Downregulation of pyruvate dehydrogenase (PDH) is critical for the aberrant preferential activation of glycolysis in cancer cells under normoxic conditions. Phosphorylation-dependent inhibition of PDH is a relevant event in this process, but it is not durable as it relies on PDH kinases that are activated ordinarily under hypoxic conditions. Thus, it remains unclear how PDH is durably downregulated in cancer cells that are not hypoxic. Building on evidence that PDH activity depends on the stability of a multi-protein PDH complex, we found that the PDH-E1β subunit of the PDH complex is downregulated to inhibit PDH activity under conditions of prolonged hypoxia. After restoration of normoxic conditions, reduced expression of PDH-E1β was sustained such that glycolysis remained highly activated. Notably, PDH-E1β silencing in cancer cells produced a metabolic state strongly resembling the Warburg effect, but inhibited tumor growth. Conversely, enforced exogenous expression of PDH-E1β durably increased PDH activity and promoted the malignant growth of breast cancer cells in vivo. Taken together, our results establish the specific mechanism through which PDH acts as an oncogenic factor by tuning glycolytic metabolism in cancer cells.

Significance: This seminal study offers a mechanistic explanation for why glycolysis is aberrantly activated in normoxic cancer cells, offering insights into this long-standing hallmark of cancer termed the Warburg effect. Cancer Res; 78(7); 1592–603. ©2018 AACR.

Cells use atmospheric oxygen to efficiently produce energy. When cells confront oxygen-limited environments, they adapt by triggering the hypoxic response, which alters cellular functions including metabolism, respiration, vascularization, and erythropoiesis (1). Hypoxia-inducible factor (HIF) plays a critical role in this process by inducing multiple genes involved in the acute response to these conditions (2). We previously showed that CREB and NFκB become activated and play important physiologic roles during prolonged hypoxia (3). Although the acute hypoxic response induces drastic changes in cellular status, the chronic response may serve to maintain the resultant status over the course of long-term hypoxia.

Many types of cancer cells exhibit a specific type of irregular metabolism characterized by high dependence on glycolysis for energy. This “Warburg effect” defined as aerobic glycolysis, is characterized by cancer phenotypes such as high glycolytic rate and elevated lactate production under normoxia (4). Hypoxia also shifts cellular metabolism into a glycolytic mode and increases lactate production. Because tumor cells are often exposed to hypoxia under physiologic conditions, it is possible that their sustained hypoxic metabolism is a direct cause of the Warburg effect; however, the underlying mechanism remains incompletely understood.

Previously, we showed that PHD3 forms a large complex in response to hypoxia (5). Pyruvate dehydrogenase (PDH), an enzyme that catalyzes conversion of pyruvate into acetyl-CoA, is a component of this complex (6). PDH consists of five major subunits, PDH-E1α, PDH-E1β, PDH-E2, PDH-E3, and PDH-E3BP, which in mammals form a complex larger than 5,000 kDa (7). Interaction of PHD3 with PDH stabilizes the PDH complex and plays a role in maintaining its enzymatic activity (6). Under hypoxic conditions, depletion of PHD3 destabilizes the complex, leading to a reduction in PDH activity. Consistent with this, PHD3−/− cells contain reduced levels of PDH complex and have diminished PDH activity.

Inhibition of PDH activity is mediated by phosphorylation of PDH-E1α by pyruvate dehydrogenase kinase 1 (PDK1), a kinase induced under hypoxic conditions (8, 9). This mechanism is important for the inhibition of PDH under hypoxia, leading to primarily glycolytic cellular metabolism, and is also considered critical for induction of the Warburg effect in cancer cells (10). However, PDH is phosphorylated primarily under hypoxic conditions, when expression of PDK is high, and becomes dephosphorylated upon reoxygenation, thus shifting cellular metabolism back to the normal mode. Because the Warburg effect is defined as aerobic glycolysis, it is possible that mechanisms other than phosphorylation are responsible for PDH inhibition under normoxia. Here, we describe a previously unknown mechanism of PDH inhibition, mediated by downregulation of the PDH-E1β subunit.

Cell culture

MCF7, MDA-MB-231, and SKBR3 breast cancer cells, and HeLa cells were obtained from ATCC and cultured in DMEM (high-glucose; Wako) containing 10% FBS and antibiotics. Mouse embryonic fibroblasts were maintained in DMEM supplemented with 10% FBS, 0.1 mmol/L nonessential amino acids, 0.2 mmol/L 2-mercaptoethanol, and antibiotics. Mycoplasma contamination was monitored with PCR or DAPI staining. Cell reauthentication test was not performed. All the cell lines were obtained between 2007 and 2012, and a frozen vial of each cell line was thawed every 2–3 months (or approximately 25 passages.)

Hypoxic treatment

Cells were treated under hypoxic conditions (1% O2 and 5% CO2, balanced with N2) in a hypoxia workstation (Hirasawa Works). An oxygen sensor was used to regulate the oxygen concentration inside the workstation, which was maintained at 1% throughout the experiment (MC-8G-S, Iijima Electrics).

Reagents and antibodies

The following antibodies were used: anti-β-actin (Sigma Japan); anti-HIF-1α (Novus Biologicals); anti-HIF-2α (Novus Biologicals); anti-CREB (Cell Signaling Technology Japan); anti-phosphoS133-CREB (Cell Signaling Technology); anti-PDH-E1α (Abcam); anti-PDH-E1β (Abcam); anti-PDH-E2 (Abcam), anti-PDH-E3 (Bio-Rad), anti-E3BP (Santa Cruz Biotechnology), anti-PINK1 (Novus Biologicals), anti-cytochrome C (Cell Signaling Technology), anti-Tom20 (BD Biosciences), anti-Tim23 (BD Biosciences), anti-β-tubulin (Wako), and anti-GAPDH (Wako). PDH complex purification beads were purchased from Abcam. MG132 and chloroquine were purchased from Sigma.

Trypan blue exclusion assay

Cells were stained with 0.5% Trypan blue solution after normoxia, hypoxia, or reoxygenation treatment. Stained and unstained cells were counted using hemocytometer.

Mitochondrial imaging with fluorescence microscopy

Cells were stained with Mitotracker (Thermo Fisher Scientific) for 30 minutes after normoxic or hypoxic treatment for 72 hours, and fixed with 4% paraformaldehyde in PBS. Samples were mounted with Vectashield containing DAPI (Vector Laboratories), and imaged with LSM510 META confocal microscope (Zeiss Japan).

FACS analysis

For cell-cycle analysis, cells were cultured under normoxic, hypoxic (72 hours), or reoxygenated (6 hours) conditions, and fixed with cold 70% ethanol. After washing with PBS twice, cells were treated with RNase at 37°C for 1 hour followed by staining with propidium iodide, and analyzed by FACSCalibur. For the apoptosis assay, cells were incubated with Annexin-V FITC (NacalaiTesque) for 15 minutes on ice after the treatment, and analyzed by FACSCalibur.

Western blotting

Cells were harvested on ice and lysed in lysis buffer [50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 1 mmol/L EDTA, 1% Triton X-100, 0.1 μg/mL PMSF, and 2 μg/mL leupeptin). Total cell lysates were subjected to SDS-PAGE (50 μg of total lysate/lane), and then transferred to nitrocellulose membranes (PALL).

qPCR

Total RNA was isolated from cultured cells using an RNeasy kit (Qiagen). First-strand cDNA synthesis was performed with 2 μg of total RNA using the PrimeScript II kit (Takara Bio). Synthesized cDNA was used for PCR analysis using SsoFast EvaGreen Supermix (Bio-Rad) on a CFX96 real-time PCR detecting system (Bio-Rad). Relative expression levels were calculated using the ΔCt method and normalized against β-actin or CPH internal control. Primer sequences are listed in Supplementary Table S1.

siRNA

Cells were transfected with negative control siRNA (#45-2001, Invitrogen), HIF-1α siRNA (Invitrogen), HIF-2α siRNA (Invitrogen), LONP siRNA (Sigma), CLPP siRNA (Sigma), or AFG3L2 siRNA (Sigma) using Lipofectamine RNAiMAX transfection reagent (Invitrogen). The day after transfection, cells were cultured in normoxic or hypoxic conditions for 24–48 hours, and then processed to extract total RNA or protein.

