Liver p53 is stabilized upon starvation and required for amino acid catabolism and gluconeogenesis.

The ability to adapt cellular metabolism to nutrient availability is critical for survival. The liver plays a central role in the adaptation to starvation by switching from glucose-consuming processes and lipid synthesis to providing energy substrates like glucose to the organism. Here we report a previously unrecognized role of the tumor suppressor p53 in the physiologic adaptation to food withdrawal. We found that starvation robustly increases p53 protein in mouse liver. This induction was posttranscriptional and mediated by a hepatocyte-autonomous and AMP-activated protein kinase-dependent mechanism. p53 stabilization was required for the adaptive expression of genes involved in amino acid catabolism. Indeed, acute deletion of p53 in livers of adult mice impaired hepatic glycogen storage and induced steatosis. Upon food withdrawal, p53-deleted mice became hypoglycemic and showed defects in the starvation-associated utilization of hepatic amino acids. In summary, we provide novel evidence for a p53-dependent integration of acute changes of cellular energy status and the metabolic adaptation to starvation. Because of its tumor suppressor function, p53 stabilization by starvation could have implications for both metabolic and oncological diseases of the liver.-Prokesch, A., Graef, F. A., Madl, T., Kahlhofer, J., Heidenreich, S., Schumann, A., Moyschewitz, E., Pristoynik, P., Blaschitz, A., Knauer, M., Muenzner, M., Bogner-Strauss, J. G., Dohr, G., Schulz, T. J., Schupp, M. Liver p53 is stabilized upon starvation and required for amino acid catabolism and gluconeogenesis.

p53 integrates a broad range of cellular stress signals to coordinate cell cycle inhibition and senescence, thereby blocking proliferative expansion of aberrant cells that could follow a neoplastic trajectory (1, 2). This universal survival mechanism, along with the fact that p53 is mutated in more than 50% of all cancers (3), has rendered p53 as one of the best-investigated molecules in medical research. More recently, and in line with the well-known Warburg effect (4), it was recognized that p53 also regulates cancer cell metabolism (5, 6) by promoting mitochondrial oxidative phosphorylation (7) while inhibiting aerobic glycolysis (8). Compared to its tumor suppressor properties, p53’s physiological role in nontransformed cells and tissues is less well investigated. Most studies that investigated the function of p53 in liver were performed either in models of hepatocellular carcinomas (9, 10) or in models with chronic p53 deficiency that are accompanied by high incidences of spontaneous tumor formation (11–13). Several recent reports involve p53 in the control of hepatic glucose homeostasis (12, 14, 15) or lipid metabolism (16) or both (17, 18). However, common to these studies is the use of germ-line or early embryonic p53 loss-of-function models that are prone to carcinogenesis in the liver and other organs (13).

Here we report that p53 protein levels in liver are highly dynamic and acutely regulated by nutrient availability. Specifically, food withdrawal robustly stabilized hepatic p53 and induced expression of its target genes. We found that this stabilization mechanism requires the known energy sensor AMPK. Moreover, acute p53 deletion in livers of adult mice led to hepatic steatosis and impaired glycogen storage. Upon food withdrawal, mice that lack p53 in liver exhibited defective amino acid catabolism and gluconeogenesis, resulting in hypoglycemia. Our findings reveal that p53 participates in AMPK-mediated nutrient sensing and that its stabilization is required for adaptation of hepatic energy metabolism to starvation.

MATERIALS AND METHODS

Mouse studies

All mouse procedures were performed in accordance with institutional guidelines and approved by the corresponding authorities. Mice were housed under standard 12 h light/12 h dark cycles. For feeding studies, 3-mo-old wild-type C57BL/6N mice were housed individually in grid-bottomed cages for 24 h and either fed at libitum (standard chow), starved, or refed for 2 h after starvation.

Acute deletion of p53 in adult mice

Adenoviruses expressing green fluorescent protein (GFP) and Cre-recombinase (CRE) were generated using the Adeno-X Expression System 2 (Clontech Laboratories, Mountain View, CA, USA) as previously described (48) and purified by standard CsCl gradients. Titers were determined by the Adeno-X Rapid Titer Kit (Clontech Laboratories). Equal titers (∼1.7 × 109 infectious units diluted in 250 µl sterile saline) were injected via the tail vein into 2-mo-old male C57BL/6 Trp53tm1Brn/J mice (49) after dilating the tail vein in a 42°C water bath for 1 min. Mice were fed standard chow (Ssniff R/M-H). Three days later, one cohort of mice that expressed either GFP or CRE was fed ad libitum, and another cohort was denied food for 24 h. The following day, mice of both cohorts were euthanized at the same time.

Quantification of serum parameters

Blood collected form the facial vein was used to determine glucose concentrations (Contour Link meter; Bayer, Leverkusen, Germany). Mice were euthanized and serum prepared by cardiac puncture. Serum parameters and hormones were quantified as previously described (50).

