Thiazolidinediones are acute, specific inhibitors of themitochondrial pyruvate carrierAjit S. Divakarunia, Sandra E. Wileya, George W. Rogersb, Alexander Y. Andreyeva, Susanna Petrosyana,Mattias Loviscachc, Estelle A. Walla, Nagendra Yadavad, Alejandro P. Heucke, David A. Ferrickb, Robert R. Henryc,f,William G. McDonaldg, Jerry R. Colcag, Melvin I. Simona,1, Theodore P. Ciaraldic,f, and Anne N. Murphya,1

Departments of aPharmacology and fMedicine, University of California at San Diego, La Jolla, CA 92093; bSeahorse Bioscience, North Billerica, MA 01862;cVeterans Affairs San Diego Healthcare System, La Jolla, CA 92161; dPioneer Valley Life Sciences Institute, Springfield, MA 01107; eDepartment of Biochemistryand Molecular Biology, University of Massachusetts, Amherst, MA 01003; and gMetabolic Solutions Development Co., Kalamazoo, MI 49007

Contributed by Melvin I. Simon, February 21, 2013 (sent for review January 28, 2013)

Facilitated pyruvate transport across the mitochondrial inner mem-brane is a critical step in carbohydrate, amino acid, and lipidmetabolism. We report that clinically relevant concentrations ofthiazolidinediones (TZDs), a widely used class of insulin sensitizers,acutely and specifically inhibit mitochondrial pyruvate carrier (MPC)activity in a variety of cell types. Respiratory inhibition was over-come with methyl pyruvate, localizing the effect to facilitatedpyruvate transport, and knockdown of either paralog, MPC1 orMPC2, decreased the EC50 for respiratory inhibition by TZDs. AcuteMPC inhibition significantly enhanced glucose uptake in humanskeletal muscle myocytes after 2 h. These data (i) report that clini-cally used TZDs inhibit theMPC, (ii) validate thatMPC1 andMPC2 areobligatory components of facilitated pyruvate transport inmamma-lian cells, (iii) indicate that the acute effect of TZDsmay be related toinsulin sensitization, and (iv) establish mitochondrial pyruvateuptake as a potential therapeutic target for diseases rooted inmetabolic dysfunction.

AMPK | pioglitazone | rosiglitazone | MSDC-0160 | XF PMP

Pyruvate uptake across the mitochondrial inner membrane isa central branch point in cellular energy metabolism with the

ability to balance glycolysis and oxidative phosphorylation andpoise catabolic and anabolic metabolism (1). Although the exis-tence of a mitochondrial pyruvate carrier has been recognized forover 40 y (2, 3), it has only recently been identified at themolecularlevel. Two small transmembrane proteins in the inner membrane,mitochondrial pyruvate carrier 1 (MPC1) and 2 (MPC2), are obli-gate components of an apparent complex that facilitates inhibitor-sensitive pyruvate transport (4, 5). This newly defined complexmay be a rational therapeutic target formodulating energy balanceand the metabolic profile.Thiazolidinediones (TZDs) are the most effective agents for

preventing the progression from hyperglycemia to type 2 diabetes(6), and the increasingly appreciated link between dysregulatedglucose metabolism and human disease has triggered the repur-posing of pioglitazone to treat neurodegenerative conditions andcertain cancers (7–9). However, significant side effects of TZDs,including volume expansion, bone loss, increased adiposity, andcardiovascular risk, have restricted broader clinical use (10, 11).TZD activity is ascribed to peroxisome proliferator-activated

receptor gamma (PPARγ), a nuclear receptor controlling geneexpression related to lipid storage, cell differentiation, and in-flammation (12, 13). However, a growing volume of data suggeststhat PPARγ-independent mechanisms, some of which are toorapid to be attributed to transcriptional events, may be relevant toeffects on metabolism (14–20). In addition, the discovery thatTZDs bind to mitochondrial membranes with low micromolaraffinity (20), mirroring the circulating concentrations in treatedpatients (21, 22), suggests that some of their metabolic effects maybe triggered by directly altering mitochondrial function.This report demonstrates that TZDs specifically inhibit MPC

activity. This inhibition can improve cellular glucose handling, as

demonstrated by the finding that both TZDs and UK5099[2-Cyano-3-(1-phenyl-1H-indol-3-yl)-2-propenoic acid], a chem-ical inhibitor of the MPC, rapidly increased glucose uptake inhuman myocytes. This discovery sets a precedent that pharma-cologic targeting of the MPC can adjust cellular metabolism.

ResultsSelective Permeabilization of the Plasma Membrane with Recom-binant Perfringolysin O. Although TZDs inhibit respiratory com-plex I at supraphysiological concentrations (23, 24), a direct effecton mitochondrial function from clinically relevant concentrationsremains to be identified. To date, the ability to interrogate specificoxidative pathways with high-throughput respirometry has beenlimited by the lack of a reagent that permeabilizes the plasmamembranes of cells in an adherent monolayer but is not injuriousto mitochondria. Historically, various natural agents with differingmechanisms of action, including cytolysins (e.g., α-toxin of Staph-ylococcus aureus and streptolysin-O) and amphipathic glycosides(saponin and digitonin), have been used for cell permeabilization(25). In the present work, recombinant perfringolysin O (rPFO)(Materials and Methods) was used to selectively permeabilize theplasma membrane to control substrate provision.rPFO is a recombinant, mutant form of a cholesterol-

dependent cytolysin derived fromClostridium perfringens (26, 27).It oligomerizes to form pores that selectively permeabilize cel-lular plasma membranes and allow passage of solutes and largeproteins of 200 kDa (28). Succinate and ADP, both of which areimpermeable to the plasmamembrane, sharply increased the rateof respiration in C2C12 myoblasts when acutely added with 1 nMrPFO (Fig. 1A). This concentration was sufficient to permeabilize10 different cell types, including cell lines and primary cultures(Fig. S1A).Unlike detergent-based methods, rPFO was not injurious to

mitochondria. A 10-fold excess concentration of rPFO affectedneither mitochondrial membrane potential nor cytochrome c re-lease from the intermembrane space, a characteristic of digitonin(25) (Fig. 1B). rPFO also had no effect on state 3 respiration inisolated mitochondria, whereas increasing concentrations of dig-itonin caused a sharp decline (Fig. S1B). When added to cells,

