Inhibition of lactate dehydrogenase A induces oxidative stress and ...

Inhibition of lactate dehydrogenase A inducesoxidative stress and inhibits tumor progressionAnne Le a , Charles R. Cooper a , Arvin M. Gouw b , Ramani Dinavahi a , Anirban Maitra b,c , Lorraine M. Deck d , Robert E. Royer e ,David L. Vander Jagt e , Gregg L. Semenza c,f,g,h,1 and Chi V. Dang a,b,c,i,j,k,2a Division of Hematology, Department of Medicine, Departments of b Pathology, c Oncology, h Pediatrics, i Molecular Biology, j Genetics, and k Cell Biology,f Institute for Cell Engineering, and g McKusick–Nathans Institute of Genetic Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD 21205;d Department of Chemistry, University of New Mexico, Albuquerque, NM 87131; and e Department of Biochemistry and Molecular Biology, University of NewMexico School of Medicine, Albuquerque, NM 87131Contributed by Gregg L. Semenza, December 14, 2009 (sent for review October 30, 2009)As the result of genetic alterations and tumor hypoxia, manycancer cells avidly take up glucose and generate lactate throughlactate dehydrogenase A (LDHA), which is encoded by a targetgene of c-Myc and hypoxia-inducible factor (HIF-1). Previousstudies with reduction of LDHA expression indicate that LDHA isinvolved in tumor initiation, but its role in tumor maintenance andprogression has not been established. Furthermore, how reductionof LDHA expression by interference or antisense RNA inhibitstumorigenesis is not well understood. Here, we report that reductionof LDHA by siRNA or its inhibition by a small-molecule inhibitor (FX11[3-dihydroxy-6-methyl-7-(phenylmethyl)-4-propylnaphthalene-1-carboxylic acid]) reduced ATP levels and induced significant oxidativestress and cell death that could be partially reversed by theantioxidant N-acetylcysteine. Furthermore, we document that FX11inhibited the progression of sizable human lymphoma and pancreaticcancer xenografts. When used in combination with the NAD +synthesis inhibitor FK866, FX11 induced lymphoma regression.Hence, inhibition of LDHA with FX11 is an achievable and tolerabletreatment for LDHA-dependent tumors. Our studies document atherapeutical approach to the Warburg effect and demonstratethat oxidative stress and metabolic phenotyping of cancers arecritical aspects of cancer biology to consider for the therapeuticaltargeting of cancer energy metabolism.glycolysis | lymphoma | pancreatic cancer | redox stress | xenograft modelsThe propensity of cancer cells to take up glucose avidly andconvert it to lactate, even under experimental conditions withadequate oxygen, has been called the Warburg effect or aerobicglycolysis (1, 2). The Warburg effect has been directly linked to theactivation of oncogenes, such as MYC, Ras, and Akt, and loss oftumor suppressor genes, which result in the deregulated conversionof glucose to lactate (3–9). As a result of the pervasivepresence of hypoxia in tumors, hypoxia-inducible factor (HIF-1) iscommonly increased. HIF-1 levels can also be increased downstreamof signal transduction pathways and by loss of the vonHippel-Lindau (VHL) tumor suppressor (10). HIF-1 is a criticaltranscription factor for hypoxic adaptation with its activation ofglycolytic enzyme genes, including lactate dehydrogenase A(LDHA), which converts pyruvate to lactate coupled with therecycling of NAD + (11).Lactate dehydrogenase is a tetrameric enzyme comprising twomajor subunits A and/or B, resulting in five isozymes (A4, A3B1,A2B2, A1B3, and B4) that can catalyze the forward and backwardconversion of pyruvate to lactate. LDHA (LDH-5, M-LDH, or A4), which is the predominant form in skeletal muscle,kinetically favors the conversion of pyruvate to lactate. LDHB(LDH-1, H-LDH, or B4), which is found in heart muscle, convertslactate to pyruvate that is further oxidized. It has long beenknown that many human cancers have higher LDHA levels thannormal tissues (12), but the link between oncogenes and glycolysiswas poorly understood. LDHA was identified as a directtarget gene of the c-Myc oncogenic transcription factor (13, 14).Because HIF also activates LDHA (11, 15), we reasoned thatreduction of LDHA expression would inhibit cellular transformationand in vivo tumorigenesis, because tumor tissues arehypoxic and normal tissues are neither profoundly hypoxic norhave activated MYC (14). Studies now document that reductionof LDHA reduced cellular transformation and markedly delayedtumor formation, indicating that LDHA is important for tumorinitiation (16, 17). These studies provide proof-of-concept thatLDHA is a tractable therapeutical target but do not establishwhether LDHA is required for tumor progression. Furthermore,the mechanisms by which LDHA inhibition leads to inhibition oftumorigenesis are poorly understood.Here, we report that reduction of LDHA causes bioenergetic andoxidative stress leading to cell death. Furthermore, using a drug-likesmall molecule (FX11 [3-dihydroxy-6-methyl-7-(phenylmethyl)-4-propylnaphthalene-1-carboxylic