PGC1α Promotes Tumor Growth by Inducing Gene Expression Programs Supporting Lipogenesis Free

Pioneering work by Warburg described the ability of tumor cells to use glycolysis to generate ATP and lactic acid, even in the presence of oxygen, that is, aerobic glycolysis, or as it is commonly called, the Warburg Effect (1). However, increased glucose utilization cannot be explained solely by increased ATP production as initially proposed by Warburg. Besides the generation of ATP, there are a number of other benefits of increased glucose metabolism. Glucose serves as a precursor for biosynthesis of molecules involved in generating biomass such as nucleic acids and lipids. Indeed, increased nucleic acid and lipid synthesis play an important role in many cancers (2–4). Therefore, the ability of cancer cells to coordinate glucose metabolism is a crucial aspect of the metabolic phenotype. This has prompted interest into understanding and targeting key molecules regulating glucose metabolism.

PPARγ coactivator 1α (PGC1α) is a major regulator of several key metabolic pathways. PGC1α was initially identified as the key factor driving thermogenesis in brown fat (5). Numerous studies have since shown a key role for PGC1α in inducing the expression of genes of oxidative phosphorylation and the tricarboxylic acid (TCA) cycle in various tissues (6–8). PGC1α also plays an important role in regulating other metabolic pathways. Recent studies show that PGC1 also promotes anabolic pathways such as de novo lipogenesis (9, 10). This is accompanied by an increase in the pentose phosphate pathway to generate NADPH for fatty acid synthesis (9). This highlights the important role that PGC1α plays in regulating multiple aspects of metabolism in addition to its ability to promote oxidative metabolism.

Initial studies on the role of PGC1α in cancer showed an association between reduced expression of PGC1α compared with normal adjacent tissue (11). The ability of PGC1 to drive mitochondrial function led to speculation that reduced PGC1α in tumors may be responsible for the Warburg effect. Indeed, several studies showed that decreased PGC1α expression was associated with reduction in mitochondrial function and increased growth (12, 13). Although the Warburg effect is a well-described phenomenon, more recent studies show that mitochondrial function is required for transformation and tumor growth (14–16). This supports several studies suggesting a potential procancer role for PGC1α (17–19). These studies highlight the conflicting data about the role of PGC1α in maintaining tumor growth. Regardless of the associations between PGC1 expression and established tumors or cell lines, whether or not PGC1α is involved in tumorigenesis is not known. Therefore, we have taken an approach using gain and loss of PGC1α expression to determine the role of PGC1α on tumorigenesis and tumor growth.

Protocols were approved by the University of Maryland Animal Care and Use Committee and conducted under veterinary supervision. PGC1α knockout (Pgc1α^−/−) mice were obtained from Dr. Bruce Spiegelman (Dana-Farber Cancer Institute/Harvard Medical School; ref. 20). These mice have been observed for more than 2 years and did not appear to be any colon or tumor development (data not shown and Dr. Bruce Spiegelman and Jiandie Lin, personal communication). Colons and livers were removed from mice, and RNA was isolated using TRIzol as previously described (21, 22). Colon carcinogenesis was induced by injecting mice once per week with 10 mg/kg azoxymethane for 8 weeks as previously described (23). Mice were monitored for 25 weeks and then euthanized. Colons were removed and fixed for tumor analysis. For liver carcinogenesis, mice were injected at 14 days of age with 25 mg/kg diethylnitrosamine. Mice were followed up to 24 or 40 weeks and then euthanized. Livers were removed for tumor analysis. Formalin-fixed liver tissue was paraffin embedded and 5 μm sections cut by the University of Maryland Greenebaum Center Pathology Core. Liver sections from mice euthanized at 24 weeks were stained with hematoxylin and eosin and pathologic analysis and tumor number determined blindly by a board certified pathologist (W. Twaddel). For liver sections from mice euthanized at 40 weeks, due to greater tumor formation in the Pgc1α^+/+ mice, it was not possible to count individual tumors, as tumors grew into each other. Therefore, tumor burden was determined by measuring tumor area. Colons were examined blinded under a dissecting scope, and gross tumor number was determined. For xenograft studies, 1 × 10⁶ cells were injected subcutaneously into the flank of severe-combined immunodeficient (SCID) mice (Taconic) in 100 μL of media. Tumor growth was monitored every 3 days using a digital caliper and volume calculated as previously described (22). At the end of the experiment, mice were euthanized, tumors harvested and processed for RNA, protein, and histopathology. Data were obtained from 8 to 12 mice per experimental group and experiments repeated at least 2 times. For studies inhibiting fatty acid synthesis, HT29 pcDNA control and pcDNA PGC1α expressing cells were inoculated into the flanks of SCID mice. As soon as tumors were palpable, mice were administered 10 mg/kg C75 (Toronto Chemical Company) twice a week and tumor growth monitored. Five mice per group were used for these experiments. For XCT790 inverse estrogen-related receptor (ERR) α agonist experiments, wild-type mice were treated with 25 mg/kg for 3 days by intraperitoneal injection. Livers were removed, RNA extracted, and real-time PCR carried out as described below.