Author contributions: A.S.D., S.E.W., G.W.R., A.Y.A., M.L., R.R.H., W.G.M., J.R.C., M.I.S.,T.P.C., and A.N.M. designed research; A.S.D., S.E.W., G.W.R., A.Y.A., S.P., M.L., E.A.W.,W.G.M., J.R.C., T.P.C., and A.N.M. performed research; G.W.R., N.Y., A.P.H., D.A.F., W.G.M.,and J.R.C. contributed new reagents/analytic tools; A.S.D., S.E.W., G.W.R., A.Y.A., M.L., R.R.H.,M.I.S., T.P.C., and A.N.M. analyzed data; and A.S.D. and A.N.M. wrote the paper.

Conflict of interest statement: G.W.R. and D.A.F. are employees of Seahorse Bioscience,which has provided a new reagent for use in the present study. J.R.C. is a cofounder andshareholder of Metabolic Solutions Development Co., and W.G.M. is an employee andshareholder of Metabolic Solutions Development Co.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: m1simon@ucsd.edu or anmurphy@ucsd.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1303360110/-/DCSupplemental.

5422–5427 | PNAS | April 2, 2013 | vol. 110 | no. 14 www.pnas.org/cgi/doi/10.1073/pnas.1303360110

Dow

nloa

ded

by g

uest

on

June

9, 2

020

digitonin caused a drop in respiration that was rescued with ex-ogenous cytochrome c, whereas respiration was again unchangedby rPFO (Fig. S1C).

TZDs Compromise Pyruvate-Driven Respiration in Permeabilized Cells.In a variety of rPFO-permeabilized cells, TZDs inhibitedpyruvate-dependent respiration (Figs. 2 and 3). In C2C12myoblasts, rosiglitazone caused a dose-dependent decrease in

pyruvate-driven, uncoupler-stimulated respiration (Fig. 2A). Theeffect occurred within minutes (making it unlikely to be mediatedby transcriptional events) and was mimicked by UK5099, a potentinhibitor of the mitochondrial pyruvate transporter (29). Bothrosiglitazone and troglitazone, a TZDwithdrawn from clinical usedue to hepatotoxicity (30), had half-maximal inhibitory concen-trations (Fig. 2B) that approximate blood concentrations reachedby TZDs in treated patients (21, 22). Metabolic Solutions De-velopment Company (MSDC)-0160, a TZD structurally similar topioglitazone (Fig. S2) and developed analogously to MSDC-0602(20), had similar effects (Fig. 2B). This compound exhibits de-creased apparent affinity to PPARγ (MSDC-0160, EC50 = 23.7μM; pioglitazone, EC50 = 1.2 μM) while retaining binding to mi-tochondrial membranes (MSDC-0160, IC50 = 1.3 μM; pioglita-zone, IC50 = 1.2 μM) (20). It has completed a promising Phase 2btrial for type 2 diabetes (31), and is currently being evaluated in anongoing trial for mild cognitive impairment (ClinicalTrials.gov; NCT01374438).Importantly, the extent of inhibition by higher concentrations of

TZDs matched that seen with UK5099 (Fig. 2B). The residualrespiration was likely driven by passive entry of pyruvate, a weakacid, across the inner membrane (1). TZD inhibition of pyruvateoxidation was also apparent in four primary cell types. In humanskeletal muscle myotubes (HSkMMs) obtained from quadricepspunch biopsies, neonatal rat ventricular myocytes (NRVMs), ratcortical neurons, and mouse brown adipocyte precursors, MSDC-0160 inhibited pyruvate-driven respiration with similar half-maximalconcentrations (Fig. 2C), suggesting that the effect is cell type-independent.Pioglitazone did not induce acute respiratory inhibition in per-

meabilized myoblasts (Fig. 2B); however, in intact myocytes, bothimmortalized (Fig. 2D) and primary (Fig. 2E), pioglitazone causedtime-dependent respiratory inhibition. Given the immediate re-spiratory inhibition with MSDC-0160 (Fig. 2D), and the fact thatpioglitazone is readily metabolized (21), it is possible that intact

A B

Fig. 1. Selective permeabilization of the plasma membrane enables controlof substrate provision and high-throughput respirometry. (A) C2C12 myo-blasts were offered 15 mM glucose in MAS buffer [with 0.2% (wt/vol) BSA]and, where indicated, also offered 10 mM succinate, 4 mM ADP, and in-creasing concentrations of rPFO (n = 4). (B) Trace: Membrane potential of ratliver mitochondria (1 mg/mL) was monitored as described in SI Materials andMethods with safranine O. Additions were 1 μg/mL oligomycin, 10 nM rPFO,and successive additions of carbonyl cyanide 4-(trifluoromethoxy)phenyl-hydrazone (FCCP) at 0.5 μM, 0.5 μM, and 1.0 μM. (Inset) Rat liver mito-chondria in KCl-based buffer were treated with rPFO or digitonin for 5 min.and then centrifuged at 21,000 × g for 1 min at 4 °C. The release of cyto-chrome c from the intermembrane space to the supernatant (Sup.) orretained in the mitochondrial pellet was measured by Western blot analysis(SI Materials and Methods).

