PDF Clinical Trial Publication U of A 2010

CANCER Glioblastoma Metabolic Modulation by Dichloroacetate 

E. D. Michelakis1, G. Sutendra1, P. Dromparis1, L. Webster1, A. Haromy1, E. Niven2, C. Maguire2, T.-L. Gammer1, J. R. Mackey3, D. Fulton3, B. Abdulkarim3, M. S. McMurtry3, and K. C. Petruk 4 

(Released on May 12, 2010, Volume 2, Issue 31, Page 31ra34) 

Solid tumors, such as the aggressive primary brain tumour glioblastoma multiforme, become resistant to cell death, in part because cytoplasmic glycolysis replaces mitochondrial oxidative phosphorylation in these tumors. As a result of this metabolic reorganization, the mitochondria become hyperpolarized. We investigated whether this cancer-specific metabolic and mitochondrial reprogramming in glioblastoma could be reversed by the small-molecule and orphan medication dichloroacetate (DCA). 49 individuals’ newly isolated glioblastomas displayed mitochondrial hyperpolarization, which DCA quickly corrected. 

In a different study, we prospectively collected baseline and serial tumor tissue from five patients with glioblastoma, created patient-specific glioblastoma cell lines and putative glioblastoma stem cells (CD133+, nestin+ cells), and administered oral DCA to each patient for up to 15 months. In vitro and in vivo, DCA depolarized mitochondria, elevated mitochondrial reactive oxygen species, and triggered apoptosis in putative GBM stem cells and GBM cells. Additionally, DCA treatment enhanced p53 activation, decreased hypoxia-inducible factor-1a, and reduced angiogenesis both in vivo and in vitro. There was no hematologic, hepatic, renal, or cardiac toxicity; rather, a dose-dependent, reversible peripheral neuropathy was the dose-limiting hazard. 

At a dose that did not result in peripheral neuropathy and at serum concentrations of DCA adequate to block the drug’s target enzyme, pyruvate dehydrogenase kinase II, which was abundantly expressed in all glioblastomas, there were signs of clinical effectiveness. When treating glioblastoma, metabolic regulation may be a suitable therapeutic strategy. INTRODUCTION An aggressive primary brain tumor called glioblastoma multiforme (GBM) responds to authorized treatments very poorly (1). After debulking surgery, temozolomide (TMZ) chemotherapy combined with radiation therapy (RT) results in an increase in median survival from 12.1 months with RT alone to 14.6 months (1). After RT and TMZ, the tumor progresses on average in just 6.9 months (1). 

The progression-free survival and TMZ response are significantly worse in recurrent gliomas (2). GBMs have exceptional molecular and genetic variability and are highly vascular tumors (1). Molecular heterogeneity should be overcome, angiogenesis should be inhibited, the blood-brain barrier should be crossed, and there should be little systemic toxicity in the ideal treatment. We predicted that the orphan small molecule dichloroacetate (DCA) meets these requirements and would be useful in treating GBM in humans based on our previous findings in animal models (3, 4). Pyruvate dehydrogenase kinase (PDK) is a mitochondrial enzyme that is inhibited by DCA (5). 

DCA increases the ratio of glucose oxidation to glycolysis by inhibiting PDK and activating pyruvate dehydrogenase (PDH), a gatekeeper enzyme that controls the flux of carbohydrates (pyruvate) into the mitochondria (3–5). Pyruvate may finish glycolysis if it stays in the cytoplasm, resulting in the production of lactic acid and the production of two moles of ATP for every glucose molecule. Pyruvate can also enter a number of anaplerotic and amino acid biosynthesis processes. But when PDH is activated, pyruvate can be decarboxylated to acetyl-coenzyme A, enter the Krebs cycle, and finish oxidizing glucose in the mitochondrial matrix, producing up to 36 moles of ATP for every molecule of glucose when oxygen is present. When pyruvate does not enter the mitochondria (such as in unhealthy mitochondria or if PDH is inhibited) or when oxygen is not present, glucose oxidation does not occur. Warburg (6) first shown that an increase in the ratio of cytoplasmic glycolysis to mitochondrial glucose oxidation characterizes the metabolism of cancer cells, even under normoxia.

There is growing interest in metabolism as a target for cancer therapy despite the fact that the mechanism underlying this “Warburg effect” is unknown and whether it is etiologically related to carcinogenesis has not been established (7). (8–11). Cancer cells may be given a growth advantage by the energy switch from mitochondrial glucose oxidation to cytoplasmic glycolysis (11). For instance, lactic acid increases angiogenesis and the breakdown of the interstitial matrix, which facilitates metastasis (11); most glycolytic enzymes also have direct antiapoptotic effects (12); and diminished mitochondrial activity is linked to the regulation of mitochondria-dependent apoptosis (3). A number of the molecular abnormalities that are present in GBM are known to reduce mitochondrial glucose oxidation and enhance cytoplasmic glycolysis (1). 

These abnormalities include the activation of the phosphatidylinositol 3-kinase-AKT or myc pathways or the repression of the p53 pathway (9, 10). Cancer cells have hyperpolarized mitochondria compared to non-cancer cells (3, 13), which is a sign of decreased mitochondrial activity. Although debatable [reviewed in (14)], mitochondrial hyperpolarization may indicate an apoptosis resistance state since it depends in part on the efflux of proapoptotic mediators through the mitochondrial transition pore (MTP) (3, 15). We have demonstrated that DCA, which inhibits PDK and increases 1 transcription, can reverse this condition in cancer cells. Edmonton, Alberta, Canada’s University of Alberta’s Department of Medicine University of Alberta’s T6G 2B7.2 Department of Biomedical Engineering and Diagnostic Imaging, Edmonton, Alberta, Canada University of Alberta’s T6G 2B7. 3 Department of Oncology, Edmonton, Alberta, Canada Neurosurgery Department, University of Alberta, Edmonton, Alberta, Canada, T6G 2B7. *To whom correspondence should be sent is T6G 2B7. Email address: em2@ualberta.ca. www.ScienceTranslationalMedicine.org

RESEARCH ARTICLE 

stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 1 

mitochondrial function is enhanced, mitochondrial hyperpolarization is reversed, and the ratio of glucose oxidation to glycolysis is decreased when downloaded from pyruvate entrance into the mitochondria (3). Therefore, DCA inhibits tumor growth both in vitro and in vivo while having no effect on healthy mitochondria and tissues (3, 16–20). The formation of mitochondrial reactive oxygen species (mROS), primarily superoxide, increases along with an increase in mitochondrial respiration. H2O2, a relatively stable mROS that can penetrate cell structures other than the mitochondria, can be produced when superoxide is dismutated. H2O2, for instance, can open plasma membrane redox-sensitive voltage-dependent potassium channels and, at least in some tissues, encourage a reduction in intracellular calcium (3, 4). 