Colony assay

Assay plates (6-well plate) were prepared by pouring DMEM/10% FBS containing 0.5% agar. Single-cell suspensions were homogeneously mixed with DMEM/10% FBS containing 0.3% agar and layered over on the bottom agar. MCF7 cells and MB231 cells were grown for 2 weeks and 4 weeks, respectively, to form colonies. Colonies were counted in three different fields.

Mitochondrial purification

Cells were suspended with mitochondrial isolation buffer [10 mmol/L HEPES (pH 7.4), 200 mmol/L d-Mannitol, 70 mmol/L Sucrose, 1 mmol/L EDTA] and disrupted with a dounce homogenizer. After centrifugation of postnuclear supernatant, isolated mitochondria and cytosolic fraction were subjected to Western blot analysis.

Metabolome analysis

MB231 cells were treated under normoxic or hypoxic (1% O2) conditions for 24 hours and then subjected to metabolome analysis. The analysis was conducted using capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS) for cation analysis and CE-tandem mass spectrometry (CE-MS/MS) for anion analysis as described previously (11, 12). Briefly, CE-TOFMS analysis was carried out using an Agilent CE capillary electrophoresis system equipped with an Agilent 6210 time-of-flight mass spectrometer (Agilent Technologies). The spectrometer scanned from m/z 50 to 1,000 (11), and peaks were extracted using the MasterHands automatic integration software (Keio University; ref. 13) and MassHunter Quantitative Analysis B.04.00 (Agilent Technologies). Principal component analysis (PCA) was performed using SampleStat.

Measurement of the oxygen consumption rate

The oxygen consumption rates (OCR) were measured using an XF24 Extracellular Flux Analyzer (Seahorse Biosciences) placed in the InvivO2 hypoxic chamber (Ruskinn Technology). MB231 cells or MCF7 cells were seeded in XF24 V7 Cell Culture microplate at a density of 4 × 104 cells/well. For hypoxic and reoxygenation experiments, medium was first changed to the assay medium, and then the OCR was measured for 12 hours [2-hour normoxia, then 8-hour hypoxia (1% O2), and then 2-hour reoxygenation]. For prolonged hypoxic experiments, cells were initially treated with hypoxia (1% O2) in growing medium for 42 hours. Then, the medium was changed to the assay medium pre-equilibrated to hypoxia (note that this process transiently exposes cells to normoxic condition), and the OCR was measured for 6 hours (until 48-hour hypoxia). The timings of hypoxic treatment are depicted in the figure. The assay medium comprised 25 mmol/L glucose, 143 mmol/L NaCl, 0.8 mmol/L MgSO4, 1.8 mmol/L CaCl2, 5.4 mmol/L KCl, 0.9 mmol/L NaH2PO4 (pH 7.4) and 1× Glutamax Supplement (Thermo Fisher Scientific).

Lactate assay

MCF7 or MB231 cells were cultured under normoxic or hypoxic conditions, and intracellular lactate levels or lactate levels in the medium were measured using a Lactate Assay Kit (Biovision).

Cell proliferation analysis

Growth of control, PDH-E1α, or PDH-E1β-KD MCF7 cells was monitored in medium containing different concentrations of glucose. Cells were cultured in DMEM [with two different glucose concentrations, low (1.0 g/L) or high (4.5 g/L), Wako] supplemented with 10% FBS. Cells were cultured for up to 96 hours, and cell numbers were counted every 24 hours using a hemocytometer.

Glucose assay

Cells were grown in low glucose DMEM supplemented with 10% FBS. Culture media were collected, and glucose concentrations were measured by Glucose Assay Kit (Abcam).

PDH assay

PDH activity was measured using the PDH enzyme activity dipstick assay kit (Abcam). Briefly, active PDH complex was captured from the cell lysate using an anti-PDH antibody, and its activity was measured in a colorimetric assay (6).

Cancer genomic data analysis

A cancer genomic datasets analysis tool, cBioPortal (14, 15), was used to assess amplification of PDH genes from different human cancer genomic datasets. Briefly, five genes encoding the PDH complex (PDHB, PDHA1, DLAT, DLD, and PDHX) were searched for in the entire 167 cancer database available in the tool. Datasets showing amplification of PDH genes were sorted, and cases showing PDHB amplification plus amplification of any of the five PDH genes were counted, and expressed as a percentage of total case. The analysis was performed using the database dated Oct 28, 2017.

Tumor formation assay in nude mice

Cells were suspended at 1 × 107 cells/mL in a 1:1 mixture of medium and Matrigel (Corning), and 100 μL of cell suspension was injected into the mammary fat pad of a BALB/cAJcl-nu/nu mouse (4-week-old female, CLEA). Tumor size was measured weekly, and mice were sacrificed 5–10 weeks after injection to collect tumors. Protein and RNA were extracted from tumors and used for subsequent analyses. All animal experiments were performed in accordance with a protocol approved by the Tokyo Medical and Dental University Animal Care Committee.

Statistical analysis

All experiments were performed at least three times, and the mean and standard deviation (SD) are shown for each experiment. Data were analyzed using Student t test, and P values smaller than 0.05 were considered significant.

Hypoxia treatment decreases tumor-forming ability of cancer cells

As hypoxia promotes malignant transformation of tumors, we tested whether hypoxic pretreatment is sufficient to enhance tumor growth. Breast tumor cell lines, MB231 and MCF7, were exposed to normoxia or hypoxia for 72 hours, and then inoculated into nude mice to generate xenografts. Pretreated cells showed normal cellular responses to prolonged hypoxia with CREB phosphorylation and HIF-2α expression (Fig. 1A). Cells were viable (Fig. 1B; Supplementary Fig. S1A and S1B), and exhibited a normal cell-cycle profile without any signs of cell death in normoxia, hypoxia, and upon reoxygenation (Fig. 1C). The colony-forming ability of hypoxic cells in soft agar was lower than that of normoxic cells; this effect was maintained after reoxygenation (Fig. 1D). Although both cell lines formed tumors in recipient mice, pretreatment with hypoxia reduced the tumor-forming ability of both cell lines (Fig. 1E and F).

Figure 1.

Pretreatment with hypoxia decreases the ability of cancer cells to form tumors. A, Expression of HIF-2α and pCREB under prolonged hypoxia. MB231 cells were pretreated with normoxia or hypoxia for 72 hours, and then proteins were extracted and subjected to Western blotting. Blotting was performed with antibodies against phospho-CREB, CREB, HIF-2α, and β-actin. B, Viability of cells exposed to prolonged hypoxia. MCF7 cells or MB231 cells were treated with normoxia or hypoxia for 72 hours, or reoxygenated for 6 hours after exposure to hypoxic conditions. Treated cells were subjected to a Trypan blue exclusion assay or stained with DAPI. The percentages of stained cells and fragmented nuclei are shown. ± indicates the SD. C, Cell-cycle distribution of MCF7 and MB231 cells treated with prolonged hypoxia. Normoxic, hypoxic (72 hours), or reoxygenated (6 hours) cells were subjected to cell-cycle analysis by flow cytometry. D, Pretreated cells were collected and subjected to a colony formation assay in soft agar. The number of colonies formed was counted and is shown on the graph. E and F, MB231 cells (E) or MCF7 cells (F) pretreated with normoxia (norm) or hypoxia (hypo) for 72 hours were inoculated into the mammary fat pad of female nude mice (n = 8). Images of tumors and tumor volumes (mm3) are shown on the graph. *, P < 0.05; **, P < 0.02.