A B C

D E

Fig. 2. TZDs inhibit pyruvate-driven respiration. (A) Representative experiment of permeabilized C2C12 myoblasts offered 5 mM pyruvate, 0.5 mM malate,2mMdichloroacetate (DCA), and 2 μMoligomycin inMAS buffer. Either rosiglitazone or UK5099was added acutely where indicated, as was 400 nM FCCP. Errorbars from technical replicates are obscured by the symbol. (B) Uncoupler-stimulated respirationwith various TZDswasmeasured as inA; half-maximal inhibitoryconcentrations are given in parentheses (n≥ 4). (C) Inhibition of pyruvate-driven respiration byMSDC-0160 in permeabilized primary cells was measured as inAwith the following FCCP concentrations: 600 nM for NRVMs and HSkMMs, 300 nM for brown adipose tissue (BAT) precursors, and 250 nM for cortical neurons.Half-maximal inhibitory concentrations are given in parentheses (n ≥ 3). (D) Intact C2C12 myoblasts were offered pyruvate in unbuffered DMEM, and 10 μMMSDC-0160 or pioglitazonewere added before 400 nM FCCP at the indicated times.MSDC-1473, a TZD negative control (EC50= 52 μM for PPARγ and>25 μMformitochondrial membranes) (Fig. S2)was added 90min before the addition of FCCP (n≥ 3). n.s., not significant. (E) Intact HSkMMswere treatedwith TZDs (10 μM)or UK5099 (2 μM) at 90 min before the addition of 600 nM FCCP. When added acutely, 10 μM pioglitazone was added at 15 min before the addition of FCCP.

Divakaruni et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5423

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

9, 2

020

cells generate an active pioglitazone metabolite. Taken together,our data show that physiologically relevant concentrations ofTZDs inhibit pyruvate oxidation.

Pyruvate Transport Is Specifically Compromised. To determine theprecisemechanismof respiratory inhibition, cells were permeabilizedand offered different oxidizable substrates (Fig. S3). In per-meabilized C2C12 myoblasts, TZDs significantly inhibited respi-ration driven by glutamate (with malate) only at 30 μM (Fig. 3A).Succinate-driven respiration was unaffected. These results areconsistent with previous reports of complex I inhibition at sup-raphysiological TZD concentrations (23, 24). In three types ofprimary cells (HSkMMs, NRVMs, and cortical neurons), pyru-vate-driven respiration was significantly compromised with bothUK5099 and MSDC-0160, but there was no effect on the oxida-tion of either glutamate or succinate (Fig. 3B). This pinpoints theTZD effect at lower, clinically relevant concentrations to in-hibition of either pyruvate transport or pyruvate dehydrogenase(PDH) activity.To distinguish between these two mechanisms, patient-derived

myotubes and cortical neurons were offered excess methyl pyru-vate, which freely crosses the inner membrane and is cleaved bymatrix esterases to generate intramitochondrial pyruvate (Fig.S3). On the addition of methyl pyruvate, the inhibitory effects ofTZDs and UK5099 were almost entirely overcome. This indicatesthat TZDs are acute, specific inhibitors of mitochondrial pyruvate

transport (Fig. 3C). However, unlike UK5099, which covalentlymodifies the pyruvate transporter via a reactive thiol group (32),the inhibitory effect of TZDs was reversible. After treatment ofintact cells with TZDs, rates of pyruvate oxidation were restoredon washing and permeabilization, whereas the effect of UK5099persisted (Fig. 3D).

The MPC Complex Is a Target of TZDs. Both MPC1 and MPC2 arerequisite components of facilitated mitochondrial pyruvate uptake(4, 5). Each paralog was stably repressed in C2C12myoblasts usinglentiviral shRNA. RT-PCR (Fig. S4A) and Western blot analysis(Fig. S4B) confirmed knockdown of transcript and protein ex-pression, and suggested that MPC1 and MPC2 are coordinatelyregulated. The functional consequence of this knockdown of eitherparalog was severely compromised pyruvate oxidation (Fig. 4A).The depressed respiration could not be attributed to differences incell number, given that total cellular protein at the time of theassay was equivalent in all lines (Fig. S4C). Moreover, respirationon complex I-linked substrates (glutamate and β-hydroxybutyrate),Q-pool substrates (succinate and glycerol-3-phosphate), or pal-mitoyl carnitine was unchanged with stable repression of eitherMPC paralog in C2C12 myoblasts (Fig. 4B). As before, methylpyruvate could significantly rescue respiration in cell lines withknockdown of either MPC paralog (Fig. 4C), indicating that re-pressed expression did not cause globalmitochondrial dysfunction.If indeed the MPC complex is a target of TZDs, then knock-

down should reduce the EC50 necessary to inhibit pyruvate-drivenrespiration. This was true forMSDC-0160 (Fig. 4D andE) and forUK5099 (Fig. 4E), a crucial positive control. This result stronglysuggests that the MPC complex is a target of TZDs at clinicallyrelevant concentrations.

Partial MPC Inhibition Increases Cellular Glucose Uptake. Both inintact muscle (16) and in vivo (33), TZDs can acutely increaseglucose uptake on a scale that suggests that their enhancedtransport activity might not be entirely explained by transcrip-tional events (≤30 min in skeletal muscle, 120 min in vivo). Todetermine whether mild MPC inhibition can account for this, wemeasured glucose uptake in L6 myotubes (Fig. 5A) and HSkMMs(Fig. 5B). After 90–120 min, both pioglitazone (10 μM; Fig. 5A)and troglitazone (11 μM; Fig. 5B), like insulin, significantly in-creased plasma membrane glucose uptake. To link this effect toMPC inhibition, we also measured uptake in response to UK5099in both cell types (Fig. 5C, blue). The degree to which glucoseuptake was stimulated in either cell type was directly proportionalto the extent of respiratory inhibition by TZDs or UK5099 (Fig.5C). Consistent with previous reports (34–36), treatment of pa-tient myotubes with 10 μM pioglitazone increased phosphoryla-tion of AMP-activated protein kinase (AMPK) (Fig. 5D). Again,this acute effect of TZDs was mimicked with UK5099, linkingchanges in MPC activity with cytoplasmic energy sensing.

