Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Mitochondrial activation by inhibition of PDKII suppresses HIF1a signaling and angiogenesis in cancer


Most solid tumors are characterized by a metabolic shift from glucose oxidation to glycolysis, in part due to actively suppressed mitochondrial function, a state that favors resistance to apoptosis. Suppressed mitochondrial function may also contribute to the activation of hypoxia-inducible factor 1α (HIF1α) and angiogenesis. We have previously shown that the inhibitor of pyruvate dehydrogenase kinase (PDK) dichloroacetate (DCA) activates glucose oxidation and induces apoptosis in cancer cells in vitro and in vivo. We hypothesized that DCA will also reverse the ‘pseudohypoxic’ mitochondrial signals that lead to HIF1α activation in cancer, even in the absence of hypoxia and inhibit cancer angiogenesis. We show that inhibition of PDKII inhibits HIF1α in cancer cells using several techniques, including HIF1α luciferase reporter assays. Using pharmacologic and molecular approaches that suppress the prolyl-hydroxylase (PHD)-mediated inhibition of HIF1α, we show that DCA inhibits HIF1α by both a PHD-dependent mechanism (that involves a DCA-induced increase in the production of mitochondria-derived α-ketoglutarate) and a PHD-independent mechanism, involving activation of p53 via mitochondrial-derived H2O2, as well as activation of GSK3β. Effective inhibition of HIF1α is shown by a decrease in the expression of several HIF1α regulated gene products as well as inhibition of angiogenesis in vitro in matrigel assays. More importantly, in rat xenotransplant models of non-small cell lung cancer and breast cancer, we show effective inhibition of angiogenesis and tumor perfusion in vivo, assessed by contrast-enhanced ultrasonography, nuclear imaging techniques and histology. This work suggests that mitochondria-targeting metabolic modulators that increase pyruvate dehydrogenase activity, in addition to the recently described pro-apoptotic and anti-proliferative effects, suppress angiogenesis as well, normalizing the pseudo-hypoxic signals that lead to normoxic HIF1α activation in solid tumors.


Apoptosis resistance and angiogenesis are both critical for tumor progression but are often seen as separate, independent processes. An ideal cancer therapy would target both and be selective to cancer cells, sparing non-cancerous tissues. There is emerging evidence that reversing the unique metabolic remodeling of cancer (that is, a strong glycolytic phenotype, the Warburg effect) is a promising therapeutic strategy that may offer relative selectivity to cancer.1, 2, 3, 4, 5 The suppressed mitochondrial function and decreased glucose oxidation (GO) in exchange for upregulated glycolysis, a much less efficient means of generating energy, may at first appear paradoxical for cancer. However, suppressing mitochondrial function can also lead to decreased mitochondria-driven apoptosis, and may allow important metabolites in the glycolytic pathway, including ribose sugars and nonessential amino acids to be shifted towards biosynthetic pathways that are needed in rapidly proliferating cells.4, 6

Metabolic targeting of key enzymes that regulate the balance between GO (that exclusively occurs in mitochondria) and cytoplasmic glycolysis, including pyruvate dehydrogenase kinase (PDK),7 lactate dehydrogenase8 or pyruvate kinase,9 has been shown to increase apoptosis, decrease proliferation and tumor growth. Dichloroacetate (DCA), a small molecule that inhibits PDK (thus activating pyruvate dehydrogenase (PDH), the gatekeeper of GO) decreases tumor growth in many cancers including non-small cell lung,7 pancreatic,10 metastatic breast,11 colon,12 prostate,13 endometrial,14 ovarian cancer15 and neuroblastoma.16 Importantly, preliminary data from a small human trial showed that DCA may hold promise in patients with advanced recurrent glioblastoma multiforme (GBM).17

Activation of the master regulator of angiogenesis hypoxia-inducible factor 1a (HIF1α) results in suppression of mitochondrial function and upregulation of glycolysis,18, 19, 20 but there is also emerging evidence that activation of mitochondrial signaling can directly inhibit HIF1α in a feed forward manner. For example, inhibition of PDK can inhibit HIF1α in cancer cells, although the mechanism has not been systematically explored.21 Mitochondria may regulate HIF1α via a number of ways including (a) direct effects of mitochondria-derived reactive oxygen species(mROS) like H2O2 on redox-sensitive units of HIF1α22, 23, 24 or prolylhydroxylases (PHDs), which hydroxylate HIF1α and signal it for ubiquination and degradation18, 25, 26; (b) direct effects of the mitochondria-derived metabolite α-ketoglutarate (αKG), a required co-factor for the HIF1α degradation by PHDs27; and (c) indirect effects mediated by a redox- and glycogen synthase kinase 3β (GSK3β)-mediated activation of p53,28, 29, 30, 31 a known inhibitor of HIF1α signaling.32, 33, 34, 35

Thus, it is surprising that the effects of this evolving new family of metabolic modulators on tumor angiogenesis are unknown. In the human glioblastoma DCA trial, there was early evidence in some of the treated patients that their tumors following DCA treatment (compared with the tumors of the same patients before DCA treatment) were unexpectedly less hemorrhagic and less vascular.17 Furthermore, it is clear from the many studies that have now reported anti-tumor effects of DCA that its effects on cancer cells in vitro in terms of apoptosis are modest, while more robust effects are observed in vivo,36 suggesting additional effects, like a potential primary inhibition of angiogenesis. Thus, we hypothesized that metabolic targeting with either molecular or pharmacologic inhibition of PDK, resulting in the reversal of the Warburg effect, can also have primary anti-angiogenic effects via a mitochondria-mediated inhibition of HIF1α, further increasing the promise of this new family of cancer therapies.


DCA reduces tumor vascularity and perfusion and inhibits HIF1α in vivo

We randomized athymic rats to receive either DCA orally or vehicle (water) one month after the subcutaneous injection of A549 non-small cell lung cancer (nSCLC) cells, when tumors become measurable. DCA increased survival significantly, causing a sustained decrease in tumor growth during a 9-month follow-up period (Figure 1a). DCA-treated animals had smaller tumors with decreased cell density and vascularization, compared with tumors from vehicle-treated animals, which were large and well-vascularized, even upon macroscopic examination (Figure 1a).

Figure 1

Activation of PDH increases survival, decreases tumor size and perfusion in nSCLC. (a) Top: DCA (50 mg/Kg) increases survival (top) and decreases tumor growth (middle) (n=12 animals per group, *P<0.05 vs control). Bottom: H&E staining shows that tumors from DCA-treated animals have decreased cell density compared with the untreated controls. On macroscopic examination tumors from DCA-treated animals have decreased vasculature compared with untreated controls. (b) DCA-treated animals have decreased tumor sestamibi uptake measured by SPECT-CT, which is indicative of decreased perfusion. Pseudo-color map indicates sestamibi levels (red is high and purple is low levels of sestamibi uptake; n=5 animals per group, *P<0.05 vs control). (c) DCA-treated animals have decreased contrast-enhanced micro-bubble perfusion compared with controls. Green represents individual contrast enhanced microbubbles measured by ultrasound (n=5 animals per group, *P<0.05 vs control). (d) DCA-treated animals have decreased tumor lectin signal (an endothelium-specific marker) compared with untreated controls (lectin: green, merged with nuclear stain DAPI: blue; n=5 animals per group, *P<0.05 vs control). (e) Left panel: DCA decreases HIF1α nuclear localization, expression and vascular density in vivo ((low magnification left; high magnification, right); HIF1α: red; vWF: green; nuclear stain DAPI: blue). Co-staining shows that in control, but not DCA-treated tumors, cells with nuclear HIF1α are found around capillary networks. Right panel: DCA decreases tumor HIF1α mRNA levels (n=5 animals per group, *P<0.01).

We quantified the DCA-induced effects in tumor perfusion with two clinically used techniques: radioisotope sestamibi-Tc99m imaging using a micro-SPECT-PET-CT and micro-bubble contrast-enhanced ultrasound perfusion imaging. DCA-treated rats had tumors with decreased sestamibi-Tc99m uptake and micro-bubble perfusion, compared with vehicle-treated controls (Figures 1b and c). We also quantified the vascular density of perfused tumors by intravenous injection of an endothelium-specific lectin (ricinus communis agglutinin) before euthanasia. Lectin co-localized with the endothelial marker von-Willebrand factor (vWF) (Supplementary Figure 1a). The tumor vessel density was decreased in the DCA-treated animals compared with the vehicle-treated controls (Figure 1d). The decrease in the amount of tumor microvessels in the DCA-treated rats was associated with a suppression of HIF1α within the tumor cells (Figure 1e). This was shown by the decrease of HIF1α mRNA, compared with control rats, in which abundant HIF1α was localized within the nucleus.

To assess the effects of DCA on a more aggressive tumor model, we injected rats with CRL-2335 mammary carcinoma cells, a cell line that has low expression of p5337 (Supplementary Figure 1b). When injected with the same number of cells, these animals developed larger tumors at a faster rate compared with nSCLC cells, but DCA decreased both the size and the vascularity of these tumors as well (Supplementary Figures 1c and d).