acid]), we document that LDHAis required not only for tumor initiation but for tumor maintenanceand progression. When combined with another metabolic inhibitor,FK866 (APO866), that inhibits NAD + synthesis throughdirect inhibition of nicotinamide phosphoribosyltransferase(NAMPT) (18, 19), FX11 is able to induce lymphoma regression.These studies indicate that targeting cancer metabolism throughsmall drug-like molecules is achievable, and hence paves the wayfor the development of unique classes of anticancer drugs.Results and DiscussionReduction of LDHA Induces Oxidative Stress and Cell Death. Wesought to understand the mechanisms of cell death followingLDHA reduction by short interfering RNA (siLDHA), which hasbeen shown to inhibit tumorigenesis. First, we determined theeffect of reduced LDHA expression on oxygen consumption byhuman Panc (P) 493 B-lymphoid cells, because reduction ofLDHA would favor the entry of pyruvate into mitochondria foroxidative phosphorylation, thereby enhancing oxygen consumption.Reduction of LDHA expression by siRNA in P493 cellsresulted in an increase in oxygen consumption (Fig. 1 A and B).Author contributions: A.L., A.M., D.L.V.J., G.L.S., and C.V.D. designed research; A.L., C.R.C.,A.M.G., and R.D. performed research; C.R.C., R.D., A.M., L.M.D., R.E.R., and D.L.V.J. contributednew reagents/analytic tools; A.L., C.R.C., A.M.G., R.D., A.M., D.L.V.J., G.L.S., andC.V.D. analyzed data; and A.L., C.R.C., G.L.S., and C.V.D. wrote the paper.The authors declare no conflict of interest. C.V.D. is member of the Scientific AdvisoryBoard of Agios Pharmaceuticals; there is no sponsored research or technology licensingactivities involving the company. An invention relating to this work was reported to theJohns Hopkins Technology Transfer office.Freely available online through the PNAS open access option.1 To whom correspondence may be addressed at: The Johns Hopkins University School ofMedicine, BRB 671, 733 N. Broadway Ave., Baltimore, MD, 21205. E-mail: gsemenza@jhmi.edu.2 To whom correspondence may be addressed at: The Johns Hopkins University School ofMedicine, Ross Building, Room 1032, 720 Rutland Avenue, Baltimore, MD 21205. E-mail:cvdang@jhmi.edu.This article contains supporting information online at www.pnas.org/cgi/content/full/0914433107/DCSupplemental.CELL BIOLOGYwww.pnas.org/cgi/doi/10.1073/pnas.0914433107 PNAS | February 2, 2010 | vol. 107 | no. 5 | 2037–2042

Fig. 2. Inhibition of LDHA by FX11 resulted in increased oxygen consumption,ROS production, and cell death. (A) Oxygen consumption of P493cells was determined by a Clark-type oxygen electrode in the presence (slope =−2.4) and absence (slope = −1.7) of FX11. Data are representative of duplicateexperiments. (B) ROS levels were determined by DCFDA fluorescence in P493cells treated with FX11 or FK866. Data are representative of triplicate samplesof two separate experiments. (C) Cell death was determined by flow cytometryof Annexin V- and 7-AAD-stained cells after 24 h of FX11 treatment as comparedwith control. The number in each figure represents the average percentageof dead cells. The FX11-treated cells compared with the control grouphave a P value of 8.92e-06. (D and E) Cell population growth of control cellscompared with cells treated with FX11 or FK866 in the presence or absence of20 mM NAC. All cells were grown at 1 × 10 5 cells/mL. Cell counts were performedin triplicate and shown as mean ± SD, and the entire experiment wasreplicated, with similar results.GAPDH is another pivotal glycolytic enzyme that converts NAD +to NADH, we also sought to determine whether FX11 can inhibitits NAD + -dependent conversion of glyceraldehyde-3-phosphateto bis-phosphoglycerate. We found that even at 74 μM FX11,GAPDH activity was not inhibited. Through formal Michaelis–Menten kinetics, the estimated K i is >>300 μM for GAPDH,indicating that FX11 was selective for LDHA among glycolyticenzymes that use the cofactor NAD.As observed with siRNA-mediated reduction of LDHA,inhibition of LDHA by FX11 also resulted in increased oxygenconsumption, ROS production, and cell death (Fig. 2 A–C). Wefound that the NAMPT inhibitor of NAD + synthesis, FK866,also increased ROS production (Fig. 2B). Because the increasein ROS levels might result from lowered NADPH productionattributable to the possible inhibition of glucose transport (andhence diminished hexose monophosphate shunt activity, whichproduces NADPH) by an indirect effect of FX11 and FK866, wemeasured and found that glucose uptake, as measured by NBDglucose[2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose],was not significantly diminished and that NADPHlevels were virtually unaltered by treatment with FX11 or FK866(Fig. S3). NAC could partially rescue the diminished proliferationof P493 cells treated with either FX11 or FK866 (Fig. 2D and E), indicating that oxidative stress contributed partially tothe inhibition of cell proliferation. Given the significant effect ofFX11 on the proliferation of P493 