Determination of liver triglycerides

Liver triglycerides were determined using a commercially available kit (Randox Laboratories, London, United Kingdom). In brief, liver samples were homogenized in a chilled buffer containing 10 mM sodium dihydrogenphosphate, 1 mM EDTA, and 1% polyoxyethylene (10) tridecyl ether set to pH 7.4. Triglycerides were measured in cleared supernatants according to the manufacturer’s instructions and normalized to total protein content of the sample.

Quantification of glycogen, glucose, and amino acids by NMR spectroscopy

Methanol, sodium phosphate, dibasic (Na2HPO4), sodium hydroxide, hydrochloric acid (32% m/v), and sodium azide (NaN3) were obtained from VWR International (West Chester, PA, USA). TSP [3-(trimethylsilyl)propionic acid-2,2,3,3-d4 sodium salt] was obtained from Alfa Aesar (Ward Hill, MA, USA). Deuterium oxide (D2O) was obtained from Cambridge Isotope Laboratories (Tewksbury, MA, USA). Deionized water was purified using an inhouse Milli-Q Advantage Water Purification System from EMD Millipore (Billerica, MA, USA). All chemicals were used with no further purification.

The phosphate buffer solution was prepared by dissolving 5.56 g of anhydrous NaH2PO4, 0.4 g of TSP, and 0.2 g NaN3 in 400 ml of deionized water and adjusted to pH 7.4 with 1 M NaOH and HCl. Upon addition of deionized water to a final volume of 500 ml, the pH was readjusted to pH 7.4 with 1 M NaOH and HCl. For metabolite extraction, the liver mixtures (30–50 mg) were homogenized in 600 µl ice-cold PBS using metal beads and the LT TissueLyser (Qiagen, Germantown, MD, USA), followed by sonication (ultrasound probe, 3 × 10 s at 10% output). Forty microliters was removed for protein concentration measurement with bicinchoninic acid assay (Thermo Fisher Scientific, Waltham, MA, USA). Samples were mixed with 2× volume of −20°C cooled methanol, vortexed, and stored at −20°C for 30 min. The mixture was centrifuged (12,000 g, 30 min, 4°C), and supernatant was lyophilized centrifugal evaporation. A total of 500 µl of phosphate buffer in D2O was added to the samples, redissolved, and transferred to 5 mm NMR tubes. All NMR experiments were performed at 310 K on a Bruker Avance III 500 MHz spectrometer equipped with a TXI probe head (Bruker Daltonics, Bremen, Germany). The 1D CPMG (Carr-Purcell-Meiboom-Gill) pulse sequence (cpmgpr1d, 73728 points in F1, 12019.230 Hz spectral width, 2048 transients, recycle delay 4 s), with water suppression using presaturation, was used for 1H 1D NMR experiments. Metabolite reference chemical shifts were taken from the Madison-Qingdao Metabolomics Consortium Database () database (51), and all metabolites were cross-checked using reference compounds. The Bruker Topspin 3.1 and MestReNova 10.0 (Mestrelab Research, Santiago de Compostela, Spain) software packages were used for NMR data acquisition, processing, and analyses. Metabolite concentrations were determined using TSP as the internal standard.

Histologic analyses and immunohistochemistry

For Oil Red O staining, 6-µm thin cryosections were mounted on poly-l-lysine-coated slides and dried overnight. Working solution of Oil Red O was prepared by mixing stock solution [0.5% (w/v) in 2-propanol] 3:2 with water followed by filtration. Working solution was incubated for 10 min. To stain liver glycogen, formalin-fixed, paraffin-embedded livers were cleared and incubated in 1% periodic acid (25 min) followed by Schiff reagent (Sigma-Aldrich, St. Louis, MO, USA; 25 min) and rinsing in SO2 water. Immunohistochemical staining of formalin-fixed, paraffin-embedded livers was performed after antigen retrieval (120°C, 7 min at pH 9) and peroxidase blocking (Dako, Glostrup, Denmark) using the UltraVision LP detection system (Thermo Fisher Scientific) according to the manual with 1 ng/µl Ki-67 antibody (M728; Dako). For color reaction, AEC (3-amino-9-ethylcarbazole) chromogen (Thermo Fisher Scientific) was used. Stainings were performed in a LabVision 2D autostainer (Thermo Fisher Scientific). Mouse IgG1 was used as negative control. Counterstaining with hematoxylin was done on all slides. All quantifications were performed by Ilastik Interactive Learning and Segmentation software (). Automatic counting of segments was done by a customized R script (R Foundation for Statistical Computing, Vienna, Austria; ).