Figure 1.

Pretreatment with hypoxia decreases the ability of cancer cells to form tumors. A, Expression of HIF-2α and pCREB under prolonged hypoxia. MB231 cells were pretreated with normoxia or hypoxia for 72 hours, and then proteins were extracted and subjected to Western blotting. Blotting was performed with antibodies against phospho-CREB, CREB, HIF-2α, and β-actin. B, Viability of cells exposed to prolonged hypoxia. MCF7 cells or MB231 cells were treated with normoxia or hypoxia for 72 hours, or reoxygenated for 6 hours after exposure to hypoxic conditions. Treated cells were subjected to a Trypan blue exclusion assay or stained with DAPI. The percentages of stained cells and fragmented nuclei are shown. ± indicates the SD. C, Cell-cycle distribution of MCF7 and MB231 cells treated with prolonged hypoxia. Normoxic, hypoxic (72 hours), or reoxygenated (6 hours) cells were subjected to cell-cycle analysis by flow cytometry. D, Pretreated cells were collected and subjected to a colony formation assay in soft agar. The number of colonies formed was counted and is shown on the graph. E and F, MB231 cells (E) or MCF7 cells (F) pretreated with normoxia (norm) or hypoxia (hypo) for 72 hours were inoculated into the mammary fat pad of female nude mice (n = 8). Images of tumors and tumor volumes (mm3) are shown on the graph. *, P < 0.05; **, P < 0.02.

Close modal

PDH-E1β expression is reduced in prolonged hypoxia

Hypoxia is typically associated with alterations in cellular metabolism. As pretreatment of breast cancer cells with hypoxia (72 hours) reduced their ability to form tumors, we searched for a possible cause by focusing on metabolic enzymes. Prolonged exposure to hypoxia for up to 72 hours resulted in a decrease in PDH-E1β protein expression in MCF7 and MB231 cells (Fig. 2A), as well as in breast cancer SKBR3 cells, HeLa cells, and mouse embryonic fibroblasts (Supplementary Fig. S2A–S2C). These results indicate that this phenomenon is a general cellular response that is conserved across multiple cell types. It is mainly PDH-E1β that was affected by hypoxia, and PDH-E1α, PDH-E3, and PDH-E3BP were also reduced to some extent, whereas PDH-E2 was not affected (Fig. 2A). Hypoxic treatment induced mitochondrial fission and mitophagy in some of MCF7 cells; however, mitochondrial structure and content in both cell lines were mostly intact after 72 hours of hypoxic treatment (Fig. 2B and C). These results suggest that reduced expression of PDH-E1β is an active process that takes place in the mitochondria. Importantly, the reduction in the PDH-E1β expression was maintained even after reoxygenation for 24 hours, when both PDH-E1α phosphorylation and HIF-2α expression decreased (Fig. 2D). Accordingly, we observed a decrease in PDH enzymatic activity when cells were exposed to hypoxia for 48 hours, that was maintained after 24 hours of reoxygenation (Fig. 2E). Furthermore, PDH activity remained inactive after reoxygenation for up to 48 hours when the cells were pretreated with hypoxia for 72 hours (Fig. 2E). These results indicate that downregulation of PDH-E1β plays a role in maintaining glycolytic metabolism upon reoxygenation, and that the effect is sustained for longer periods when the cells are pretreated with hypoxia for longer times.

Figure 2.

Expression of PDH-E1β decreases under prolonged hypoxia. A, MCF7 or MB231 cells were treated with hypoxia for the indicated time and then subjected to Western blotting. Blotting was performed with antibodies against PDH-E1β, PDH-E1α, PDH-E2, PDH-E3, PDH-E3BP, and β-actin. p-PDH-E1α, phosphorylated PDH-E1α. B, MCF7 or MB231 cells were treated with normoxia or hypoxia for 72 hours and then stained with MitoTracker. Representative images of stained mitochondria under each condition are shown. Scale bar, 20 μm. C, MCF7 or MB231 cells treated with normoxic or hypoxic conditions for 72 hours were harvested and subjected to Western blotting. Blotting was performed with antibodies against Tom20 (mitochondrial outer membrane), cytochrome C (intermembrane space), Tim23 (inner membrane), PDH-E1β (matrix), PINK1, and β-tubulin. N, normoxia; H, hypoxia. D, MCF7 or MB231 cells were treated with hypoxia for the indicated times, followed by 24 hours of reoxygenation (re). Cells were then subjected to Western blotting with antibodies specific for PDH-E1β, PDH-E1α, PDH-E2, HIF-1α, HIF-2α, and β-actin. p-PDH-E1α, phosphorylated PDH-E1α. E, MCF7 cells were treated with hypoxia, followed by reoxygenation (re) for the indicated times and then harvested. PDH activity was then measured. Representative image of dipsticks used for the PDH assay (top) and relative PDH activity (bottom) are shown. *, P < 0.05 and **, P < 0.02.

Figure 2.

Expression of PDH-E1β decreases under prolonged hypoxia. A, MCF7 or MB231 cells were treated with hypoxia for the indicated time and then subjected to Western blotting. Blotting was performed with antibodies against PDH-E1β, PDH-E1α, PDH-E2, PDH-E3, PDH-E3BP, and β-actin. p-PDH-E1α, phosphorylated PDH-E1α. B, MCF7 or MB231 cells were treated with normoxia or hypoxia for 72 hours and then stained with MitoTracker. Representative images of stained mitochondria under each condition are shown. Scale bar, 20 μm. C, MCF7 or MB231 cells treated with normoxic or hypoxic conditions for 72 hours were harvested and subjected to Western blotting. Blotting was performed with antibodies against Tom20 (mitochondrial outer membrane), cytochrome C (intermembrane space), Tim23 (inner membrane), PDH-E1β (matrix), PINK1, and β-tubulin. N, normoxia; H, hypoxia. D, MCF7 or MB231 cells were treated with hypoxia for the indicated times, followed by 24 hours of reoxygenation (re). Cells were then subjected to Western blotting with antibodies specific for PDH-E1β, PDH-E1α, PDH-E2, HIF-1α, HIF-2α, and β-actin. p-PDH-E1α, phosphorylated PDH-E1α. E, MCF7 cells were treated with hypoxia, followed by reoxygenation (re) for the indicated times and then harvested. PDH activity was then measured. Representative image of dipsticks used for the PDH assay (top) and relative PDH activity (bottom) are shown. *, P < 0.05 and **, P < 0.02.

Close modal

HIF-1 regulates PDH-E1β expression in hypoxia

Next, we investigated the mechanism underlying PDH-E1β downregulation under prolonged hypoxia. qPCR analysis of hypoxic cultured cells revealed that PDH-E1β expression was reduced, whereas PDH-E1α expression was largely unaffected, at the mRNA level (Fig. 3A). Because HIF is a key regulator of the hypoxic response, we next investigated whether HIF is involved in this process. Knockdown (KD) of HIF-1α increased PDH-E1β expression in normoxia and hypoxia (Fig. 3B, lanes 1 and 2; 4 and 5; Supplementary Fig. S3). Furthermore, knockdown of HIF-1α induced the expression of PDH-E1β mRNA to some extent, whereas a typical HIF-1 target gene GLUT1 was efficiently downregulated (Fig. 3C). Moreover, HIF-1α KD induced PDH activity in normoxic MCF7 cells when its activity is not inhibited by phosphorylation (Fig. 3D). A similar effect was observed in HeLa and MB231 cells (Supplementary Figs. S4A–S4D and S5A–S5D). However, increase of PDH-E1β mRNA by the depletion of HIF-1 was observed only under hypoxic conditions. These results indicate that PDH-E1β is commonly regulated at the level of protein in a HIF-1–dependent manner both under normoxic and hypoxic conditions, whereas a regulation at the level of mRNA is limited to hypoxic conditions. Knockdown of HIF-2α also moderately increased PDH-E1β expression in hypoxic MCF7 and MB231 cells, which is likely due to functional overlap with HIF-1α (Fig. 3B; Supplementary Fig. S5A). However, this effect was not seen in HeLa cells. Taken together, these results suggest that HIF-1 is the main factor responsible for downregulating PDH-E1β, resulting in a reduction in its enzymatic activity in MCF7, MB231, and HeLa cells. We also examined the induction of transcriptional repressors under hypoxic conditions. Transcriptional repressors, DEC1, DEC2, and Cited2 were induced markedly in MCF7 and HeLa cells, whereas DEC1 and Cited2 were induced in MB231 cells (Supplementary Fig. S6A–S6C). Thus, these factors could also be involved in downregulation of PDH-E1β at the mRNA level during prolonged hypoxia.