cells, which are dependent onMyc, we ruled out the trivial possibility that FX11 could inhibitMyc expression itself (Fig. S1C). In aggregate, these studiesdocument that reduction of LDHA levels or activity triggersoxidative stress and cell death.FX11 Inhibits Glycolysis and Alters Cellular Energy Metabolism. Inaddition to oxidative stress induced by inhibition of LDHA, wesought to determine how FX11 affects cellular bioenergetics.First, we observed that both FX11 and FK866 decreased mitochondrialmembrane potential and that the combination accentuatedthe abnormality (Fig. 3A). In this regard, the combinationof FX11 and FK866 was more toxic to P493 cells than either onealone, causing a more profound inhibition of cell proliferation(Fig. 3B). A dose–response study using a combination of FX11and FK866 doses revealed a combination index (CI-50) of 0.78,suggesting a slightly synergistic effect of the combination on cellproliferation (Fig. S4A). After 20 h of exposure to FX11 orFK866, a decrease in ATP levels was accompanied by activationof AMP kinase and phosphorylation of its substrate acetyl-CoAcarboxylase (Fig. 3 C and D and Fig. S1D), suggesting that, inaddition to induction of oxidative stress, these agents depletedcellular energy levels. Decreased ATP levels, despite an increasein cellular oxygen consumption, suggests that FX11 treatmentincreased nonproductive mitochondrial respiration as reportedwith shRNA-mediated knock-down of LDHA (16). Becauseinhibition of LDHA decreases NAD + recycling, treatment ofP493 cells with FX11 was associated with increases in theNADH/NAD + ratio (Fig. 3E) and cellular autofluorescence,reflective of elevated cellular NADH levels (Fig. S4 B and C). Incontrast to FK866, FX11 significantly diminished but did notcompletely inhibit cellular production of lactate (Fig. 3F). Theseobservations suggest that FX11 targets LDHA, inhibits glycolysis,and shunts pyruvate into the mitochondrion.FX11 Inhibits Cells that Are Dependent on Glycolysis. It stands toreason that if FX11 targets LDHA, cells that depend on LDHAfor glycolysis would be more susceptible to FX11 inhibition thanthose that primarily use oxidative phosphorylation. In thisregard, we sought to determine whether metabolic phenotypescould affect the sensitivity of cancer cells to FX11. We used thehuman RCC4 renal cell carcinoma cell line and the RCC4 cellline reconstituted with VHL (RCC4-VHL). Loss of VHL inRCC4 rendered these cells constitutively glycolytic because ofthe stabilization and expression of HIF-1 and HIF-2. Reconstitutionwith VHL resulted in degradation of HIF-1α and HIF-2α and increased mitochondrial biogenesis and oxygen consumption(23). Given the metabolic differences in these isogeniccell lines, we expect a more significant influence of FX11 on theRCC4 cells as compared with the RCC4-VHL cells. Indeed, adose–response study revealed that RCC4 is more sensitive toFX11 compared with RCC4-VHL (Fig. S5 A and B).To corroborate these findings further, we studied the glycolyticMCF-7 and the oxidative MDA-MB-453 breast carcinoma celllines (24). We confirmed that MCF-7 was more dependent onglucose, whereas MDA-MB-453 was more dependent on glutamineoxidation (Fig. S6 A and B), such that deprivation of glucosehas a more profound growth inhibitory effect on MCF-7. Adose-dependent study further revealed that MCF-7 is moresensitive to FX11 (Fig. S5 C and D). Although there are manyother differences between these cell lines, the correlation ofFX11 sensitivity and glucose dependency of MCF-7 supports thenotion that glycolysis predisposes cancer cells to growth inhibitionby FX11.CELL BIOLOGYLe et al. PNAS | February 2, 2010 | vol. 107 | no. 5 | 2039

further slowed by FX11 (Fig. S5F). Although we noted thatsiControl cells had a modestly diminished proliferation rate ascompared with untreated cells, the siLDHA treatment remarkablydisabled cell proliferation contemporaneous with reducedLDHA expression (Fig. S6 C and D). These observations collectivelyindicate that the growth inhibitory effect of FX11 isconsistent with its ability to inhibit LDHA.We surmised that the human P493 B-lymphoma cells would besensitive to FX11 because they express LDHA (25) but that thesensitivity would be heightened under hypoxia when glycolysis isfavored. This is particularly important, because P493 cellsdepend on both glucose and glutamine metabolism when culturedat 20% (vol/vol) O 2 (26). When P493 cells were subjectedto a dose–response study with FX11, we found that growthinhibition by 9 μM FX11 was increased when the cells werecultured at 1% O 2 (Fig. S7 A and B). Hypoxia also sensitized thehuman P198 pancreatic cancer cell line to inhibition by FX11(Fig. S7 C and D), suggesting that reliance on LDHA for hypoxicmetabolism caused cancer cells to be susceptible to the growthinhibitoryeffects of LDHA inhibition by FX11.Fig. 3. (A) FK866 enhances FX11-induced loss of mitochondrial membranepotential. P493 cells treated with control vehicle, FK866, FX11, or bothinhibitors