Activation of PDH increases mitochondrial function and inhibits HIF1α

To assess mitochondrial function in vivo, we measured mitochondrial membrane potential and mROS in tumor tissues immediately after resection, as described previously.17 The generation of the negative potential across the mitochondrial membranes (that is, the export of hydrogen cations as electrons are donated to the electron transport chain) is linked to respiration and can be used as a surrogate for mitochondrial function. Mitochondrial hyperpolarization is associated with a decrease in the production of mROS (that is, superoxide) and resistance to apoptosis, because it promotes a closure of the redox- and voltage-sensitive mitochondrial transition pore, through which pro-apoptotic mediators efflux to the cytoplasm.38 Thus, it is not surprising that most solid tumors have more hyperpolarized mitochondria compared with non-cancerous tissues.7, 39 DCA-treated nSCLC tumors had decreased mitochondrial membrane potential (measured by TMRM, a mitochondria-selective voltage-sensitive rhodamine-based dye) and increased mROS (that is, mitochondrial superoxide measured by MitoSOX) compared with controls (Figures 2a and b). These data show that the mitochondrial effects of DCA, previously shown in short-term in vivo models, are sustained in a long-term model and parallel the decrease in tumor angiogenesis in vivo. Importantly, these effects were associated with an increase in PDH activity measured by two separate techniques (Figure 2c).

Figure 2

Activation of PDH increases mitochondrial activity in nSCLC. (a) Freshly excised nSCLC tumors from DCA-treated animals show depolarized mitochondrial membrane potential compared with untreated controls (TMRM: red, nuclear stain DAPI: blue; n=5 animals per group, *P<0.05 vs control). (b) Freshly excised nSCLC tumors from DCA-treated animals show increased levels of mitochondrial-derived reactive oxygen species (that is, superoxide) measured by MitoSOX (red) (nuclear stain DAPI: blue; n=5 animals per group, *P<0.05 vs control). (c) DCA increases PDH activity (dipstick and spectrophotometric assays) in nSCLC cells, compared with vehicle-treated controls (n=3 different experiments, *P<0.05 vs control). (d) DCA increases whole-cell respiration in nSCLC cells compared with vehicle-treated controls. In the representative trace, DCA increases respiration as shown by the increase in the rate of oxygen consumption (n=3 different experiments, *P<0.05 vs control). (e) DCA increases endogenous H2O2 levels measured by Amplite fluorescence (green) in nSCLC cells compared with vehicle-treated controls (nuclear stain DAPI in blue; n=100 cells per group in 3 different experiments; *P<0.05 vs control). DCA increases nSCLC H2O2 production in the supernatant of intact cells compared with vehicle-treated controls (*P<0.05 vs control), which is negated by the SOD inhibitor diethyl-dithio-carbamate (10 μM) (#P<0.01 vs DCA). (f) DCA increases IDH activity and α-ketoglutarate levels in nSCLC cells compared with vehicle-treated controls (*P<0.05 vs control; n=5 different experiments, data are normalized to protein concentration).

As predicted, the increase in PDH activity, and thus the increase in the entry of pyruvate into the mitochondria, caused an increase in nSCLC cell respiration in both whole-cell (Figure 2d) and mitochondria preparations (+28%±3.5%, n=5; P<0.05). Furthermore, in keeping with the increase in mitochondrial superoxide, DCA increased both intracellular and extracellular nSCLC H2O2 levels (Figure 2e). H2O2 has a longer effective diffusible radius that (in contrast to superoxide) allows it to reach extra-mitochondrial redox-sensitive targets. The superoxide dismutase inhibitor (SODi) diethyl-dithio-carbamate40 eliminated most of the H2O2, suggesting that it is mostly derived from superoxide (Figure 2e). DCA also increased the activity of the Krebs’ cycle enzyme isocitrate dehydrogenase (IDH) and levels of its product α-ketoglutarate (αKG) (Figure 2f). These results were reproduced in the CRL-2335, mammary carcinoma cells, where DCA decreased mitochondrial membrane potential and increased mROS, PDH activity, mitochondrial respiration, H2O2 (in an SOD-dependent manner), IDH activity and αKG levels (Supplementary Figure 2).

The increase in αKG levels was associated with increased prolyl-hydroxylase (PHD) activity in DCA-treated nSCLC cells shown by both immunofluorescence and immunoblots with an antibody against hydroxylated HIF1α (Supplementary Figures 3a and b). This was associated with decreased nuclear levels of HIF1α in DCA-treated nSCLC cells (Supplementary Figure 3c). In addition, a cell-permeable form of αKG (octyl-αKG, 0.01–0.1 mM) mimicked DCA and decreased nuclear levels of HIF1α (Supplementary Figure 3c). The DCA-induced decrease in nuclear HIF1α levels was also observed in mammary carcinoma cells (Supplementary Figure 3d).

To confirm the decrease in HIF1α levels observed in vivo in Figure 1e, we treated nSCLC cells with DCA and detected a small but significant decrease in HIF1α protein levels at 12 and 24 h after treatment (Figure 3a). In addition, DCA decreased HIF1α mRNA levels in nSCLC cells in vitro (Supplementary Figure 4). In a separate experiment, we treated nSCLC cells with DCA in the presence of the 20S proteasome inhibitor MG132 and showed that the DCA-induced decrease in HIF1α protein levels was mostly negated by MG132 (Figure 3a), suggesting that DCA promotes HIF1α degradation pathways.

Figure 3

Activation of PDH increases degradation and decreases activity of HIF1α in nSCLC. (a) DCA treatment for 12 and 24 h decreases HIF1α protein levels using immunoblots (n=3 different experiments, *P<0.05 vs control); left immunoblot. DCA treatment for 12 h decreases HIF1α protein levels, which is negated by the proteosome inhibitor MG132 (10 μM)(n=3 different experiments); right immunoblot. (b) DCA decreases endogenous, but not a mutated form of HIF1α (CA5) that is resistant to proline hydroxylation and proteosomal degradation by PHDs (n=3 different experiments, *P<0.05 vs control). (c) DCA decreased HIF1α activity (as measured by a dual luciferase reporter assay) in nSCLC cells. The PHD inhibitors DMOG and cobalt chloride increased HIF1α activity compared with control nSCLC cells. DCA partially reversed the effects of both DMOG and cobalt chloride (n=12 wells per group in 3 different experiments, *P<0.05 vs control, #P<0.05 vs non-treated control cells, P<0.05 vs non-treated DCA cells). (d) DCA and a PDKII siRNA both inhibit HIF1α activity as shown by decreased luminescence in a dual luciferase reporter assay. However, in the presence of PDKII siRNA, DCA does not further inhibit HIF1α activity. PDHE1α siRNA does not alter HIF1α activity, but negates the effects of DCA (n=12 wells per group in 3 different experiments, *P<0.05 vs control). (e) PDHE1α siRNA increased HIF1α activity as measured by a dual luciferase reporter assay in SAEC cells (n=12 wells per group in 3 different experiments, *P<0.05 vs control).

We then treated nSCLCs with an adenovirus carrying a well-studied mutated form of HIF1α (CA5), which is resistant to proline hydroxylation and proteosomal degradation.41 Because AdCA5 encodes an altered form of human HIF1α with deletion of amino acids 392–520, we can separate endogenous HIF1α from CA5 using immunoblots and an antibody to HIF1α that recognizes amino acids 610–727.42 Although DCA decreased endogenous HIF1α protein levels, DCA did not decrease the CA5 mutant form of HIF1α (Figure 3b), suggesting that the primary effects of DCA on HIF1α-mediated degradation are through PHDs.

We then measured the effects of DCA on HIF1α activity by using a standard dual-luciferase reporter assay. Because we would be unable to separate the effects of DCA between endogenous and the CA5 form of HIF1α in terms of HIF1α transcriptional activity, we did not use the AdCA5 but instead treated nSCLCs with two different PHD inhibitors, dimethyloxalyglycine N-(methoxyoxoacety)-glycine methylester (DMOG) and cobalt chloride.43 DMOG inhibits the αKG-binding site on PHDs, while cobalt chloride inhibits PHDs by displacing Fe(II) from the catalytic center.20, 44 DMOG and cobalt chloride further increased HIF1α activity in nSCLCs, suggesting that PHD activity is not maximally suppressed in this cancer cell line; but addition of DCA partially decreased HIF1α activity in both, although not to the level of DCA-treated control nSCLC cells (Figure 3c). This suggests that DCA has additional effects on HIF1α transcriptional activity, independent of its effects on PHD-dependent degradation of HIF1α.

We then treated nSCLC cells with a PDKII and a PDHE1α short interfering RNA (siRNA) (the target for PDKII) and effectively inhibited the expression of both enzymes (Supplementary Figure 5a). PDKII siRNA-treated cells had decreased HIF1α activity (using the HIF1α-specific luciferase construct) and nuclear protein similar to the levels caused by DCA, with no additional inhibition seen with DCA (Figure 3d and Supplementary Figure 5b). PDHE1α siRNA-treated nSCLC cells did not show any increase in nuclear HIF1α and activity, suggesting that PDH is maximally inhibited in this cell line, but were completely resistant to DCA (Figure 3d and Supplementary Figure 5b). On the other hand, PDHE1α siRNA treatment of normal small airway epithelial cells (SAECs), where PDH activity may not be inhibited like in cancer, caused an increase in HIF1α activity and nuclear levels (Figure 3e and Supplementary Figures 5c and d).