Figure 3.

Downregulation of PDH-E1β occurs in a HIF-1–dependent manner. A, MCF7 or MB231 cells were treated with hypoxia for the indicated times, and expression of PDH-E1α and PDH -E1β was monitored by qPCR (n = 3). B, MCF7 cells were transfected with control, HIF-1α, or HIF-2α-targeting siRNA, and treated with normoxia or hypoxia for 48 hours. Harvested cells were subjected to Western blotting. Blotting was performed with antibodies against PDH-E1β, PDH-E1α, HIF-1α, HIF-2α, and β-actin. #, a nonspecific band detected by the antibody. Bands in the PDH-E1β blot were quantified, and their relative intensities are represented by the numbers below. C and D, MCF7 cells transfected with control, HIF-1α, or HIF-2α siRNA were treated with normoxia or hypoxia for 48 hours. Expression of PDH-E1β and GLUT1 was monitored by qPCR (C), and PDH activity was measured by PDH assay (D). E, MCF7 cells were transfected with siRNA targeting three mitochondria matrix proteases (LONP, CLPP, and AFG3L2) and treated with normoxia or hypoxia for 48 hours. Harvested cells were subjected to Western blotting with antibodies against PDH-E1β, PDH-E1α, and β-actin. Bands in the PDH-E1β blot were quantified, and their relative intensities are represented by the numbers below. *, P < 0.05; **, P < 0.02.

Figure 3.

Downregulation of PDH-E1β occurs in a HIF-1–dependent manner. A, MCF7 or MB231 cells were treated with hypoxia for the indicated times, and expression of PDH-E1α and PDH -E1β was monitored by qPCR (n = 3). B, MCF7 cells were transfected with control, HIF-1α, or HIF-2α-targeting siRNA, and treated with normoxia or hypoxia for 48 hours. Harvested cells were subjected to Western blotting. Blotting was performed with antibodies against PDH-E1β, PDH-E1α, HIF-1α, HIF-2α, and β-actin. #, a nonspecific band detected by the antibody. Bands in the PDH-E1β blot were quantified, and their relative intensities are represented by the numbers below. C and D, MCF7 cells transfected with control, HIF-1α, or HIF-2α siRNA were treated with normoxia or hypoxia for 48 hours. Expression of PDH-E1β and GLUT1 was monitored by qPCR (C), and PDH activity was measured by PDH assay (D). E, MCF7 cells were transfected with siRNA targeting three mitochondria matrix proteases (LONP, CLPP, and AFG3L2) and treated with normoxia or hypoxia for 48 hours. Harvested cells were subjected to Western blotting with antibodies against PDH-E1β, PDH-E1α, and β-actin. Bands in the PDH-E1β blot were quantified, and their relative intensities are represented by the numbers below. *, P < 0.05; **, P < 0.02.

Close modal

We next examined the molecular mechanism by which PDH-E1β expression is reduced in hypoxic cells. Although mRNA regulation was involved as demonstrated above, regulation at the protein level was also indicated by the observation that PDH-E1β was almost undetectable in PDH-E1α-KD cells, even though the PDH-E1β mRNA level was normal (Supplementary Fig. S7A–S7C). To investigate the cause of the depletion, we tested inhibitors of the proteasome and lysosome, the two major proteolytic pathways in the cell. Neither class of compound inhibited the decrease in PDH-E1β level (Supplementary Fig. S8A and S8B). Because PDH proteins localize in the mitochondrial matrix, and LONP is induced in hypoxia and controls energy metabolism in mitochondria (16), we next focused on three mitochondrial matrix proteases: LONP, CLPP, and i-AAA (AFG3L2). Depletion of LONP, CLPP, or AFG3L2 restored the PDH-E1β expression, which was reduced in hypoxic-treated cells, suggesting that these proteases are involved in degrading PDH-E1β in MCF7, MB231, and HeLa cells (Fig. 3E, lanes 5–8; Supplementary Figs. S4D, S5D, and S9).

Depletion of PDH-E1α/E1β diminishes PDH activity and increases lactate production

Because PDH-E1β expression decreases while PDH activity is repressed under prolonged hypoxia, we mimicked this situation by depleting PDH-E1β expression with shRNA. For this purpose, MCF7 and MB231 cells were infected with retrovirus carrying shRNA targeting control, PDH-E1α or PDH-E1β (#1 and #2). Knockdown of both genes was efficient (Fig. 4A). Importantly, knockdown of PDH-E1β decreased expression of PDH-E1α, and vice versa; however, downregulation at the mRNA level was specific for the shRNA target, and did not affect the other subunit (Supplementary Fig. S7A), indicating that this coregulatory effect occurred at the protein level. The effect was not inhibited by proteasome or lysosome inhibitors (Supplementary Fig. S7B and S7C). These results indicate that a coregulatory machinery, distinct from major degradative pathways, synchronizes the expression of PDH-E1α and PDH-E1β in cells. Notably, expression of other PDH subunits (PDH-E2, PDH-E3, and PDH-E3BP) was not altered in PDH-E1α- or PDH-E1β-KD cells (Fig. 4A). PDH activity was almost completely inhibited in both types of shRNA-expressing cells (Fig. 4B).

Figure 4.

PDH-E1β knockdown promotes lactate production and glycolytic metabolism. A, MB231 cells stably expressing shRNA against PDH-E1α or PDH-E1β (E1α-KD, E1β-KD) were established. Inhibition of PDH-E1α or PDH-E1β was examined by Western blotting with antibodies against PDH-E1β and PDH-E1α. PDH-E2, PDH-E3, PDH-E3BP, and β-actin blots are also shown. B, PDH activities in E1α-KD and E1β-KD MB231 cells were examined by PDH assay. C, Lactate production in E1α-KD and E1β-KD MB231 cells. Images of culture medium from control, E1α-KD, and E1β-KD cells are shown (top). Intracellular lactate concentrations were measured in a lactate assay (n = 3). D, The metabolic status of control, E1α-KD, and E1β-KD MB231 cells was analyzed using a flux analyzer. The oxygen consumption rate (OCR) and lactate production rate were measured in two experimental settings (i) hypoxia and reoxygenation and (ii) prolonged hypoxia as in the scheme. The ratio of the OCR to the lactate production rate was plotted (n = 6). The OCR and the lactate production (lactate levels in the medium) were measured at the time points indicated by arrows. Arrowhead, the point of medium change to the assay medium in ii. *, P < 0.05; **, P < 0.02.

Figure 4.

PDH-E1β knockdown promotes lactate production and glycolytic metabolism. A, MB231 cells stably expressing shRNA against PDH-E1α or PDH-E1β (E1α-KD, E1β-KD) were established. Inhibition of PDH-E1α or PDH-E1β was examined by Western blotting with antibodies against PDH-E1β and PDH-E1α. PDH-E2, PDH-E3, PDH-E3BP, and β-actin blots are also shown. B, PDH activities in E1α-KD and E1β-KD MB231 cells were examined by PDH assay. C, Lactate production in E1α-KD and E1β-KD MB231 cells. Images of culture medium from control, E1α-KD, and E1β-KD cells are shown (top). Intracellular lactate concentrations were measured in a lactate assay (n = 3). D, The metabolic status of control, E1α-KD, and E1β-KD MB231 cells was analyzed using a flux analyzer. The oxygen consumption rate (OCR) and lactate production rate were measured in two experimental settings (i) hypoxia and reoxygenation and (ii) prolonged hypoxia as in the scheme. The ratio of the OCR to the lactate production rate was plotted (n = 6). The OCR and the lactate production (lactate levels in the medium) were measured at the time points indicated by arrows. Arrowhead, the point of medium change to the assay medium in ii. *, P < 0.05; **, P < 0.02.