were stained with JC-1 and subjected to flow cytometric analysis,with FL2 representing red fluorescence intensity and FL1 representing greenfluorescence intensity, which is reflective of cells with decreased mitochondrialmembrane. The average percentage (±SEM) of cells with decreasedmembrane potential is indicated in each panel. The P values were 0.03,0.002, and 0.0008 when comparing FX11-treated cells, FK866-treated cells, orcells treated with both, respectively, with the control group. (B) Effect ofFK866 and FX11 on P493-6 cell proliferation. Live cells were counted usingtrypan blue dye exclusion. Data are shown as the mean ± SD of triplicatesamples. (C) Effect of FX11 or FK866 on ATP levels. P493 cells were treatedwith 9 μM FX11 or 0.5 nM FK866 for 20 h and counted. ATP levels (mean ±SEM, n = 5 experiments) were determined by luciferin–luciferase-based assayon aliquots containing an equal number of live cells. *P = 0.008; **P = 0.003.(D) Immunoblot of phosphor-AMP kinase (PAMPK) in lysates of cell treatedwith FX11 or FK866. Tubulin serves as a loading control. AICAR, an AMPanalogue that activates AMPK, was used to treat the cells as a positivecontrol. (E) FX11 increases the NADH/NAD + ratio. NADH/NAD + ratio in P493cells treated with 9 μM FX11 for 24 h as compared with vehicle control.*P = 0.028. (F) FX11 inhibits lactate production. Lactate levels in the media ofP493 human B cells treated with 9 μM FX11 or 0.5 nM FK866 for 24 h ascompared with control. Control RPMI contained 10.7 mmol/L glucose and nodetectable lactate. *P = 6.9E-06.We further tested whether the inhibition of human P493 Bcells by FX11 depended on glucose or on LDHA. The growth ofP493 was inhibited by about 60% when depleted of glucose ascompared with growth in normal medium (Fig. S5E). Addition ofFX11 could not inhibit P493 cells further in the absence ofglucose, suggesting that the effect of FX11 on cell proliferationwas glucose-dependent. Furthermore, knock-down of LDHAexpression by two sequential electroporations with siRNAcaused a markedly diminished proliferative rate that was notFX11 Inhibits Tumorigenesis in Vivo. Although the renewed interestin the Warburg effect is accompanied by a greater understandingof its molecular underpinnings, targeting it for therapeuticalpurposes remains a major challenge (27). By characterizing asmall-molecule inhibitor of LDHA, we found that it is effectivein inhibiting cellular growth and triggering cell death by bothinducing ROS production and depleting ATP. We observed thathypoxia further sensitized human P493 lymphoma cells toLDHA inhibition by FX11. In this regard, the pervasive hypoxictumor microenvironment, as compared with normal tissues,ought to force a further dependency of these lymphoma cells onglycolysis, and particularly on LDHA (Fig. 4A and Fig. S8A). Ofnote, primary human lymphomas have been documented to haveelevated LDHA expression particularly in hypoxic regions (28).Hence, we sought to determine whether in vivo efficacy could bedemonstrated with FX11 as an inhibitor of LDHA. We calculateda desired FX11 dose of 42 μg for daily i.p. injection. Weexpected an initial serum level of ∼100 μM, assuming a uniformand immediate distribution in the vascular system withoutaccounting for the drug half-life or drug metabolism. It should benoted that solubility was a significant dose-limiting factor,because we could only double the dose further before reachingthe limited solubility of FX11 in aqueous solution.First, we sought to determine whether FX11 could inhibitP493 tumor initiation after a palpable tumor developed. Ascontrols, we injected animals with vehicle (2% (vol/vol) DMSO)or with 0.8 mg of doxycycline to inhibit MYC expression in thesetransformed human B cells. As expected, doxycycline profoundlyinhibited palpable tumor xenograft growth as compared withvehicle-injected control animals (Fig. 4B). Intriguingly, daily i.p.injection of 42 μg of FX11 also resulted in a remarkable inhibitionof tumor growth. It is notable that we have ruled out thetrivial possibility that FX11 might directly inhibit Myc expressionto mediate this profound effect (Fig. S1C). These observationsindicate that LDHA is necessary for tumor initiation.To test whether LDHA is required for tumor maintenance orprogression, we challenged the ability of FX11 to inhibit tumorxenograft growth by treating P493 lymphomas or human P198pancreatic tumors that reached the size of 200 mm 3 beforetreatment commenced. For comparison, we treated animals withvehicle control or a compound related to FX11, called E, thatlacks the benzyl group and has a K i for LDHA of >90 μM, ormore than 10-fold higher than that of FX11. We found that Ehad no antitumor activity as compared with vehicle. At thissolubility-limiting dose, FX11 displayed a static but significanteffect over 10 days (Fig. 4C). Intriguingly, we observed that thehypoxic regions associated with untreated control tumors were2040 | www.pnas.org/cgi/doi/10.1073/pnas.0914433107 Le et al.