The oxidized target p53 is another example of a redox-sensitive target (21, 22). In GBM, the p53 axis is blocked, which increases the proliferative status of GBM cells (1). Due to their competition for the same cotranscription factor, p53 and HIF-1a inhibit hypoxia-inducible factor-1a (HIF-1a)-stimulated transcription (23, 24). HIF-1a boosts the expression of PDK, many glycolytic enzymes, and glucose transporters, maintaining the glycolytic phenotype (25, 26). HIF-1a also promotes angiogenesis by upregulating the production of vascular endothelial growth factor (VEGF). Normoxic HIF-1a activation may also promote angiogenesis. Inhibited mitochondria may transmit pseudohypoxic redox signals and activate HIF-1a even in normoxia because mitochondria are crucial oxygen sensors (27). (28–30). Additionally, because a-ketoglutarate is a cofactor for the prolyl hydroxylation event that breaks down HIF-1a, a decrease in this Krebs’ cycle direct product may also encourage HIF activation (30). 

We postulated that DCA, which can pass the blood-brain barrier when taken orally, would inhibit the growth of GBM in vivo. In addition, we proposed three additional ways for this to happen: I reversing the glycolytic phenotype and normalizing DYm, which would encourage mitochondria-dependent apoptosis; (ii) raising mROS and encouraging p53 activation; and (iii) raising a-ketoglutarate levels. The latter two outcomes would result in a reduction in VEGF, suppression of angiogenesis, and inhibition of HIF-1a. RESULTS DCA’s effects on 49 newly isolated GBM tumors’ mitochondria We examined 49 recently removed consecutive primary GBMs (60% male, 48 11 years) to ascertain whether human GBM could be a target for metabolic treatment with DCA. 

Along with the clinical and neuropathology data, we used Fig. 1 to confirm the diagnosis of GBM. PDK and DYm in GBM and healthy brain. Freshly removed human GBM tissue’s DYm and mROS levels were compared to normal brain tissue taken after an epilepsy surgery (Figure 1A) (control). In comparison to GBM or noncancer cells treated with DCA, the DYm-sensitive dye TMRM accumulated at a higher concentration in untreated GBM cells. DCA only affects GBM cells; non-cancerous brain tissue is unaffected. In the DCA-treated tissue compared to untreated tissue, the concentration of the mitochondria-specific dye mitoSOX increased. P 0.001 and P 0.05, respectively, as compared to the GBM vehicle. Arbitrary fluorescence units, or AFUs. PDKII expression in GBM and healthy brain as determined by confocal immunohistochemistry (B). 

A blue dye called 4′,6-diamidino-2- phenylindole is used to stain the nuclei (DAPI). Only the tissue surrounding tumors expresses PCNA. Arrows indicate cells that express both PCNA and PDKII. (C) Human GBM tumors express more of the PDK isoenzyme PDKII (which has the lowest Ki for DCA), but not PDKI, than does healthy brain tissue. These immunoblots display the positive control as P. (cell lysate with high expression of PDK, provided commercially in the antibody kit). Different tumors have varying levels of PDK, which may not accurately predict the enzyme’s activity in vivo (n = 3 for control; n = 8 for GBM tissue). 

Research article available at www.ScienceTranslationalMedicine.org, *P 0.05. stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 2 GFAP (glial fibrillary acidic protein) 

expression was seen in immunohistochemistry, but neither bIII-tubulin nor oligodendrocyte markers were present (fig. S1). In contrast to noncancer brain tissues acquired before epilepsy surgery, DYm was elevated in the recently isolated GBMs (n = 3). (Fig. 1A). In GBM but not in healthy brain tissue, DCA but not the vehicle (normal saline) led to mitochondrial depolarization. DCA elevated GBM mROS as well (Fig. 1A). This revealed that PDK is at least largely responsible for the metabolic and mitochondrial remodeling in GBM, which is partially reversible. According to immunohistochemistry and immunoblots, the response to DCA is consistent with a larger concentration of PDKII [the most widely expressed isoform and the one with the lowest Ki for DCA (31)] in GBM than in noncancer brain tissue (Fig. 1, B and C). 

Proliferating cell nuclear antigen (PCNA) was present in cells with the greatest PDKII concentrations, indicating that these cells were proliferating (Fig. 1B). These findings, which were gathered over a two-year period, improved the case for later giving DCA to patients with GBM (4). Clinical outcomes of DCA in five GBM patients Five patients with primary GBM who had been referred to us by our brain cancer program and whose tissue was still available from the previous debulking surgery were then treated with DCA. Three patients (patients 1 to 3) had recurrent GBM with disease progression after numerous chemotherapies (in addition to the conventional treatment with surgery, RT, and TMZ) and were regarded acceptable for palliative therapy. 

Two additional patients (patients 4 and 5) were newly diagnosed, and following the initial debulking operation, DCA was provided in addition to the conventional treatment of RT and TMZ. In patient 4, a 3-month DCA pretreatment was followed by the addition of RT and TMZ, whereas in patient 5, DCA was started after debulking surgery at the same time as RT and TMZ. Tissue from the previous debulking surgery (before to DCA administration) was compared to the post-DCA therapy tissue to determine whether the patients needed reoperation or autopsy. Table S1 has a clinical summary of them. Pharmacokinetic and pharmacodynamic data are available from the administration of DCA to patients for more than 30 years, mostly for the therapy of inborn defects in mitochondrial metabolism (5, 32–34). After treating patients for a month with a starting dose of 12.5 mg/kg orally twice daily, we increased the dose to 25 mg/kg orally twice daily. When dose-limiting toxicity manifested, we next followed a dose de-escalation approach, reducing the dose by 50%. 

