p53 Regulates Mitochondrial Respiration

See allHide authors and affiliations

Science  16 Jun 2006:
Vol. 312, Issue 5780, pp. 1650-1653
DOI: 10.1126/science.1126863


The energy that sustains cancer cells is derived preferentially from glycolysis. This metabolic change, the Warburg effect, was one of the first alterations in cancer cells recognized as conferring a survival advantage. Here, we show that p53, one of the most frequently mutated genes in cancers, modulates the balance between the utilization of respiratory and glycolytic pathways. We identify Synthesis of Cytochrome c Oxidase 2 (SCO2) as the downstream mediator of this effect in mice and human cancer cell lines. SCO2 is critical for regulating the cytochrome c oxidase (COX) complex, the major site of oxygen utilization in the eukaryotic cell. Disruption of the SCO2 gene in human cancer cells with wild-type p53 recapitulated the metabolic switch toward glycolysis that is exhibited by p53-deficient cells. That SCO2 couples p53 to mitochondrial respiration provides a possible explanation for the Warburg effect and offers new clues as to how p53 might affect aging and metabolism.

Cancer is a genetic disease caused by the dysregulation of various cellular pathways that orchestrate cell growth and death (1). It is clear that some of these pathways must modulate cellular metabolism. As described by Otto Warburg in 1931, cancer cells preferentially utilize glycolytic pathways for energy generation while down-regulating their aerobic respiratory activity (2). A number of mechanisms have been proposed to explain the Warburg effect (37), but there have not been any reports of a genetically defined pathway that couples a tumor suppressor gene to mitochondrial aerobic respiration.

Critical to aerobic life is the cytochrome c oxidase (COX) complex in the mitochondrion, where most of the molecular oxygen is consumed in the eukaryotic cell. In cancer cells mitochondrial respiratory activity is decreased in association with changes in the expression levels of COX complex subunit proteins, but the genetic mechanisms that underlie their modulation are unclear (8, 9). We reasoned that because the metabolic alterations of cancer cells are so widespread, an explanation for this phenomenon must lie, at least in part, in a pathway that is commonly altered in cancer cells. We therefore examined whether alteration of p53, the gene most commonly mutated in human cancer, might affect COX complex assembly and activity.

Aerobic respiration was assessed in liver mitochondria preparations from mice with wild-type (+/+), heterozygous (+/–), or homozygous (–/–) disruption of TP53. There was a significant decrease in oxygen consumption that closely correlated with p53 deficiency (p53+/+ versus p53–/–, P < 0.001) (Fig. 1A). We confirmed a similar p53-dependent reduction (∼25%) in cellular oxygen consumption using isogenic human colon cancer HCT116 cells with a targeted disruption of TP53 (Fig. 1A) (10). Direct measurement of COX enzymatic activity showed a comparable reduction of 25 ± 3% in the p53–/– cells (fig. S1). To estimate the relative contributions of glycolysis and aerobic respiration, we calculated total adenosine 5′-triphosphate (ATP) generated by measuring whole-cell lactate production and oxygen consumption, respectively (11). The total amount of ATP measured was similar in all three genotypes, but the p53-deficient cells produced significantly higher levels of lactate, indicating a change in the mode of energy production to one favoring glycolysis (Fig. 1B). The ratio of ATP produced by glycolysis versus ATP produced by aerobic respiration increased with p53 deficiency: p53+/+, 0.81 ± 0.12; p53+/–, 1.04 ± 0.20; and p53–/–, 1.72 ± 0.16 (Fig. 1B).

Fig. 1.

Dependence of aerobic respiration, glycolysis, and exercise capacity on p53. (A) Oxygen consumption (mean ± SD, nmol min–1 mg–1 protein) was measured by a Clark-type oxygen microelectrode in mouse liver mitochondria preparations and in HCT116 cells with three TP53 genotypes: wild-type (+/+), heterozygous (+/–), and homozygous disruption (–/–). (B) The amount of ATP (mean ± SD, nmol min–1 mg–1 protein) produced by aerobic respiration (gray bars, O2) and glycolysis (light bars, Lac) was calculated by measuring oxygen consumption and lactate production in the three different TP53 genotypes of HCT116 cells. (C) Body mass composition of muscle, fat, and free fluid (percentage of total) was determined in nonanesthetized age- and sex-matched mice using a nuclear magnetic resonance analyzer. Body mass (mean ± SD) for the p53+/+ and p53–/– mice was 20.4 ± 1.2 g and 20.1 ± 1.9 g, respectively. p53 genotypes indicated for wild-type (+/+) (dark bars, n = 5) and homozygous knockout (–/–) (light bars, n = 5) mice. (D) Maximum duration of swimming exercise measured in a temperature-controlled adjustable-current tank. Exercise time (mean ± SD, s) to the point of exhaustion, as indicated by inability to stay afloat, is shown for each genotype (P = 0.008, n = 5 for each group). Note that the p53–/– mouse (foreground) assumes a more vertical posture consistent with early fatigue, in contrast to the wild-type mouse (background) with horizontal body and tail positions (arrowheads) (13).