Fig. 4. In vivo efficacy of FX11 as an antitumor agent. (A) Immunohistochemicalstaining to detect hypoxic regions (dark brown) of spleen, liver,and P493 lymphoma by pimonidazole labeling. (B) Effect of FX11 on growthof palpable human P493 B-cell xenografts. Control animals were treatedwith daily i.p. injection of vehicle (2% (vol/vol) DMSO), and doxycycline(0.8 mg/day) was used as a positive control because it inhibits Myc expressionand tumorigenesis in P493 cells. (C) Effect of FX11 and/or FK866 dailytreatment as compared with control or compound E (a weak LDHA inhibitor)on established human lymphoma xenografts. (Inset) Photographs of representativeanimals treated with control vehicle or FX11. (D) FX11 inhibitedP198 human pancreatic cancer xenografts as compared with compound E.For experiments in all panels, 2.0 × 10 7 P493 cells or 5 × 10 6 P198 cells wereinjected s.c. into SCID mice or athymic nude mice, respectively. When thetumor volume reached 200 mm 3 ,42μg of FX11 and/or 100 μg of FK866 wasinjected i.p. daily and observed for 10–14 days. The tumor volumes weremeasured using digital calipers every 4 days and calculated using the followingformula: [length (mm) × width (mm) × width (mm) × 0.52]. Theresults represent the average ± SEM.relatively diminished in the FX11-treated tumors (Fig. S8B). Wespeculated that hypoxic tumor cells are more dependent onglycolysis, and hence are diminished by FX11 relative to thenonhypoxic tumor cells. We also observed a significant responseof human P198 tumor xenografts to FX11 as compared with E(Fig. 4D). Although a more aggressive human pancreatic tumorxenograft LZ10.7 grew significantly faster than P198, it was alsosensitive to FX11 as a single agent (29) (Fig. S8C). The structureand activity relation of FX11 and E in vivo correlated with theinhibition and binding of LDHA by these compounds in vitro,further supporting the notion that FX11 targets LDHA. Theseresults collectively indicate that LDHA plays a role in tumorprogression and maintenance.On the basis of our studies of cultured cells (Fig. 3B), wehypothesized that FX11 might accentuate the effect of FK866 inthe treatment of the P493 human lymphomas. Hence, weselected a dose of 100 μg of FK866, which gave a static outcome,and, indeed, we found remarkable tumor regression when animalswere treated with both FX11 and FK866 (Fig. 4C). Thesefindings underscore the fact that targeting cancer metabolism isfeasible and that LDHA is a significant candidate target forfurther development.Given the significant effects of FX11 as an LDHA inhibitor invivo, we studied a group of treated animals to begin to understandthe potential side effects of FX11. It is notable thathumans lacking LDHA develop normally but have been shownto display exertional myopathy (30). In this regard, although wedid not formally exercise the animals to examine exertional tolerance,we did not note lethargy or the inability to eat and drink.In fact, animals treated with FX11 did not lose weight. In initialstudies of the hematology and blood chemistry, we did not seecytopenia in animals treated with FX11 alone; however, two (offive studied) animals treated with FK866 did show mild thrombocytopenia(Fig. S9). The average leukocyte count in the controlgroup was skewed upward by two animals that hadleukocytosis with >15 K/μL (normal range: 1.8–to 10.7 K/μL).The blood chemistries did not reveal any evidence of kidney[blood urea nitrogen (BUN) or creatinine] or liver (aspartateaminotransferase, alanine aminotransferase, and alkaline phosphatase)toxicity in animals treated with FX11 or FK866 alone atdoses that affected tumor growth in vivo (Fig. S9). However, thecombination of FX11 and FK866, as compared with control,increased BUN.ConclusionsTargeting cancer energy metabolism has been partly elusivebecause of our poor understanding of metabolic phenotypes ofdifferent cancers, compounded by the lack of understanding ofcellular responses to inhibition of specific enzymes involved inenergy metabolism. Here, we document that LDHA inhibition,which was previously shown through genetically engineered cellsto reduce tumor initiation (16, 17), not only resulted indecreased ATP levels and reduced mitochondrial membranepotential but in a remarkable increase in oxidative stress that islinked to cell death. Furthermore, inhibition of LDHA by FX11inhibited tumor xenograft progression, confirming that LDHA isrequired for tumor maintenance.We found that similar to siRNA reduction of LDHA expression,FX11 could increase cellular oxygen consumption, increase ROSproduction, and induce cell death that could be partially rescuedby the antioxidant NAC. Seemingly paradoxical to the antiapoptoticeffect of NAC in vitro observed in the current work, wehave documented previously that NAC has an antitumorigeniceffect in vivo because of its inhibition of HIF-1 in tumors (31). Theillusion of a conundrum with NAC treatment arises only whenresults of experiments performed in vitro under normoxic conditionsare compared with the results of NAC effects on in vivotumorigenesis when HIF-1 is induced within the hypoxic tumormicroenvironment. Our