for each sample in a 20-μl reaction using an iScript cDNA synthesis kit (Bio-Rad). One microliter of the cDNA library was used in a 25-μl PCR. Fast SYBR Green Master Mix (Bio-Rad) was used to determine the threshold cycle (Ct) value of each sample using a CFX96 real-time PCR detection system (Bio-Rad). β-actin served as the normalization gene in these studies. The relative expression levels for the target genes were given by 2^ΔCt (the Ct of β-actin minus the Ct of the target gene). Primer sequences used for PCR were list in Table S4.

Mass spectrometric analysis

A protein band visualized via Coomassie Brilliant Blue staining was excised from SDS-PAGE and digested in gel in 50 mM ammonium bicarbonate buffer containing RapiGest (Waters Corporation) overnight at 37 °C with 200 ng of modified sequencing-grade trypsin (Promega). The digested samples were analyzed using high-sensitivity liquid chromatography tandem mass spectrometry with an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific). Proteins were identified by searching the fragment spectra against the Swiss-Prot protein database (EBI) using the Mascot search engine (version 2.3; Matrix Science) with the Proteome Discoverer software program (version 1.4; Thermo Fisher Scientific).

GST pull-down assay

Equal amount of His-tagged purified protein (200 ng/sample) were incubated with 100 ng of GST fusion proteins together with glutathione agarose beads in a modified binding buffer (50 mM Tris-HCl [pH 7.5], 1% Triton X-100, 150 mM NaCl, 1 mM dithiothreitol, 0.5 mM EDTA, 100 M PMSF, 100 M leupeptin, 1 M aprotinin, 100 M sodium orthovanadate, 100 M sodium pyrophosphate, and 1 mM sodium fluoride) at 4^°C. The glutathione beads were then washed four times with binding buffer and the bound proteins were boiled with 4 × Laemmli buffer prior to electrophoresis on SDS-polyacrylamide gels.

Cell viability analysis

2 × 10⁵ cells were plated in DMEM with 10% bovine calf serum with or without glucose deprivation. The viable cells were counted using trypan blue (0.5%) staining 24 h after seeding. The data were presented as the mean ± SD from three independent experiments.

DNA constructs and mutagenesis

PCR-amplified human ACSS1, ACSS2, ACSS3, importin α5, importin α1, importin α4, importin α3, importin α6, importin α7, TFEB, TFEC, TFE3, and MITF were subcloned into a pCold I, pGEX-4T-1, or pcDNA 3.1 vector. PX458-HF plasmid containing a high-fidelity Streptococcus pyogenes Cas9 bearing mutations of N497/R661/Q695/Q926A(Kleinstiver et al., 2016), pcDNA3.1 Flag-ACSS2 S659A, pcDNA3.1 Flag-ACSS2 R664/665A, pcDNA3.1 Flag-TFEB R245/248A, pCold I ACSS2 S659A, pCold I ACSS2 R664/665A, and pCold I ACSS2 T363K were constructed using a QuikChange site-directed mutagenesis kit (Stratagene). PcDNA3.1 Flag-TFEB contained nonsense mutations of G959A, C960A, and A962G.

pGIPZ shRNA was constructed by ligation of oligonucleotide targeting human ACSS2, KPNA1, or TFEB into the Xho I/Mlu I digested pGIPZ vector and TRIPZ human ACSS2 inducible shRNA was purchased from GE Dharmacon (Lafayette, CO).

ChIP assay

ChIP was performed using a SimpleChIP plus Enzymatic Chromatin IP kit. Chromatin prepared from cells in a 10-cm dish was used to determine total DNA input and for overnight incubation with specific antibodies or normal mouse IgG. Primer sequences used for PCR were list in Table S5.

ChIP–seq analyses

Sequencing libraries were prepared using the NEBNext DNA Library Prep Master Mix Set according to the Illumina’s instructions. The libraries were single-end sequenced on an Illumina HiSeq 2000 (The University of Texas MD Anderson Cancer Center). Sequencing reads were aligned to hs37d5 reference genome using BWA v0.7.12 followed by duplicate reads removal using sambamba v0.6.5 (Tarasov et al., 2015). MACS2′s callpeak function was used with default parameters to call peaks from alignment results (Zhang et al., 2008). To annotate the peaks called by MACS2, the annotatePeaks.pl function of HOMER v4.8.3 was used with GENCODE19 reference gene annotation (Harrow et al., 2012). We then filtered the annotated peaks using the following criteria: each peak must have the “promoter-TSS” annotation, be within the canonical chromosomes as well as having a peak-score ≥ 2. The remaining peaks were further filtered according to desired distance from TSS.