Clinical monitoring of the patients lasted up to 15 months. There was no hematologic, hepatic, renal, or cardiac damage in any of the individuals (table S1). The only harm that was obvious was peripheral neuropathy. Patients’ peripheral neuropathy levels varied, depending on dose. Fig. 2. Effects of DCA in patients with GBM in vivo. (A) T1 gadolinium-enhanced axial MRI scans of patients 1 (left) and 2 (right) collected before to and following DCA therapy (left, midventricular level; right, supraventricular level); and combined PET-MRI images (right). Patient 2’s tumor appears to have resolved after 15 months of treatment with oral DCA as the sole therapeutic agent. After receiving DCA treatment for 9 months in patient 1, the metastatic paraventricular tumor mass regressed.

 Baseline: third month of DCA therapy; after nine months: twelfth month of DCA therapy In patient 1, the primary tumor site, which is visible in fig. S2 but cannot be seen at the level of these images, did not change during this time. (B) Sample micrographs of tissue from patient 3 (clinical details in text) showing tumor proliferation (% PCNA-positive cells) and apoptosis (% TUNEL-positive cells) before and after chronic DCA therapy. After DCA treatment, there is a large rise in apoptosis, a considerable drop in PCNA expression, and a decrease in the number of cells (shown by the number of nuclei, in blue (DAPI)). A minimum of three slides were utilized each experiment, and the percentage of PCNA- or TUNEL-positive cells were counted blindly in eight random fields on each slide (n = 350–400 cells per patient). *P < 0.01. (C) (C) When compared to the baseline tissues from the same patients collected prior to DCA treatment, PDH activity is significantly higher in the GBM tissues from patients treated with DCA. This shows efficient in vivo PDK inhibition in tumor tissue (n = 3 patients). *P < 0.001. www.ScienceTranslationalMedicine.org 

RESEARCH ARTICLE 

stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 3 

neuropathy was reversible, supporting other findings, and it was downloaded from (35–37). None of the patients experienced clinically significant peripheral neuropathy when the dosage was lowered to 6.25 mg/kg orally twice a day (table S1). The initial half-life of DCA is about one hour. DCA blocks its own metabolism, causing serum concentrations to rise and eventually plateau (34). For the first two to three months, DCA plasma trough concentrations in our patients were undetectable, but after that, they reached therapeutic levels. Trough DCA concentrations at a dosage of 6.25 mg/kg orally twice daily for at least three months were 0.44 0.16 mM (mean SD; n = 4). (table S1). These values fall within the same range as the Ki of DCA for PDKII (0.2 mM) and are comparable to those seen in prolonged DCA therapy of persons with mitochondrial abnormalities (34) (31). 

Magnetic resonance imaging (MRI) revealed some signs of radiologic regression in patients 1, 4, and 5. (Fig. 2A and figs. S2 to S4). Despite taking high doses of steroids, Patient 3 had a very large tumor and brain edema at baseline (fig. S5). He also had a low Karnofsky score, and his condition worsened with time. Three months after beginning DCA therapy, he passed away from problems related to brain edema. In month 11 of DCA therapy, Patient 2 required debulking and cyst drainage. After three months of DCA therapy, Patient 4 demonstrated radiologic improvement, at which point further debulking was carried out and RT plus TMZ was administered in addition to DCA. All but patient 3 were still alive and in clinical stability at month 15 of DCA therapy (telephone follow-up). 

The Supplementary Material offers further clinical information. Effects of DCA on primary GBM cell lines, putative GBM-SC generated from DCA-treated patients, and in vivo GBM tumors We performed research using tissues from these five patients and were able to compare tissues from patients 2 to 4 before and after DCA treatment; we only had “before” tissues from patients 1 and 5. Post-DCA GBM tissue in all three patients displayed lower cell density per unit volume, reduced proliferation, and enhanced apoptosis compared to pre-DCA tissue (Fig. 2B), as well as higher tissue PDH enzymatic activity, indicating successful in vivo PDK inhibition (Fig. 2C). 

Potentially responsible for post-treatment resistance and GBM recurrence are putative GBM cancer stem cells (GBM-SCs) (38–43). These cells surround capillaries in niches and are identified as CD133+/Nestin+ GBM-SC (41). In these vascular GBM-SC units, GBM-SC can cause angiogenesis, but their availability to circulating growth factors maintains their molecular stem cell nature (44). Proliferation of GBM-SC is linked to a very bad clinical outcome (42). In all pre-DCA tumors, CD133+/nestin+ GBM-SC expressed PCNA in vivo, demonstrating that they are proliferating, however in patients 2 to 4, the proportion of CD133+/nestin+ cells that expressed PCNA drastically decreased after DCA therapy (Fig. 3A). 

Comparing CD133+ cells to nearby non-GBM-SC in vivo, simultaneous labeling with a CD133 antibody and tetramethyl rhodamine methyl ester (TMRM) revealed that CD133+ cells had the highest DYm (fig. S6). Compared to the histopathology of GBM, only 10% of cells in tumor-derived primary cell lines exhibited both CD133 and nestin, whereas >90% of the cells displayed the mature marker GFAP (but not bIII-tubulin or oligodendrocyte) (fig. S7) (fig. S1). From GBM tumors, we extracted putative GBM-SC and cultivated them with the necessary growth factors (20 ng/ml each of human fibroblast growth factor and human epidermal growth factor). 

These cells generated distinctive neurospheres (Fig. 4 and fig. S7), exhibited extremely high levels of CD133 and nestin expression, and very low levels of mature glial markers, which is a reliable indicator of a bad clinical outcome (43). In newly excised tumors, primary cell lines, GBM-SC isolated from those tumors, as well as differentiated cells produced from GBM-SC, we assessed DYm (Fig. 3B). In Fig. 3, the greatest potential was discovered. Effects of DCA on presumptive GBM-SC and GBM cells in vivo and in vitro. (A) In the three patients for whom we had tissues from both before and after chronic DCA therapy, representative examples of quadruple staining of GBM tissues and summary data demonstrate that the percentage of putative GBM-SC (CD133+ and nestin+ cells) expressing PCNA is decreased by chronic DCA therapy in vivo. 