N-Acetylcysteine (NAC) is an antioxidant that has recently been shown to inhibit HIF1α activity in a PHD-dependent manner.45 NAC did not completely reverse the increase in mROS by DCA in nSCLCs, suggesting that this antioxidant may not be as important in regulating mROS as it is in decreasing cytoplasm-derived ROS (Supplementary Figure 6a). Similar to the work by Gao et al.,45 NAC decreased HIF1α activity in nSCLCs, while the addition of DCA further decreased HIF1α activity (Supplementary Figure 6b). Again, the further decrease in HIF1α activity suggests that DCA is inhibiting HIF1α transcriptional activity in addition to its effects on PHD-dependent HIF1α degradation.

DCA increases p53 and GSK3β-mediated inhibition of HIF1α

p53 is redox-sensitive28, 29, 30 and has been shown to degrade HIF1α in a PHD-independent manner35 and inhibit HIF1α transcriptional activity.32, 33 DCA increased p53-activity (shown by a p53-specific luciferase assay) and nuclear localization in nSCLC cells and this was prevented by an SODi and mimicked by exogenous H2O2 (Figures 4a and b). In addition, DCA increased p53 nuclear localization in vivo (Figure 4b). The increase in p53 activity was supported by the increase in the p53-target gene product p21 in vitro and in vivo (Figures 4b and c). p21 is a cyclin-dependent kinase inhibitor and can halt proliferation by locking cells in the G0/G1 phase, an effect that may contribute to the anti-proliferative properties of DCA in nSCLC cells, as shown by a decrease in the percentage of cells expressing proliferating cell nuclear antigen (PCNA) (Figure 4d). In addition, DCA increased p53 activity and nuclear localization, increased p21 protein and mRNA levels and decreased proliferation in the CRL-2335 mammary carcinoma cells as well (Supplementary Figures 7a–c).

Figure 4

Activation of PDH increases p53 activity in nSCLC. (a) DCA increases p53 activity measured by a dual luciferase reporter assay. The effects of DCA on p53 activity are negated by the SOD inhibitor diethyl-dithio-carbamate (#P<0.05 vs DCA) and mimicked by exogenous H2O2 (t-butyl-H2O2) (n=12 wells per group in 3 different experiments, *P<0.05 vs control). (b) DCA increases nuclear localization of p53 (p53: red, nuclear stain DAPI: blue) and the levels of the p53-target gene p21 (p21: green, nuclear stain DAPI: blue) in vitro and in vivo measured by immunofluorescence. (c) DCA increases p21 protein levels in vitro (n=3 different experiments, *P<0.05 vs control) and in vivo measured by immunoblots (n=5 for control and 4 for DCA, *P<0.05 vs control). Representative immunoblots are shown on top. (d) DCA decreases proliferation measured by the percentage of PCNA-positive (PCNA: red, nuclear stain DAPI: blue) nSCLC cells (n=100 cells from 3 different experiments, *P<0.05 vs control).

Glycogen synthase kinase-3β (GSK3β) is a metabolic kinase that is inhibited in the glycolytic environment of cancer46, 47 or pulmonary hypertension48 and has been shown to activate p5331 and inhibit HIF1α.49 DCA increased GSK3β activity in nSCLC and mammary carcinoma cells in vitro (in a dose-dependent manner) and in nSCLC tumors in vivo (Figure 5a and Supplementary Figures 8a and b). Thus, GSK3β activation may also contribute to the DCA-mediated activation of p53 and inhibition of HIF1α.

Figure 5

Activation of PDH increases GSK3β activity and decreases glucose transporter expression and activity in nSCLC in vitro and in vivo. (a) DCA increases GSK3β activity in a dose-dependent manner in vitro, as shown by the increase in the ratio of the active to the nonactive (phosphorylated) form of the enzyme both in vitro (n=3 experiments, *P<0.05 vs control; #P<0.05 vs 0.5 mM DCA) and in vivo (n=5 for control and 4 for DCA treated tumors, *P<0.05 vs control). Representative immunoblots are shown in Supplementary Figure 8a. (b) DCA decreases the expression of GLUTs 1, 3 and 4 in vitro (n=3 different experiments) and in vivo (n=5 animals per group, *P<0.05 vs control) using qRT–PCR. DCA decreases the expression of VEGF in vivo (n=5 animals per group,*P<0.05 vs control). Right box: DCA decreases glucose transport in nSCLCs compared with vehicle-treated controls (n=5 different experiments, *P<0.05 vs control). (c) DCA-treated animals have decreased tumor glucose uptake measured by micro-PET-CT imaging with [18F]-deoxy-glucose. Pseudo-color map indicates FDG levels (red is high and purple is low levels of FDG uptake, n=5 animals per group, *P<0.05 vs control).

These effects of DCA were also present in GBM cells, derived from a patient with a new diagnosis of advanced GBM at the time of his initial surgery. DCA increased H2O2 levels, IDH activity and αKG levels, similar to nSCLC and mammary carcinoma cells, although the effects on IDH activity and αKG levels were more robust in GBM cells compared with the other two cancers (Supplementary Figure 9a). Once again, DCA decreased nuclear levels of HIF1α, increased p53 activity, increased p21 protein and messenger RNA (mRNA) levels and decreased proliferation (Supplementary Figures 9b–e).

DCA-induced activation of PDH decreases glucose transporter expression and glucose uptake

The mRNA levels of several glucose transporters (GLUTs) and VEGF, all gene products of HIF1α were decreased in DCA-treated nSCLCs in vitro and in vivo (Figure 5b). To assess the functional consequences of the decrease in GLUTs expression, we measured radiolabeled glucose transport. DCA decreased functional glucose transport in nSCLC and mammary carcinoma cells in vitro, with the later being more pronounced (Figure 5b and Supplementary Figure 10). Furthermore, tumor fluoro-deoxy-glucose (FDG) uptake, measured by positron emission tomography (PET) using a micro-SPECT-PET-CT, was decreased in the nSCLC DCA-treated rats (Figure 5c).

DCA decreases the recruitment of bone marrow-derived angiogenic cells

In addition to VEGF,50, 51 HIF1α activation results in increased expression of stromal-derived factor 1 (SDF1),52 a signaling chemokine involved in the recruitment of bone marrow-derived angiogenic cells to ischemic tissues53 and tumors.54 Recruitment of these cells to tumors contributes to angiogenesis, because these cells (as well as other inflammatory cells) can differentiate into tumor-associated cells (for example tumor-associated fibroblasts55), which can be additional sources of angiogenic factors.56 DCA decreased SDF1 levels within nSCLC cells both in vitro and in vivo (Figure 6a) and decreased SDF1 and VEGF levels in the media of nSCLC cells in a dose-dependent manner (Figure 6b).

Figure 6

Activation of PDH decreases SDF1, VEGF and the recruitment of bone marrow-derived mesenchymal stem cells (BMDMSC) in nSCLC. (a) DCA decreases nuclear localization of HIF1α and SDF1 expression in vitro and in vivo using immunohistochemistry and confocal microscopy (HIF1α: green; SDF1: red; nuclear stain DAPI: blue). Increased levels of SDF1 are seen within the control cells that have nuclear HIF1α whereas in DCA-treated cells HIF1α is not in the nucleus and SDF1 levels are decreased. (b) DCA decreases secreted SDF1 and VEGF levels, measured by ELISA, in nSCLC in a dose-dependent manner (n=5 different experiments, *P<0.05 vs control; #P<0.05 vs 0.5 mM DCA). (c) DCA decreases the recruitment of cKit+ BMDMSCs in a dose-dependent manner in vitro, as shown by a Boyden Chamber migration assay, as described in the Results section (n=3 different experiments, *P<0.05 vs control; #P<0.05 vs 0.5 mM DCA). Strong cKit expression is seen in cells used in the assay (cKit: red, nuclear stain DAPI: blue), which were characterized as BMDMSCs by standard criteria shown in Supplementary Figures 11a and b. (d) DCA-treated animals have decreased protein levels of SDF1 and cKit in their tumors, compared with the control tumors, using immunoblots (each band corresponds to a separate rat tumor; n=5 for control and 4 for DCA; *P<0.05 vs control). (e) Immunohistochemistry and quadruple staining identify the presence of recruited cKit+ rat cells (staining positive for an anti-rat antibody) near the vessels of the human nSCLC xenotransplant control tumors, but not in DCA-treated tumors (vessels stained green by lectin; rat cells are shown in red using an anti-rat antibody, cKit is purple; nuclear stain DAPI is blue).