HT29 and Colo205 cells were obtained from American Type Culture Collection and maintained in Dulbecco's Modified Eagle's Media (Cellgro) supplemented with 10% FBS and penicillin/streptomycin. American Type Culture Collection characterizes cell lines by short tandem repeat profiling. Experiments were carried out with cells at less than 25 passages after receipt. Lentiviral expression short hairpin RNA (shRNA) against PGC1α was obtained from Sigma. Lentiviral particles expressing shRNA against PGC1 were produced according to manufacturer's directions in 293T cells. Virus was transduced into Colo205 cells along with 8 μg/mL polybrene and cells selected with puromycin. PGC1α knockdown was confirmed by RT-PCR and Western blotting for PGC1α (Calbiochem). For PGC1α gain-of-function studies, control pcDNA and pcDNA expressing PGC1α were transfected into HT29 cells and cells selected in G418 to obtain PGC1α expressing stable cells. PGC1α overexpression was confirmed by RT-PCR and Western blotting for PGC1α (Calbiochem).

Tissues and cells were lysed in RIPA buffer and proteins harvested. Hundred micrograms of protein was separated using SDS-PAGE and transferred to nitrocellulose. Membranes were incubated with PGC1α (Calbiochem), SREBP1c (BD Bioscience), and actin (Sigma) antibodies and secondary anti-rabbit or anti-mouse horseradish peroxidase–conjugated secondary antibody (Jackson ImmunoResearch). Proteins were visualized using enhanced chemiluminescence.

Cells were plated at 10,000 cells per well in a 6-well plate, and cells were counted every 2 days using a hemocytometer and trypan blue exclusion as previously described (22).

RNA was extracted from cells and tissues using TRIzol as previously described (21, 22). cDNA was synthesized and real-time RT-PCR conducted using SYBR Green as previously described using gene-specific primers (Supplementary Table SI) and normalized to actin as a control (21, 22).

For total triacylglycerol (TAG) determination and TAG synthesis, the lipids were extracted by the Folch method from individual liver and tumor tissues. Total upper phase was dried down, resuspended in isopropanol, and assayed with triglyceride kit (Sigma) by the University of Maryland Nutrition and Obesity Research Core. Equivalent of 5 mg for liver tissue and 1 mg for tumor tissue were analyzed with thin layer chromatography extractions. Tissue lipids were separated with chloroform/acetone/acetic acid (96:4:1) as solvent. The lipids were visualized with phosphomolybdenum vapor.

For metabolic flux analysis, we used stable isotope–based tracer analysis. [U6-¹³C₆]glucose (>99% purity and 99% isotope enrichment for each carbon position; Cambridge Isotope Labs) was used as a tracer. Mice with HT29 pmscv and HT29 PGC1α expressing xenografts were administered ¹³C glucose, and tumors and plasma were collected 3 hours later. Specific extractions and analysis were conducted as previously described and below (10, 24, 25). Fatty acids were extracted by saponification of TRIzol cell extracts after removal of the RNA containing supernatant with 30% KOH and 100% ethanol using petroleum ether. Fatty acids were then converted to their methylated derivatives using 0.5 N methanolic HCl. Palmitate was monitored at m/z of 270. The enrichment of acetyl units and the synthesis of new lipid fraction were determined using the mass isotopomers of palmitate with the enrichment of ¹³C-labeled acetyl units used to reflect synthesis of the new lipid fraction as determined by mass isotopomer distribution analysis (MIDA). Media C¹³/C¹² ratios in released CO₂ were used as the direct measure of glucose oxidation.