Transcriptome analyses

Hepatic transcriptome analyses were performed with 3 mice in each of the following 4 groups: fed GFP, fed CRE, starved GFP, and starved CRE. Liver pieces (∼20 mg) were homogenized in Qiazol using metal beads with the LT TissueLyser (all Qiagen). Total RNA was isolated with miRNeasy (Qiagen) according to the manual and quality checked with an Agilent Technologies Bioanalyzer (Santa Clara, CA, USA). Total RNA (100 ng) with a RNA integrity number of >7.5 was used for microarray hybridization (Mouse Gene 2.0 ST; Affymetrix, Santa Clara, CA, USA) performed according to standard Affymetrix protocols at the core facility for molecular biology, Medical University Graz. Basic analysis to obtain fold changes was done with the Partek Genomics Suite software (St. Louis, MO, USA). A 2-way ANOVA was applied, and the resulting P values were corrected for multiple testing using a false discovery rate (5%) cutoff. In-depth bioinformatics analyses were done on 1.5× differentially expressed genes [Database for Annotation, Visualization, and Integrated Discovery (DAVID) web tool; ] or with the entire annotated data set [Gene Set Enrichment Analysis (GSEA) software; ] as previously described (19). Data were deposited in National Center for Biotechnology Information Gene Expression Omnibus (GSE81226; ). Heatmaps and hierarchical clustering were performed with Genesis academic software ().

RNA isolation and quantitative PCR

Total RNA was isolated with either miRNeasy (Qiagen, p53fl/fl) or PeqGOLD total RNA kit (Peqlab, Erlangen, Germany; wild-type mice and cell culture) according to the manuals and quantitated with Nanodrop (Peqlab). cDNA was prepared using cDNA-Synthese Kit H Plus (Peqlab) and diluted to 1 ng/µl. Quantitative PCR (qPCR) was done as previously described (52) on a Bio-Rad CFX cycler with SYBR Green chemistry (Bio-Rad, Hercules, CA, USA). Primer sequences are listed in Supplemental Table S4. For mouse experiments, 36b4 (Rplp0) was used as reference gene, while Gapdh was used with HepG2 cells. For expression analyses in p53 fl/fl livers, the same 3 replicates per group were used for microarray and qPCR analyses.

Immunoblotting

Cultured cells were scraped and collected in RIPA buffer including PIC and PhosStop (Roche, Basel, Switzerland). Liver samples were homogenized in RIPA buffer using metal beads and the LT TissueLyser (Qiagen). All samples were sonicated (ultrasound probe, 3 × 10 s at 10% output) and centrifuged. Clear supernatant was used to measure protein concentration with a bicinchoninic acid assay (Thermo Fisher Scientific). Immunoblotting was performed as previously described (19, 52). The following antibodies were used: human p53 (sc-126; Santa Cruz Biotechnology, Santa Cruz, CA, USA), mouse p53 (2524; Cell Signaling Technology, Danvers, MA, USA), β-actin (ab6276; Abcam, Cambridge, MA, USA), ras-related nuclear protein, (RAN) (610340; BD Biosciences, San Jose, CA, USA), cyclin-dependent kinase inhibitor 1A (CDKN1A; p21) (2947; Cell Signaling Technology), and AMPK/acetyl-CoA carboxylase 1 (ACC1) total and phosphorylated from Cell Signaling Technology (9957; AMPK and ACC1 antibody sampler kit). Densitometric quantification was done by Image Studio Lite software (Li-Cor Biosciences; Lincoln, NE, USA).

Primary mouse hepatocyte isolation, culture, treatment, and small interfering RNA-mediated depletion of p53

Isolation of hepatocytes was performed as previously described (53). In brief, livers of anesthetized male C57BL/6J mice were perfused with digestion buffer containing 5000 U collagenase (Worthington Biochemical, Lakewood, NJ, USA). After filtration and separation by Percoll gradients (Biochrom, Cambourne, United Kingdom), cells were seeded on collagen-coated 12-well plates in DMEM containing 10% fetal bovine serum and 1% penicillin/streptomycin (both Thermo Fisher Scientific). The next day, hepatocytes were washed and incubated with starvation medium [HBSS and 10 mM HEPES (all Thermo Fisher Scientific), as described in Lee et al.] or incubated with 1 µM dexamethasone (Dex; Sigma-Aldrich), 5 nM glucagon (Sigma-Aldrich), or 0.5 mM 3-isobutyl-1-methylxanthine (IBMX; Sigma-Aldrich). Ethanol was used as vehicle control. At the indicated time points cells were collected for protein and RNA analysis. For small interfering RNA (siRNA)-mediated depletion, medium of culture plate–attached hepatocytes was replaced by 500 µl of DMEM, and cells were transfected with 1 nM of siRNA (si_Control: 5′–UAG-CGA-CUA-AAC-ACA-UCA-AUU–3′ or si_TP53: 5′–GAA-UGA-GGC-CUU-AGA-GUU-AUU–3′, Eurogentec, Ougrée, Belgium) and 4 µl of Lipofectamine 2000 (Thermo Fisher Scientific) per well overnight. The next morning, media were replaced by complete DMEM.

Oxygen consumption rate of primary mouse hepatocytes

Ninety-six-well microplates were coated by collagen, type I (2 mg/ml collagen in 20 µM sterile acetic acid), overnight. Primary mouse hepatocytes were seeded at a density of 5000 cells per well and in a total volume of 50 µl culture medium, and p53 knockdown was performed as described above. Three days later, oxygen consumption rate (OCR) was determined in the presence or absence of 500 nM of carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (commonly known as FCCP) using a Seahorse Analyzer (Seahorse Bioscience, North Billerica, MA, USA) according to the manufacturer's instructions.