A minimum of three slides were utilized each experiment, and the percentage of PCNA- or TUNEL-positive cells were counted in eight random fields on each slide (n = 70–75 cells per group). *P < 0.05. (B) TMRM uptake was measured in freshly excised GBM tissue (pre-DCA treatment tumors in patients 1 to 5), primary GBM cell lines created from the same tumors, putative GBM-SC (CD133+ cells removed from these tumors and expanded in appropriate media), and secondary GBM cell lines created from the GBM-SC differentiation using the proper media and growth factor adjustments. Shown are representative photos on the left and mean summary statistics on the right. Five people each group made up the sample size for the tissue experiments, representing tissues from five patients (see the Supplementary Material). 

For the cell experiments, there were 70 participants per group (see the Supplementary Material). DCA was administered to the tissues for 90 minutes at a concentration of 5 mM, and to the cells for 24 hours at a concentration of 0.5 mM. (normal saline was used for vehicle). In Fig. 4 and fig. S7, the development of neurospheres from GBM-SC is also depicted. Compared to baseline, *P 0.05; compared to primary GBM cells, #P 0.01.

 www.ScienceTranslationalMedicine.org

RESEARCH ARTICLE 

stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 4 

retrieved from the alleged GBM-SC. The mitochondrial potential of the main and secondary GBM cells produced from GBM-SC (15 days of differentiation) was comparable to that of the parent tumors. In all cell types, DCA (0.5 mM for 24 hours) reduced potential. Although the exact origin of the increased DYm in cancer (3, 13) is still unknown, it has been suggested that a critical glycolytic enzyme called hexokinase II (HXKII) moving from the cytoplasm to the outer mitochondrial membrane may be a contributing factor (45, 46). There, the voltage-dependent anion channel (a part of the MTP) may be bound to and inhibited by HXKII, raising the DYm and the apoptotic threshold. 

The cancer DYm is reduced and the resistance to apoptosis is reversed by inhibiting this translocation (45, 46). Our main cell lines derived from pre-DCA malignancies demonstrated a sustained HXKII mitochondrial translocation, which may be the cause of the rise in DYm. After DCA treatment, HXKII translocation was absent in primary cell lines from tumors (fig. S8), which is consistent with the idea that DCA suppressed glycolysis and reduced DYm. The GBM cell lines produced from patients 2 to 4 had significant levels of PDKII, just like in the tumors, but they also expressed the other known isoenzymes (fig. S9A). 

The percentage of cells containing GBM-SC markers reduced to a level corresponding to that of the primary cell lines (10%) when GBM-SCs were permitted to differentiate into secondary GBM cell lines. However, when allowed to develop in the presence of DCA (0.5 mM), the percentage of cells expressing GBM-SC markers fell to just 5%. (fig. S7). In fact, DCA promoted apoptosis in GBM-SC in vitro as well as in GBM primary cell lines (Fig. 4 and fig. S9B) (fig. S9C). Combination therapy makes sense given that DCA plus TMZ boosted apoptosis in GBM-SCs much more (Fig. 4 and fig. S9B). The colocalization of nestin, CD133, and terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining in the post-DCA treatment tumors demonstrated that GBM-SC apoptosis also occurred in vivo (Fig. 4 and fig. S9D). 

Effects of DCA on angiogenesis and the GBM-SC microvessel unit in vivo and in vitro According to a publication using von Willebrand factor (vWF) staining, niches of GBM-SC were discovered surrounding microvessel beds in the untreated, pre-DCA tissues (41). Because microvascular endothelial cells’ death was also elevated in addition to GBM-SC (Fig. 4 and fig. S9D), suggesting a potential suppression of angiogenesis, this GBM-SC microvessel unit (44) was eliminated by DCA treatment. In fact, the post-DCA therapy tumors’ lower vWF staining indicated that there was a reduction in vascularity (Fig. 5A). Before DCA, HIF-1a was substantially expressed or activated (nuclear localization), whereas after DCA, HIF-1a was suppressed (Fig. 5B). 

When compared to the tumors from the same patients before DCA treatment, those from patients 2 to 4 had significantly higher levels of mROS in vivo (superoxide evaluated by mitoSOX) (Fig. 5C). Acute DCA elevated mROS in the pre-DCA tumors to levels seen in the tumors after DCA therapy. Acute DCA, on the other hand, barely slightly elevated mROS in the post-DCA tumors, indicating an almost maximal impact in vivo. Additionally, DCA elevated mROS in GBM-SC (Fig. 5C). Low ROS levels in cancer stem cells may indicate apoptosis resistance, and therapies that boost ROS in cancer stem cells may be more effective (47). Although we looked at mROS (mitochondrial superoxide), there is debate about whether activating HIF-1a in cancer leads to an overall rise or fall in ROS [reviewed in (28)]. 

We also measured whole-cell H2O2 using a different method. H2O2 in GBM cells was elevated by DCA in a dose-dependent manner (Fig. 6). Additionally, in a dose-dependent way, DCA raised intracellular a-ketoglutarate concentrations (Fig. 6). This is plausible given that a-ketoglutarate is a byproduct of the Krebs cycle and that glucose oxidation increases after PDH activation (3). The electron donors that feed into the electron transport chain during respiration are created via the Krebs cycle. Accordingly, DCA promoted an overall increase in mitochondrial activity by raising respiration rates in GBM cells by 44 4% (mean SEM; n = 3; P 0.05). H2O2 and a-ketoglutarate levels are rising (Fig. 4). Both presumptive GBM-SC and microvascular endothelial cells undergo apoptosis when exposed to DCA. (Left) In vitro studies on GBM-SC demonstrate apoptosis in response to DCA (0.5 mM for 72 hours) alone and DCA with TMZ (100 mM). Apoptosis is evaluated by the TUNEL assay. (Right) The tissue from tumors, both before and after DCA therapy, demonstrates apoptosis in GBM-SC. Examples that serve as good examples are provided, and fig. S9, B and D presents summary mean data. In GBM-SC developing in vitro, neurosphere development is evident. In tumor tissue, simultaneous quadruple labeling reveals that the putative GBM-SCs express significant levels of PDKII. Additionally, by labeling them with the endothelial marker vWF, they can be seen to create niches around capillary networks. The post-DCA tissues show increased TUNEL labeling in both GBM-SC and endothelial cells, in contrast to the pre-DCA tissues where there is no TUNEL staining. 