In order to search for motifs present in the surrounding region of the peaks, we used the findMotifsGenome.pl program of HOMER to search for five motifs using the -S 5 option and constrained the search region to be +/−50bp of a peak position using the -size 50 option.

Total annotated, lysosomal, and autophagosomal gene sets, extracted from ENSEMBL GRCh37 (n=63,737), hLGDB database (http://lysosome.unipg.it [n=443]) and HADb database (http://autophagy.lu [n=223]), were used to determine whether the lysosome and autophagy pathways in the ACSS2- or TFEB-binding regions of the genes were significantly enriched via hypergeometric test (Brozzi et al., 2013; Cunningham et al., 2015; Moussay et al., 2011). The hypergeometric test was performed by using the R statistical program. Pearson correlation coefficient was calculated using R for the peaks scores corresponding to the union of the filtered binding sites found in ACSS2 and TFEB ChIP-seqs. Permutation test was performed 12 million times to determine the significance of the correlation. Heatmaps were plotted using bamCoverage, computeMatrix and plotHeatmap tools from the deepTools software suite (Ramirez et al., 2016).

Immunofluorescent analysis

Cells were fixed and incubated with primary antibodies at a dilution of 1:100, fluorescence dye-conjugated secondary antibodies, and DAPI, according to standard protocols. Cells were examined using a deconvolutional microscope (Zeiss, Thornwood, NY) with a 40 × oil immersion objective. Axio Vision 4 software from Zeiss was used to deconvolve Z-series images. AxioVision 4 module co-localization (Zeiss, Thornwood, NY) was used for the co-localization analysis of deconvoluted images (Fang et al., 2007).

Autophagy analyses

Total numbers of LC3B or GFP-LC3B puncta per cell were quantified from confocal images using ImageJ and the Analyse Particles Plugin (a constant threshold for all of the images within each experiment was applied). Twenty cells per condition were used for quantification.

Cell permeabilization

GBM cells (3 × 10⁶) were seeded onto 10-cm dishes. Cell permeabilization then was performed as described previously (Salabei et al., 2014). Briefly, cells were washed with 3× PBS and then incubated in 30 μg/ml digitonin/PBS for 5 min followed by treatment with DMEM containing indicated concentrations of acetyl-CoA.

Subcellular fractionation

Cells at 60% confluence in 60-mm well dishes were treated with or without glucose starvation or 2-DG for indicated times. Nuclear and cytosolic fractions were isolated from the cells using a nuclei/cytosol fractionation kit (BioVision). Nuclear and cytosolic proteins were used in immunoblot analyses.

Measurement of nuclear acetyl-CoA concentrations

Isolation of nuclei was performed using Nuclei PURE Prep kit. Briefly, adherent cells (3 × 10⁶) washed with PBS were scraped from the dishes in the presence of lysis buffer. Cells suspended in lysis buffer were placed on top of a 1.8 M sucrose gradient and then centrifuged at 30,000 × g for 45 min in a precooled swinging bucket ultracentrifuge. Nuclei at the bottom of the centrifuge were collected and washed with the nucleus storage buffer provided in Nuclei PURE Prep kit. Purity of nuclei was assessed by immunoblot, and the acetyl-CoA concentrations in the isolated nuclei were immediately measured using an acetyl-CoA fluorometric assay kit.

Histone acid extraction

Cells (3 × 10⁶) suspended in 10 ml of medium were seeded in 10-cm dishes and treated with or without glucose deprivation. Cells were harvested and lysed in Triton Extraction Buffer (PBS containing 0.5% Triton X-100 [v/v], 2 mM PMSF, 0.02% NaN₃ [w/v]). The pellet was collected using a centrifuge and washed twice with Triton Extraction Buffer, and 200 μl of 0.2 N HCl was used to extract the histones overnight at 4 °C. Acid-extracted histones were used in Coomassie Brilliant Blue staining or immunoblot analysis.