Mass spectral data were obtained on the HP5973 mass selective detector connected to an HP6890 gas chromatograph. The settings are as follows: gas chromatography (GC) inlet, 230°C; transfer line, 280°C; mass spectrometry (MS) source, 230°C; and MS Quad, 150°C. An HP-5 capillary column (30 m length, 250 μm diameter, 0.25 μm film thickness) was used for glucose, ribose, and lactate analysis. A Bpx70 column (25 m length, 220 μm diameter, 0.25 μm film thickness; SGE Incorporated) was used for fatty acid analysis with specific temperature programming for each compound studied as previously described.

For growth and gene expression analysis, the Student t test was used to determine statistical significance. The Fisher exact test was applied to colon cancer incidence with significance defined as P < 0.05.

One of the first studies to show an association between PGC1α and cancer showed that PGC1α levels are reduced in colon-derived tumor tissue compared with normal adjacent tissue (11). PGC1α is abundantly expressed throughout the small intestine and colon and in the stem cell crypt compartment and at the top of intestinal crypts (26). Therefore initially, we examined the role of PGC1α in colon tumorigenesis. Mitochondrial gene targets of PGC-1α involved in the TCA cycle and oxidative phosphorylation were downregulated from the colons of PGC1α^−/− mice compared with Pgc1α^+/+ mice (Fig. 1A). We also examined whether there was a compensatory increase in PGC1β due to loss of PGC1α but found a decrease in expression (Supplementary Fig. S1A). We then induced tumorigenesis using a colon-specific carcinogen, azoxymethane. Azoxymethane-induced tumors originate from epithelial cells lining the colon and grow as polyps or adenomas which are similar to colon carcinoma in humans. Mice were examined for colorectal tumors 25 weeks following the last azoxymethane injection. Despite the reduction in oxidative phosphorylation gene expression in the Pgc1α^−/− mice, there was a significant reduction in tumor incidence in the Pgc1α^−/− mice. Eighty-seven percent of Pgc1α^+/+ mice had colonic polyps, whereas less than 30% of the Pgc1α^−/− mice had polyps (Fig. 1B, P < 0.01). In addition, in mice with tumors, loss of PGC1α reduced tumor multiplicity by more than 50% (Fig. 1C). Therefore, despite studies showing reduced PGC1α expression in colon-derived tumors compared with normal tissue, loss of PGC1α protects against colon tumorigenesis.

Next, we wanted to determine whether the ability of PGC1α to promote tumorigenesis was specific for the colon. PGC1α plays a key role in regulating glucose homeostasis in the liver and represents one of the most well-studied sites of action of PGC1α (7). Initially we examined the livers of Pgc1α^+/+ and Pgc1α^−/− mice for the expression of PGC1α targets. Similar to previous studies, loss of PGC1α was associated with a reduction in the expression of oxidative phosphorylation and TCA cycle genes (Fig. 2A; ref. 8). We also observed a decrease in PGC1β; however, it was not statistically significant (Supplementary Fig. S1B). Next, we examined the role of Pgc1α on liver tumorigenesis using Pgc1α^+/+ and Pgc1α^−/− mice. Liver carcinogenesis was induced in 14-day-old mice using the liver-specific carcinogen, diethylnitrosamine. Diethylnitrosamine is a DNA-alkylating agent that induces pericentral foci and small dysplastic hepatocytes leading to multifocal hepatocellular carcinoma displaying characteristic similar to that observed in human hepatocellular carcinoma. After 24 weeks, the number of liver tumors in Pgc1α^−/− mice was reduced by approximately 60% compared with Pgc1α^+/+ mice (Fig. 2B). After 40 weeks following diethylnitrosamine treatment, we were not able to distinguish individual tumors in Pgc1α^+/+ mice. Therefore, we examined mice for tumor burden. We observed a significant decrease in tumor burden in the livers of Pgc1α^−/− mice compared with Pgc1α^+/+ mice 40 weeks following diethylnitrosamine treatment (Fig. 2C, right). These data show that despite the reduction in oxidative phosphorylation and TCA cycle gene expression in Pgc1α knockout mice, loss of PGC1α protects against carcinogenesis.