Close modal

PDH catalyzes the conversion of pyruvate to acetyl-CoA, thus connecting glycolysis to the TCA cycle. Inhibition of PDH activity should suppress this conversion, resulting in accumulation of pyruvate, leading in turn to the conversion of pyruvate into lactate by lactate dehydrogenase. The culture medium of PDH-E1α-KD and PDH-E1β-KD cells rapidly became more acidic (yellow in Phenol red–containing medium) than that of control cells plated at a similar density, reflecting elevated production of lactate in the knockdown cells (Fig. 4C). Measurement of intracellular lactate levels in these cells confirmed that PDH-KD cells contained higher lactate concentrations (Fig. 4C).

Inhibition of PDH expression induces glycolytic metabolism

High lactate production is characteristic of cancer cells exhibiting the Warburg effect. Accordingly, we tested the metabolic state of PDH-E1α-KD and PDH-E1β-KD MB231 cells by measuring their OCR using a flux analyzer and lactate concentration in the medium using a Lactate Assay Kit. The ratio of the OCR to lactate production rate under normoxic conditions was around 1300 pmol/min/mmol/L in control cells, but only one-half of that value in PDH-KD cells (Fig. 4D), suggesting that PDH-KD cells are in a metabolic state resembling the Warburg effect (Fig. 4D; Supplementary Fig. S10A and S10B). Exposure to hypoxia for 8 hours reduced the OCR and increased the lactate production rate in both control and PDH-KD cells, resulting in comparable metabolic states (Fig. 4D). Reoxygenation for 2 hours did not change the OCR/lactate production ratio (Fig. 4D). Furthermore, prolonged exposure to hypoxia for 48 hours also resulted in OCR/lactate production ratios that were comparable among these three cell types, indicating that they all utilize a similar glycolytic metabolism under prolonged hypoxic conditions. Similar changes in the OCR/lactate production ratio were observed in another breast cancer cell line, MCF7, when examined under the same hypoxic conditions (Supplementary Figs. S10B and S11). Reoxygenation for 2 hours moderately recovered the OCR in MCF7 cells, but not in MB231 cells, suggesting that some additional time is required to fully recover the OCR once cells are exposed to hypoxia extensively (Supplementary Fig. S10A and S10B). To further characterize the metabolic state of PDH-KD cells, we performed metabolome analysis to compare control and PDH-E1β-KD MB231 cells cultured under normoxic or hypoxic conditions for 24 hours. As PDH catalyzes conversion of pyruvate to acetyl-CoA, PDH-E1β-KD cells accumulated pyruvate, contained less acetyl-CoA, and more lactate than control cells (Fig. 5A). Similar changes were observed in hypoxic treated cells, although the levels of acetyl-CoA and lactate became higher in both control and PDH-KD cells. Higher lactate production is an indication that these cells use anaerobic metabolism in response to hypoxia. Higher acetyl-CoA level may indicate an active conversion of acetate into acetyl-CoA in these cancer cells (17). Furthermore, PDH-E1β-KD cells showed elevated glycolysis under normoxic conditions, with higher levels of glycolytic metabolites compared with control cells such as fructose-1,6-bisphosphate (F1,6P), 3-phosphoglycerate (3-PG), 2-phosphoglycerate (2-PG), and phosphoenolpyruvate (PEP), another indication of Warburg effect-like metabolism (Fig. 5B). The difference between control and the KD cells was no longer detectable under hypoxic conditions since glycolytic metabolism is induced both in control and KD cells. Furthermore, the levels of citric acid and cis-aconitic acid generated by the TCA cycle were reduced markedly in PDH-E1β-KD cells, although parts of the cycle beyond 2-oxoglutarate (2-OG) remained mostly normal under normoxic conditions (Fig. 5C). Under hypoxic conditions, levels of metabolites formed beyond 2-OG in the TCA cycle were significantly higher in KD cells than in control cells, which might implicate activation of a glutamine synthesis pathway that would compensate for metabolites beyond 2-OG. All of the 116 metabolites analyzed are shown in Supplementary Table S2. Principal component (PC) analysis was also performed on this data set to visualize differences between control and PDH-E1β-KD cells (Fig. 5D). The first PC distinguished between normoxia- and hypoxia-treated samples, whereas the second PC clearly separated PDH-E1β-KD cells from control cells. Glutathione (GSH), pyruvate, malic acid, and fumaric acid were the top four contributors to the second PC.

Figure 5.

Enhanced glycolysis and reduced TCA cycle activation in PDH-E1β-KD cells. A–C, Metabolome analysis comparing control and E1β-KD MB231 cells. Cells were treated with normoxia or hypoxia (1% O2) for 24 hours, and the concentrations of metabolites representative of pyruvate metabolism (A), glycolysis (B), and the TCA cycle (C) were measured (n = 3). *, P < 0.05 and **, P < 0.02. D, Results of principal component analysis, shown as a plot of the first principal component (accounting for 62.9% of total variance) against the second principal component (accounting for 18.7% of total variance).

Figure 5.

Enhanced glycolysis and reduced TCA cycle activation in PDH-E1β-KD cells. A–C, Metabolome analysis comparing control and E1β-KD MB231 cells. Cells were treated with normoxia or hypoxia (1% O2) for 24 hours, and the concentrations of metabolites representative of pyruvate metabolism (A), glycolysis (B), and the TCA cycle (C) were measured (n = 3). *, P < 0.05 and **, P < 0.02. D, Results of principal component analysis, shown as a plot of the first principal component (accounting for 62.9% of total variance) against the second principal component (accounting for 18.7% of total variance).

Close modal

The ability to form tumors is reduced in PDH-KD cells

Many cancer cells exhibit glycolytic metabolism, which is often linked to malignant transformation. Since PDH-KD cells also exhibited glycolytic metabolism, we examined their tumor-forming ability in immunodeficient nude mice. Two clones from PDH-E1α shRNA and PDH-E1β shRNA #1, and a clone from control shRNA and PDH-E1β shRNA #2 were examined. Both PDH-E1α-KD and PDH-E1β-KD MCF7 cells formed significantly smaller tumors than the control cells (Fig. 6A and B). Reduced expression of PDH-E1α or PDH-E1β in tumors was confirmed by qPCR and Western blotting (Supplementary Fig. S12A and S12B). Similar reduction of tumor growth was observed in another breast cancer cell line, MB231 (Fig. 6C). The gene expression profile was altered in tumors formed by PDH-E1α or PDH-E1β-KD cells, especially in regard to hypoxic target genes such as VEGF, GLUT1, and MMP1 (Fig. 6D). Because these cells exhibit reduced expression of GLUT1, we assessed their sensitivity to glucose concentration by measuring their growth rates in two different culture media, high- and low-glucose. Although growth of control, PDH-E1α-KD, and PDH-E1β-KD cells did not differ significantly when they were cultured in high-glucose medium, growth was significantly slower in PDH-E1α-KD and PDH-E1β-KD cells in low-glucose medium (Fig. 6E). Glucose was absorbed more rapidly from the culture medium by PDH-KD cells, possibly because these cells depend mostly on glycolysis for energy production; these cells require more glucose to produce the same amount of ATP as control cells during growth (Fig. 6F). However, expression of GLUT1 and 4 was not higher in the KD cells than in control cells, which might imply that the GLUTs expressed in PDH-KD cells are sufficient to uptake glucose at least in this experimental setting (Fig. 6G). Altogether, these results indicate that PDH-E1α-KD and PDH-E1β-KD cells require a larger supply of glucose to grow as rapidly as control cells.