ASSESSMENT ARTICLE

 www.ScienceTranslationalMedicine.org 

stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 5 the decrease in HIF-1a activity (Fig. 5B), which is supported by the dose-dependent reduction in VEGF production by GBM cells, can be used to explain (Fig. 6). These findings support the decrease in angiogenesis observed in vivo (Figs. 4 and 5) and imply that GBM cells communicate with endothelial cells in a paracrine fashion. We employed the common method of human endothelial cell tube creation in Matrigel to investigate whether DCA can directly decrease angiogenesis. In vitro angiogenesis was directly inhibited by DCA in a dose-dependent manner under physiological mild hypoxia (fig. S10). Patients 2 to 4’s post-DCA therapy tumors displayed enhanced nuclear translocation activity of the mROS-sensitive p53, which was further supported by increased activity and abundance of its downstream target p21 (fig. S11).

The reduction in HIF-driven transcription can also be attributed to these effects on p53 or p21, which are similarly consistent with the antiproliferative and proapoptotic actions of DCA in GBM (fig. S12).

DISCUSSION DCA, a metabolic regulator, has anticancer properties in both rodents and cultured cells (3, 16–20). The utility of DCA in patients with GBM has now been demonstrated. A prolonged radiologic stability or tumor regression was linked to DCA treatment in some GBM patients, and the drug generally showed favorable safety characteristics. This preliminary, first-in-human report offers justification for further research with this generic small chemical in GBM patients. Our findings suggest that metabolic intervention is a good choice for treating GBM. In GBM tumors and cell lines, PDKII is the target of DCA, and DCA can decrease its activity in vivo. In line with the metabolic remodeling (Warburg effect) and associated apoptosis resistance that distinguish GBM and other solid tumors, GBM is characterized by mitochondrial hyperpolarization (11).

In GBMs that develop from lowergrade gliomas (secondary GBMs), changes in the genes for cytoplasmic and mitochondrial isocitrate dehydrogenases have been reported (48), but the mechanism by which these mutations connect to carcinogenesis is yet unknown (49, 50). The mitochondrial remodeling in our patients’ primary GBMs was at least partially reversible with DCA, indicating that it was not caused by an irreparable malfunction. As a result of putative GBM-SC having the most hyperpolarized mitochondria both in vivo and in vitro, we also demonstrate that putative GBM-SC may experience the same metabolic and mitochondrial remodeling, but to a greater extent. Both in vitro and in vivo, DCA’s induction of apoptosis in GBM-SC caused by the reversal of this mitochondrial remodeling.

Despite the fact that DCA induces apoptosis to a small degree Fig. 5. Effect of DCA on in vivo angiogenesis, HIF-1a, and mROS in GBM. (A) Heavy vWF expression and enhanced vascularity are observed in pre-DCA therapy tumor tissues (representative micrograph and summary data from patients 2 to 4). In contrast, the capillary network and vascularity are drastically reduced in post-DCA tumor tissues. In the post-DCA micrograph, a single little vessel can be detected, which is typical of various regions inside these tumors. A minimum of three slides per subject each experiment were used to calculate the mean vWF arbitrary fluorescence unit. *P < 0.01. (B) In pre-DCA tumors, active (nuclear) HIF-1a is highly expressed. HIF-1a is confined to the nuclei in the higher-magnification insets. HIF-1a expression is shown to be significantly downregulated in post-DCA malignancies. HIF-1a is either absent or dispersed in nonnuclear regions of the cell at higher magnifications. By detecting the fluorescence signal contained within the nucleus, nuclear HIF-1a was quantified. Also included are the means for patients 2 through 4. In each experiment, a minimum of three slides per patient were used to measure the mean HIF-1a arbitrary fluorescence unit, which was derived from eight random fields each slide. *P < 0.01. (C) (C) Acute DCA (5 mM for 90 min) was applied to newly separated tumors from patients 2 to 4 (both before and after DCA therapy), as opposed to a vehicle (normal saline). Because these patients are receiving continuous oral DCA therapy, the level of mROS (superoxide measured by mitoSOX) in the post-DCA tissues is significantly larger than it was in the pre-DCA tissues at baseline (before acute DCA). Acute DCA significantly increased mROS in pre-DCA tumors but only slightly in post-DCA tumors, indicating that oral DCA therapy has almost its full potential in vivo. The mean pooled data are displayed and analyzed as in Fig. 1B. In pre-DCA tumors, *P 0.05, # P 0.01 compared to vehicle Acute DCA (0.5 mM for 90 min) similarly raised the level of mROS in cultured putative GBM-SC neurospheres. The pre-DCA treated GBM tissues were used to isolate GBM-SCs. (n = 40 clusters per group) The mean mitoSOX arbitrary fluorescence unit per neurosphere is displayed. *P < 0.05.

ASSESSMENT ARTICLE

www.ScienceTranslationalMedicine.org

stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 6

It is relatively selective, sparing noncancer cells (as compared to cytotoxic drugs), and because it engages GBM-SC, it might have a longer-lasting therapeutic effect. Patient 3 died and patient 4 experienced a recurrence during the first 3 months when DCA had not acquired sustained therapeutic values (as indicated by the fact that trough plasma concentrations were undetectable) (as shown by the fact that trough plasma concentrations were undetectable). Patients may therefore receive inadequate care when receiving DCA at first, increasing their likelihood of developing the condition. DCA might be able to enhance the effectiveness of conventional treatments by lowering resistance to mitochondrialdependent apoptosis. Indeed, the long-term advantages felt by patient 5 may be due to the effects of DCA with TMZ on GBMSC apoptosis.