DiscussionThis study provides unequivocal evidence that TZDs are acute,specific inhibitors of the mitochondrial pyruvate carrier at clini-cally relevant concentrations. In addition to demonstrating thatthe recently defined MPC is a target of these effective insulinsensitizers, it also provides evidence that acute inhibition of MPCactivity can regulate cellular glucose metabolism. Of course, thesedata do not obviate involvement of PPARγ in the insulin-sensi-tizing effects of TZDs. Rather, we suggest that TZDs workthrough a previously undefined, pleiotropic mechanism in whichboth transcriptional regulation and acute MPC inhibition en-hance the metabolic profile.This work was empowered by the use of rPFO-permeabilized

cells to measure mitochondrial respiration in situ. Cells per-meabilized with rPFO were not subject to the drawbacks associ-ated with detergent-based methods, such as mitochondrial outer

A B

C D

Fig. 3. TZDs specifically inhibit mitochondrial pyruvate transport. (A) Un-coupler-stimulated respiration was measured acutely in permeabilizedC2C12 myoblasts as in (Fig. 2A), but cells were offered either 10 mM gluta-mate + 5 mM malate or 10 mM succinate + 2 μM rotenone. The TZD con-centrations were 3 μM, 10 μM, and 30 μM, and the corresponding UK5099concentrations were 300 nM, 3 μM, and 10 μM (n = 4). (B) HSkMM, NRVMs,and cortical neurons were permeabilized and offered respiratory substrates,after which maximal FCCP-stimulated respiration (oligomycin present) wasmeasured. Either 10 μM MSDC-0160 or 300 nM UK5099 was added 6 minbefore the addition of FCCP. Pyr., 5 mM pyruvate + 0.5 mM malate + 2 mMDCA; Glu., 10 mM glutamate + 5 mM malate; Succ., 10 mM succinate + 2 μMrotenone (n = 4). (C) FCCP-stimulated respiration was measured in HSkMMsand cortical neurons as in B. MSDC-0160 and rosiglitazone, 10 μM; troglita-zone, 5 μM; UK5099, 300 nM. Pyr/Mal, 5 mM pyruvate + 0.5 mMmalate + 2 mMDCA;MePyr + P/M, 20mMmethyl pyruvate + 5mMpyruvate + 0.5mMmalate +2 mM DCA. ‡Significant rescue compared with matched treatment withoutmethyl pyruvate (P < 0.05); ‡‡P < 0.01. (D) Intact C2C12 myoblasts were giveneither 10 μM TZD or 2 μM UK5099. After 90 min, cells were washed andpermeabilized, and then offered pyruvate, malate, DCA, oligomycin, andFCCP as before.

5424 | www.pnas.org/cgi/doi/10.1073/pnas.1303360110 Divakaruni et al.

Dow

nloa

ded

by g

uest

on

June

9, 2

020

membrane damage and cell detachment, possibly attributable tothe threshold cholesterol content required for pore formation(37). Such reliability allowed us to perform a mechanistic analysisdespite a limited sample size in primary cells, and gave us theability to interrogate mitochondrial function in genetically modi-fied cells without isolation-induced mitochondrial damage.Rigorous bioenergetic analysis coupled with genetic suppression

of either obligatory MPC paralog demonstrated that TZDs specif-ically inhibited mitochondrial pyruvate uptake. The effect occurredwithin minutes in permeabilized cells, rendering transcriptionalactivation an unlikely mechanism, and at single-digit micromolarconcentrations that reflect circulating levels in treated patients (21,22). Although previous work documented that TZDs can inhibitcomplex I (23, 24), the concentrations used exceeded physiologicalrelevance; indeed, some TZDs in the present study significantlyinhibited complex I at 30 μM.One preliminary report has suggestedthat TZDs exert substrate-specific effects on respiration of isolatedbrain mitochondria (15), but an in-depth mechanistic analysis wasnot reported. MitoNEET has also been put forth as a direct mito-chondrial target of TZDs (38), but to date, there are no datashowing that altered protein expression can modulate TZD bindingor efficacy.Our work can also further define the explicit function of MPC1

and MPC2. Questions have been raised about whether existingevidence can discriminate between a distinct role for the MPCcomplex in pyruvate transport as opposed to a separate role inoverall pyruvate metabolism (39). Although flux through the PDHcomplex certainly can affect the rate of pyruvate transport, thedemonstration that excess methyl pyruvate can almost entirelyrescue respiration from both MPC inhibition and knockdownindicates that the function of the MPC complex is discrete fromPDH activity. Furthermore, the yeast glycerol-3-phosphate de-hydrogenase was reported to physically associate with the MPCcomplex (5). However, we found that the stable knockdown ofeither paralog had no effect on respiration rates in permeabilizedC2C12 myoblasts oxidizing glycerol-3-phosphate (Fig. 4B).Although it may seem counterintuitive that reduced mitochon-

drial pyruvate uptake may be a mechanism of insulin sensitization,it is important to note that the restriction of pyruvate-drivenrespiration by TZDs is never complete and is readily reversible