Our results indicated that p53–/– mice exhibit decreased mitochondrial respiration, which could translate into a decrease in functional aerobic capacity. The general appearance and behavior of p53+/+ and p53–/– mice were indistinguishable, as previously observed (12), and their body mass and composition were essentially identical (Fig. 1C). However, the endurance of p53–/– mice was reduced, as measured by a swimming stress test (Fig. 1D) (13).

Because p53 is a transcriptional activator, we reasoned that a gene regulated by p53 might mediate this effect on mitochondrial respiration. We examined a serial analysis of gene expression (SAGE) database of potential p53 target genes (14) that may influence mitochondrial function. SCO2 (Synthesis of Cytochrome c Oxidase 2) was induced in a p53-dependent manner in HCT116 cells, as evidenced by a ninefold increase in transcripts. On the basis of more recent genetic and biochemical studies (15, 16), we reasoned that SCO2 could potentially mediate the effect of p53 on aerobic respiration. SCO2 is required for the assembly of mitochondrial DNA-encoded COX II subunit (MTCO2 gene) into the COX complex, and inactivating mutations of this gene in humans result in fatal cardioencephalomyopathy due to aerobic respiratory failure (17).

To confirm that SCO2 transcription is directly transactivated by p53, we used human colon cancer DLD1 cell lines with either inducible wild-type (WT) or mutant (MUT) p53 for expression studies (18). SCO2 mRNA expression increased within 3 to 18 hours of induced WT p53 expression but was not affected by mutant p53 (Fig. 2A and fig. S2). We also measured the mRNA expression level of SCO1, another COX assembly factor. SCO1 mRNA expression remained unchanged with induced p53 expression, serving as a negative control and demonstrating the specificity of the observed SCO2 mRNA increase (Fig. 2A).

Fig. 2.

Dependence of SCO2 mRNA transactivation and protein levels on p53. (A) SCO2 and SCO1 transactivation by wild-type (WT, dark bar) or mutant (MUT, light bar) p53 were measured using a modified tetracycline-inducible system in human colon cancer DLD1 cell line. mRNAs were quantified by real-time reverse transcription (RT)–PCR and normalized to a control gene EIF3S5 (mean ± SD, relative value). Time (in hours) indicates induction by removal of doxycycline. (B) p53 transactivates luciferase reporter construct containing the SCO2 gene p53 binding site. Normalized luciferase activity under the control of a ∼1.1-kb fragment of the SCO2 gene containing wild-type (dark bars) or mutated (light bars) p53 binding site in response to varying amounts of wild-type (pCMV-p53) or mutant (pCMV-p53mt135) p53 transfection. (C) Dependence of SCO2 protein expression levels on p53 gene dosage in mouse liver mitochondria and HCT116 cells. Western blot analyses were done with specific antibodies against SCO2, SCO1, and VDAC proteins in mouse liver mitochondria and human HCT116 cell lines.

The time course of SCO2 transcription and the presence of p53 DNA-binding consensus sequence in intron 1 (–322 to –302 from the start codon) suggested that SCO2 may be directly transactivated by p53. We tested this hypothesis by cloning a ∼1.1-kb fragment of the SCO2 gene containing the p53-binding sequence into a luciferase reporter plasmid. To control for the nonspecific effects of endogenous p53, we used the p53–/– HCT116 cell line to express the reporter and exogenous p53 plasmids. As predicted for a specific p53 binding site, we observed increases in luciferase activity that were dependent on the dose of exogenous WT p53 (pCMV-p53) but not mutant p53 (pCMV-p53mt135) (Fig. 2B). When the p53 binding sequence in the reporter construct was mutated, the transcriptional activation by exogenous WT p53 was abrogated, establishing its in vitro specificity (Fig. 2B).

Antibodies against SCO2 were generated and used to determine the expression of SCO2. SCO2 protein expression decreased in both p53-deficient mouse liver mitochondria and human cell lines, whereas the amount of mitochondrial anion transporter (voltage-dependent anion channel, or VDAC) remained constant (Fig. 2C). Although p53 transactivates SCO2, the amounts of SCO2 protein detected in the p53–/– mice and human cells indicate that p53-independent pathways exist for its regulation, as is the case for other p53-target genes (Fig. 2C) (19). In contrast to SCO2 expression, the level of SCO1 protein was unaffected by p53 genotype status, consistent with transcriptional data (Fig. 2C and fig. S2). These results suggested a direct interaction between p53 and SCO2 to regulate mitochondrial aerobic respiration.

Expression of WT SCO2 cDNA in p53–/– HCT116 cells increased mitochondrial oxygen consumption within 48 hours in a dose-dependent manner (Fig. 3A). The amount of SCO2 protein needed to rescue the deficit in mitochondrial respiration of the p53–/– cells corresponded well to the physiological levels observed in the p53+/+ cells (Fig. 3A). Oxygen consumption also increased with SCO2 expression in human colon cancer cell lines with naturally occurring p53 mutations (DLD1 and SW480 cells), suggesting that SCO2 expression can generally rescue oxygen consumption in p53-deficient cells (Fig. 3A).

Fig. 3.