We then isolated and characterized rat bone marrow-derived mesenchymal stem cells (BMDMSC) based on currently proposed criteria57 in terms of surface markers (CD-73, -90 and -105 positive and CD-14, -19, -31 and -34 negative), plastic adherence and differentiation potential (for example, to osteocytes) (Supplementary Figures 11a and b). In addition, BMDMSCs also express cKit58, 59 and CXCR4, the receptor for SDF1.60 To determine whether DCA decreases the recruitment of BMDMSCs in vitro, we plated BMDMSCs in the top insert of a Boyden-Chamber assay and the supernatant from vehicle-treated versus DCA-treated nSCLC cells in the bottom insert. DCA decreased the migration of BMDMSC in a dose-dependent manner (Figure 6c). The effect of DCA on BMDMSC was strictly on cell migration, as DCA did not induce apoptosis in these cells (Supplementary Figure 11c). In vivo, DCA-treated nSCLC tumors had a decrease in SDF1 as well as cKit levels (Figure 6d), in agreement with a decrease in cKit+-cells infiltrating the tumors (Figure 6e). These data are in keeping with a recent report implicating SDF1 in the homing of cKit+ stem cells to ischemic tissues.61 To confirm that the cKit+-cells had migrated from the rat bone marrow into the human xenotransplant tumor, we double stained them with an anti-rat antibody. The recruitment of rat cKit+-cells was evident mostly in areas around blood vessels in the vehicle-treated nSCLC tumors (Figure 6e and Supplementary Figure 11d) and was diminished in the DCA-treated tumors (Figure 6e).

DCA decreases paracrine angiogenic signaling in vitro

To study the direct paracrine effects of DCA treatment on tumor angiogenesis, we studied endothelial cell tubule/network formation using a matrigel assay. Compared with normoxic non-cancer-induced control medium, hypoxic (pO230 mm Hg) non-cancer-induced medium increased angiogenesis, as hypoxia is a direct signal for angiogenesis (Figure 7a). The supernatant of normoxic untreated nSCLC cells and mammary carcinoma cells (cancer-induced media) increased angiogenesis to a level similar to the hypoxic medium, reflecting the importance of the pseudo-hypoxic activation of HIF1α in cancer cells (Figure 7a). In addition, the supernatant from DCA-treated nSCLC and mammary carcinoma cells returned angiogenesis to levels comparable to non-cancer induced control medium (Figure 7a).

Figure 7

Activation of PDH decreases angiogenesis in vitro in nSCLC and mammary carcinoma. (a) Matrigel assays show that hypoxic medium (mild, physiologic hypoxia: PO230–40 mm Hg; with appropriate gas mixtures and buffers so that the pH and PCO2 remain normal at 7.4 and 40 respectively) increases human endothelial cell tubule formation (n=5 different experiments, *P<0.05 vs normoxic controls, PO2120 mm Hg). The effluent of normoxic nSCLC and mammary carcinoma cells (normoxic cancer-induced medium) increases tubule formation to a similar level with hypoxic non-cancer-induced medium. In contrast, the effluent of DCA-treated normoxic nSCLC and mammary carcinoma cells does not increase tubule formation (n=5 different experiments, *P<0.05 vs normoxic control, #P<0.05 vs untreated effluent). (b) A proposed metabolic basis for cancer angiogenesis: In cancer (at least in several solid tumors), the mitochondria-based GO and respiration are decreased, in part due to inhibition of pyruvate dehydrogenase (PDH) activity. Resulting pseudo-hypoxic signals from mitochondria can contribute to a ‘feed forward’ HIF1α activation, which in turn increases PDK expression, inhibiting PDH and closing a feedback loop that causes a sustained glycolytic environment. DCA activates PDH by inhibiting PDK. Increased PDH activity reverses the pseudo-hypoxic signals by increasing mROS, including H2O2, and Krebs’ cycle substrates (that is, α-ketoglutarate; αKG). The increase in αKG results in activation of prolyl-hydroxylases (PHD), while the increase in H2O2 results in activation of p53 and direct inhibition of HIF1α transcriptional activity. Activation of p53 results in HIF1α degradation and decreased transcriptional activity, independent of PHDs and H2O2. Inhibition of HIF1α, results in decreased recruitment of bone-marrow derived mesenchymal stem cells (BMDMSC) and decreased tumor angiogenesis and perfusion. Under certain conditions, H2O2 could inhibit PHDs, resulting in HIF1α stabilization (see discussion).


Here, we describe a metabolic strategy for inhibiting cancer angiogenesis. We extend the growing evidence that metabolic modulators can promote apoptosis in cancer cells directly, to a paradigm in which they can have additional beneficial effects, limiting tumor angiogenesis. It is well established that HIF1α activation in cancer can induce many of the metabolic signatures of cancer, including the Warburg effect. However, although there is emerging evidence that mitochondria signaling can directly activate HIF1α in a feed forward manner,21 perhaps contributing to the normoxic activation of HIF1α in cancer, the potential effects of mitochondria-targeting drugs on tumor angiogenesis have not been studied. Given the momentum that the metabolic theory of cancer is gaining and the interest in developing metabolic modulators as cancer therapies, our work may be important in translational cancer research and therapeutics.

Suppressed GO/Gly ratio and mitochondrial hyperpolarization are partially reversible in cancer cells and contribute to a resistance to apoptosis.7, 17 Mitochondrial hyperpolarization and a decrease in GO/Gly ratio seem to also be critical events in the mitochondrial response to hypoxia in oxygen-sensitive tissues like the pulmonary arteries.48 We have recently shown that the cause for this mitochondrial hyperpolarization and PDH inhibition in hypoxia-exposed pulmonary arteries was a disruption of the ‘mitochondria-endoplasmic reticulum unit’, resulting in a decrease in intra-mitochondrial calcium and a subsequent decrease in the activity of PDH, a calcium-dependent enzyme.62 Inhibition of PDH, GO and their downstream mitochondrial signaling can lead to HIF1α activation in the oxygen-sensitive pulmonary circulation. Forced or sustained activation of PDH and GO during hypoxia prevents hypoxic signaling: the mitochondria do not hyperpolarize, mROS do not decrease, HIF1α is not activated and the acute and chronic response to hypoxia (that is, pulmonary hypertension) are absent in the pulmonary circulation.48 It is unknown whether a similar disruption of the ‘mitochondria-endoplasmic reticulum unit’ that could support a decrease in PDH function also occurs in cancer. There are however, examples of very vascular tumors like paragangliomas63 or renal cell carcinomas,64 which result from disruption of GO and mitochondrial signaling due to loss-of-function mutations in mitochondrial enzymes. It remains to be determined whether PDH may also be genetically abnormal in some vascular tumors. Certainly, if a suppression of GO can occur in a cancer cell during normoxia (for example, by primary inhibition of PDH), pseudo-hypoxic signaling could be initiated, leading to HIF1α activation and angiogenesis, initiating a positively reinforcing feedback loop with further suppression of mitochondrial function by HIF1α.

The DCA-induced increase in αKG and H2O2 that we describe can inhibit HIF1α activity in both a PHD-dependent and -independent manner (Figure 7b), as our experiments with cobalt chloride or DMOG-treated cells show (Figure 3c).

PHD-dependent HIF1α degradation

αKG is a cofactor in the PHD-dependent hydroxylation and degradation of HIF1α.27 In conditions that αKG production may be suppressed (for example, because of a functional suppression of mitochondria), the tonic PHD-dependent inhibition of HIF1α that occurs in normoxic conditions may be inhibited. Therefore, a DCA-induced restoration of αKG levels may promote PHD-dependent HIF1α inhibition, as we show with the increase in HIF1α hydroxylation (Supplementary Figures 3a and b).

However, several papers in the literature suggest that H2O2 (and other oxidative signals) inhibits PHD activity, stabilizing HIF1α.25, 26 Thus, the decrease in H2O2 levels that is associated with suppressed mitochondrial function would be predicted to activate PHD and inhibit HIF1α, and in such a case, a DCA-induced increase in H2O2 would promote HIF1α stabilization, in contrast to our findings.

PHD-independent inhibition of HIF1α

On the other hand, there are several pathways through which H2O2 and other metabolic signals may indeed decrease HIF1α activity, independent of PHDs. These include (a) H2O2 and other oxidants can directly oxidize sulfhydryl groups on HIF1α itself, resulting in inhibition of its binding to DNA, thus suppressing HIF1α transcriptional activity,22, 24 (b) H2O2 can also activate p53, which is known to be redox-sensitive28, 29, 30 and to promote HIF1α ubiquination and degradation via a PHD-independent mechanism,34, 35 (c) the increase in p53 activity can also indirectly decrease HIF1α transcriptional activity as p53 and HIF1α compete for a common co-transcription factor, that is, p30032 and (d) an increase in GSK3β activity, which is known to occur in conditions of normal metabolism (and is inhibited in glycolytic environments like those in cancer or pulmonary hypertension46, 47, 48, 65) is known to both activate p5331 and directly inhibit HIF1α.49, 66 We provide strong evidence that DCA normalizes the metabolic remodeling in cancer by promoting GO and normalizing the suppressed metabolic-mitochondrial signals that may activate HIF1α in cancer cells, even in normoxia. We also show that DCA increases p53 activity in an H2O2-dependent manner and GSK3β activity. In summary, these mechanisms support a PHD-independent activation of HIF1α as well as an αKG-based PHD-dependent activation of HIF1α in the normoxic cancer cells, which were reversed by DCA (Figure 7b).