(Figs. 2 and 3D). Clinically relevant drug concentrations cannotentirely block pyruvate transport and oxidation, which presumablywould be a toxic effect inconsistent with the clinical utility of TZDs.Moreover, unlike inhibition by UK5099, the reversibility of the in-teraction may induce a conditioning effect whereby mild, transientMPC inhibition potentiates reliance on alternative substrates. Al-though the formal possibility exists that this interaction mediatestoxicity, the demonstration that MPC inhibition by TZDs can acutelyincrease glucose uptake and increase AMPK phosphorylationproves the principle that partial MPC inhibition can acutely im-prove cellular glucose handling.Substantial inhibition of pyruvate entry into the matrix might be

expected to increase lactate levels (40). However, unlike metfor-min treatment, lactic acidosis has not generally been reported asa consequence of chronic TZD administration (41), suggestinga tonic degree of MPC inhibition in treated patients. It is crucial tonote that MPC inhibition still allows the oxidation of amino acids,fatty acids, and other complex I-linked substrates (Fig. 4B andFig. S3).Many of the beneficial effects of TZDs on whole-body metab-

olism may, to some degree, be attributable to MPC inhibition aswell. Restricted mitochondrial pyruvate uptake might suppressflux through pyruvate carboxylase, limiting the fuel available forhepatic gluconeogenesis (42). This mechanism also might helpexplain why TZDs can decrease lipid accumulation in the liverand skeletal muscle (43, 44). MPC inhibition likely would di-minish the pool of intramitochondrial citrate, potentially reducingits efflux and, in turn, lipogenesis. If so, then the associated pro-duction of malonyl CoA would decrease as well. This would re-lieve malonyl CoA-mediated inhibition of carnitine palmitoyltransferase I and accelerate fatty acid oxidation, a characteristic ofskeletal muscle myocytes exposed to chronic TZD treatment (35,45, 46). Furthermore, reduced intramitochondrial pyruvate likelywould enhance amino acid oxidation to maintain tricarboxylicacid cycle activity and ATP production. It also may stimulatemitochondrial malic enzyme activity, producing pyruvate frommalate and hence enhancing NAD(P)H levels.Perhaps the strongest evidence that mild MPC inhibition can

be insulin-sensitizing is the increase in glucose uptake observedin L6 myotubes and HSkMMs. Enhanced glucose transport

A B C

D E

Fig. 4. Knockdown of MPC1 and MPC2 specifically compromises pyruvate oxidation and increases sensitivity to MPC inhibitors. (A) A representative ex-periment in which permeabilized, transduced cells were offered 5 mM pyruvate (with malate, DCA, and oligomycin as before) followed by 400 nM FCCP.UK5099, 300 nM. Where not visible, error bars from technical replicates are obscured by the symbol. OCR, oxygen consumption rate. (B) Uncoupler-stimulatedrespiration was measured in permeabilized, transduced cells offered different oxidizable substrates. Abbreviations and concentrations are as in Fig. 3B. Palm,40 μM palmitoyl carnitine + 0.5 mM malate. β-OH, 10 mM β-hydroxybutyrate + 0.5 mM malate; G-3-P, 10 mM glycerol-3-phosphate + 2 μM rotenone. (C)Uncoupler-stimulated respiration was measured as in A, with abbreviations as in Fig. 3C. ‡Significant rescue relative to matched treatment without methylpyruvate, P < 0.05; ‡‡P < 0.01. (D) Concentration-response curves of pyruvate-driven, uncoupler-stimulated respiration were generated with permeabilized,transduced cells. An aggregate curve of six biological replicates acutely given MSDC-0160 is presented. Substrate and uncoupler concentrations are as in Fig.2A. (E) EC50 values for UK5099 and MSDC-0160 inhibition were calculated for each replicate experiment (n = 6).

Divakaruni et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5425

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

9, 2

020

occurred within 90 min of TZD treatment in patient-derived myo-tubes, and could be reproduced by the MPC inhibitor UK5099.Previous work has in fact reported that 30 μM TZD enhancedthe rate of glucose metabolism in rat cortical astrocytes (47), al-though this concentration can cause respiratory inhibition ofcomplex I. Although others have noted that TZD administrationcan acutely activate AMPK (34–36) and subsequently stimulateglucose uptake through a PPARγ-independent mechanism (16),this report demonstrates that these effects can be reproducedwith UK5099 (Fig. 5C), a specific inhibitor of the MPC (Fig. 3A).MPC inhibition also may trigger signaling via protein acetylationon either side of the mitochondrial inner membrane, given thatimpaired mitochondrial pyruvate uptake would increase theconcentration of pyruvate, and thus of acetyl units, in the cyto-plasm. Acetylation as a posttranslational modification is likelyimportant to the regulation of cell metabolism (48).We propose that mild inhibition of pyruvate transport by TZDs

induces a beneficial, hormetic effect on whole-body metabolism.This mechanism can potentially explain their acute insulin-sensi-tizing effects by initiating a cascade of events including increasedglucose uptake and enhanced oxidation of alternative fuels, such asfatty and amino acids. Modulation of this insulin-independentmechanism, potentially mediated by AMPK, could be of tremen-dous benefit for the treatment of metabolic syndrome and type 2diabetes (49, 50). Moreover, dysregulated glucose metabolismoccurs not only in type 2 diabetes, but also in pathologies, includingcancer, neurodegenerative disease, and heart failure. As such, thedemonstration that pharmacologic modulation of MPC canregulate the pattern of cellular glucose metabolism establishes

an important avenue for drug development centered aroundthis target.

Materials and MethodsAnimals and Human Subjects. All animal protocols were approved by theUniversity of California at San Diego’s Institutional Animal Care and UseCommittee. Human skeletal muscle biopsy specimens were obtained fromsubjects with approval of the University of California at San Diego’s Committeeon Human Investigation. Informed written consent for biopsy was obtainedfrom all subjects after explanation of the protocol. Samples were obtainedfrom those who displayed normal glucose tolerance in response to a standard75-g oral glucose tolerance test and lacked a familial history of type 2 diabetes.

Cell Culture. C2C12 mouse and L6 rat myoblasts were obtained from AmericanTypeCultureCollectionandculturedassuggestedbythesupplier.Myocytes fromhuman skeletal muscle biopsy specimens were prepared as described previously(51). NRVMs were prepared according to published methods (52). After iso-lation, cell suspensions were preplated for 2 h to reduce fibroblast contamina-tion. Murine brown adipocyte tissue precursors were prepared as describedpreviously using CD1mice (21–28d old) (53). Rat cortical neuronswere preparedfrom E18 Sprague–Dawley rats according to published methods (54).