Figure 6.

PDH-KD cells exhibit reduced tumor formation ability. A, MCF7 cells stably expressing shRNA against PDH-E1α or PDH-E1β were inoculated into the mammary fat pad of female nude mice (n = 7). Images of tumors are shown. B, The size of each tumor was monitored weekly, and the volume of tumors (mm3) formed by two clones for shRNA (E1α, E1β#1) and a clone for shRNA (con, E1β#2) were plotted. C, Volume (mm3) of control, E1α-KD, and E1β-KD MB231 tumors (n = 6). D, Tumors from MB231 cells were collected at week 5 and subjected to qPCR analysis. Expression levels of VEGF, GLUT1, and MMP1 are shown. E, Growth rate of control, E1α-KD, and E1β-KD MCF7 cells. Cells were cultured in high-glucose (4.5 g/L; hi) or low-glucose (1 g/L; low) medium for up to 96 hours, and cell numbers were counted at the indicated time points (n = 3). F, Glucose consumption by control, E1α-KD, and E1β-KD MCF7 cells. Cells were cultured for 48 hours in low-glucose medium and the amount of glucose remaining in the medium was measured. G, Expression of glucose transporters in cells. Cells at 0- and 48-hour time-points were collected in F and subjected to qPCR analysis. Expression of GLUT1 and GLUT4 is shown. *, P < 0.05; **, P < 0.02.

Figure 6.

PDH-KD cells exhibit reduced tumor formation ability. A, MCF7 cells stably expressing shRNA against PDH-E1α or PDH-E1β were inoculated into the mammary fat pad of female nude mice (n = 7). Images of tumors are shown. B, The size of each tumor was monitored weekly, and the volume of tumors (mm3) formed by two clones for shRNA (E1α, E1β#1) and a clone for shRNA (con, E1β#2) were plotted. C, Volume (mm3) of control, E1α-KD, and E1β-KD MB231 tumors (n = 6). D, Tumors from MB231 cells were collected at week 5 and subjected to qPCR analysis. Expression levels of VEGF, GLUT1, and MMP1 are shown. E, Growth rate of control, E1α-KD, and E1β-KD MCF7 cells. Cells were cultured in high-glucose (4.5 g/L; hi) or low-glucose (1 g/L; low) medium for up to 96 hours, and cell numbers were counted at the indicated time points (n = 3). F, Glucose consumption by control, E1α-KD, and E1β-KD MCF7 cells. Cells were cultured for 48 hours in low-glucose medium and the amount of glucose remaining in the medium was measured. G, Expression of glucose transporters in cells. Cells at 0- and 48-hour time-points were collected in F and subjected to qPCR analysis. Expression of GLUT1 and GLUT4 is shown. *, P < 0.05; **, P < 0.02.

Close modal

Elevated tumor formation ability in PDH-E1β–expressing cells

Next, we established a MCF7-derived cell line that ectopically expressed PDH-E1β (MCF7+E1β cells). These cells expressed higher levels of PDH-E1β expression under both normoxia and hypoxia (Fig. 7A), possibly because the degradative machinery involving mitochondria proteases is restricted by elevated expression of PDH-E1β under hypoxic conditions. Furthermore, expression of PDH-E1α was higher in these cells (Fig. 7A), which is due in part to the induction of PDH-E1α at the mRNA level (Supplementary Fig. S13). This might be caused by the cells balancing the amount of PDH-E1α and PDH-E1β subunits; the expression of one is associated with that of the other (Fig. 4A). Ectopically expressed PDH-E1β is processed properly, and localized in mitochondria to form the PDH complex (Fig. 7B; Supplementary Fig. S14). MCF7+E1β cells had higher PDH activity than the parental line under normoxia, but the activity sharply dropped when they were cultured under hypoxia (Fig. 7C). However, MCF7+E1β cells sustained high PDH-E1β levels even under hypoxia, and reoxygenation for 24 hours was sufficient to restore PDH activity close to the level in normoxia, in contrast to the control cells, which still exhibited reduced activity (Fig. 7C). When these cells were transplanted into nude mice, MCF7+E1β cells formed significantly larger tumors than control cells (Fig. 7D and E). Furthermore, MCF7+E1β cells pretreated with hypoxia initially formed smaller tumors than their normoxic counterparts, but growth accelerated after 5 weeks, and by 8 weeks the tumors derived from hypoxia-treated cells were as large as those derived from normoxia-treated cells (Fig. 7E). This was in clear contrast to parental MCF7 cells, which exhibited persistently slower tumor growth when pretreated with hypoxia, even though the tumor-bearing animals were maintained for longer period (Fig. 1F; Supplementary Fig. S15A and S15B). Importantly, the hypoxic target genes, GLUT1 and GLUT4, which are barely expressed in parental cells, were induced in tumors derived from MCF7+E1β cells (Fig. 7F), whereas GLUT1 was expressed at low levels in tumors derived from PDH-E1β-KD cells (Fig. 6D).

Figure 7.

MCF7 cells ectopically expressing PDH-E1β exhibit elevated tumor formation ability. A, MCF7 cells stably expressing FLAG-tagged PDH-E1β (MCF7+E1β) were treated with normoxia (norm) or hypoxia (hypo) for 48 hours, harvested, and subjected to Western blotting with antibodies against PDH-E1β, FLAG, PDH-E1α, or β-actin. B, A mitochondrial fraction was prepared from MCF7 and MCF7+E1β cells. Samples were subjected to Western blotting with antibodies specific for FLAG, PDH-E1β, PDH-E1α, Tim23, or β-actin. C, cytoplasmic; M, mitochondria. C, Control or MCF7+E1β cells were exposed to normoxic, hypoxic (48 hours), or hypoxic (48 hours) plus reoxygenation (24 hours) conditions. Cell lysates were prepared and subjected to a PDH assay and Western blot analysis. Representative image of the dipsticks used in the assay and relative PDH activity are shown. Expression of PDH-E1β was detected using anti-PDH-E1β and anti-GAPDH antibodies (bottom right). D and E, Control (norm) and MCF7+E1β cells treated with normoxia (norm+E1β) or hypoxia (hypo+E1β) for 72 hours were inoculated into the mammary fat pad of female nude mice (n = 8). Images of tumors are shown (D). Tumor size (mm3) was monitored weekly and plotted (E). F, Tumors were collected at week 5 and subjected to qPCR analysis. Expression of GLUT1, GLUT4, and PDH-E1β was detected. G, Expression of the five main subunits of the PDH enzyme (PDH-E1α, PDH-E1β, PDH-E2, PDH-E3, and PDH-E3BP combined; gray bar) and PDH-E1β alone (black bar) in samples from human cancer patients. Cases showing amplification of total PDH or PDH-E1β, as a percentage of the total number of specimens from each organ. Ad, adenocarcinoma; NE, neuroendocrine tumor. *, P < 0.05; **, P < 0.02.

Figure 7.