In hypoxic normal cells, an mROS (H2O2)-dependent PHD inhibition may drive HIF1α activation, but this mechanism may be less relevant to normoxic activation of HIF1α in cancer cells. Although we did not study hypoxic signaling, our data and the literature suggest that this mitochondria-HIF1α axis maybe regulated differently between pathologic normoxic conditions (that is, cancer) and physiologic hypoxic conditions.18, 66 The multitude of mechanisms suggests that a) the final result on HIF1α transcriptional activity may be a balance of different and even opposing signals at any given time, and b) diversity in the mitochondria-HIF1α signaling among tissues and conditions or among different cancers is also possible.

Although we observed a significant decrease in HIF1α mRNA levels in our in vivo models, the decrease in mRNA levels in vitro was modest. In addition, our experiment with the proteosome inhibitor MG132 preventing most of the effects of DCA-mediated decreases in HIF1α protein levels (Figure 3a) suggests that in vitro, the effects of DCA on HIF1α levels are driven mostly by an increase in its degradation rather than by a decrease in its mRNA levels. The differences between the mRNA levels of HIF1α after DCA treatment in vivo and in vitro may be due to (a) the fact that these animals were treated with DCA for several months whereas, in our in vitro studies, DCA treatment was for 12 h, and (b) the differences between the in vivo conditions of a tumor that cannot be recapitulated in vitro; specifically, the studied material from the tumors included many non-cancer cells, like inflammatory or angiogenic circulating cells, in which the mechanisms for HIF1α expression regulation may be different than in cancer cells. For example, a recent report showed that the HIF promoter region does contain a HIF1α-binding site, suggesting that HIF1α may be able to regulate its own expression.67 Therefore, the decrease in HIF1α mRNA by DCA may be an extension of its direct inhibition on HIF1α transcriptional activity. In summary, while the early and direct effects of DCA on HIF1α levels are due to an increase in its degradation, the chronic effects of DCA may include an additional decrease in its transcription.

Several downstream targets of HIF1α were decreased in DCA-treated cells confirming the effectiveness of this anti-angiogenic strategy, including glucose transporters, VEGF and SDF1. Direct effects on angiogenesis were shown in vitro and effective inhibition of angiogenesis in vivo was shown with several techniques. We also provided evidence that the inhibition of SDF1 could contribute to decreased recruitment of stem cells within the tumor, although more work is needed to determine the relative importance of this mechanism in different cancer models.

Along with recent publications by others, our work suggests that positively reinforced feedback loops exist between inhibition of GO and activation of HIF1α. While we show that the decrease in GO leads to HIF1α activation, others have shown that HIF1α activation inhibits mitochondria enzymes that would further suppress GO. Specifically, strong evidence exists that HIF1α increases the expression of PDK168, 69 and pyruvate kinase M2 (PKM2).70 Such positively reinforcing feedback loops underlie not only the importance of this mechanism for the survival of tumors, but also the difficulty of its therapeutic targeting.

Nevertheless, the inhibition of angiogenesis by a metabolic modulator like DCA (or other potential inhibitors of PDK or pyruvate kinase isoforms) may prove to be more attractive than VEGF inhibitors. This is because the signaling pathway is inhibited more proximally, at the level of oxygen sensors like the mitochondria. This may allow less opportunity for tumors to escape, as is often the case with VEGF inhibitors. The persistent activation of HIF and other pro-angiogenic factors allows for recruitment of other pathways, including the attraction of bone marrow or circulating pro-angiogenic cells within the tumors, and eventual resistance to VEGF inhibition.71 Reversing the original pseudo-hypoxic signals from the mitochondria may be a logical alternative approach to inhibit angiogenesis. In addition, this approach may provide more tumor selectivity. Although direct inhibition of HIF1α or VEGF may adversely affect many physiologic or regenerative functions requiring appropriate activation of this pathway, reversal of the primary pseudo-hypoxic signal may restrict the effects to the tissues where this is primarily taking place, that is, the tumors. It is important to remember that HIF1α can be activated by mild hypoxia, not enough to inhibit mitochondrial function.18 However, severe hypoxia/anoxia or primary deficiencies in mitochondrial enzymes will cause a suppression of GO, inducing the signals that we discuss above. Thus, DCA, by reversing this mitochondrial suppression, may preferentially affect tumors and not interfere with the HIF1α signaling in physiologic conditions (that is, wound healing, erythropoietin activation, and so on). In keeping with this concept, the GBM patients treated with DCA showed decrease in tumor vascularity, but after many months of therapy, did not develop anemia or wound-healing abnormalities.17

At the same time, such approaches, that is, targeting mitochondrial enzymes, may allow the tumor to eventually express alternate isoforms of these enzymes, perhaps less sensitive to these drugs. For example, there are four isoforms of PDK, all with variable sensitivities to DCA, with PDKII being the most sensitive.72 Very recently, PDKII and DCA were crystallized together revealing molecular interactions that explain the relatively specific inhibition of PDKII by DCA.73, 74 Our data using siRNA to inhibit PDKII and PDH in this and our previous work7 support the fact that DCA works primarily by inhibition of PDKII. It is possible, however, that after chronic DCA therapy PDKII may be downregulated, or compensated by PDK isoforms that are less sensitive to DCA. Another mechanism of potential resistance maybe the recently reported decrease of the DCA transporter SLC5A8, via methylation, is cancer tissues.75 Babu et al. reversed this by increasing the expression of SLC5A8 with the DNA methylation inhibitor 5′-Azadc, potentiating the anticancer effects of DCA. The fact that the effects of DCA were sustained in our xenotransplant nSCLC model for 9 months is reassuring (we had previously shown tumor regression in short-term, 3–4-week, studies in xenotransplant models, without exploring any effects on angiogenesis7), but chronic and large studies in humans will be needed to address this potential concern.

In summary, the ability of metabolic modulators to suppress several mitochondria-driven paracrine pro-angiogenic pathways, in addition to their direct effects on apoptosis of the tumor cells, may open a new window in the rapidly expanding chapter of metabolic modulators in cancer therapy.

Materials and methods

The University of Alberta Animal Ethics Committee approved all animal experiments.

Cell culture conditions, cell lines, confocal imaging

Immunobloting, qRT–PCR, commercially available ELISAs for VEGF and SDF, Hydrogen Peroxide Assay, PDH activity assays, IDH activity assay, α-ketogultarate assay, mitochondrial and whole cell respiration assays, mesencheymal stem cell isolation, culture and characterization. See Supplement information.

In vivo tumorigenicity assays

A cell suspension of non-small cell lung cancer (A549) or mammary carcinoma cells (CRL-2335) in growth media (3 × 106 cells per injection) was injected subcutaneously into nude athymic rats. Rats were divided into controls (water) and DCA-treatment groups (DCA 0.070 g/l in drinking water) at one month for non-small cell lung cancer and 10 days for mammary carcinoma, post tumor cell injection, a time at which tumors become clearly measurable. Endpoints were mortality or severe morbidity due to very large tumors, requiring euthanasia, based on pre-specified standard criteria by our animal ethics committee. By measuring the amount of water consumed, we calculated and adjusted the concentration of DCA required to achieve a daily dose similar to that used clinically (50 mg/kg), as previously described.7 Rats were observed weekly for the appearance of tumors at injection sites, and tumor size was measured weekly with calipers by measuring the maximum length and width to assess tumor area (mm2).

Positron emission tomography (PET) imaging

Glucose uptake was imaged in vivo in rats anesthetized using isoflurane gas anesthesia (20% oxygen) using FDG as a radiotracer. PET imaging was performed using a Triumph PET/SPECT/CT system (, Northridge, CA, USA). Rats were injected with 500 uCi of FDG and imaged 60 min after FDG injection. PET data acquisition images are shown using a pseudocolor map (red=high glucose uptake). Data were quantified as mean PET signal by segmentation of the tumor using computed tomography (CT).

SPECT imaging

SPECT imaging was performed using a Triumph PET/SPECT/CT system (GE Healthcare). Rats were injected with 5 mCi of Tc-99-sestamibi under isoflurane gas anesthesia (20% oxygen) and imaged immediately after radioisotope injection. SPECT data acquisition images are shown using a pseudocolor map (red indicates increased perfusion). Tumors were excised from animals and Tc-99-sestamibi was directly measured using a Tc-99 m radioisotope detector.

Lectin perfusion assay

Rats were anesthetized, heparinized using 1000 IU/kg of heparin (American Pharmaceuticals Partners, Schaumbert, IL, USA) and injected with 5 mg of FITC-conjugated Ricinus Communis Agglutinin I (Vector Laboratories, Burlinghame, CA, USA) in the jugular vein. After circulation of lectin for 20 min, the rats were euthanized and the tumor tissue was harvested. Tissue was sliced 5 μm thick and fixed in 4% paraformaldehyde and imaged using confocal microscropy. Lectin signal was quantified using Ziess software and displayed as arbitrary fluorescence units (AFU).