current observations are corroborated bystudies in Saccharomyces cerevisiae with enhanced or defectiverespiration (32). Genetic knock-down or glucose repression ofrespiration in yeast reduced apoptosis and enhanced clonogenicsurvival, whereas forced enhancement of respiration increasedROS production and reduced colony growth that could be partiallyrescued by the antioxidant glutathione. In this regard, a recentperspective on cancer energy metabolism emphasizes the importanceof redox homeostasis in cancer cell survival (6). FX11 alsoreduced ATP levels, suggesting that inhibition of LDHA causedbioenergetic and oxidative stress, which, together, inhibits tumorxenograft maintenance and progression.Because FX11 has a catechol moiety, it could hypotheticallybe converted in vivo to a dihydroquinone that is reactive andcould cause effects other than inhibition of LDHA. Although theCELL BIOLOGYLe et al. PNAS | February 2, 2010 | vol. 107 | no. 5 | 2041

Supporting InformationLe et al. 10.1073/pnas.0914433107SI Experimental ProceduresCell Lines and Hypoxic Exposure. P493 human lymphoma B cells weremaintained in RPMI 1640 with 10% (vol/vol) FBS and 1% (wt/vol)penicillin–streptomycin. P198 human pancreatic cancer cells,RCC4 and RCC4-VHL human renal carcinoma cells, and MCF-7and MDA-MB-453 breast cancer cells were maintained in highglucose(4.5 mg/mL) DMEM with 10% (vol/vol) FBS and 1% (wt/vol) penicillin–streptomycin. Nonhypoxic cells (20% (vol/vol) O 2 )were maintained at 37 °C in a 5% (vol/vol) CO 2 and 95% (vol/vol)air incubator. Hypoxic cells (1% O 2 ) were maintained in a controlledatmosphere chamber (PLAS-LABS) with a gas mixturecontaining 1% O 2 , 5% (vol/vol) CO 2 , and 94% (vol/vol) N 2 at37 °C for the indicated time. Bright live cells were counted dailyin a hemacytometer using trypan blue dye to exclude dead cells. Allcells were grown at a concentration of 10 5 cells /mL. All drug treatmentsbegan at day 0. FX11 was added daily.RNA Interference Experiments. siRNAs targeting human LDHA(ON-TARGETplus SMARTpool) were purchased from DharmaconResearch, Inc. Targeting sequences for LDHA were a poolof the following four target sequences: sequence 1, GGAGAA-AGCCGUCUUAAUU; sequence 2, GGCAAAGACUAUA-AUGUAA; sequence 3, UAAGGGUCUUUACGGAAUA; andsequence 4, AAAGUCUUCUGAUGUCAUA. For P493 humanlymphoma B cells, transfection of siRNAs was performed using anAmaxa Nucleotransfection device according to the manufacturer’sinstructions. Briefly, 2 μg of siLDHA ON-TARGETplus SMARTpoolor ON-TARGETplus Nontargeting Pool (Dharmacon Research,Inc.) was transfected into 2 × 10 6 cells at 0 h. At 24 h, 10 5 cellswere treated with 0.1% DMSO or FX11 for 48 more hours. Theremaining cells were harvested for immunoblot analysis. For P198human pancreatic cancer cells, transfection of siRNAs was performedusing X-tremeGENE siRNA Transfection Reagent (Roche)according to the manufacturer’s instructions.Western Blot Analysis. Cell pellets were harvested after washingwith PBS. Protein concentration was determined by BCA assay(Pierce) and 30 μg of protein per well was separated by SDS/PAGE and transferred by iBlot gel transfer stacks (Invitrogen).Rabbit monoclonal anti-LDHA (Epitomics) was used to detecthuman LDHA; phosphor-AMP kinase α and phospho-acetyl-CoA carboxylase antibodies were purchased from Cell Signaling.Membranes were reprobed with anti-α-tubulin as a loadingcontrol. The enhanced chemiluminescence reagent ECL (GEHealthcare) was used for detection.Oxygen Consumption. Oxygen consumption was measured using aClark-type oxygen electrode (Oxytherm System; HansatechInstruments Ltd). Then, 5 × 10 6 cells in 1 mL of medium wereplaced in the chamber above a membrane that is permeable tooxygen. Oxygen diffuses through the membrane and is reduced atthe cathode surface so that a current flows through the circuit,which is completed by a thin layer of KCl solution. The current thatis generated bears a direct stoichiometric relation to the oxygenreduced and is converted to a digital signal. Determinations weredone in triplicate, and the entire experiment was done twice.ROS Measurement by Flow Cytometry. Intracellular ROS productionwas measured by staining cells with 5-(and-6)-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (carboxy-H 2 DCFDA;Molecular Probes) according to the manufacturer’s instructions.Next, 10 5 cells/mL were treated with 9 μM FX11 or 0.5 nMFK866 for 24 h. Stained cells were analyzed in FACScan flowcytometers (BD Bioscience).Annexin V Assay. After 24 h of FX11 treatment, cells were harvestedand washed twice with cold PBS and the assay was performedusing the Annexin V–7-AAD apoptosis detection Kit I(BD Biosciences Pharmingen) according to the manufacturer’sinstructions.Determination of K i and Enzyme Kinetics. The reaction velocity ofpurified human LDHA or GAPDH was determined by a decreaseor increase, respectively, in absorbance at 340 nm of NADH. TheLDHA activity was assessed using the protocol described in theWorthington Enzyme Manual, with varying concentrations ofNADH. The GADPH activity was assessed using the protocoldescribed in the Worthington Enzyme Manual with