MCF7 cells ectopically expressing PDH-E1β exhibit elevated tumor formation ability. A, MCF7 cells stably expressing FLAG-tagged PDH-E1β (MCF7+E1β) were treated with normoxia (norm) or hypoxia (hypo) for 48 hours, harvested, and subjected to Western blotting with antibodies against PDH-E1β, FLAG, PDH-E1α, or β-actin. B, A mitochondrial fraction was prepared from MCF7 and MCF7+E1β cells. Samples were subjected to Western blotting with antibodies specific for FLAG, PDH-E1β, PDH-E1α, Tim23, or β-actin. C, cytoplasmic; M, mitochondria. C, Control or MCF7+E1β cells were exposed to normoxic, hypoxic (48 hours), or hypoxic (48 hours) plus reoxygenation (24 hours) conditions. Cell lysates were prepared and subjected to a PDH assay and Western blot analysis. Representative image of the dipsticks used in the assay and relative PDH activity are shown. Expression of PDH-E1β was detected using anti-PDH-E1β and anti-GAPDH antibodies (bottom right). D and E, Control (norm) and MCF7+E1β cells treated with normoxia (norm+E1β) or hypoxia (hypo+E1β) for 72 hours were inoculated into the mammary fat pad of female nude mice (n = 8). Images of tumors are shown (D). Tumor size (mm3) was monitored weekly and plotted (E). F, Tumors were collected at week 5 and subjected to qPCR analysis. Expression of GLUT1, GLUT4, and PDH-E1β was detected. G, Expression of the five main subunits of the PDH enzyme (PDH-E1α, PDH-E1β, PDH-E2, PDH-E3, and PDH-E3BP combined; gray bar) and PDH-E1β alone (black bar) in samples from human cancer patients. Cases showing amplification of total PDH or PDH-E1β, as a percentage of the total number of specimens from each organ. Ad, adenocarcinoma; NE, neuroendocrine tumor. *, P < 0.05; **, P < 0.02.

Close modal

Expression of PDH-E1β in human cancers

Finally, we looked for changes in the expression level of PDH in human cancer patient samples using data from cBioPortal (Fig. 7G; Supplementary Table S3; refs. 14, 15). Gene amplification of PDHs was detected in a variety of cancer types, including breast, esophagus-stomach, and lung. Prostate (adenocarcinoma) showed a high rate of amplification, up to 50% (all five PDH subunits combined), and bladder, melanoma, ovarian, and prostate (neuroendocrine) showed higher ratios of PDH-E1β amplification.

One of the major pathways so far reported to regulate the PDH activity is mediated by phosphorylation of the PDH-E1α subunit. Here, we demonstrated that PDH-E1β is downregulated under hypoxic conditions and constitutes another layer of PDH inhibition in cells. Phosphorylation is responsible for quick and timely inhibition so that PDH activity can be altered in a relatively short time, for example, during acute hypoxia. In contrast, during long periods of hypoxia, it may be important to assure inhibition of PDH by downregulating the protein. Thus, cells might utilize two independent systems to securely regulate the activity of this central metabolic enzyme and adapt to both acute and prolonged phases of hypoxia. Importantly, the lower PDH activity in MCF7 cells was maintained even after reoxygenation for up to 48 hours when they were pretreated with hypoxia for 72 hours (Fig. 2E). Therefore, this novel mechanism of PDH regulation could serve as a basis for establishing aerobic glycolysis in cancer cells.

Depletion of HIF-1α increased PDH-E1β expression mainly at the protein level in three different cell lines examined, indicating that the HIF-1 pathway plays a negative role on the PDH-E1β expression. As HIF-1 induces LONP, induction of mitochondrial matrix proteases might be one of the negative regulatory mechanisms (Fig. 3E; Supplementary Figs. S4D and S5D; ref. 16). Of note, as knockdown of mitochondrial matrix proteases only moderately affected PDH-E1β under normoxic conditions, we cannot exclude a possibility that there might be some additional machineries to regulate PDH-E1β at the protein level in normoxia. As PDH-E1β promoter harbors a putative HRE, we used a ChIP assay to examine the possibility that HIF-1 binds to the promoter region of PDH-E1β. Whereas interaction between HIF1 and the GAPDH HRE was clear, there was no interaction with the PDH-E1β promoter (Supplementary Fig. S16).

PDH-E1α and PDH-E1β are tightly associated in cells, and play a role in pyruvate dehydrogenation. Our data indicate that, when one subunit is knocked down, expression of the other also decreases (Fig. 4A). We also showed that PDH-E1α becomes transcriptionally induced when PDH-E1β is ectopically expressed (Supplementary Fig. S13). Thus, cellular machinery might coregulate these two subunits at the mRNA and protein levels, which would strictly and precisely control enzymatic activity. Importantly, PDH-E3 and PDH-E3BP are also downregulated by prolonged hypoxia, but are not altered in PDH-E1α and PDH-E1β-KD cells (Figs. 2A and 4A). These results indicate that PDH-E3 and PDH-E3BP are not simply downregulated when PDH-E1α and PDH-E1β are depleted, but there would rather be an active mechanism to downregulate these proteins in response to hypoxia.

Although the molecular mechanism underlying the Warburg effect is not completely understood, reports show that dysregulation of molecules functioning in glycolysis or the TCA cycle is involved (18). For example, EGFR signaling, which is often activated in cancer cells, induces phosphorylation of pyruvate kinase PKM2. Phosphorylated PKM2 then activates transcription factors in the nucleus to induce GLUT1 and LDH, which in turn play critical roles in establishing the Warburg effect (19). Inhibition of AMPK in cells also induces glycolytic metabolism (20). Patients carrying an AMPK mutation are often associated with a malignant form of cancer.

In this study, we showed that knockdown of a single PDH-E1 subunit is sufficient to cause Warburg effect–type metabolism in breast cancer cell lines (Fig. 4D; Supplementary Fig. S11). However, it did not promote tumor growth of these cells; rather, PDH-depleted cells formed smaller tumors than control cells (Fig. 6A–C). This phenomenon could be in part due to reduced expression of GLUT1, which would limit glucose uptake by PDH-KD tumors (Fig. 6D). In contrast, GLUT1 and GLUT4 were highly induced in tumors derived from cells ectopically expressing PDH-E1β (Fig. 7F). Upregulation of GLUTs might be critical for cancer cells that reside in a microenvironment with a limited glucose supply.

The current understanding is that the Warburg effect creates a favorable metabolic state for tumor progression (21). However, our results indicate that sustained Warburg effect-type metabolism rather prevents cancer cells from growing. Dysregulation of mitochondria metabolism and lack of oxidative phosphorylation reduces tumorigenicity in K-ras–driven lung cancer and breast cancer cells, respectively (22, 23). Moreover, whether a tumor relies primarily on glycolysis or oxidative phosphorylation for energy production is not necessarily related to tumor malignancy; rather it depends on the environment within the organ of origin or the tissue to which the cancer cells have metastasized (24). Thus, flexible switching between glycolysis and oxidative phosphorylation, depending on the circumstances that the tumor faces at any given time, might be required for progression. PDH might coordinate these two phases of tumor metabolism, thereby promoting tumor growth. However, it may be possible that the PDH-KD cells we generated cannot induce the same changes in metabolites and/or gene expression that exist in tumor cells exhibiting the Warburg effect, even though they, under normoxic conditions, consume less oxygen and produce high levels of lactate, both of which are hallmark criteria of the Warburg effect.

Importantly, even after establishing stable glycolytic metabolism in prolonged hypoxic culture by downregulating PDH-E1β, tumor growth can readily recover if PDH-E1β is constitutively expressed at a higher level (Fig. 7E). As the tumor microenvironment is constantly changing from a normoxic to a hypoxic state and back again (25), this change will potentially lead to repetitive up- and downregulation of PDH-E1β, which could alter the metabolic mode accordingly under pathophysiologic conditions. This could, in turn, promote tumor growth by providing tumor cells with the opportunity to use both glycolysis and oxidative phosphorylation to meet their metabolic needs (26).

Patients harboring mutations in PDH exhibit severe lactic acidosis, which results in neurological disorders such as Leigh syndrome, which often causes lethality at a young age (27). However, the relationship between PDH mutation and the occurrence of cancer is not well understood. A search of the cBioPortal database revealed that PDH genes are amplified in a wide range of cancer types (Fig. 7G). Thus, amplification of PDH could represent a tactical strategy by which cancer cells flexibly switch between aerobic and anaerobic metabolism, regardless of the oxygen environment to maintain a favorable metabolic state.