Non-targeted micro-marker ultrasound

The VeVo 770 (Visual Sonics, Toronto, ON, Canada) imaging system with non-targeted contrast agent to assess tumor perfusion. Animals were canulated at the jugular vein and a bolus injection of 2 × 109 microbubbles/ml was administered. Data acquisition of tumor perfusion was initiated 20 frames before the bolus injection and data were continuously acquired for an additional 780 frames. Maximum perfusion was quantified in the tumor by Visual Sonics Imaging Software, which calculates the maximum amount of contrast agent detected over the full 800 frames.

Adenoviral infection

Non-small cell lung cancer cells were grown on T75 flasks (Sigma-Aldrich, Oakville, ON, Canada) and infected with AdCA5 (a generous gift from Dr Gregg Semenza) at a multiplicity of infection of 150 in serum-free DMEM (Gibco, Invitrogen, Burlington, ON, Canada) for 6 h allowing an infection rate of 50%. Medium was then exchanged for F12K containing 10% FBS.

Matrigel assay

Matrigel (BD Biosciences, San Jose, CA, USA) was prepared as a 1:1 mixture with serum free, supplemented Medium 200 (Cascade Biologics, Portland, OR, USA) and plated (300 μl per well) on a 24-well dish. Once the matrigel had solidified, human aortic endothelial cells were plated at 60 000 cells/well.

Synthesis of α-ketoglutarate derivative

Cell-permeable α-ketoglutarate was synthesized as previously described.27 The octyl-α-ketoglutarate ester was prepared as follows: octyl-chloroformate was added drop by drop to α-ketoglutaric acid (10 mmol) and triethylamine (1.0 eq) in dichloromethane (50 ml). The mixture was stirred at room temperature for 16 h and diluted with dichloromethane (50 ml), washed with 0.5 N aqueous HCl (50 ml) and dried. Octyl-α-ketoglutarate ester was prepared to >98% purity as assessed by gas chromatography.

Cell migration assay

Cell migration assay was performed according to protocol using the CHEMICON QCMTM 24-well Migration Assay (Millipore, Billerica, MA, USA). MSCs (1 × 106 cells per ml); also see supplement.

Glucose transport

Cancer cells were grown to confluence in 12-well dishes and treated with DCA (0.5 mM) for 4 days. Two hours before performing radioactive flux studies the media was removed, the cells washed twice with glucose-free Krebs-Ringer’s Solution (120 mM NaCl, 25 mM NaHCO3, 4 mM KCl, 1.2 mM KH2PO4, 2.5 mM MgSO4, 1.25 mM CaCl2, pH 7.4 in sterile water) and incubated at 37 °C for two hours in the glucose-free Krebs-Ringer’s. Radioactive glucose (0.1 mM and 5 mM), radiolabeled [14C]-d-glucose (Amersham, Baie d'Urfe, QC, Canada) and [3 H]-L-glucose (Amersham) was added to the cells, incubated on a heating pad for 25 min, then washed twice with ice-cold ‘Stop’ solution (glucose-free Krebs-Ringer’s solution with 200 μM phloretin added). The cells were then lysed by adding 500 ml of 5% trichloracetic acid and placed on a shaker table overnight. Next, three 150 μl samples from each well were added to 4 ml of ScintiSafe liquid scintillation fluid (Fisher, Toronto, ON, Canada) in scintillation vials. The vials were then placed in a Beckman LS 6500 multi-purpose liquid scintillation counter for quantification. All radioactivity was normalized to standards and corrected for background levels of isotope.

HIF1α and p53 luciferase assays

HIF1α and p53 dual-luciferase reporter kits (SA Biosciences, Mississauga, ON, Canada) were used to assess HIF1α and p53 activity in nSCLC and mammary carcinoma cells as described in the commercially available protocols. Briefly, cells (10 000 cells/well) were plated on a 96-well plate containing the transfection agent (Xfect; CloneTech, Mountain View, CA, USA) and the HIF1α and p53 Cignal reporter (which contained a control Renilla reporter as well). After 16 h of transfection, the transfection media was replaced with regular growth media and exposed to treatment for 48 h. HIF1α and p53 activity was assessed using the Promega dual luciferase reporter assay system from Promega (SA Biosciences Cells were lysed with the passive lysis buffer (provided in the kit) before luminescence was detected with a illuminometer. HIF1α and p53 activity was standardized to the transfection of each cell by normalizing data to Renilla luminescence.

siRNA treatment

Non-small cell lung cancer cells and SAECs were grown to 80% confluence in 96-well culture-plates and confocal dishes. The transfection agent siPORTamine (Ambion siRNA Transfection II kit 1631, Ambion, Faster city, CA, USA) was pre-incubated at room temperature for 10 min at a ratio of 1:12 in OptiMEM1 culture medium (Invitrogen-GIBCO) before being mixed with 50 nM PDKII or PDHE1α siRNA (Ambion) or scrambled RNA (Ambion). The culture media was aspirated from the cells, and the transfection agent-RNA complex mixture was allowed to spread over the monolayer of cells. Plates were incubated at 37 °C for 48 h.


We used unpaired two sample t tests for before and after DCA comparisons and ANOVA with post hoc analysis (Tukey’s test) for comparisons among three or more groups and Kaplan Meier for survival studies using standard SPSS software (Statview 4.02, SAS Institute, Cary, NC, USA). Data are presented as mean ±s.e.m. and P<0.05 was considered significant.


  1. 1

    Gatenby RA, Gillies RJ . Why do cancers have high aerobic glycolysis? Nat Rev Cancer 2004; 4: 891–899.

    CAS  Article  Google Scholar 

  2. 2

    Michelakis ED, Webster L, Mackey JR . Dichloroacetate (DCA) as a potential metabolic-targeting therapy for cancer. Br J Cancer 2008; 99: 989–994.

    CAS  Article  Google Scholar 

  3. 3

    Pan JG, Mak TW . Metabolic targeting as an anticancer strategy: dawn of a new era? Sci STKE 2007; 2007: pe14.

    PubMed  Google Scholar 

  4. 4

    Vander Heiden MG, Cantley LC, Thompson CB . Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 2009; 324: 1029–1033.

    CAS  Article  Google Scholar 

  5. 5

    Dromparis P, Sutendra G, Michelakis ED . The role of mitochondria in pulmonary vascular remodeling. J Mol Med 2010; 88: 1003–1010.

    CAS  Article  Google Scholar 

  6. 6

    DeBerardinis RJ, Lum JJ, Hatzivassiliou G, Thompson CB . The biology of cancer: metabolic reprogramming fuels cell growth and proliferation. Cell Metab 2008; 7: 11–20.

    CAS  Article  Google Scholar 

  7. 7

    Bonnet S, Archer SL, Allalunis-Turner J, Haromy A, Beaulieu C, Thompson R et alA mitochondria-K+ channel axis is suppressed in cancer and its normalization promotes apoptosis and inhibits cancer growth. Cancer Cell 2007; 11: 37–51.

    CAS  Article  Google Scholar 

  8. 8

    Fantin VR, St-Pierre J, Leder P . Attenuation of LDH-A expression uncovers a link between glycolysis, mitochondrial physiology, and tumor maintenance. Cancer Cell 2006; 9: 425–434.

    CAS  Article  Google Scholar 

  9. 9

    Christofk HR, Vander Heiden MG, Harris MH, Ramanathan A, Gerszten RE, Wei R et alThe M2 splice isoform of pyruvate kinase is important for cancer metabolism and tumour growth. Nature 2008; 452: 230–233.

    CAS  Article  Google Scholar 

  10. 10

    Chen Y, Cairns R, Papandreou I, Koong A, Denko NC . Oxygen consumption can regulate the growth of tumors, a new perspective on the Warburg effect. PLoS One 2009; 4: e7033.

    Article  Google Scholar 

  11. 11

    Sun RC, Fadia M, Dahlstrom JE, Parish CR, Board PG, Blackburn AC . Reversal of the glycolytic phenotype by dichloroacetate inhibits metastatic breast cancer cell growth in vitro and in vivo. Breast Cancer Res Treat 2010; 120: 253–260.

    CAS  Article  Google Scholar 

  12. 12

    Sanchez-Arago M, Chamorro M, Cuezva JM . Selection of cancer cells with repressed mitochondria triggers colon cancer progression. Carcinogenesis 2010; 31: 567–576.

    CAS  Article  Google Scholar 

  13. 13

    Cao W, Yacoub S, Shiverick KT, Namiki K, Sakai Y, Porvasnik S et alDichloroacetate (DCA) sensitizes both wild-type and over expressing Bcl-2 prostate cancer cells in vitro to radiation. Prostate 2008; 68: 1223–1231.

    CAS  Article  Google Scholar 

  14. 14

    Wong JY, Huggins GS, Debidda M, Munshi NC, De Vivo I . Dichloroacetate induces apoptosis in endometrial cancer cells. Gynecol Oncol 2008; 109: 394–402.

    CAS  Article  Google Scholar 

  15. 15

    Saed GM, Fletcher NM, Jiang ZL, Abu-Soud HM, Diamond MP . Dichloroacetate induces apoptosis of epithelial ovarian cancer cells through a mechanism involving modulation of oxidative stress. Reprod Sci 2011; 18: 1253–1261.