Pgc1α whole body knockout mice exhibit multiple metabolic abnormalities (20). In addition, given the ability of PGC1α to control the expression of metabolic genes, it may be altering the metabolism of the carcinogens used. Therefore, we wanted to determine the effect of PGC1α in a more defined setting. We reduced the expression of PGC1α in the Colo205 human colorectal cancer cell line using a lentiviral-based shRNA against PGC1α (Supplementary Fig. S1C). Knockdown of PGC1α led to a reduction in oxidative phosphorylation and PGC1β gene expression (Fig. 3A and Supplementary Fig. S1D). Interestingly, despite the decrease in mitochondrial gene expression, we did not observe a difference in cell proliferation in vitro (Fig. 3B). Because PGC1α plays a major role in nutrient sensing and homeostasis, we examined the effect of loss of PGC1α on tumor growth in vivo. We inoculated Colo205 cells expressing nontargeting-shRNA or PGC1α-shRNA into the flank of SCID mice and measured tumor xenograft growth. Growth of PGC1α-shRNA expressing cells was reduced almost 60% compared with control nontargeting-shRNA expressing cells (Fig. 3C). Next, we wanted to determine whether PGC1 could promote tumor growth by overexpressing PGC1α in the HT29 colon cancer cell line (which expresses little PGC1α; Supplementary Fig. S1E). Ectopic expression of PGC1α increased the expression of genes driving oxidative phosphorylation (Fig. 3D). In addition, we observe that PGC1α promotes the expression of PGC1β (Supplementary Fig. S1F). Similar to the knockdown data, altering PGC1α expression did not appear to alter cell proliferation in vitro (Fig. 3E). We then examined the effect of PGC1α expression on tumor growth in vivo. Control and Pgc1α expressing HT29 cells were inoculated into the flank of SCID mice, and tumor growth was measured. As shown in Fig. 3F, PGC1α-overexpressing tumors grew almost 3 times as large as control tumors. Although PGC1α did alter cell growth in vitro, these studies show a direct role for PGC1α in promoting tumor growth in vivo.

Recent studies have shown that despite the well-known role of Pgc1α in driving oxidative metabolism, it can also promote de novo fatty acid synthesis (9, 10). Lipogenesis has become recognized as playing an important role in tumorigenesis and cancer cell growth (2). Indeed, the key proteins controlling fatty acid synthesis from acetyl-coA, acetyl-Co carboxylase (ACC) and fatty acid synthase (FASN), play an important role in promoting cancer growth (27, 28). As shown in Fig. 4A, the gene expression of Acc and Fasn were reduced in the colons of Pgc1α^−/− mice compared with Pgc1α^+/+ mice. Similarly, the expression of Acc and Fasn was also reduced in the livers of Pgc1α^−/− mice compared with Pgc1α^+/+ mice (Fig. 4B). We then examined lipogenic gene expression from the tumor xenografts with loss and gain of PGC1α expression. Knockdown of PGC1α in Colo205 tumors led to significant reduction in expression of both ACC and FASN (Fig. 4C). Conversely, expression of PGC1α in HT29 tumors increased ACC and FASN expression (Fig. 4D).

The induction of the TCA cycle and oxidative phosphorylation by PGC1 raises the question as to how these pathways are linked to fatty acid synthesis. Acetyl-CoA is required for de novo fatty acid synthesis from glucose. However, acetyl-CoA produced from glucose is generated in the mitochondria, whereas fatty acid synthesis occurs in the cytoplasm. To use acetyl-CoA for fatty acid synthesis, it is converted to citrate in the TCA cycle and shuttled out of the mitochondria by the mitochondrial citrate transporter, solute carrier family 25, member 1 (SLC25A1). In the cytoplasm, ATP citrate lyase (ACLY) converts the mitochondria-produced citrate into oxaloacetate and acetyl-CoA. Loss of PGC1α expression in PGC1α^−/− mice or knockdown of PGC1 in Colo205 cells led to a reduction in SLC25A1 and ACLY (Fig. 5A–C). In contrast, the expression of SLC25A1 and ACLY was increased in HT29 tumors overexpressing PGC1α (Fig. 5D). Together, these data suggest that PGC1α coordinately regulates gene expression promoting metabolic pathways required for converting glucose into fatty acids.