N. Takeda reports receiving a commercial research grant from Daiichi Sankyo Company, Limited and Bayer Yakuhin, Ltd. No potential conflicts of interest were disclosed by the other authors.

Conception and design:K. Nakayama

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.):R. Yonashiro, K. Eguchi, M. Wake, N. Takeda, K. Nakayama

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis):R. Yonashiro, K. Eguchi, M. Wake, N. Takeda, K. Nakayama

Writing, review, and/or revision of the manuscript:R. Yonashiro, M. Wake, N. Takeda, K. Nakayama

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):K. Eguchi

Study supervision:K. Nakayama

We are grateful to Dr. Yoshio Miki for materials and Takahide Enomoto for technical assistance. We also thank Drs. Johji Inazawa, Hiroshi Shibuya, Hiroshi Nishina and members of his laboratory for helpful suggestions. K. Nakayama was supported by the Takeda Science Foundation, the Japan Foundation for Applied Enzymology, the Kato Memorial Bioscience Foundation, and the Ichiro Kanehara Foundation. This study was also supported by a Grant-in-Aid for Scientific Research C (JSPS KAKENHI Grant Number 15K08260), a Grant-in-Aid for Scientific Research on Innovative Areas “Oxygen Biology” (Grant Number 17H05523, MEXT, Japan), and the Integrated Research Projects on Intractable Disease Program, Medical Research Institute, TMDU.

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.

1.
Semenza
GL
. 
Targeting HIF-1 for cancer therapy
.
Nat Rev Cancer
2003
;
3
:
721
32
.
2.
Weidemann
A
,
Johnson
RS
. 
Biology of HIF-1alpha
.
Cell Death Differ
2008
;
15
:
621
7
.
3.
Nakayama
K
. 
cAMP-response element-binding protein (CREB) and NF-kappaB transcription factors are activated during prolonged hypoxia and cooperatively regulate the induction of matrix metalloproteinase MMP1
.
J Biol Chem
2013
;
288
:
22584
95
.
4.
Warburg
O
. 
On the origin of cancer cells
.
Science
1956
;
123
:
309
14
.
5.
Nakayama
K
,
Gazdoiu
S
,
Abraham
R
,
Pan
ZQ
,
Ronai
Z
. 
Hypoxia-induced assembly of prolyl hydroxylase PHD3 into complexes: implications for its activity and susceptibility for degradation by the E3 ligase Siah2
.
Biochem J
2007
;
401
:
217
26
.
6.
Kikuchi
D
,
Minamishima
YA
,
Nakayama
K
. 
Prolyl-hydroxylase PHD3 interacts with pyruvate dehydrogenase (PDH)-E1beta and regulates the cellular PDH activity
.
Biochem Biophys Res Commun
2014
;
451
:
288
94
.
7.
Patel
MS
,
Nemeria
NS
,
Furey
W
,
Jordan
F
. 
The pyruvate dehydrogenase complexes: structure-based function and regulation
.
J Biol Chem
2014
;
289
:
16615
23
.
8.
Kim
JW
,
Tchernyshyov
I
,
Semenza
GL
,
Dang
CV
. 
HIF-1-mediated expression of pyruvate dehydrogenase kinase: a metabolic switch required for cellular adaptation to hypoxia
.
Cell Metab
2006
;
3
:
177
85
.
9.
Papandreou
I
,
Cairns
RA
,
Fontana
L
,
Lim
AL
,
Denko
NC
. 
HIF-1 mediates adaptation to hypoxia by actively downregulating mitochondrial oxygen consumption
.
Cell Metab
2006
;
3
:
187
97
.
10.
Hitosugi
T
,
Fan
J
,
Chung
TW
,
Lythgoe
K
,
Wang
X
,
Xie
J
, et al
Tyrosine phosphorylation of mitochondrial pyruvate dehydrogenase kinase 1 is important for cancer metabolism
.
Mol Cell
2011
;
44
:
864
77
.
11.
Ohashi
Y
,
Hirayama
A
,
Ishikawa
T
,
Nakamura
S
,
Shimizu
K
,
Ueno
Y
, et al
Depiction of metabolome changes in histidine-starved Escherichia coli by CE-TOFMS
.
Mol Biosyst
2008
;
4
:
135
47
.
12.
Ooga
T
,
Sato
H
,
Nagashima
A
,
Sasaki
K
,
Tomita
M
,
Soga
T
, et al
Metabolomic anatomy of an animal model revealing homeostatic imbalances in dyslipidaemia
.
Mol Biosyst
2011
;
7
:
1217
23
.
13.
Sugimoto
M
,
Wong
DT
,
Hirayama
A
,
Soga
T
,
Tomita
M
. 
Capillary electrophoresis mass spectrometry-based saliva metabolomics identified oral, breast and pancreatic cancer-specific profiles
.
Metabolomics
2010
;
6
:
78
95
.
14.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
15.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
16.
Fukuda
R
,
Zhang
H
,
Kim
JW
,
Shimoda
L
,
Dang
CV
,
Semenza
GL
. 
HIF-1 regulates cytochrome oxidase subunits to optimize efficiency of respiration in hypoxic cells
.
Cell
2007
;
129
:
111
22
.
17.
Eales
KL
,
Hollinshead
KE
,
Tennant
DA
. 
Hypoxia and metabolic adaptation of cancer cells
.
Oncogenesis
2016
;
5
:
e190
.
18.
Cairns
RA
,
Harris
IS
,
Mak
TW
. 
Regulation of cancer cell metabolism
.
Nat Rev Cancer
2011
;
11
:
85
95
.
19.
Yang
W
,
Zheng
Y
,
Xia
Y
,
Ji
H
,
Chen
X
,
Guo
F
, et al
ERK1/2-dependent phosphorylation and nuclear translocation of PKM2 promotes the Warburg effect
.
Nat Cell Biol
2012
;
14
:
1295
304
.
20.
Faubert
B
,
Boily
G
,
Izreig
S
,
Griss
T
,
Samborska
B
,
Dong
Z
, et al
AMPK is a negative regulator of the Warburg effect and suppresses tumor growth in vivo
.
Cell Metab
2013
;
17
:
113
24
.
21.
Vander Heiden
MG
,
Cantley
LC
,
Thompson
CB
. 
Understanding the Warburg effect: the metabolic requirements of cell proliferation
.
Science
2009
;
324
:
1029
33
.
22.
Fogal
V
,
Richardson
AD
,
Karmali
PP
,
Scheffler
IE
,
Smith
JW
,
Ruoslahti
E
. 
Mitochondrial p32 protein is a critical regulator of tumor metabolism via maintenance of oxidative phosphorylation
.
Mol Cell Biol
2010
;
30
:
1303
18
.
23.
Weinberg
F
,
Hamanaka
R
,
Wheaton
WW
,
Weinberg
S
,
Joseph
J
,
Lopez
M
, et al
Mitochondrial metabolism and ROS generation are essential for Kras-mediated tumorigenicity
.
Proc Natl Acad Sci U S A
2010
;
107
:
8788
93
.
24.
Dupuy
F
,
Tabaries
S
,
Andrzejewski
S
,
Dong
Z
,
Blagih
J
,
Annis
MG
, et al
PDK1-dependent metabolic reprogramming dictates metastatic potential in breast cancer
.
Cell Metab
2015
;
22
:
577
89
.
25.
Dewhirst
MW
,
Cao
Y
,
Moeller
B
. 
Cycling hypoxia and free radicals regulate angiogenesis and radiotherapy response
.
Nat Rev Cancer
2008
;
8
:
425
37
.
26.
Wallace
DC
. 
Mitochondria and cancer
.
Nat Rev Cancer
2012
;
12
:
685
98
.
27.
Smeitink
J
,
van den Heuvel
L
,
DiMauro
S
. 
The genetics and pathology of oxidative phosphorylation
.
Nat Rev Genet
2001
;
2
:
342
52
.

Supplementary data