Culture and treatment of primary human hepatocytes

Primary human hepatocytes from 2 male nondiseased donors were kindly provided by QPS Hepatic Biosciences (Newark, DE, USA) and cultivated according to the provider’s protocols. Briefly, cells were reconstituted and centrifuged in 50 ml recovery medium (QPS Hepatic Biosciences) and counted. A total of 0.8 × 106 cells/ml were plated in plating medium (QPS Hepatic Biosciences) in collagen I–coated 24-well plates (BD BioCoat) and switched to maintenance medium (QPS Hepatic Biosciences) the next day. Three days after plating, hepatocytes were washed and incubated with starvation medium (HBSS and 10 mM HEPES) or incubated with 1 µM Dex, 5 nM glucagon, 10 µM nutlin-3a (Biomol, Plymouth Meeting, PA, USA), or 0.5 mM 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) (Biomol). DMSO was used as vehicle control. Hepatocytes were collected for protein and RNA analysis at the indicated timepoints.

Culture and treatment of HepG2 cells

HepG2 cells were cultured in DMEM containing 1 mg/ml glucose with 10% fetal bovine serum and 1% penicillin/streptomycin (all Thermo Fisher Scientific). Treatments were applied as indicated 30 min after medium change. Starvation medium consisted of HBSS and 10 mM HEPES. For knockdown experiments, 105 cells were seeded in 24-well plates and transfected with Metafectene si+ (Biontex, Munich, Germany) using 60 nM siRNA. Control and p53 siRNAs (5′–GAAAUGUUCUUGCAGUUAA–3′) were obtained from Microsynth (Balgach, Switzerland) and siAMPK from Santa Cruz (sc-29673). Forty-eight hours after transfection, cells were treated as indicated.

Statistical analyses

All cell culture experiments were performed at least in triplicates unless indicated otherwise. Bar graph data are shown as means ± sem. Significant differences between 2 groups were assessed by 2-tailed unpaired or 1-sample Student’s t test. For comparison of more than 2 groups, a 2-way ANOVA was used to determine significant differences followed by a post hoc test between each 2 groups. Significance computation for microarray comparison was done using a 2-way ANOVA followed by multiple-testing correction. Corrected values of P < 0.05 were considered differentially expressed. For DAVID analyses, Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were considered as significantly enriched with Benjamini-Hochberg’s corrected value of P < 0.05.

RESULTS

p53 protein in liver is stabilized under starvation

We previously reported the starvation-dependent induction of p53 target genes in several murine tissues relevant to metabolic homeostasis (19)—that is, liver, adipose tissue, and skeletal muscle. We corroborated our initial observations and found that expression of p53 target genes like p21, DNA damage inducible transcript 4 (Ddit4), lipin 1 (Lpin1), and sestrin 2 (Sesn2) were strongly up-regulated in livers of starved mice and partially repressed after 2 h of refeeding (). Strikingly, we detected a robust increase in liver p53 protein upon food withdrawal that decreased after refeeding (Fig. 1 and its densitometric quantification in Fig. 1). This p53 accumulation was not due to up-regulation of p53 mRNA or reduced expression of its negative regulator mouse double minute 2 homolog (MDM2) proto-oncogene, Mdm2, which mediates p53 degradation (20) (Fig. 1). Hence, these data show that starvation induces a profound accumulation of p53 protein in liver, most likely by posttranscriptional p53 stabilization.

Figure 1.

p53 protein in liver is stabilized by starvation. Four-month-old C57BL/6N mice were either fed ad libitum, starved for 24 h, or refed for 2 h after 24-h starvation. Livers were sectioned and shock frozen for subsequent RNA or protein extraction. A) qPCR of RNA extracted from shock-frozen liver samples to determine expression of p53 target genes. Mean expression in fed group was set to 1. Ddit4, DNA damage inducible transcript 4; Lpin1, lipin 1; Sesn2, sestrin 2. B) Liver lysates were subjected to Western blot analysis to detect p53 protein in fed, starved, and refed mice. Expression of β-actin (ACTB) served as loading control. C) Densitometric quantification of signals (B). Value in fed group was set to 1 (n = 3). D) qPCR of RNA extracted from liver samples to determine expression of p53 and Mdm2 (mouse double minute 2 homolog). Mean expression in fed group was set to 1. All qPCR data above are derived from 5 to 7 mice per group. *P < 0.05 compared to fed group, #P < 0.05 compared to starved group.

p53 is stabilized by starvation in a hepatocyte-autonomous manner

To investigate whether starvation-induced p53 protein accumulation is cell autonomous and to elucidate potential upstream regulators, we first tested whether starvation-associated hormones induce p53 levels in cultured hepatocytes. Starvation or fasting signals like glucagon and the synthetic glucocorticoid Dex failed to affect p53 protein levels in primary mouse or human hepatocytes (, respectively). Similarly, there was no effect of these hormones on p53 levels in HepG2 cells (Fig. 2), a p53 competent cell line derived from a human hepatocellular carcinoma (21). Also, IBMX, a phosphodiesterase inhibitor that activates protein kinase A by increasing cAMP levels (22), did not lead to p53 protein accumulation in primary mouse hepatocytes (Fig. 2) or HepG2 cells (Fig. 2). Treatment with the known p53 activators nutlin-3a [a potent MDM2 inhibitor (23)] (Fig. 2) and quinacrine (24) (Fig. 2), on the other hand, led to the expected accumulation of p53 protein. These data indicate that the p53 accumulation upon starvation in liver is unlikely to be caused by the tested fasting hormones.