SREBP1 is one of the most well-studied transcription factors driving the lipogenic gene expression program in the liver (29). A number of studies show that SREBP1c and its ability to promote lipogenesis play a role in increased tumor development and growth (30). SREPB1c exists as a precursor in the cytoplasm and is activated by cleavage and subsequent nuclear localization of the mature form (2, 29). This prompted us to examine the livers and colons of wild-type and knockout mice for the expression of mature SREPB1c. Loss of PGC1α did not alter the expression of cleaved SREBP1c in the liver and colons from mice (Supplementary Fig. S2A and S2B). In addition, we did not detect SREBP1c protein expression in HT29 and Colo205 xenografts (data not shown). This further suggests that PGC1α does not mediate its effects by regulating mature SREBP1c expression.

PGC1α directs programs of gene expression by interacting with transcription factors. One of the most well-characterized transcription factors mediating the effects of PGC1α on energy metabolism is ERRα (31, 32). Interestingly, ERRα has also been shown to promote tumor growth and is associated with reduced survival in several cancers (33–37). We examined the effect of ERRα inhibition in the liver of wild-type C57Bl/6J using an ERRα inverse agonist, XCT790. Inhibition of ERRα decreased the expression of cytochrome c expression, a typical target of PGC1α and ERRα (Supplementary Fig. S2A). However, we did not observe a difference in SLC25A1, ACLY, ACC, and FASN gene expression following treatment with the ERRα antagonist (Supplementary Fig. S2B–S2E). While ERRα may be responsible for the effects of PGC1α on gene expression driving cellular bioenergetics, the data suggest that ERRα is not responsible for the effects of PGC1α on lipogenic gene expression.

Given the ability of PGC1α to promote lipogenic gene expression, we next examined whether PGC1 promoted lipid accumulation. Initially, we examined TAG levels in Pgc1α^+/+ and Pgc1α^−/− liver and HT29 xenograft tumor tissue. TAG content was significantly reduced in the livers of Pgc1α^−/− mice (Supplementary Fig. S4A). In HT29 tumors expressing Pgc1α, TAG levels were significantly increased (Supplementary Fig. S4B). Measuring TAG levels shows mainly the steady-state accumulation of lipids and is not a direct measure of synthesis. Therefore, we used stable ¹³C isotope tracer studies to determine the effect of PGC1α on de novo fatty acid synthesis. Mice with control or PGC1α-expressing xenografts were established and administered ¹³C glucose, and plasma and tumors were harvested 3 hours later. Plasma from mice bearing PGC1-expressing tumors showed increased ¹³CO₂ concentration (Fig. 6A). This shows the increased glucose utilization in mice with PGC1α-expressing tumors even over the course of a short incubation time (3 hours). We then directly examined fatty acid synthesis in tumors by measuring the incorporation of ¹³C from glucose into palmitate, the product of FASN. Despite the short incubation, approximately 2% of the palmitate derived from tumors was ¹³C labeled. This increased by more than 15% in the Pgc1α-expressing tumors (Fig. 6B). We also examined plasma ¹³C-labeled palmitate to determine whether the labeled palmitate was derived from nontumor tissue and subsequently taken up by tumors. The percentage of ¹³C-labeled palmitate in plasma was much less than 1% of the total palmitate and did not change in plasma from mice bearing PGC1α-expressing tumors, confirming that palmitate was being produced by tumor (Supplementary Fig. S4C). Subsequent positional mass isotope analysis showed that the increase in labeled palmitate was due to increased de novo synthesis, which was increased more than 50% compared with control tumors (Fig. 6C).

These studies show that Pgc1α expression is associated with induction of a program of lipogenic gene expression and de novo lipogenesis. However, it does not show that increased fatty acid synthesis per se is mediating the effects of Pgc1α on tumor growth. To test this, we established pcDNA control and PGC1α-expressing xenografts in mice and inhibited fatty acid synthesis with the FASN inhibitor C75, once tumors had formed (38). We used a lower dose of C75 than previous studies, given the ability of C75 to inhibit tumor growth. Similar to the studies above, Pgc1α-expressing tumors grew about 3 times as large as control tumors (Fig. 6D). C75 reduced the growth of the control tumors at about 20%, although it was not statistically significant. In contrast, C75 treatment of mice with tumors expressing PGC1α significantly reduced the growth of tumors by approximately 50%. This shows that the tumor growth promoted by PGC1 is mediated, in part, via induction of fatty acid synthesis.

Altered cancer metabolism has become recognized as a hallmark of cancer. Although the Warburg effect and glycolysis are recognized as key aspects of tumor metabolism, tumors cells need to be able to coordinate energy-generating and biosynthetic pathways to effectively promote cell proliferation (2–4). PGC1α is a key metabolic regulator that controls multiple aspects of glucose metabolism. Our studies show a novel role for PGC1α in promoting carcinogenesis and tumor growth. This effect appears to be mediated via coordinating the induction of a gene expression program that facilitates the conversion of glucose to fatty acids.