    CAS  Article  Google Scholar 

  16. 16

    Vella S, Conti M, Tasso R, Cancedda R, Pagano A . Dichloroacetate (DCA) inhibits neuroblastoma growth by specifically acting against malignant undifferentiated cells. Int J Cancer 2011; 130: 1484–1493.

    Article  Google Scholar 

  17. 17

    Michelakis ED, Sutendra G, Dromparis P, Webster L, Haromy A, Niven E et alMetabolic modulation of glioblastoma with dichloroacetate. Sci Transl Med 2010; 2: 31ra34.

    CAS  Article  Google Scholar 

  18. 18

    Denko NC . Hypoxia, HIF1 and glucose metabolism in the solid tumour. Nat Rev Cancer 2008; 8: 705–713.

    CAS  Article  Google Scholar 

  19. 19

    Semenza GL . Hypoxia-inducible factor 1 (HIF-1) pathway. Sci STKE 2007; 2007: cm8.

    Article  Google Scholar 

  20. 20

    Semenza GL . Hypoxia-inducible factors in physiology and medicine. Cell 2012; 148: 399–408.

    CAS  Article  Google Scholar 

  21. 21

    McFate T, Mohyeldin A, Lu H, Thakar J, Henriques J, Halim ND et alPyruvate dehydrogenase complex activity controls metabolic and malignant phenotype in cancer cells. J Biol Chem 2008; 283: 22700–22708.

    CAS  Article  Google Scholar 

  22. 22

    Huang LE, Arany Z, Livingston DM, Bunn HF . Activation of hypoxia-inducible transcription factor depends primarily upon redox-sensitive stabilization of its alpha subunit. J Biol Chem 1996; 271: 32253–32259.

    CAS  Article  Google Scholar 

  23. 23

    Salceda S, Caro J . Hypoxia-inducible factor 1alpha (HIF-1alpha) protein is rapidly degraded by the ubiquitin-proteasome system under normoxic conditions. Its stabilization by hypoxia depends on redox-induced changes. J Biol Chem 1997; 272: 22642–22647.

    CAS  Article  Google Scholar 

  24. 24

    Wang GL, Jiang BH, Semenza GL . Effect of altered redox states on expression and DNA-binding activity of hypoxia-inducible factor 1. Biochem Biophys Res Commun 1995; 212: 550–556.

    CAS  Article  Google Scholar 

  25. 25

    Brunelle JK, Bell EL, Quesada NM, Vercauteren K, Tiranti V, Zeviani M et alOxygen sensing requires mitochondrial ROS but not oxidative phosphorylation. Cell Metab 2005; 1: 409–414.

    CAS  Article  Google Scholar 

  26. 26

    Mansfield KD, Guzy RD, Pan Y, Young RM, Cash TP, Schumacker PT et alMitochondrial dysfunction resulting from loss of cytochrome c impairs cellular oxygen sensing and hypoxic HIF-alpha activation. Cell Metab 2005; 1: 393–399.

    CAS  Article  Google Scholar 

  27. 27

    MacKenzie ED, Selak MA, Tennant DA, Payne LJ, Crosby S, Frederiksen CM et alCell-permeating alpha-ketoglutarate derivatives alleviate pseudohypoxia in succinate dehydrogenase-deficient cells. Mol Cell Biol 2007; 27: 3282–3289.

    CAS  Article  Google Scholar 

  28. 28

    Huang C, Zhang Z, Ding M, Li J, Ye J, Leonard SS et alVanadate induces p53 transactivation through hydrogen peroxide and causes apoptosis. J Biol Chem 2000; 275: 32516–32522.

    CAS  Article  Google Scholar 

  29. 29

    Wang S, Leonard SS, Ye J, Ding M, Shi X . The role of hydroxyl radical as a messenger in Cr(VI)-induced p53 activation. Am J Physiol Cell Physiol 2000; 279: C868–C875.

    CAS  Article  Google Scholar 

  30. 30

    Xie S, Wang Q, Wu H, Cogswell J, Lu L, Jhanwar-Uniyal M et alReactive oxygen species-induced phosphorylation of p53 on serine 20 is mediated in part by polo-like kinase-3. J Biol Chem 2001; 276: 36194–36199.

    CAS  Article  Google Scholar 

  31. 31

    Watcharasit P, Bijur GN, Song L, Zhu J, Chen X, Jope RS . Glycogen synthase kinase-3beta (GSK3beta) binds to and promotes the actions of p53. J Biol Chem 2003; 278: 48872–48879.

    CAS  Article  Google Scholar 

  32. 32

    Schmid T, Zhou J, Kohl R, Brune B . p300 relieves p53-evoked transcriptional repression of hypoxia-inducible factor-1 (HIF-1). Biochem J 2004; 380: 289–295.

    CAS  Article  Google Scholar 

  33. 33

    Vousden KH, Ryan KM . p53 and metabolism. Nat Rev Cancer 2009; 9: 691–700.

    CAS  Article  Google Scholar 

  34. 34

    Kaluzova M, Kaluz S, Lerman MI, Stanbridge EJ . DNA damage is a prerequisite for p53-mediated proteasomal degradation of HIF-1alpha in hypoxic cells and downregulation of the hypoxia marker carbonic anhydrase IX. Mol Cell Biol 2004; 24: 5757–5766.

    CAS  Article  Google Scholar 

  35. 35

    Ravi R, Mookerjee B, Bhujwalla ZM, Sutter CH, Artemov D, Zeng Q et alRegulation of tumor angiogenesis by p53-induced degradation of hypoxia-inducible factor 1alpha. Genes Dev 2000; 14: 34–44.

    CAS  PubMed  PubMed Central  Google Scholar 

  36. 36

    Papandreou I, Goliasova T, Denko NC . Anti-cancer drugs that target metabolism, is dichloroacetate the new paradigm? Int J Cancer 2011; 128: 1001–1008.

    CAS  Article  Google Scholar 

  37. 37

    Muangnoi P, Lu M, Lee J, Thepouyporn A, Mirzayans R, Le XC et alCytotoxicity, apoptosis and DNA damage induced by Alpinia galanga rhizome extract. Planta Med 2007; 73: 748–754.

    CAS  Article  Google Scholar 

  38. 38

    Zamzami N, Kroemer G . The mitochondrion in apoptosis: how Pandora's box opens. Nat Rev Mol Cell Biol 2001; 2: 67–71.

    CAS  Article  Google Scholar 

  39. 39

    Chen LB . Mitochondrial membrane potential in living cells. Annu Rev Cell Biol 1988; 4: 155–181.

    CAS  Article  Google Scholar 

  40. 40

    Chandel NS, Vander Heiden MG, Thompson CB, Schumacker PT . Redox regulation of p53 during hypoxia. Oncogene 2000; 19: 3840–3848.

    CAS  Article  Google Scholar 

  41. 41

    Kelly BD, Hackett SF, Hirota K, Oshima Y, Cai Z, Berg-Dixon S et alCell type-specific regulation of angiogenic growth factor gene expression and induction of angiogenesis in nonischemic tissue by a constitutively active form of hypoxia-inducible factor 1. Circ Res 2003; 93: 1074–1081.

    CAS  Article  Google Scholar 

  42. 42

    Okuyama H, Krishnamachary B, Zhou YF, Nagasawa H, Bosch-Marce M, Semenza GL . Expression of vascular endothelial growth factor receptor 1 in bone marrow-derived mesenchymal cells is dependent on hypoxia-inducible factor 1. J Biol Chem 2006; 281: 15554–15563.

    CAS  Article  Google Scholar 

  43. 43

    Chan DA, Sutphin PD, Denko NC, Giaccia AJ . Role of prolyl hydroxylation in oncogenically stabilized hypoxia-inducible factor-1alpha. J Biol Chem 2002; 277: 40112–40117.

    CAS  Article  Google Scholar 

  44. 44

    Epstein AC, Gleadle JM, McNeill LA, Hewitson KS, O'Rourke J, Mole DR et alC. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation. Cell 2001; 107: 43–54.

    CAS  Article  Google Scholar 

  45. 45

    Gao P, Zhang H, Dinavahi R, Li F, Xiang Y, Raman V et alHIF-dependent antitumorigenic effect of antioxidants in vivo. Cancer Cell 2007; 12: 230–238.

    CAS  Article  Google Scholar 

  46. 46

    Pastorino JG, Hoek JB, Hexokinase II . the integration of energy metabolism and control of apoptosis. Curr Med Chem 2003; 10: 1535–1551.

    CAS  Article  Google Scholar 

  47. 47

    Pastorino JG, Hoek JB, Shulga N . Activation of glycogen synthase kinase 3beta disrupts the binding of hexokinase II to mitochondria by phosphorylating voltage-dependent anion channel and potentiates chemotherapy-induced cytotoxicity. Cancer Res 2005; 65: 10545–10554.

    CAS  Article  Google Scholar 

  48. 48

    Sutendra G, Bonnet S, Rochefort G, Haromy A, Folmes KD, Lopaschuk GD et alFatty acid oxidation and malonyl-CoA decarboxylase in the vascular remodeling of pulmonary hypertension. Sci Transl Med 2010; 2: 44ra58.