Figure 2.

p53 is stabilized by starvation in a hepatocyte-autonomous way. A–C) Western blot analysis of p53 protein in primary mouse (A) and human (B) hepatocytes and HepG2 cells (C) treated with indicated compounds and appropriate vehicle controls for 24 h. Concentrations: 5 nM glucagon, 1 µM Dex, 0.5 mM IBMX, 10 µM nutlin-3a, and 20 µM quinacrine. Quinacrine and nutlin-3a were used as positive controls and human β-actin (ACTB) or mouse RAN served as loading controls. D, E) Hepatocytes were starved in HBSS/HEPES for 6 h [primary mouse hepatocytes (D)] or 24 h [primary human hepatocytes (E)]. F) HepG2 cells were starved for indicated times or treated for 24 h with 10 µM nutlin-3a or 20 µg/ml etoposide as positive controls for p53 accumulation. Western blot analysis was performed to determine p53 and p21 protein levels, and ACTB or RAN served as loading controls. G–I) qPCR analysis of p53 mRNA expression after 12 h [primary mouse hepatocytes (G)] or 24 h [primary human hepatocytes (H), HepG2 (I)] of starvation. Expression levels are set to 1 in control.

To further dissect the underlying mechanisms, we subjected hepatocytes to nutrient-free starvation conditions (25) and detected a robust accumulation of p53 protein in all 3 in vitro models (Fig. 2). Similar to p53 accumulation in livers of starved mice (Fig. 1), this was not due to increased mRNA expression (Fig. 2). In HepG2 cells, starvation for 1, 6, and 24 h resulted in a time-dependent induction of p53 protein and its canonical target, p21 (Fig. 2 and Supplemental Fig. S1). The extent of induction was robust and comparable to the effect of the known p53-stabilizing compounds nutlin-3a and etoposide (26) (Fig. 2). Taken together, we found that starvation induces a hepatocyte-autonomous accumulation of p53 protein in different in vitro cell systems, including nontransformed primary hepatocytes.

Full p53 stabilization under starvation requires AMPK signaling

We next investigated alternative hormone-independent pathways upstream of p53 stabilization. Starvation of HepG2 cells led to a concomitant increase in phosphorylation of AMPK α and β subunits, suggesting that this signaling cascade might be involved in regulation of p53 (). Consistent with this notion, treatment with the AMPK activator AICAR (27) increased p53 protein levels in human primary hepatocytes (Fig. 3) and HepG2 cells (Fig. 3). Further, AICAR induced the expression of p53 target genes p21, Trp53 induced glycolysis regulatory phosphatase (Tigar), and sestrin 1 (Sesn1) in HepG2 cells by magnitudes comparable to 24-h starvation or nutlin-3a treatment (Fig. 3). Because HepG2 cells closely mirrored the AICAR-induced p53 accumulation in primary human hepatocytes, we used HepG2 cells to further investigate the interaction of AMPK signaling and p53 during starvation. First, we addressed whether the induction of p53 target genes by AICAR is indeed mediated by p53 stabilization rather than other AMPK downstream effectors. We found that the AICAR-induced p21 up-regulation was blunted in HepG2 cells that were depleted of p53 by siRNA (Supplemental Fig. S1). Moreover, depletion of AMPKα, validated by reduced AMPK protein levels and reduced phosphorylation of its known substrate ACC1 (Fig. 3), partially inhibited the starvation-induced accumulation of p53 protein (Fig. 3). In summary, our data implicate AMPK activation in the starvation-induced accumulation of p53 protein in primary hepatocytes and HepG2 cells.

Figure 3.

Full p53 stabilization under starvation requires AMPK signaling. A) HepG2 cells were starved for indicated times, and Western blot analysis was performed to determine protein levels of AMPK subunits α and β and their respective phosphorylated forms. β-Actin (ACTB) served as loading controls. B) Primary human hepatocytes from 2 donors were treated for 24 h with 0.5 mM AICAR or with DMSO as vehicle control and analyzed for p53 protein expression. Phosphorylated form of ACC1 is shown to validate AMPK activation; ACTB served as loading control. C) HepG2 cells were treated for 24 h with 10 µM nutlin-3a, 0.5 mM AICAR, or respective vehicle controls and analyzed for p53 protein expression. D) qPCR analysis of p53 target gene expression after starvation or treatment with 0.5 mM AICAR or 10 µM nutlin-3a. Expression levels are set to 1 in control. Sesn1, sestrin 1; Tigar, Trp53 induced glycolysis regulatory phosphatase. *P < 0.05 compared to control. E) Western blot analysis validating knockdown of AMPKα and decreased AMPK activity by reduced phosphorylation of downstream target ACC1. F) Two days after transfection with control or AMPKα siRNA, HepG2 cells were starved for 6 h or received fresh medium. Western blot analysis was used to determine p53 protein levels. G) Quantification of p53 protein expression from 3 independent experiments (F). Protein abundance was set to 1 for starved cells transfected with control siRNA. *P < 0.05 compared to siControl.