Lentiviral infections and luciferase assays

To construct a CLEAR sequence-containing lentiviral reporter vector, an oligonucleotide containing four tandem copies of the CLEAR sequence were inserted into pGreenFire1 lentivector (System Biosciences, CA) containing an EF1-Puro selection cassette by digestion of pGreenFire1 with EcoRI and SpeI restriction enzymes. Lentiviruses were produced by cotransfecting 293FT cells with pGreenFire1-CLEAR and three packaging plasmids (pLP1, pLP2, and pLP/VSVG). Lentiviruses were harvested 72 h after transfection, centrifuged to remove cell debris, and filtered through 0.45 μm membrane (Millipore, Bedford, MA). To avoid multiple genomic integrations of the target gene, a total of 6 × 10⁵ cells seeded in a 6-well plate were infected by 6 × 10⁵ infectious units of lentiviruses (multiplicity of infection [MOI] = 1). The cell pools stably expressing luciferase under promoter of 4×CLEAR-SV40 were generated by treating lentivirus-infected cells with puromycin (5 μg/ml) for 7 days. To determine the ability of the nuclear ACSS2 to promote transcription, the cells were treated with or without glucose starvation, and luciferase assays were performed using a Luciferase Assay System (Promega). Luciferase activities were normalized to the cell number obtained from the duplicated samples.

Electron microscopy

Samples were fixed with a solution containing 3% glutaraldehyde plus 2% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.3, and then washed in 0.1 M sodium cacodylate buffer and treated with 0.1% Millipore-filtered cacodylate-buffered tannic acid, postfixed with 1% buffered osmium tetroxide for 30 min, and stained en bloc with 1% Millipore-filtered uranyl acetate. The samples were dehydrated in increasing concentrations of ethanol, infiltrated, and embedded in LX-112 medium. The samples then were polymerized in a 60 °C oven for approximately 3 days. Ultrathin sections of the samples were cut using an Ultracut microtome (Leica), stained with uranyl acetate and lead citrate in an EM Stainer (Leica), and examined under a JEM-1010 transmission electron microscope (JEOL) at an accelerating voltage of 80 kV. Digital images were obtained using an AMT Imaging System (Advanced Microscopy Techniques).

Genomic editing

Genomic mutations in U87 and U251 cells were created using the CRISPR/Cas9 system as described previously (Sander and Joung, 2014). Single guide RNAs (sgRNAs) were designed to target the genomic areas adjacent to human ACSS2 mutation sites. Annealed guide oligos containing overhangs were inserted into the high-fidelity PX458-HF vector digested with the BbsI restriction enzyme. In a 24-well plate, U87 and U251 cells at 60% confluence were co-transfected with a single-stranded donor oligonucleotide (20 pmoles) used as a template to introduce the mutations and a vector (0.5 g) able to co-express an sgRNA targeting the ACSS2 gene and WT hSpCas9 tagged with green fluorescent protein. Twenty-four hours after transfection, cells were trypsinized and diluted in a medium for single-cell seeding into a 96-well plate, and green fluorescent protein-positive cells were marked and subjected to genomic DNA extraction. Genotyping was performed by sequencing PCR products amplified from primers spanning the mutation area. Primer sequences used for PCR were list in Table S6.

Measurement of ACSS2 activity

ACSS2 converts ATP, acetate, and CoA into AMP, pyrophosphate, and acetyl-CoA. The generation of pyrophosphate was measured using a coupled enzymatic assay. The reaction mixture contained pyrophosphatase, which converts pyrophosphate to phosphate, and purine nucleoside phosphorylase, which converts nucleoside analog (MESG) to a compound that absorbs light at 360 nm in the presence of phosphate. A typical 100-μl reaction mixture in a 96-well plate includes 50 M CoA, 100 μM ATP, 5 mM sodium acetate, 0.2 mM MESG, 0.2% bovine serum albumin, 20 nM ACSS2, 0.4 M purine nucleoside phosphorylase, 1 U/ml yeast pyrophosphatase, 1 mM MgCl₂, 150 mM NaCl, and 50 mM HEPES (pH 7.5). The increasing absorbance was read at 360 nm in a kinetic mode for 1 h using a microplate reader (BMG LABTECH).

Quantification and statistical analysis

All data represent the mean ± SD of at least three independent experiments. Sample number (n) indicates the number of independent biological samples in each experiment. Sample numbers and experimental repeats are indicated in figure legends. Pair-wise comparisons were performed using a two-tailed Student T-test unless specifically indicated. P values less than 0.05 were considered significant. Analyses were performed using the Microsoft Excel.

We thank Kenneth Dunner and Alicia Ledoux for technical support and Don Norwood for critical reading of this manuscript. This work was supported by National Cancer Institute grants 2R01 CA109035, 1R0 CA169603, and 1R01 CA204996 (Z.L.), National Institute of Neurological Disorders and Stroke grant 1R01 NS089754 (Z.L.), the National Institutes of Health/National Cancer Institute MD Anderson Support Grant (P30CA016672), the National Natural Science Foundation of China (No. 81672872 and No. 81472386 to C.Q.). Z.L. is a Ruby E. Rutherford Distinguished Professor.