PDKII small interfering RNA mimics the actions of DCA on cancer cells, and DCA has no further effects beyond those detected following PDKII knockdown (3). This shows that the suppression of PDKII, an enzyme that is present in higher amounts in GBM, is the mechanism via which DCA exerts its anticancer effects. DCA restores the elevated glycolysis to glucose oxidation ratio in cancer cells, improving mitochondrial activity, via blocking PDKII (3). Our most recent research shows that DCA increases a-ketoglutarate levels and GBM cell respiration, which is consistent with this mechanism. The effects of DCA on cancer cell metabolism and apoptosis in vivo, however, may be mediated by other processes as opposed to the tightly controlled circumstances of cell culture.

Our data on HXKII, mROS, HIF-1a, p53, and p21 are consistent with DCA-induced inhibition of angiogenesis, induction of apoptosis, and suppression of proliferation in both GBM and GBM-SC, as summarized in the Supplementary Material and fig. S12. This is in addition to the effects of DCA on cancer cells previously described (3). Other than the previously mentioned dosage-dependent, reversible nondemyelinating neurotoxicity (32, 34), which was minimal or nonexistent at the 6.25 mg/kg oral, twice-daily dose, DCA exhibited no other obvious toxicities. This dosage produced plasma concentrations that were necessary for PDK inhibition as well as biologic and clinical effectiveness (31). Due to the small number of treated patients in our study, it is impossible to draw any clear conclusions about DCA’s efficacy as a treatment for GBM. Our research backs up the necessity for more DCA trials in GBM, with a focus on combination therapy methods. Indicating a novel strategy in the treatment of this incurable illness, GBM may also be sensitive to other medications in the newly developed family of metabolic modulators.

ADDITIONAL INFORMATION

https://sciencetranslationalmedicine.org/cgi/content/full/2/31/31ra34/DC1

Resources and Procedures Findings Discussion Molecular characterisation of GBM tumors, Figure S1. Figure S2. Tumor response evolution in patient 1. Figure S3. Tumor response evolution in patient 4. Figure S4. Tumor response evolution in patient 5. Figure S5. GBM MRI from patient 3. Figure S6. Mitochondrial membrane potential in GBM-SC from recently resected GBM tissue. Characterization of primary GBM cells and GBM-SC is shown in Fig. S7. HXKII in GBM cells taken from patients before and after continuous DCA treatment is shown in Fig. S8.

Effects of DCA therapy on vascular apoptosis and GBM-SC are shown in Fig. S9. Fig. S10. In vitro effects of DCA on angiogenesis Effects of DCA therapy on p53 and p21 activity in vivo are shown in Fig. S11. A potential thorough mechanism for DCA’s anticancer actions in GBM is shown in Figure S12 (see Supplementary Discussion). Table S1. Clinical and laboratory characteristics of five GBM patients both before and after DCA therapy. References QUOTATIONS AND NOTES 1. P. Y. Wen and S. Kesari, Adult malignant gliomas. 359, 492-507 N. Engl. J. Med (2008). The following individuals: 2. W. K. Yung, R. E. Albright, J. Olson, R. Fredericks, K. Fink, M. D. Prados, M. Brada, A. Spence, R. J. Hohl, W. Shapiro, M. Glantz, H. Greenberg, R. G. Selker, N. A. Vick, R. Rampling, H. Friedman, P. Phillips, J. Bruner, N. a phase II trial comparing procarbazine and temozolomide in those with glioblastoma multiforme at first relapse. 83, 588-593, Br. J. Cancer (2000). 3. L. Puttagunta, S. Bonnet, G. Harry, K. Hashimoto, C. J. Porter, M. Andrade, B. Thebaud, E. D. Michelakis, S. L. Archer, J. Allalunis-Turner, A. Haromy, C. Beaulieu, R. Thompson, C. T. Lee, G. D. Lopaschuk, Cancer suppresses a mitochondria-K+ channel axis, and normalizing it encourages apoptosis and slows the spread of the disease. 11; Cancer Cell; 35–51 (2007). 4. Dichloroacetate (DCA) as a potential metabolic targeting therapy for cancer. E. D. Michelakis, L. Webster, and J. R. Mackey. Br. Journal of Cancer 99, 989–994 (2008). The Pharmacology of Dichloroacetate by P. W. Stacpoole. 38, 1124-1144 in Metabolism (1989). 6. O. Warburg, “On the Structural Changes in Tumors” (Constable, London, 1930). The Warburg hypothesis fifty years later, S. Weinhouse. Cancer Res. Clin. Oncol. 87, 115–126 Z. Krebsforsch. Klin. Onkol (1976). Understanding the Warburg effect: The metabolic requirements of cell proliferation, by M. G. Vander Heiden, L. C. Cantley, and C. B. Thompson. 1029-1033 in Science 324 (2009). 9. T. W. Mak, J. G. Pan A new era for metabolic targeting as an anti-cancer strategy? Science STKE 2007, p. (2007). 10. J. W. Kim, C. V. Dang, The Warburg effect and cancer’s molecular sweet appetite. 66, 8927-8930 (Cancer Research) (2006). R. J. Gillies, R. A. Gatenby, Why is aerobic glycolysis more prevalent in cancers? National Cancer Review 4, 891-899 (2004). 12. C. V. Dang, J. W. Kim, many functions of glycolytic enzymes. Biochem. Trends 30, 142-150 (2005). L. B. Chen, potential of the mitochondrial membrane in live cells. 4, 155-181 Annu. Rev. Cell Biol (1988). 14. L. Galluzzi, C. Brenner, G. Kroemer, permeabilization of the mitochondrial membrane during cell death. Biology Reviews 87, 99–163 (2007). 15. The Mitochondrion in Apoptosis: How Pandora’s Box Opens, by N. Zamzami and G. Kroemer 67–71 in Nat. Rev. Mol. Cell Biol (2001). Figure 6 shows how DCA affects GBM cells’ levels of H2O2, a-ketoglutarate, and VEGF. DCA increases H2O2 generation and intracellular a-ketoglutarate concentrations in cultured GBM cells while decreasing secreted VEGF in a dose-dependent manner (n = 3 studies). # P 0.01, compared to 0.5 mM DCA; *P 0.01, compared to control.