Expression of SCO2 rescues alterations in mitochondrial respiration. (A) Transient transfection of SCO2 (pcDNA-SCO2, ng of DNA) into p53–/– HCT116 cells (light bars) increased SCO2 protein expression and oxygen consumption (mean ± SD, nmol min–1 mg–1 protein) to p53+/+ (dark bar) levels. Control empty vector (pcDNA, ng of DNA) had no effect. Oxygen consumption was increased to wild-type HCT116 levels by SCO2 in p53-mutant (MUT) cell lines DLD1 and SW480 (gray bars). (B) Transient knockdown of p53 in wild-type HCT116 cells by sequence-specific siRNA (p53) decreases SCO2 protein and oxygen consumption (mean ± SD, nmol min–1 mg–1 protein). Mock (no siRNA) and nonspecific (NS) siRNA transfections had no effect.

To rule out the possibility that decreased mitochondrial respiration in the p53–/– HCT116 cells may result from chronic metabolic or other unrelated genetic changes that may have occurred during clonal selection, we examined oxygen consumption after reducing p53 expression by small interfering RNA (siRNA). Decreased p53 expression in parental HCT116 cells resulted in decreased SCO2 protein expression and reduced oxygen consumption within 48 hours, phenocopying the p53-deficient cells (Fig. 3B).

To genetically test whether p53 regulates mitochondrial respiration through SCO2, we disrupted the SCO2 locus by homologous recombination in the human HCT116 cell line. Elimination of one of the two SCO2 alleles could phenocopy the p53-deficient cells through the direct reduction of SCO2 expression. A targeting vector was designed to remove about 500 base pairs of SCO2 coding sequence upon homologous recombination into the SCO2 locus. This construct was assembled into a recombinant adeno-associated virus (rAAV) (20). After viral transduction and screening of 1200 drug-selected clones, we identified three independently derived clones that gave the appropriately sized knockout (KO) polymerase chain reaction (PCR) product, indicating successful disruption of the SCO2 allele (Fig. 4A). The amounts of SCO2 protein in each of the three SCO2+/– (H1 to H3) knockout clones were reduced relative to that in the WT parental (P) and sister clones (S1 and S2) that had also undergone parallel viral transduction and selection processes (Fig. 4A). Cell viability and growth rates were similar for the WT and SCO2+/– clones. However, the SCO2+/– cells displayed decreased oxygen consumption (O2) and increased glycolytic activity (Lac) compared to WT cells, whereas total ATP concentrations remained the same, similar to the results for p53-deficient cells (Fig. 4B).

Fig. 4.

Targeted disruption of one SCO2 allele reproduces the p53-deficient metabolic phenotype. (A) Three clones (H1 to H3) with heterozygous disruption of SCO2 were identified by the appearance of the 3.8-kb KO band accompanied by a decrease in the intensity of the 2.7-kb WT band. P is the parental cell line, and S1 and S2 are two sibling clones without SCO2 disruption. SCO2 protein expression levels were assessed by Western blotting. (B) The amount of ATP (mean ± SD, nmol min–1 mg–1 protein) produced by aerobic respiration (gray bars, O2) and glycolysis (light bars, Lac) was calculated by measuring oxygen consumption and lactate production in the three wild-type (+/+) and three SCO2 heterozygous (+/–) knockout clones.

The direct regulation of aerobic respiration by p53 has important implications for understanding tumorigenesis. To date, much of the research on cellular processes driven by oncogenes and tumor suppressor genes has focused on the regulation of cell birth and death (1). Our study identifies a direct transcriptional target of p53 that modulates aerobic respiration and, by its haploinsufficiency, recapitulates the Warburg effect. It also defines a genetic pathway, not involving cell cycle regulation, by which the inactivation of p53 may promote tumorigenesis by decreasing cellular dependence on oxygen, potentially permitting growth in more hypoxic environments. Two recent studies have implicated p53 in the control of glycolysis by its negative regulation of phosphoglycerate mutase (PGM) and Akt (6, 7, 21). Interestingly, despite the marked increase in glycolysis stimulated by Akt expression, cellular oxygen consumption remained stable, reflecting the tight homeostatic controls governing mitochondrial respiration (22). Additional functions of the TP53 gene continue to be uncovered (23), and the effect of p53 status on exercise tolerance suggests that it has global functions that extend beyond processes related to cell birth and death (24). The COX complex is critical for aerobic eukaryotes, and SCO2 and MTCO2 (COX II) are ancient genes conserved in organisms as diverse as yeast and humans. The regulation of mitochondrial genome-encoded COX II by p53 suggests that our observations pertain to a fundamental control point in metabolism. Recent studies have implicated p53 as a mediator of senescence and aging (2527), although specific aspects of this model remain controversial (28, 29). Given the mounting evidence supporting a role for metabolism and oxidative stress in aging (30, 31), the functional relationship between p53 and the COX complex assembly may underlie some aspects of organismal aging. In filamentous fungus, the genetic disruption of COX II markedly increases life-span (32). Future studies examining the role of p53 in mitochondrial regulation may clarify how a tumor suppressor gene can have such diverse and global effects on cellular and organismal functions.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

References and Notes

References and Notes

Stay Connected to Science

Navigate This Article