Elevated fatty acid synthesis has become recognized as an important pathway in cancer (2). In addition to generating membranes for biomass, lipids are used for signaling pathways that are often elevated in many cancers. Lipids play an important role in transmitting signals from the plasma membrane via second lipid messengers and eicosanoids. In addition, lipid modification of a number of oncogenes, including RAS and AKT, is required for full oncogenic activation (39, 40). Therefore, the ability of PGC1α to modulate fatty acid synthesis would also provide cancer cells with precursors for signal transduction pathways regulating cell growth.

Although increased mitochondrial function in terms of the TCA cycle and oxidative phosphorylation are typically associated with reduced growth, recent studies highlight the need for oxidative phosphorylation and the TCA cycle in promoting tumor growth (14–16). Indeed, several studies show a potential role for PGC1α in this process. Ectopic expression of KRAS in NIH3T3 fibroblasts leads to increased proliferation, which is associated with increased PGC1α and its downstream target genes (17). The ability of breast cancer cells to metastasize to the brains of mice was also associated with increased PGC1α expression and its target genes (18). Another study showed that activation of PPARδ induced cell proliferation, which was associated with increased PGC1α expression (41). Despite the association between PGC1α and cell growth, a direct role for PGC1α was not shown. A more recent report showed that knockdown of PGC1α in prostate cancer cells reduced growth in vitro (42). However, the prostate cell lines used have very little PGC1α, raising questions about PGC1α knockdown.

These studies highlight the observation that multiple metabolic pathways regulate tumor cell growth and that increased mitochondrial function per se does not necessarily inhibit growth. The coordinated induction of TCA cycle and oxidative phosphorylation by PGC1α would provide cells with a strong metabolic advantage. Making a daughter cell is a bioenergetically costly endeavor whereby glucose is used for both energy and biosynthetic precursors. Induction of oxidative phosphorylation and the TCA cycle by PGC1α would enable cells to make more glucose available for biomass generation, as oxidative phosphorylation and the TCA cycle are more efficient at generating energy. The importance of lipogenesis in tumor metabolism highlights another need for the induction of the TCA cycle by PGC1α. De novo fatty acid synthesis from glucose requires acetyl-CoA. However, acetyl-CoA is produced in the mitochondria, whereas fatty acid synthesis occurs in the cytosol. Therefore, the TCA cycle is required for converting mitochondria-generated acetyl-CoA into citrate. Our studies show that PGC1α plays an additional role in this process by (directly or indirectly) inducing the expression of genes involved in these pathways and bridging the known mitochondrial and lipogenic functions of PGC1α. ACLY promotes the conversion of citrate back to oxaloacetate and acetyl-CoA, providing substrates for fatty acid synthesis by ACC and FASN. Therefore, our studies suggest that PGC1 is coordinating energy production and mitochondrial function with biosynthetic pathways to fuel cancer growth (Fig. 7).

The ability of PGC1 to regulate energy metabolism occurs, in part, via coactivation of the transcription factor ERRα (31, 32). Recent studies highlight an important role for ERRα in promoting cancer growth in several different cancer types (35, 36, 43). It has also been reported that higher expression of ERRα is associated with a worse prognosis for several cancers (33, 34, 37). A stronger connection between ERRα and PGC1α was suggested in a recent study which showed that tumorigenesis of fibroblasts by KRAS is mediated, in part, by PGC1α and ERRα (44). However, using an inverse agonist of ERRα, we did not observe an alteration to lipogenic gene expression in livers of mice. Therefore, although PGC1α might regulate the expression of genes driving energy metabolism, it most likely regulates lipogenic gene expression independently of ERRα. In contrast to ERRα, SREBP1c is a key transcription factor mediating the program of lipogenesis. We did not observe a difference in expression of mature SREBP1c in livers or colons of Pgc1α^−/− mice. Further ruling out a role for SREBP1c was the lack of SREBP1c in the Colo205 and HT29 xenografts. However, the possibility exists that PGC1α may be interacting and coactivating SREBP1c to increase lipogenic gene expression without altering the expression of SREBP1c. Future studies will elucidate the mechanism(s) by which PGC1α promotes lipogenesis in cancer and whether the ability of PGC1α to promote gene expression programs regulating energy metabolism and lipogenesis is the result of distinct or related transcriptional programs. In addition, it remains to be determined whether the effect of PGC1α on SLC25A1 and ACLY expression is a direct transcriptional effect or whether it is secondary to induction of fatty acid synthesis.