ASSESSMENT ARTICLE

www.ScienceTranslationalMedicine.org

stm.sciencemag.org on May 13, 2010 Vol 2 Issue 31 31ra34 7

retrieved from 16. I. Papandreou, P. D. Sutphin, N. C. Denko, R. A. Cairns, In solid tumors, metabolic targeting of hypoxia and HIF1 can improve cytotoxic treatment. 104, 9445-9450 Proc. Natl. Acad. Sci. U.S.A. (2007). 17. Dichloroacetate (DCA) sensitizes both wild-type and over-expressing Bcl-2 prostate cancer cells to radiation in vitro, according to W. Cao, S. Yacoub, K. T. Shiverick, K. Namiki, Y. Sakai, S. Porvasnik, C. Urbanek, and C. J. Rosser. 1223–1231 in Prostate 68 (2008). 18. S. Dhar and S. J. Lippard created Mitaplatin, a powerful combination of the orphan medicine dichloroacetate and cisplatin. 106, 22199–22204 Proc. Natl. Acad. Sci. U.S.A. (2009). 19. A. C. Blackburn, P. G. Board, M. Fadia, J. E. Dahlstrom, R. C. Sun, Dichloroacetate suppresses the development of metastatic breast cancer cells both in vitro and in vivo by reversing the glycolytic phenotype. 120, 253-260 Breast Cancer Res. Treat (2010). 20. G. S. Huggins, M. Debidda, N. C. Munshi, I. De Vivo, J. Y. Wong, Endometrial cancer cells undergo apoptosis when exposed to dichloroacetate. 109, 394–402 Gynecol. Oncol (2008). 21. J. Ye, M. Ding, X. Shi, S. Wang, S. S. Leonard, The function of hydroxyl radical as a messenger in p53 activation caused by Cr(VI). Cell Physiology and Am. J. Physiol. 279, C868-C875 (2000). 22. V. Castranova, V. Vallyathan, X. Shi, Y. Lu, M. Costa, Y. Rojanasakul, J. Li, J. Ye, S. S. Leonard, H. M. Shen, L. Butterworth, and Z. Zhang. Vanadate stimulates p53 transactivation by hydrogen peroxide and causes apoptosis. 275, 3251–32522, J. Biol. Chem (2000). 23. “P300 relieves p53-evoked transcriptional suppression of hypoxia-inducible factor-1,” by T. Schmid, J. Zhou, R. Köhl, and B. Brüne (HIF-1). Biochemical Journal 380, 289–295 (2004). 24. M. V. Blagosklonny, W. G. An, L. Y. Romanova, J. Trepel, T. Fojo, and L. Neckers found that p53 suppresses transcription that is triggered by hypoxia-inducible factor. Biological Chemistry 273, 11995-11998 (1998). 25. “HIF-1-mediated expression of pyruvate dehydrogenase kinase: A metabolic switch necessary for cellular adaptation to hypoxia,” by J. W. Kim, I. Tchernyshyov, G. L. Semenza, and C. V. Dang 177–185 Cell Metab. 3, (2006). 26. “The metabolism of tumors”: 70 years later, by G. L. Semenza, D. Artemov, A. Bedi, Z. Bhujwalla, K. Chiles, D. Feldser, E. Laughner, R. Ravi, J. Simons, P. Taghavi, and H. Zhong. 251-260 Novartis Found. Symp (2001). 27. S. L. Archer, K. J. Buckler, J. López-Barneo, acute oxygen-detecting systems. 2042-2055 N. Engl. J. Med (2005). N. C. Denko, “Hypoxia, HIF1, and glucose metabolism in the solid tumor.” National Cancer Review 8, 705-713 (2008). Targeting HIF-1 for cancer therapy, G. L. Semenza. National Cancer Review 3, 721-732 (2003). Cell-permeating a-ketoglutarate derivatives cure pseudohypoxia in succinate dehydrogenase-deficient cells, according to research by E. D. MacKenzie, M. A. Selak, D. A. Tennant, L. J. Payne, S. Crosby, C. M. Frederiksen, D. G. Watson, and E. Gottlieb. Cellular Biology 27, 3282-3289 (2007). 31. P. Wu, R. A. Harris, M. M. Bowker-Kinley, W. I. Davis, K. M. Popov, evidence that the mammalian pyruvate dehydrogenase complex is regulated in certain tissues. Biochemical Journal 329, 191–196 (1998). The following individuals: 32. P. W. Stacpoole, D. S. Kerr, C. Barnes, S. T. Bunch, P. R. Carney, E. M. Fennell, N. M. Felitsyn, R. L. Gilmore, M. Greer, G. N. Henderson, A. D. Hutson, R. E. Neiberger, R. G. O’Brien, L. A. Perkins, R. G. Dichloroacetate has undergone a controlled clinical trial for the treatment of congenital lactic acidosis in children. 117, 1519–1531, Pediatrics (2006). 33. E. M. Harman, R. G. Thomas, A. C. Lorenz, P. W. Stacpoole Lactic acidosis treatment using dichloroacetate. 108, 58–63 Ann Intern Med (1988). 34. T. Langaee, Z. Han, T. L. Kurtz, P. W. Stacpoole, Dichloroacetate’s role in treating disorders caused by mutations in the mitochondria. Drug Delivery Review, 60, 1478–1487. (2008). 35. Dichloroacetate produces toxic neuropathy in MELAS: A randomized, controlled clinical trial, P. Kaufmann, K. Engelstad, Y. Wei, S. Jhung, M. C. Sano, D. C. Shungu, W. S. Millar, X. Hong, C. L. Gooch, X. Mao, J. M. Pascual, M. Hirano, P. W. Stacpoole, S. DiMaur Brain 66, 324–330 (2006). 36. L. R. Gilbert, P. R. Carney, E. Valenstein, D. W. Theriaque, J. J. Shuster, P. W. Stacpoole Dichloroacetate treatment for children with congenital lactic acidosis is being evaluated. e1223–e1228 in Pediatrics 121 (2008). Clinical pharmacology and toxicity of dichloroacetate, by P. W. Stacpoole, G. N. Henderson, Z. Yan, and M. O. James, 37. Environmental Health Perspectives, vol. 106, no. 4, pp. 989–994 (1998). 38. Neural stem cells and the origin of gliomas, N. Sanai, A. Alvarez-Buylla, and M. S. Berger. N Engl J Med 353, 811-822 (2005). 