Mitochondrial Isolation, Membrane Potential Measurements, and Cytochrome cRelease.Mitochondria from rat skeletal muscle, rat liver, and C2C12 cells wereisolated by differential centrifugation (55). Rat liver mitochondrial mem-brane potential was monitored with 5 μM safranine O at 495 nm excitation/586 nm emission. Cytochrome c release was measured in supernatants andpellets from incubations of rat liver mitochondria in KCl-based medium, asdescribed in SI Materials and Methods.

Respirometry. Respiration in intact and permeabilized cells was measuredusing a Seahorse XF24 or XF96 Extracellular Flux Analyzer (Seahorse Bio-science). Unless specified otherwise, intact cells were offered 10 mM glucose,10 mM Na+ pyruvate, and 2 mM GlutaMAX (Invitrogen) in unbufferedDMEM (D5030; Sigma-Aldrich), pH 7.4, at 37 °C.

Respirometry with permeabilized cells was conducted in MAS buffer (70mM sucrose, 220 mM mannitol, 10 mM KH2PO4, 5 mM MgCl2, 2 mM Hepes,and 1 mM EGTA; pH 7.2) at 37 °C without BSA unless stated otherwise. rPFO[XF Plasma Membrane Permeabilizer (XF PMP); Seahorse Bioscience] isa recombinant perfringolysin O derivative (PFOC459A) that requires a higherthreshold level of cholesterol than native PFO (37), optimal for selectiveplasma membrane permeabilization. rPFO was added at 1 nM to selectivelypermeabilize the plasma membrane. The ATP synthase inhibitor oligomycin(2 μM) and oxidizable substrates were provided as indicated.

Lentiviral shRNA Knockdown, RT-PCR, and Western Blot Analysis. Cells withstable repression of MPC1 and MPC2 were generated with MISSION lentiviralshRNA plasmids under puromycin selection. Cells were lysed, and mRNA wasextracted using the RNeasy Kit (Qiagen) with on-column DNase digestion(Qiagen). Mitochondrial protein and cell lysates were solubilized and run ona Laemmli gel, transferred to PVDF, and analyzed by immunoblotting forMPC1, MPC2, cytochrome c, AMPK, and phosphorylated AMPK (pAMPK).

Glucose Uptake. Glucose uptake was measured as described previously (56).Differentiated L6 or patient derived myocytes were washed in serum-freemedium and incubated ± insulin (32 nM) or drug (10 μM TZD or 2 μMUK5099) for 90 min at 37 °C in a 5% (vol/vol) CO2 incubator. Glucose uptake,quantified using the nonmetabolized radiolabeled analog 2-deoxyglucose(10 μM final concentration), was measured in triplicate over 10 min at roomtemperature. Data were normalized to the protein content in each well. Theuptake of labeled L-glucose was used to correct samples for the nonspecificdiffusion of tracer.

Statistics. Statistical analysis and curve fitting were conducted usingGraphPad Prism. Significance was assessed by ANOVA for repeated measureswith Dunnett’s posttest (95% confidence interval). When data were ex-pressed as a percentage of control values, significance was calculated on thesquare root of the normalized data. A P value < 0.05 (*) was consideredstatistically significant (**P < 0.01; ***P < 0.001). Data are presented asmean ± SEM.

Note Added in Proof. While this report was in press, the observation thatinitiated this study, demonstrating that thiazolidinediones can directly binda protein complex containing MPC2, was accepted for publication (57).

A B

C D

Fig. 5. Mild MPC inhibition increases plasma membrane glucose uptake andactivates AMPK. (A) Glucose uptake was measured in L6 myotubes after a 2-htreatment with either 32 nM insulin or 10 μMpioglitazone (n = 6). (B) Glucoseuptake was measured in HSkMMs after a 90-min treatment with either 32 nMinsulin or 11 μM troglitazone (n = 12). (C) The increase in the rate of glucoseuptake is plotted against the degree of respiratory inhibition for eachtreatment. Respiration was measured in intact L6 myotubes (circles) andHSkMMs (triangles) as in Fig. 2E. Gray, basal (no drug added); blue, 2 μMUK5099; red, 10 μM pioglitazone; purple, 11 μM troglitazone; broken line,regression analysis (r2 = 0.92). (D) HSkMMs were treated for 2 h with either10 μMpioglitazone or 2 μMUK5099 as in Fig. 5A, and prepared forWestern blotanalysis. (Upper) Sample immunoblots of pAMPK and AMPK from matchedsamples. (Lower) Densitometry analysis from four different patient samples.

5426 | www.pnas.org/cgi/doi/10.1073/pnas.1303360110 Divakaruni et al.

Dow

nloa

ded

by g

uest

on

June

9, 2

020

ACKNOWLEDGMENTS. We thank the laboratory of Dr. Joan Heller Brown(Department of Pharmacology, University of California at San Diego) forproviding isolated NRVMs (Grant P01HL085577), and Dr. Morton P. Printz (De-partment of Pharmacology, University of California at San Diego) for helpfuldiscussions of our work. This work was supported by the National Institutes ofHealth (Grant R42DK081298); the American Diabetes Association (Grant 1-08-

RA-139); Seahorse Bioscience (A.N.M.); Center for Excellence in Apoptosis Re-search translational funds from Massachusetts Technology Collaborative[Grant A00000000004448 (to N.Y. and A.P.H.)]; National Institutes of HealthGrant R24DK092154, Defense Security Grant 7-05-DCSA-04, the Department ofVeterans Affairs Medical Research Service (to R.R.H.); and the Ellison MedicalFoundation [Grant AG-SS-2190-08 (to M.I.S.)].