Acute p53 deletion in livers of adult mice impairs glycogen storage and induces steatosis

The marked dynamics of p53 protein levels in livers of fed and starved mice implied a physiologic role in the hepatic starvation response. We therefore generated an acute liver-specific p53 loss-of-function model that excluded developmental compensations due to germline deletion or metabolic disturbances caused by tumor formation (13). To this end, we used adenoviruses to express either GFP as control or CRE in livers of adult mice that were homozygous for a floxed Trp53 allele and characterized these animals within a short period to circumvent compensatory mechanisms. p53 mRNA expression in whole liver was reduced by 80% by CRE expression, and p21 was down-regulated to a similar extent, although not in a statistically significant manner (). We found that most of the measured serum parameters were unchanged (alanine transaminase, triglycerides, free fatty acids, cholesterol, LDL, hydroxybutyric acid, and insulin), with the exception of serum lactate, which was decreased in mice with liver-specific p53 deletion (Supplemental Table S1). While blood glucose in ad libitum–fed mice was unchanged (Fig. 4), glucose concentrations in liver lysates were reduced (Fig. 4). Periodic acid–Shiff staining of liver sections and NMR measurements of liver lysates demonstrated a ∼50% reduction of hepatic glycogen in p53-deleted livers (Fig. 4, respectively). Triglyceride-specific Oil Red O staining showed increased lipid accumulation (Fig. 4). While computer-assisted quantification of histologic images did not reach significance because of a high variation in CRE-expressing mice, determining liver triglycerides by an alternative colorimetric assay showed a statistically significant increase in mice with hepatic p53 deletion (Fig. 4). Liver steatosis has been observed previously in chronic models with loss of p53 (16). These data indicate that in the fed state, acute depletion of liver p53 leads to disturbances in glycogen storage and induces hepatic steatosis.

Figure 4.

Effects of acute and liver-specific p53 deletion on hepatic glucose and lipid metabolism. p53 fl/fl mice were tail-vein injected with adenoviral constructs expressing GFP (control) or CRE (p53 deletion). A) qPCR to detect p53 and p21 transcript levels in liver during fed state. Expression was set to 1 in GFP group. B) Blood glucose measured in GFP and CRE mice fed ad libitum (n = 4–5). C) Glucose concentrations were measured by NMR spectroscopy in methanol-extracted liver lysates of GFP and CRE mice fed ad libitum (n = 4–5). D) Periodic acid–Schiff staining of GFP and CRE liver sections with subsequent segmentation and quantification to determine liver glycogen content (n = 4–5). E) Glycogen content measured by NMR spectroscopy in methanol-extracted liver lysates of GFP and CRE mice fed ad libitum (n = 4–5). F) Oil Red O staining of GFP and CRE liver cryosections with subsequent segmentation and quantification to determine neutral lipids (n = 4–5). G) Liver triglycerides of GFP and CRE mice measured by colorimetric assay (n = 4–5). *P < 0.05 compared to GFP. Scale bars, 100 μm.

Global gene expression profiling suggests a role for p53 in hepatic amino acid metabolism

To identify potential regulatory mechanisms of metabolite handling affected by p53, we next investigated liver transcriptomes after acute p53 deletion in ad libitum–fed mice. Because p53 is mainly described as a transcriptional activator (28, 29), we first focused on genes down-regulated after p53 depletion. Using DAVID functional annotation (30) to map genes down-regulated ≥1.5-fold onto GO terms and KEGG pathways almost exclusively yielded terms related to amino acid metabolism and/or catabolism (). Transcripts comprising these categories are shown as heat map in Fig. 5. Similarly, when performing GSEA of whole transcriptomes (31), genes in pathways related to catabolism of glucogenic and ketogenic aromatic amino acids were among the strongest deenriched pathways in livers lacking p53 (Supplemental Table S2). Genes up-regulated after p53 deletion primarily mapped to cell cycle regulation (Supplemental Fig. S2) with some key cell cycle regulators, such as cyclins, cyclin-dependent kinases, and E2f factors affected (Supplemental Fig. S2). This is consistent with the known function of p53 as a cell cycle inhibitor (2) and is in accordance with increased Ki-67 staining as a measure for cell proliferation in liver sections of mice that lack hepatic p53 expression (Supplemental Fig. S2). These results demonstrate that p53 in hepatocytes is not only required to limit cell cycle progression but also for the expression of genes controlling amino acid metabolism.

Figure 5.