    Article  Google Scholar 

  49. 49

    Mottet D, Dumont V, Deccache Y, Demazy C, Ninane N, Raes M et alRegulation of hypoxia-inducible factor-1alpha protein level during hypoxic conditions by the phosphatidylinositol 3-kinase/Akt/glycogen synthase kinase 3beta pathway in HepG2 cells. J Biol Chem 2003; 278: 31277–31285.

    CAS  Article  Google Scholar 

  50. 50

    Forsythe JA, Jiang BH, Iyer NV, Agani F, Leung SW, Koos RD et alActivation of vascular endothelial growth factor gene transcription by hypoxia-inducible factor 1. Mol Cell Biol 1996; 16: 4604–4613.

    CAS  Article  Google Scholar 

  51. 51

    Pugh CW, Ratcliffe PJ . Regulation of angiogenesis by hypoxia: role of the HIF system. Nat Med 2003; 9: 677–684.

    CAS  Article  Google Scholar 

  52. 52

    Karshovska E, Zernecke A, Sevilmis G, Millet A, Hristov M, Cohen CD et alExpression of HIF-1alpha in injured arteries controls SDF-1alpha mediated neointima formation in apolipoprotein E deficient mice. Arterioscler Thromb Vasc Biol 2007; 27: 2540–2547.

    CAS  Article  Google Scholar 

  53. 53

    Ceradini DJ, Kulkarni AR, Callaghan MJ, Tepper OM, Bastidas N, Kleinman ME et alProgenitor cell trafficking is regulated by hypoxic gradients through HIF-1 induction of SDF-1. Nat Med 2004; 10: 858–864.

    CAS  Article  Google Scholar 

  54. 54

    Aghi M, Cohen KS, Klein RJ, Scadden DT, Chiocca EA . Tumor stromal-derived factor-1 recruits vascular progenitors to mitotic neovasculature, where microenvironment influences their differentiated phenotypes. Cancer Res 2006; 66: 9054–9064.

    CAS  Article  Google Scholar 

  55. 55

    Spaeth EL, Dembinski JL, Sasser AK, Watson K, Klopp A, Hall B et alMesenchymal stem cell transition to tumor-associated fibroblasts contributes to fibrovascular network expansion and tumor progression. PLoS One 2009; 4: e4992.

    Article  Google Scholar 

  56. 56

    Crawford Y, Kasman I, Yu L, Zhong C, Wu X, Modrusan Z et alPDGF-C mediates the angiogenic and tumorigenic properties of fibroblasts associated with tumors refractory to anti-VEGF treatment. Cancer Cell 2009; 15: 21–34.

    CAS  Article  Google Scholar 

  57. 57

    Dominici M, Le Blanc K, Mueller I, Slaper-Cortenbach I, Marini F, Krause D et alMinimal criteria for defining multipotent mesenchymal stromal cells. The International Society for Cellular Therapy position statement. Cytotherapy 2006; 8: 315–317.

    CAS  Article  Google Scholar 

  58. 58

    Colter DC, Class R, DiGirolamo CM, Prockop DJ . Rapid expansion of recycling stem cells in cultures of plastic-adherent cells from human bone marrow. Proc Natl Acad Sci USA 2000; 97: 3213–3218.

    CAS  Article  Google Scholar 

  59. 59

    Rebelatto CK, Aguiar AM, Moretao MP, Senegaglia AC, Hansen P, Barchiki F et alDissimilar differentiation of mesenchymal stem cells from bone marrow, umbilical cord blood, and adipose tissue. Exp Biol Med (Maywood) 2008; 233: 901–913.

    CAS  Article  Google Scholar 

  60. 60

    Honczarenko M, Le Y, Swierkowski M, Ghiran I, Glodek AM, Silberstein LE . Human bone marrow stromal cells express a distinct set of biologically functional chemokine receptors. Stem Cells 2006; 24: 1030–1041.

    CAS  Article  Google Scholar 

  61. 61

    Askari AT, Unzek S, Popovic ZB, Goldman CK, Forudi F, Kiedrowski M et alEffect of stromal-cell-derived factor 1 on stem-cell homing and tissue regeneration in ischaemic cardiomyopathy. Lancet 2003; 362: 697–703.

    CAS  Article  Google Scholar 

  62. 62

    Sutendra G, Dromparis P, Wright P, Bonnet S, Haromy A, Hao Z et alThe role of nogo and the mitochondria-endoplasmic reticulum unit in pulmonary hypertension. Sci Transl Med 2011; 3: 88ra55.

    CAS  Article  Google Scholar 

  63. 63

    Douwes DPB, Hogendoorn PC, Kuipers-Dijkshoorn N, Prins FA, van Duinen SG, Taschner PE et alSDHD mutations in head and neck paragangliomas result in destabilization of complex II in the mitochondrial respiratory chain with loss of enzymatic activity and abnormal mitochondrial morphology. J Pathol 2003; 201: 480–486.

    Article  Google Scholar 

  64. 64

    Tomlinson IP, Alam NA, Rowan AJ, Barclay E, Jaeger EE, Kelsell D et alGermline mutations in FH predispose to dominantly inherited uterine fibroids, skin leiomyomata and papillary renal cell cancer. Nat Genet 2002; 30: 406–410.

    CAS  Article  Google Scholar 

  65. 65

    Sutendra G, Dromparis P, Bonnet S, Haromy A, McMurtry MS, Bleackley RC et alPyruvate dehydrogenase inhibition by the inflammatory cytokine TNFalpha contributes to the pathogenesis of pulmonary arterial hypertension. J Mol Med 2011; 89: 771–783.

    CAS  Article  Google Scholar 

  66. 66

    Semenza GL . Targeting HIF-1 for cancer therapy. Nat Rev Cancer 2003; 3: 721–732.

    CAS  Article  Google Scholar 

  67. 67

    Bhattacharyya A, Chattopadhyay R, Hall EH, Mebrahtu ST, Ernst PB, Crowe SE . Mechanism of hypoxia-inducible factor 1 alpha-mediated Mcl1 regulation in Helicobacter pylori-infected human gastric epithelium. Am J Physiol Gastrointest Liver Physiol 2010; 299: G1177–G1186.

    CAS  Article  Google Scholar 

  68. 68

    Lu CW, Lin SC, Chen KF, Lai YY, Tsai SJ . Induction of pyruvate dehydrogenase kinase-3 by hypoxia-inducible factor-1 promotes metabolic switch and drug resistance. J Biol Chem 2008; 283: 28106–28114.

    CAS  Article  Google Scholar 

  69. 69

    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: 177–185.

    Article  Google Scholar 

  70. 70

    Luo W, Hu H, Chang R, Zhong J, Knabel M, O'Meally R et alPyruvate kinase M2 is a PHD3-stimulated coactivator for hypoxia-inducible factor 1. Cell 2011; 145: 732–744.

    CAS  Article  Google Scholar 

  71. 71

    Francia G, Emmenegger U, Kerbel RS . Tumor-associated fibroblasts as ‘Trojan Horse’ mediators of resistance to anti-VEGF therapy. Cancer Cell 2009; 15: 3–5.

    CAS  Article  Google Scholar 

  72. 72

    Bowker-Kinley MM, Davis WI, Wu P, Harris RA, Popov KM . Evidence for existence of tissue-specific regulation of the mammalian pyruvate dehydrogenase complex. Biochem J 1998; 329 (Pt 1): 191–196.

    CAS  Article  Google Scholar 

  73. 73

    Knoechel TR, Tucker AD, Robinson CM, Phillips C, Taylor W, Bungay PJ et alRegulatory roles of the N-terminal domain based on crystal structures of human pyruvate dehydrogenase kinase 2 containing physiological and synthetic ligands. Biochemistry 2006; 45: 402–415.

    CAS  Article  Google Scholar 

  74. 74

    Li J, Kato M, Chuang DT . Pivotal role of the C-terminal DW-motif in mediating inhibition of pyruvate dehydrogenase kinase 2 by dichloroacetate. J Biol Chem 2009; 284: 34458–34467.

    CAS  Article  Google Scholar 

  75. 75

    Babu E, Ramachandran S, Coothankandaswamy V, Elangovan S, Prasad PD, Ganapathy V et alRole of SLC5A8, a plasma membrane transporter and a tumor suppressor, in the antitumor activity of dichloroacetate. Oncogene 2011; 30: 4026–4037.

    CAS  Article  Google Scholar 

Download references


This study was funded by grants from the Canadian Institutes for Health Research (CIHR) and Alberta Innovates Health Solutions (AIHS) to EDM. We would like to thank Dr Gregg Semenza for his help, providing materials and advice.

Author information



Corresponding author

Correspondence to E D Michelakis.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies the paper on the Oncogene website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Sutendra, G., Dromparis, P., Kinnaird, A. et al. Mitochondrial activation by inhibition of PDKII suppresses HIF1a signaling and angiogenesis in cancer. Oncogene 32, 1638–1650 (2013).

Download citation


  • mitochondria
  • angiogenesis
  • hypoxia-inducible-factor
  • stem cell
  • tumor perfusion
  • vascularity

Further reading


Quick links