1. Divakaruni AS, Murphy AN (2012) Cell biology: A mitochondrial mystery, solved.Science 337(6090):41–43.

2. Papa S, Francavilla A, Paradies G, Meduri B (1971) The transport of pyruvate in ratliver mitochondria. FEBS Lett 12(5):285–288.

3. Halestrap AP, Denton RM (1974) Specific inhibition of pyruvate transport in rat livermitochondria and human erythrocytes by alpha-cyano-4-hydroxycinnamate. BiochemJ 138(2):313–316.

4. Herzig S, et al. (2012) Identification and functional expression of the mitochondrialpyruvate carrier. Science 337(6090):93–96.

5. Bricker DK, et al. (2012) A mitochondrial pyruvate carrier required for pyruvate up-take in yeast, Drosophila, and humans. Science 337(6090):96–100.

6. DeFronzo RA, Abdul-Ghani MA (2011) Preservation of β-cell function: The key to di-abetes prevention. J Clin Endocrinol Metab 96(8):2354–2366.

7. Cunnane S, et al. (2011) Brain fuel metabolism, aging, and Alzheimer’s disease. Nu-trition 27(1):3–20.

8. Colmers IN, Bowker SL, Johnson JA (2012) Thiazolidinedione use and cancer incidencein type 2 diabetes: A systematic review and meta-analysis. Diabetes Metab 38(6):475–484.

9. Miller BW, Willett KC, Desilets AR (2011) Rosiglitazone and pioglitazone for thetreatment of Alzheimer’s disease. Ann Pharmacother 45(11):1416–1424.

10. Colca JR, Kletzien RF (2006) What has prevented the expansion of insulin sensitisers?Expert Opin Investig Drugs 15(3):205–210.

11. Graham DJ, et al. (2010) Risk of acute myocardial infarction, stroke, heart failure, anddeath in elderly Medicare patients treated with rosiglitazone or pioglitazone. JAMA304(4):411–418.

12. Lehmann JM, et al. (1995) An antidiabetic thiazolidinedione is a high-affinity ligandfor peroxisome proliferator-activated receptor gamma (PPAR gamma). J Biol Chem270(22):12953–12956.

13. Choi JH, et al. (2010) Anti-diabetic drugs inhibit obesity-linked phosphorylation ofPPARgamma by Cdk5. Nature 466(7305):451–456.

14. Norris AW, et al. (2003) Muscle-specific PPARgamma-deficient mice develop increasedadiposity and insulin resistance but respond to thiazolidinediones. J Clin Invest 112(4):608–618.

15. Feinstein DL, et al. (2005) Receptor-independent actions of PPAR thiazolidinedioneagonists: Is mitochondrial function the key? Biochem Pharmacol 70(2):177–188.

16. LeBrasseur NK, et al. (2006) Thiazolidinediones can rapidly activate AMP-activatedprotein kinase in mammalian tissues. Am J Physiol Endocrinol Metab 291(1):E175–E181.

17. Wei S, Kulp SK, Chen CS (2010) Energy restriction as an antitumor target of thiazo-lidinediones. J Biol Chem 285(13):9780–9791.

18. Birnbaum Y, Long B, Qian J, Perez-Polo JR, Ye Y (2011) Pioglitazone limits myocardialinfarct size, activates Akt, and upregulates cPLA2 and COX-2 in a PPAR-γ-independentmanner. Basic Res Cardiol 106(3):431–446.

19. Thal SC, et al. (2011) Pioglitazone reduces secondary brain damage after experi-mental brain trauma by PPAR-γ-independent mechanisms. J Neurotrauma 28(6):983–993.

20. Chen Z, et al. (2012) Insulin resistance and metabolic derangements in obese mice areameliorated by a novel peroxisome proliferator-activated receptor γ-sparing thiazo-lidinedione. J Biol Chem 287(28):23537–23548.

21. Eckland DA, Danhof M (2000) Clinical pharmacokinetics of pioglitazone. Exp ClinEndocrinol Diabetes 108(Suppl 2):S234–S242.

22. Kirchheiner J, et al. (2006) Pharmacokinetics and pharmacodynamics of rosiglitazonein relation to CYP2C8 genotype. Clin Pharmacol Ther 80(6):657–667.

23. Brunmair B, et al. (2004) Thiazolidinediones, like metformin, inhibit respiratorycomplex I: A common mechanism contributing to their antidiabetic actions? Diabetes53(4):1052–1059.

24. Nadanaciva S, Dykens JA, Bernal A, Capaldi RA, Will Y (2007) Mitochondrial impair-ment by PPAR agonists and statins identified via immunocaptured OXPHOS complexactivities and respiration. Toxicol Appl Pharmacol 223(3):277–287.

25. Schulz I (1990) Permeabilizing cells: Some methods and applications for the study ofintracellular processes. Methods Enzymol 192:280–300.

26. Ramachandran R, Heuck AP, Tweten RK, Johnson AE (2002) Structural insights intothe membrane-anchoring mechanism of a cholesterol-dependent cytolysin. Nat StructBiol 9(11):823–827.

27. Heuck AP, Moe PC, Johnson BB (2010) The cholesterol-dependent cytolysin family ofgram-positive bacterial toxins. Subcell Biochem 51:551–577.

28. Sanyal S, Claessen JH, Ploegh HL (2012) A viral deubiquitylating enzyme restoresdislocation of substrates from the endoplasmic reticulum (ER) in semi-intact cells. JBiol Chem 287(28):23594–23603.

29. Halestrap AP (1975) The mitochondrial pyruvate carrier: Kinetics and specificity forsubstrates and inhibitors. Biochem J 148(1):85–96.

30. Watkins PB, Whitcomb RW (1998) Hepatic dysfunction associated with troglitazone. NEngl J Med 338(13):916–917.

31. Colca JR, et al. (2013) Clinical proof of concept with MSDC-0160, a prototype mTOTmodulating insulin sensitizer. J Clin Pharm Ther 93: in press.