Acute and liver-specific p53 deletion blunts expression of genes related to hepatic amino acid metabolism. Microarray analysis of RNA isolated from livers of ad libitum fed GFP and CRE mice (n = 3 per group). A) DAVID functional annotation using GO terms (biologic processes) and KEGG pathways of genes down-regulated more than 1.5-fold in p53-deleted livers compared to GFP control. Only terms that are significant after correction for multiple testing are shown. P < 0.05 by Benjamini-Hochberg correction. B) Heat map showing differential expression of down-regulated genes comprising DAVID categories (A) after normalization per gene and hierarchical clustering using Genesis software.

Loss of hepatic p53 impairs glucose homeostasis and amino acid catabolism during starvation

We next studied the effects of acute loss of hepatic p53 in starved mice. As in ad libitum–fed mice, mRNA expression of p53 in liver of CRE-expressing mice was strongly reduced compared to GFP controls (). Hepatic glycogen content was almost completely exhausted after 24 h of food withdrawal in both controls and mice with liver-specific p53 deletion (Supplemental Fig. S3). Liver triglycerides showed the expected increase upon starvation (32) without differences between GFP control and CRE mice (Supplemental Fig. S3). While most measured serum parameters were similar (Supplemental Table S3), fasting blood glucose was reduced (Fig. 6), and glucose in liver lysates trended toward lower levels in mice lacking p53 in liver (Fig. 6). We next analyzed transcriptome data with respect to pathways enriched by starvation in control and p53-deleted mice. Performing GSEA with GO terms and KEGG pathways highlighted pathways involved in amino acid metabolism in general, and amino acid catabolism in particular, as major starvation responses in control but not p53-deleted mice, as shown by the 20 top-ranking terms/pathways (, blue bars, sorted by decreased normalized enrichment score). Consistently, mRNA expression of genes comprising these terms was overall reduced in livers of starved mice lacking hepatic p53 (Supplemental Fig. S3). Thus, p53 is required not only for basal expression of amino acid metabolism-regulating genes in the fed state but also for the starvation-induced up-regulation of genes controlling amino acid catabolism. Consistent with these gene expression profiles, we found that only control mice catabolized amino acids like lysine, valine, and isoleucine in liver upon starvation, while mice that lack p53 in liver exhibited no or a more subtle reduction (Fig. 7). Liver contents of both isoleucine and alanine were significantly higher under starvation, further supporting a defect in their hepatic catabolism in the absence of p53 (Fig. 7). Interestingly, transcriptomes of p53-deleted livers showed a strong shift toward mitochondria-related terms during starvation at the expense of amino acid metabolism (Fig. 7, orange bars). In accordance with the induction of mitochondria-related terms upon p53 deletion in liver, we found that basal and maximal oxygen consumption of primary hepatocytes was increased after depleting p53 by siRNA (Fig. 7). In summary, these findings indicate that p53 is required for starvation-induced amino acid catabolism and the maintenance of glucose homeostasis during starvation, and that loss of p53 in hepatocytes induces a shift to mitochondrial respiration.

Figure 6.

Effects of acute and liver-specific p53 deletion on hepatic starvation response. p53 fl/fl mice were tail-vein injected with adenoviral constructs expressing GFP (control) or CRE (p53 deletion). A) qPCR to detect p53 and p21 transcript levels in liver after 24 h of starvation. Expression is set to 1 in GFP group. B) Blood glucose in GFP and CRE mice starved for 24 h (n = 5–6). C) Glucose concentration were measured by NMR spectroscopy in methanol extracted liver lysates of GFP and CRE mice starved for 24 h (n = 5–6). *P < 0.05 compared to GFP.

Figure 7.

Acute and liver-specific p53 deletion impairs starvation-adapted gene expression. A) GSEA of whole-transcriptome data ranked by differential expression between fed and 24 h starved states of CRE and GFP groups (n = 3 per group). Overlay of normalized enrichment scores of top 20 enriched GO_BP terms and KEGG pathways in GFP (blue bars) and CRE (orange bars) are sorted by decreased normalized enrichment score in GFP group. B) Indicated amino acids in methanol extracts from liver lysates were determined by NMR spectroscopy (n = 4 per group). Mean value in GFP fed group is set to 100%. One-way ANOVA followed by post hoc test was performed. For pairwise comparisons, *P < 0.05 compared to respective fed group, and #P < 0.05 and between GFP and CRE group under starvation. C) Validation of siRNA-mediated p53 knockdown in primary hepatocytes isolated from wild-type mice. Representative results of 3 independent experiments are shown. D) OCR in primary hepatocytes was measured 3 d after siRNA application using Seahorse Bioscience XF96 extracellular flux analyzer. Representative results of 3 independent experiments are shown. *P < 0.05 compared to si_Control.

DISCUSSION

The liver is a central coordinator of glucose and lipid metabolism and dynamically adapts to nutrient availability. To our knowledge, we are first to report that hepatic p53 protein is stabilized by food withdrawal. Because nutrient deprivation represents a profound stress stimulus, it appears plausible that p53 as a primary, well-conserved stress response mediator activated during tumorigenesis (1) is also induced in physiologic situations that constitute acute bioenergetic stress (i.e., starvation) ().