39. M. Valentine, J. E. Rehg, R. J. Gilbertson, C. J. Sherr, M. F. Roussel, T. Uziel, O. Ayrault, C. Calabrese, F. Zindy, Mouse medulloblastomas have genetic mutations, and primary cerebellar granule neuron progenitors can become tumors on their own. 67, 2676–2684. Cancer Research (2007). 40. H. Nakamura, J. Kuratsu, T. Kondo, T. Hide, T. Takezaki, Targets for research and therapy are brain tumor stem cells. 25, 67–72 Brain Tumor Pathology (2008). 41. M. W. Gaber, D. Finklestein, M. Allen, A. Frank, I. T. Bayazitov, S. S. Zakharenko, A. Gajjar, A. Davidoff, R. J. Gilbertson, H. Poppleton, M. Kocak, T. L. Hogg, C. Fuller, B. Hamner, E. Y. Oh, a perivascular habitat for stem cells in brain tumors. 69–82 in Cancer Cell 11, (2007). 42. Cancer stem cell analysis and clinical outcome in patients with glioblastoma multiforme by R. Pallini, L. Ricci-Vitiani, G. L. Banna, M. Signore, D. Lombardi, M. Todaro, G. Stassi, M. Martini, G. Maira, L. M. Larocca, and R. De Maria. Cancer Research 14, 8205-8212 (2008). 43. M. Masterman-Smith, D. R. Laks, K. Visnyei, B. Angenieux, N. M. Orozco, I. Foran, W. H. Yong, H. V. Vinters, L. M. Liau, J. A. Lazareff, P. S. Mischel, T. F. Cloughesy, S. Horvath, and H. I. Kornblum In malignant gliomas, neurosphere development is a standalone predictor of clinical prognosis. Cells of Stem 27, 980-987 (2009). 44. J. N. Rich, R. J. Gilbertson Glioblastoma stem cells and the vascular niche: making a tumor’s bed. National Cancer Review 7, 733-736 (2007). Hexokinase II: The integration of energy metabolism and control of apoptosis, by J. G. Pastorino and J. B. Hoek, 45. Current Medical Chemistry 10, 1535-1551 (2003). 46. N. Shulga, J. B. Hoek, and J. G. Pastorino By phosphorylating voltage-dependent anion channels, glycogen synthase kinase 3b activation inhibits hexokinase II’s ability to attach to mitochondria and increases chemotherapy-induced cytotoxicity. 65, 10545-10554; Cancer Research (2005). In addition to M. Diehn, R. W. Cho, N. A. Lobo, T. Kalisky, M. J. Dorie, A. N. Kulp, D. Qian, J. S. Lam, L. E. Ailles, M. Wong, B. Joshua, M. J. Kaplan, I. Wapnir, F. M. Dirbas, G. Somlo, C. Garberoglio, B. Paz, Reactive oxygen species concentrations and radioresistance in cancer stem cells are related. The Nature 458: 780–783 (2009). 48. IDH1 and IDH2 mutations in gliomas, by H. Yan, D. W. Parsons, G. Jin, R. McLendon, B. A. Rasheed, W. Yuan, I. Kos, I. Batinic-Haberle, S. Jones, G. J. Riggins, H. Friedman, A. Friedman, D. Reardon, J. Herndon, K. W. Kinzler, V. E 360, 765-773 N. Engl. J. Med (2009). Metabolic enzymes as oncogenes or tumor suppressors, C. B. Thompson, 49. 360, 813-815 N. Engl. J. Med (2009). 50. Cancer-associated IDH1 mutations result in the production of 2-hydroxyglutarate, according to L. Dang, D. W. White, S. Gross, B. D. Bennett, M. A. Bittinger, E. M. Driggers, V. R. Fantin, H. G. Jang, S. Jin, M. C. Keenan, K. M. Marks, R. M. Prins, P. S. Ward, K. E. Y Biology 462, 739–744 (2009). 51. Funding: The Hecht Foundation (Vancouver, British Columbia, Canada; E.D.M.), the Canada Institutes for Health Research, the Canada Research Chairs Program (E.D.M.), as well as contributions from the general public to the DCA program, sponsored this work (received and managed by the Regents of the University of Alberta and the Faculty of Medicine). The authors would like to thank the Alberta Health Services for their assistance (D. Gordon, Senior VicePresident, Major Tertiary Hospitals). Contributions of the author: The investigations were planned, the mechanistic experiments were overseen, money was procured, the data were evaluated, and the report was written by E.D.M. All clinical investigations were overseen by K.C.P., who also co-designed the studies and co-wrote the report. All of the mechanistic studies were carried out and the text was revised by G.S. and P.D. All investigations were coordinated by L.W., who also helped with data collection, clinical data analysis, and paper editing. A.H., E.N., C.M., T.-L.G., and M.S.M. all helped with data collection, analysis, and manuscript editing. The clinical investigations were co-designed by J.R.M., D.F., and B.A. They also helped with data collection and paper editing. contrasting objectives: A pending usage patent on the use of DCA as a cancer treatment belongs to E.D.M. as a co-owner. This patent hasn’t been actively or proactively commercialized. Posted on November 11, 2009 approved on April 23, 2010 10.1126/scitranslmed.3000677 published on May 12, 2010 Citation: Metabolic modulation of glioblastoma with dichloroacetate by E. D. Michelakis, G. Sutendra, P. Dromparis, L. Webster, A. Haromy, E. Niven, C. Maguire, T.-L. Gammer, J. R. Mackey, D. Fulton, B. Abdulkarim, M. S. McMurtry, and K. C. Petruk. 2, 31ra34 Sci. Transl. Med (2010). On May 13, 2010, stm.sciencemag.org published the following research article: www.ScienceTranslationalMedicine.org 12 May 2010 Vol 2 Issue 31 31ra34 8. retrieved from