32. Hildyard JC, Ammälä C, Dukes ID, Thomson SA, Halestrap AP (2005) Identification andcharacterisation of a new class of highly specific and potent inhibitors of the mito-chondrial pyruvate carrier. Biochim Biophys Acta 1707(2-3):221–230.

33. Lee MK, Olefsky JM (1995) Acute effects of troglitazone on in vivo insulin action innormal rats. Metabolism 44(9):1166–1169.

34. Fryer LG, Parbu-Patel A, Carling D (2002) The anti-diabetic drugs rosiglitazone andmetformin stimulate AMP-activated protein kinase through distinct signaling path-ways. J Biol Chem 277(28):25226–25232.

35. Coletta DK, et al. (2009) Pioglitazone stimulates AMP-activated protein kinase sig-nalling and increases the expression of genes involved in adiponectin signalling,mitochondrial function and fat oxidation in human skeletal muscle in vivo: A rand-omised trial. Diabetologia 52(4):723–732.

36. Hawley SA, et al. (2010) Use of cells expressing gamma subunit variants to identifydiverse mechanisms of AMPK activation. Cell Metab 11(6):554–565.

37. Moe PC, Heuck AP (2010) Phospholipid hydrolysis caused by Clostridium perfringensα-toxin facilitates the targeting of perfringolysin O to membrane bilayers. Bio-chemistry 49(44):9498–9507.

38. Colca JR, et al. (2004) Identification of a novel mitochondrial protein (“mitoNEET”)cross-linked specifically by a thiazolidinedione photoprobe. Am J Physiol EndocrinolMetab 286(2):E252–E260.

39. Halestrap AP (2012) The mitochondrial pyruvate carrier: Has it been unearthed atlast? Cell Metab 16(2):141–143.

40. Brivet M, et al. (2003) Impaired mitochondrial pyruvate importation in a patient anda fetus at risk. Mol Genet Metab 78(3):186–192.

41. Stang M, Wysowski DK, Butler-Jones D (1999) Incidence of lactic acidosis in metforminusers. Diabetes Care 22(6):925–927.

42. Natali A, Ferrannini E (2006) Effects of metformin and thiazolidinediones on sup-pression of hepatic glucose production and stimulation of glucose uptake in type 2diabetes: A systematic review. Diabetologia 49(3):434–441.

43. Bajaj M, et al. (2010) Effects of pioglitazone on intramyocellular fat metabolism inpatients with type 2 diabetes mellitus. J Clin Endocrinol Metab 95(4):1916–1923.

44. Teranishi T, et al. (2007) Effects of pioglitazone and metformin on intracellular lipidcontent in liver and skeletal muscle of individuals with type 2 diabetes mellitus.Metabolism 56(10):1418–1424.

45. Bandyopadhyay GK, Yu JG, Ofrecio J, Olefsky JM (2006) Increased malonyl-CoA levelsin muscle from obese and type 2 diabetic subjects lead to decreased fatty acid oxi-dation and increased lipogenesis; thiazolidinedione treatment reverses these defects.Diabetes 55(8):2277–2285.

46. Cha BS, et al. (2005) Impaired fatty acid metabolism in type 2 diabetic skeletal musclecells is reversed by PPARgamma agonists. Am J Physiol Endocrinol Metab 289(1):E151–E159.

47. Dello Russo C, et al. (2003) Peroxisome proliferator-activated receptor gamma thia-zolidinedione agonists increase glucose metabolism in astrocytes. J Biol Chem 278(8):5828–5836.

48. Guan KL, Xiong Y (2011) Regulation of intermediary metabolism by protein acety-lation. Trends Biochem Sci 36(2):108–116.

49. Ruderman N, Prentki M (2004) AMP kinase and malonyl-CoA: Targets for therapy ofthe metabolic syndrome. Nat Rev Drug Discov 3(4):340–351.

50. Friedrichsen M, Mortensen B, Pehmøller C, Birk JB, Wojtaszewski JF (2013) Exercise-induced AMPK activity in skeletal muscle: Role in glucose uptake and insulin sensi-tivity. Mol Cell Endocrinol 366(2):204–214.

51. Henry RR, Abrams L, Nikoulina S, Ciaraldi TP (1995) Insulin action and glucose me-tabolism in nondiabetic control and NIDDM subjects: Comparison using humanskeletal muscle cell cultures. Diabetes 44(8):936–946.

52. Rubio M, et al. (2009) Cardioprotective stimuli mediate phosphoinositide 3-kinase andphosphoinositide dependent kinase 1 nuclear accumulation in cardiomyocytes. J MolCell Cardiol 47(1):96–103.

53. Cannon B, Nedergaard J (2001) Respiratory and thermogenic capacities of cells andmitochondria from brown and white adipose tissue. Methods Mol Biol 155:295–303.

54. Kushnareva YE, Wiley SE, Ward MW, Andreyev AY, Murphy AN (2005) Excitotoxicinjury to mitochondria isolated from cultured neurons. J Biol Chem 280(32):28894–28902.

55. Chappell JB, Hansford RG (1972) Preparation of mitochondria from animal tissues andyeasts. Subcellular Components: Preparation and Fractionation, ed Birnie GD (But-terworths, London), pp 77–91.

56. Ciaraldi TP, Abrams L, Nikoulina S, Mudaliar S, Henry RR (1995) Glucose transport incultured human skeletal muscle cells: Regulation by insulin and glucose in non-diabetic and non-insulin-dependent diabetes mellitus subjects. J Clin Invest 96(6):2820–2827.

57. Colca JR, et al. (2013) Identification of a mitochondrial target of thiazolidinedioneinsulin sensitizers (mTOT) – relationship to newly identified mitochondrial pyruvatecarrier proteins. PLoS One, in press.

Divakaruni et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5427

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

9, 2

020