varying concentrationsof NAD + . K i values were determined from doublereciprocalplots by linear regression analysis using SigmaPlotEnzyme Kinetic software.CarboxyLink Immobilization and Affinity Columns. FX11 and Emolecules that contain carboxyl groups were coupled to immobilizeddiaminodipropylamine resins according to the manufacturer’sinstructions (Pierce). About 1.9 mg (95% couplingefficiency) of FX11 or E was coupled to 2 mL of resin as estimatedby the amount of either molecule recovered after conjugation.An equal amount of cell lysate was loaded onto thesebeads and eluted with 1 mM NADH after washing with six columnvolumes of high salt (1 M NaCl). LDHA activity was performedfrom the eluates to assess the binding affinity of themolecules with LDHA.Mitochondrial Membrane Potential Measurement by Flow Cytometry.The lipophilic cation dye [JC-1(5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolcarbocyanine iodide); Invitrogen) wasused to detect the loss of the mitochondrial membrane potential.The negative charge established by the intact mitochondrialmembrane potential lets the lipophilic dye stain the mitochondriabright red, which emits in channel 2 (FL2). When the mitochondrialmembrane potential collapses, JC-1 remains in thecytoplasm in a green fluorescent monomeric emission in channel1 (FL1). JC-1 reversibly changes its color from green to orangeas membrane potentials increase (over values of 80–100 mV).Then, 10 5 cells/mL were treated with 9 μM FX11 and/or FK866for 24 h. Stained cells were analyzed in FACScan flow cytometers(BD Bioscience).Measurement of ATP. P493 cells were treated with 9 μM FX11 or0.5 nM FK866 for 20 h and counted. ATP levels were determinedby a luciferin–luciferase-based assay (Promega) onaliquots containing equal number of cells according to a standardprotocol.Determination of NADH/NAD + and NADPH/NADP + Ratios. The NADH/NAD + and NADPH /NADP + ratios were assessed using theprotocol described by the manufacturer. The assay is based on anenzyme-catalyzed kinetic reaction in which a tetrazolium dye 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide is reducedby NAD(P)H in the presence of phenazine methosulfate.The intensity of the reduced product color, measured at 565 nm, isproportionate to the NAD(P)H /NAD(P) + concentration in thesamples. The NAD(P)H or NAD(P) + content of the cells washarvested separately using a corresponding extraction buffer.Le et al. www.pnas.org/cgi/content/short/09144331071of11

Fig. S1. (A) Effect of reduced LDHA expression by siRNA on oxygen consumption in P198 human pancreatic cancer cells. Oxygen consumption was determinedby the use of a Clark-type oxygen electrode at 72 h posttransfection with siLDHA (slope = −2.1) or siControl (slope = −1.1). (B) Immunoblotting was performedon whole-cell lysates, probed with rabbit monoclonal anti-LDHA antibody, and reprobed with anti-α-tubulin as a loading control. (C) FX11 does not affect c-Myc levels in P493 cells. Tetracycline, known to repress c-Myc expression, was used as a control. Equivalent amounts of proteins were immunoblotted with antic-Mycantibody, and α-tubulin served as a loading control. (D) Activation of phospho-acetyl-CoA carboxylase (P-ACC) in lysates of cells treated with FX11 orFK866. Western blot analysis was performed after 24 h of treatment. Equivalent amounts of proteins were immunoblotted with anti-P-ACC, and α-tubulinserved as a loading control.Le et al. www.pnas.org/cgi/content/short/09144331073of11

Fig. S2. (A and B) FX11 and its derivative E are competitive inhibitors of LDHA with NADH as substrate. Lineweaver–Burk plots were determined from triplicateexperiments using averages of activities. K i determination was performed with 13.5 μM FX11 or 27 μME.(C)Affinity chromatography of LDHA using sepharoseimmobilizedFX11 or E. Equal volumes of P493 human B cell lysates were chromatographed with six column volumes of high salt (1 M NaCl) wash, followed byelution with 1 mM NADH. LDHA activity was determined for each fraction, and the experiment was replicated with a representative experiment shown.Le et al. www.pnas.org/cgi/content/short/09144331074of11

Fig. S3. (A) FX11 and/or FK866 effects on NADPH/NADP + ratio. P493 cells were treated with 10 μM FX11 and/or 0.1 nM FK866 for 24 h before the NADPH/NADP ratio was assayed. FX11 treatment compared with the control (P = 0.8) and FK866 treatment compared with the control (P = 0.9). (B) Effects of FX11treatment on glucose uptake by P493 cells. A fluorescent analogue of 2-deoxyglucose (2-NBDG) was used to assay the glucose uptake of P493 cells, followed byflow cytometry. Viable cells taking up glucose were determined by negative 7-AAD and positive 2-NBDG. The number in each panel indicates the geometricalmean of green fluorescence intensity (520 nm) of the cell population.Le et al. www.pnas.org/cgi/content/short/09144331075of11