Figure 8.

Schematic summary of p53 action in liver. Starvation-induced stabilization of liver p53 is hepatocyte autonomous and, at least in part, mediated by AMPK activation. While basal p53 activity is necessary for glycogen storage and to prevent hepatic steatosis during feeding, p53 protein stabilization is required for gluconeogenesis and amino acid catabolism in starved state, thus establishing p53 as key mediator of metabolic flexibility.

With regard to potential upstream mechanisms, we show that p53 protein stabilization by starvation is dependent on the cellular energy sensor AMPK (27) in a hepatocyte-autonomous manner. While AMPK activation has previously been shown to stabilize p53 in cancer cell lines (33, 34), we describe a physiologic mechanism to reprogram the metabolic response of nutrient-deprived mouse and human hepatocytes. AMPK-dependent phosphorylation of p53 at Ser15 was shown to increase stability by disrupting the binding of p53 to its endogenous inhibitor MDM2, thereby reducing p53 proteasomal turnover (20). Further studies are needed to determine whether starvation-induced p53 stabilization in liver involves Ser15 phosphorylation by AMPK. However, because AMPK-depleted cells were still partially responsive to starvation-induced p53 stabilization (Fig. 3), additional posttranslational mechanisms (e.g., changes in phosphorylation or acetylation levels by mammalian target of rapamycin (33) or by sirtuins (34), respectively) may also play a role in this process.

Our observation that acute loss of p53 impairs the starvation response and homeostasis of a variety of macronutrients (lipids, carbohydrates, and amino acids) renders p53 a nodal point for metabolic flexibility of the liver. Context-dependent roles for p53 in the control of gluconeogenic gene expression were reported: Zhang et al. (12) showed a p53-dependent down-regulation of gluconeogenic genes [phosphoenolpyruvate carboxykinase 1 (PCK1) and glucose-6 phosphatase (G6PC)] in human colon cancer cells via mechanisms involving SIRT6 and FOXO1, while Goldstein et al. (14) demonstrated that p53 is necessary for PCK2 and G6PC expression and glucose production in HepG2 cells. Although we detected a nonsignificant ∼50% reduction of hepatic Pck1 expression after p53 deletion (data not shown), this is unlikely to be a major contributor to the observed hypoglycemia upon starvation considering the weak correlation between Pck1 expression and gluconeogenic flux (35). However, our results strongly support a role for hepatic p53 in glucose homeostasis in vivo, potentially via defective amino acid catabolism.

Besides disturbed glucose homeostasis, we observed hepatic steatosis in the absence of p53. Similar observations were made by others in models of chronic p53 reduction (14, 36, 37), and a role of p53 in enhancing lipid catabolism has been suggested (38). Kung et al. (17) reported development of nonalcoholic fatty liver disease in a humanized p53 codon 72 polymorphism mouse model, and our data suggest a similar, although milder, phenotype after acute p53 depletion in liver. Whether hepatic steatosis in our model is due to increased triglyceride synthesis, higher hepatic fatty acid uptake, or disturbed fatty acid oxidation is unknown. Another possible mechanism leading to steatosis in mice with liver-specific p53 deletion could be defective autophagy of lipids [macrolipophagy (39, 40)] because p53 was shown to be a positive regulator of autophagy in some cell systems (41). Hence, these data warrant further in-depth studies to clarify the exact role of p53 in hepatic lipid metabolism.

The major transcriptional effect of acute p53 deletion in liver was strongly reduced expression of genes controlling amino acid catabolism. This was in accordance with the levels of hepatic amino acids, decreasing upon food withdrawal in control mice while being less affected in the absence of p53 (Fig. 7). Seen in the light of earlier publications that established the contribution of certain amino acids to liver glucose and lipid metabolism (42, 43), it could be concluded that diminished amino acid metabolism is the underlying cause for the observed metabolic disturbances in our model, including hypoglycemia under starvation. However, this needs to be formally proven. Further, we observed increased hepatocyte proliferation upon acute p53 deletion that may contribute to the metabolic changes observed, as cell cycle and metabolism are strongly interlinked (44, 45).

Notably, acute p53 deletion not only affects macronutrient metabolism but also promotes a mitochondrial biogenic gene expression profile under starvation and mitochondrial respiration in primary hepatocytes. This shift away from glycolytic metabolism is in contradiction to the known anti-Warburg effect of p53 in cancer cells (7, 46). This could be seen as an adaptive, compensatory response to maintain cellular ATP levels.

In summary, our findings highlight novel functions of hepatic p53 with regard to the physiologic adaption to starvation in the liver (Fig. 8). p53 is stabilized by an AMPK-dependent mechanism and accumulates in the liver under starvation. It coordinates catabolic pathways essential to maintain metabolic flexibility. It is therefore tempting to speculate that some of the beneficial effects of fasting and caloric restriction in regard to aging, cancer, and metabolic diseases (47) may indeed be mediated by p53, hence providing novel insights into the complex biology of metabolic homeostasis and a healthy lifestyle.