The gain and loss of PGC1α expression studies presented here help to resolve the conflicting data about the role of PGC1 in cancer. However, a recent study overexpressing PGC1α in a breast cancer–derived xenograft model did not observe a difference in growth in control tumors versus tumors expressing PGC1α (45). Because we observed an effect of PGC1α on tumor growth using both gain and loss of PGC1α expression in colon cancer cell lines, tissue-specific differences may explain the contradictory results. Indeed, recent studies suggest that PGC1α displays tissue-specific differences in function (46). Hence, PGC1α may be a target in colorectal and liver cancer but not breast cancer.

Despite alterations in oxidative metabolism gene expression in vitro, we primarily observed an effect of PGC1α on tumor growth in vivo. The difference between in vitro and in vivo effects may be explained by the important role that PGC1α plays in nutrient response and signaling. In cell culture, most nutrients, such as glucose and oxygen, are not limiting. However, in vivo, the tumor microenvironment is an area of intense metabolic stress where nutrients are more limiting and therefore PGC1α may play a more important role. Our in vitro data also disagree with a study showing that knocking down PGC1α in prostate-derived cancer cells reduces growth in vitro (42). In addition to questions about expression of PGC1α in prostate as mentioned above because PGC1α is primarily a transcriptional coactivator, these differences may be attributable to the presence of cofactors that are expressed in a tissue- or cell type–specific manner. A recent article described an antigrowth role for PGC1α in the colon (26). It is unclear as to the reasons for the contradictory result. Our studies used stable Lentiviral- and retroviral-based technologies, whereas the recent study used adenoviral PGC1α injections directly into tumors. In addition, D'Errico and colleagues found that loss of PGC1 protects against tumor formation. Possible differences may be attributed to the chemical and genetic models used and strain differences. In addition, a question that is raised in general with regard to the studies by D'Errico and colleagues is that they show that PGC1 prevents tumorigenesis by promoting ROS. However, studies show that PGC1α protects against generation of reactive oxygen species and upregulates antioxidant defense (47, 48).

Previous studies show that PGC1α is reduced in tumors (12, 13, 18). Most of these studies primarily show an association between PGC1α expression and tumor growth and did not directly determine the role of PGC1α on cell growth. Therefore, PGC1α may play a role during carcinogenesis and then its expression decreases as tumors develop. Indeed, our data suggest that PGC1α represents a potential therapeutic target for chemoprevention. Obesity and diabetes are independent risk factors for developing liver and colon cancer (49, 50). Importantly, a number of studies show that PGC1α expression is elevated in the livers of obese/diabetic mice and patients. We also observe an increase of PGC1α in the colons of obese mice with type II diabetes (data not shown). Obese and diabetic individuals can be readily identified and therefore suggests that in these identifiable at-risk patients, targeting PGC1α may be a useful cancer prevention strategy. In addition, although our data point toward PGC1α playing a role in the early stages of cancer, our data using established cancer cell lines suggest that the presence of PGC1α is sufficient to promote tumor growth. Therefore, notwithstanding the reduced expression of PGC1 in established tumors, PGC1α may be a therapeutic target in tumors where it is present.

Most studies showing that metabolism is altered in cancer have been done in established cancers. Therefore, the role of metabolism on carcinogenesis is less well defined. These studies provide support for metabolism and its regulation by PGC1α as an important component of tumorigenesis and tumor growth. Importantly, PGC1 appears to accomplish this via inducing the expression of a gene expression program that coordinates the conversion of glucose into fatty acids. In conclusion, these studies suggest that reducing PGC1 expression/activity represents a potential therapeutic approach for targeting multiple aspects of altered cancer metabolism.

No potential conflicts of interest were disclosed.

The authors thank the University of Maryland School of Medicine Nutrition and Obesity Research Core for technical assistance.

This work was supported by the University of Maryland Greenebaum Cancer Center CRF, Susan G. Komen Breast Cancer Foundation (KG081400), and NIDDK (DK064685) to Geoffrey D. Girnun.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.