Fig. S4. (A) Effect of FK866 and FX11 on P493-6 cell proliferation. Live cells were counted using trypan blue dye exclusion. Experiments were done in duplicate,and the entire experiment was repeated two times with similar results. The bars represent the averages of duplicate samples from one experiment. Thecombination of FX11 and FK866 doses revealed a CI-50 = 0.78 calculated using R 2.9.1 software. (B) FX11 treatment increases cellular NADH fluorescence ofP493-6 cells compared with control, compound E, or FK866 treatment. The number in the P2 box represents the average percentage of cellular NADH autofluorescence(n = 3). FX11 treatment compared with control (P = 0.005), FK866 compared with control (P = 0.144), and E treatment compared with control(P = 0.38). (C) Fluorescence emission spectrum of NADH (green line), FX11 (red line), or E (blue line). Note that the emission maxima of FX11 or E (390 nm) donot coincide with that of NADH (460 nm), excluding the direct contribution of FX11 to the autofluorescence of FX11-treated cells at 460 nm.Le et al. www.pnas.org/cgi/content/short/09144331076of11

Fig. S5. (A–D) RCC4 cells and MCF-7 cells are more sensitive to FX11 than RCC4-VHL and MDA-MB-453 cells. Cell population growth of human renal carcinomaRCC4 and RCC4-VHL cells or human breast cancer MCF-7 and MDA-453 cells in response to different doses of FX11. (E) Effect of FX11 on the proliferation ofP493 cells is glucose-dependent. Cell population growth of P493 cells in the absence of glucose and in the presence or absence of 9 μM FX11. (F) LDHAdependenteffect of FX11. Cell population growth of P493 cells with siRNA-mediated reduced LDHA expression in the presence or absence of 9 μM FX11. Cellnumbers are shown as mean ± SD.Le et al. www.pnas.org/cgi/content/short/09144331077of11

Fig. S6. (A and B) Characterization of glucose, glutamine, and pyruvate dependency of different human breast cancer cell lines, MCF-7 and MDA-MB-453.Cells were cultured in media with different availability of glucose, glutamine, and pyruvate. All media were supplemented with 10% (vol/vol) bovine fetalserum and 1% penicillin–streptomycin. Averages of cell numbers from triplicate experiments are shown ± SD. (C) Growth curves for control, electroporated,and siControl cells compared with cells treated with siLDHA. ON-TARGETplus pool was used as siRNAs targeting human LDHA, and Nontargeting Pool was usedas siControl. The siRNAs were transfected into 2 × 10 6 cells at 0 and 24 h by electroporation. (D) Cells were harvested for immunoblot analysis. Rabbitmonoclonal anti-LDHA was used to detect LDHA, and the membrane was reprobed with anti-α-tubulin as a loading control.Le et al. www.pnas.org/cgi/content/short/09144331078of11

Fig. S7. Hypoxia accentuates the sensitivity of human P493 B cells (A and B) and human P198 pancreatic cancer (C and D) cells to FX11 inhibition of growth. Cellpopulation growth of P493 B cells or pancreatic cancer P198 cells in normoxia or hypoxia with different doses of FX11. Normoxic cells were grown at 37 °C in a5% (vol/vol) CO 2 and 95% (vol/vol) air incubator. Hypoxic cells (1% O 2 ) were maintained for the indicated time in a controlled atmosphere chamber with a gasmixture containing 1% O 2 , 5% (vol/vol) CO 2 , and 94% (vol/vol) N 2 at 37 °C. There is no significant difference in cell numbers between the 0% and 0.1% DMSOgroups.Le et al. www.pnas.org/cgi/content/short/09144331079of11

Fig. S8. (A) Pimonidazole-labeling of hypoxic regions (dark brown) of spleen, liver, and P493-6 lymphoma. (B) FX11 treatment diminishes hypoxic regions ofP493 lymphoma. Rabbit anti-Hypoxyprobe antibody was used as the primary antibody. Texas-red anti-rabbit and DakoCytomation EnVision+ System-HRP anti-Rabbit antibody were used as secondary antibodies for immunofluorescence and immunohistochemistry, respectively. Samples were analyzed under anAxiovert 200 (Zeiss) fluorescence microscope at ×0 magnification. Ten random fields from an untreated tumor and a treated tumor were photographed. (C)Effect of FX11 treatment vs. vehicle control on LZ10.7 human pancreatic cancer xenografts (P = 0.03). A total of 5 × 10 6 LZ10.7 cells were injected s.c. intoathymic nude mice. When the tumor volume reaches 150 mm 3 ,42μg of FX11 or DMSO, which is used as vehicle control, was injected i.p. daily and observed for10 days. The tumor volumes were measured using digital calipers and calculated using the following formula: [length (mm) × width (mm) × width (mm) × 0.52].The results represent the average ± SEM.Le et al. www.pnas.org/cgi/content/short/0914433107 10 of 11

Fig. S9. Effect of FX11 treatment on blood chemistries and hematology. Average values ± SD are shown for five animals in each group. ALB, albumin; ALP,alkaline phosphatase; ALT, alanine aminotransferase; AMYL, amylase; AST, aspartate aminotransferase; BA, basophils; CAL, calcium; Chol, cholesterol; CL,chloride; CREAT, creatinine; EO, eosinophils; GLU, glucose; HB, hemoglobin; HCT, hematocrit; K, potassium; LY, lymphocytes; MCH, mean corpuscular hemoglobin;MCHC, mean corpuscular hemoglobin concentration; MCV, mean corpuscular volume; MO, monocytes; NA, sodium; NE, neutrophils; RBC, red bloodcells; RDW, red cell distribution width; TBILI, total bilirubin; TPROT, total protein; Tri, triglyceride; UA, uric acid; WBC, white blood cells.Le et al. www.pnas.org/cgi/content/short/0914433107 11 of 11